US 5,596,504 AGrant
Apparatus and Method for Layered Modeling of Intended Objects Represented in STL Format and Adaptive Slicing Thereof
Issue Date:1997-01-21
•22 Claims
•20 Drawing Sheets
Abstract
A device for automating operation of a stereolithography apparatus uses an STL file as an input and includes a programmable computer, a facet processor that sorts the facets of the STL file according to a predetermined slice axis. The facet processor also groups the sorted facets according to those having common minimum vertex values with respect to the slice axis. The facet processor also subgroups the grouped facet file according to facets having common maximum vertex values with respect to the slice axis. A key characteristic identifier identifies key characteristics of the STL file. A thickness calculator determines the thickness of each layer of the model according to a geometrical error of preselected magnitude. A slicer calculates the intersection of each sliced plane by the calculated thickness. A directional ordering device insures uniformity with the direction of each other contour that defines the intersection. A model generator uses the layer thickness and intersection information to generate a portion of a model. An interface device controls the operation of the machine based on the model that is generated.
Metadata
Assignee
- Clemson University
Inventors
- Kamesh Tata
- Amit Bagchi
- Nadim M. Aziz
Application Information
Application Number:US 4197119
Filing Date:1995-04-10
Priority Date:1995-04-10
Art Unit:236
Classifications
IPC:
G06F 1900G06F 1750G06T 1710
Field of Search:
364395264468;474.05;474.24;476;468.25;468.26;468.27;468.04119;120;123;155;161401
Patent Drawings (20 sheets)
Description
Background of the Invention
The present invention relates to a method of modeling an intended object's three-dimensional surface and more particularly to a method of generating and employing a two-dimensional slice from a cross-section of a three-dimensional tesselated model of the intended object's three-dimensional surface.
The advent of Solid Freeform Fabrication (SFF) technologies (a.k.a. Rapid Prototyping technologies) has reduced the cycle time for prototyping. Typically, an SFF process starts with a solid model of an intended object, such as a part to be machined for example, in a Computer-Aided Design (CAD) system. The model of the intended object in the CAD system is then converted into a computer file in the stereolithographic (STL) format. The STL file defines the three-dimensional exterior surface of the intended object in a tesselated form as: (1) a plurality of flat planar sections known as facets and (2) the outward pointing normals to these planar sections.
In general, a "facet model" represents a three-dimensional surface of an intended object by spatial boundaries, which are defined by a set of planar faces. The facet model is a special case of the more general boundary representation, which does not require the boundaries of the object to be planar. The term "facet" can be considered to denote a planar region being used to define a model boundary wherein the outline of the planar region is defined by any polygonal shape. However, in the Solid Freeform Fabrication (SFF) community, the term "facet" is typically understood to have an outline shaped as a triangle, and that is the understanding followed in the present application.
In a typical Rapid Prototyping (RP) application for an STL file, the faceted representation of the intended object is processed by a so-called "slicing engine," which produces a two dimensional representation of the cross-section of the intended object. The slicing engine provides the two-dimensional representation in the form of 2-D layers (a.k.a. slices). Each layer is bounded top and bottom by a so-called "slicing plane," and the edge of the layer disposed between the top and bottom slicing planes represents to a varying degree of accuracy, the commensurate cross-section of the surface of the intended object. The intended object is then "built" by stacking these layers one on top of the other layer-by-layer from bottom to top. Some slicing engines have difficulty dealing with certain key characteristics in the surface of the intended object and sudden changes in the complexity of the surface of the intended object. Moreover, providing the slicing engine with a data file in the Stereolithographic format (an STL file) as the input, negatively impacts the performance of the slicing engine in rapid prototyping systems.
As shown in FIGS. 1A-1C, a typical CAD software package (e.g., 3-D Systems, 1988) models a solid object, such as a sphere, as a set of triangular facets. Unfortunately, the typical CAD package independently stores the data representing each of these facets, as if each facet were created and tossed into a bucket with no particular ordering and without information relating a given facet to any other facet in the bucket.
In stereolithography, about 60% of the total time spent preparing the model is devoted to generating the slice data, i.e., slicing. Moreover, far more of the time devoted to slicing is consumed by searching for the facets that will have an intersection with a particular slice plane, than is consumed in performing the plane-facet intersection calculations required to generate the slice data.
To reduce the time required to generate the slice data, sorting of triangles is recommended by 3-D Systems. However, this does not fully alleviate the problem because triangle sorting is uni-directional, either descending or ascending. The benefits of sorting are often diminished because of the sheer complexity of the model and because there are too many variations in the sizes of triangles.
Moreover, under certain circumstances such as shown in an E-shaped block in FIG. 2, sorting may even increase the computation time. Ignoring the front and rear faces of the block for the sake of convenience, each of the 12 faces F.sub.1 through F.sub.12 of the block will be represented in the STL format by at least two triangular facets. When the faces of the block are sorted in ascending order, with the minimum Z coordinate (Z-min) of each facet F.sub.i as the sorting criterion, faces of the block will be ordered in the following First Sequence: F.sub.1, F.sub.2, F.sub.3, F.sub.4, F.sub.5, F.sub.6, F.sub.7, F.sub.8, F.sub.9, F.sub.10, F.sub.11, F.sub.12. In this FIG. 2 example, only faces F.sub.2 and F.sub.11 intersect the slice plane A--A. Since these two faces are separated by eight other faces (assuming that each face is composed of two facets in the STL format), sixteen failed searches will occur before the correct facets that intersect the slice plane A--A can be found. If the sorting criterion is changed to the maximum Z coordinate (Z-max) of each facet, then this Second Sequence will be given by: F.sub.1, F.sub.3, F.sub.4, F.sub.5, F.sub.6, F.sub.7, F.sub.8, F.sub.9, F.sub.10, F.sub.11, F.sub.2, F.sub.12. The Second Sequence is the same as the First Sequence except for the new position of F.sub.2. Since slice plane B--B intersects only facets F.sub.2 and F.sub.3 in the FIG. 2 example, there will be again sixteen failed searches. As is evident from this example, as the complexity of the STL model increases, the computational cost of searching for facets that intersect the slice plane, quickly becomes prohibitive.
In addition, STL files contain a significant amount of redundant information, which can waste processing time. For example, vertex coordinates are stated explicitly for each facet. Since a vertex is always shared by a minimum of three facets, this results in each vertex being stored in the STL file data at least three times.
As shown in FIGS. 1A-1C, another of the limitations of an STL file is its inability to represent curved surfaces accurately. This results from the unavoidable error that arises whenever a three-dimensional surface of an object featuring curved surfaces is approximated by planar triangular facets. As illustrated in FIG. 1, choosing smaller and smaller triangles can reduce the error, but cannot eliminate the error.
In a 1991 article, Chalasani et al, have proposed two approaches for layered modeling of 3-D objects. In the first approach, each of the slice planes defining the layers is determined as the geometric intersection between the intended object and a plane horizontal to the platform of a stereolithography apparatus. Each facet that intersects the slice plane will form a directed line segment (vector) on the slice plane. Since intersections are not found in any particular order, the vectors must be sorted in head-to-tail fashion in order to derive a closed contour on that slice plane. This makes the process slow and inefficient.
Chalasani et al's second approach uses a scan-line-type search to find an intersection between the slice plane and a facet, and the search starts at the global origin. When the search point reaches the boundary of the faceted intended object for the first time since the search began, the first intersection is recorded. This first intersection will trigger a contour-tracing procedure in which the "search point becomes a draw point" and follows the plane of the facet until it reaches another edge. An adjacency list is consulted for the facet that shares this second edge of the first facet. Then the draw point will move along the plane of this facet. The process is repeated until the draw point comes back to the first intersection.
However, because Chalasani et al's methods produce many failed searches before an intersection (match) is found, it is slow. It is not clear how Chalasani et al's search point is incremented. Also, apart from stating that the layer thickness should be varied according to the local curvature of the object, Chalasani et al fails to demonstrate or suggest how this can be achieved.
In a December 1992 article, Vouzelaud and Bagchi reported their use of slices to obtain 2-D contours from a 3-D CAD model in a way that varies the slice thickness to better represent the surface of the intended object. A user defined "Quality Index," , which is the maximum peak to valley height of the profile in the assessment length, was used by Loney and Ozsoy (1987) to calculate the side step or step size for parametric cubic patches and therefore yield a constant surface attribute. However, by obtaining the 2-D contours from the CAD systems themselves, Vouzelaud and Bagchi eliminated the need to have an STL file and the limitations associated with the STL file. Also, unlike the contours obtained from the tesselated model represented by the STL file, the contours obtained from the CAD system are more accurate representations of the original model. Despite their limitations, STL files are widely used by the Rapid Prototyping industry because of the easy availability of STL translators and the simplicity of the STL format.
In a 1993 article, Dolenc and Makela suggested implementing back-tracking as part of a procedure for varying the thickness of the layers used to represent different portions of the surface of the intended object. However, the details of their implementation are not known, and the robustness of their back-tracking procedure is suspect.
Objects and Summary of the Invention
It is a principal aim of the present invention to reduce the time needed to employ an STL file in any process employing a profile or cross-section of the boundary of an intended object wherein such profile is decomposed by incremental parallel planes of intersection.
It is another principal aim of the present invention to increase the accuracy in any process employing an STL file that must be decomposed by incremental parallel planes of intersection.
It is a further principal aim of the present invention to reduce the time needed to employ an STL file in any process employing a profile or cross-section of the boundary of an intended object wherein the profile is decomposed by incremental parallel planes of intersection without losing key characteristics in the surface of the intended object.
It is still another principal aim of the present invention to employ an STL file in any process employing a profile or cross-section of the boundary an intended object wherein the profile includes multiple shapes that are simultaneously decomposed by incremental parallel planes of intersection separated by thicknesses determined by the most complex shape of the profile of the boundary of the intended object.
It is an additional principal aim of the present invention to reduce the time needed to employ an STL file in any process employing a profile or cross-section of the boundary of an intended object wherein though the profile includes sudden changes in the complexity of the boundary of the intended object, the profile is decomposed by incremental parallel planes of intersection without losing accuracy in modeling the boundary of the intended object.
It is yet another principal aim of the present invention to provide a method of modeling the boundary of an intended object from an STL file that must be decomposed into a plurality of layers stacked atop one another wherein the thickness of the layers is uniform except in portions of the model where the thickness of each layer is adapted by being increased or decreased by an amount that depends on the knowledge of the local geometry and the constraint of keeping the geometrical error between the object and the model from exceeding a predetermined value.
It is a further principal aim of the present invention to reduce the time needed to employ an STL file in any process employing a profile or cross-section of the boundary of an intended object wherein the profile is decomposed by incremental parallel planes of intersection separated by thicknesses falling within a range of thicknesses between a maximum and minimum which can be selected by the operator.
It is another principal aim of the present invention to employ an STL file in any process employing a profile or cross-section that is decomposed by incremental parallel planes of intersection to produce a model that can be viewed by the operator either before or after being subjected by the operator to various transformations such as scaling and rotation.
Additional aims and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The aims and advantages of the invention may be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims.
To achieve the aims and in accordance with the purpose of the invention, as embodied and broadly described herein, an apparatus and method is provided for effecting automatic operation of a machine with respect to an intended object having a desired profile characterized by at least one key characteristic and at least two local complexity levels, wherein the operation of the machine is controlled based on a two-dimensional sliced model of the local profile of the cross-section of the intended object with each slice plane disposed transversely with respect to and along a slice axis of the intended object, which is represented by a tesselated model of the intended object and provided as an input to the apparatus in the form of an STL file specifying triangular facets obeying the vertex-to-vertex rule. The present invention can include a slicing engine, either alone or in combination with a model generating means and an interface means as well as a variety of machines being controlled with respect to a boundary of an intended object.
In accordance with the present invention, the slicing engine of the present invention can include four main constituents: (1) a facet processor means, (2) a key characteristic identifier (KCI) means, (3) a thickness calculator means, and (4) a slicer means.
The facet processor means can include a facet sorting means, a facet grouping means and in a preferred embodiment, a facet subgrouping means. The facet sorting means is configured to produce a sorted facet file by sorting the facets of the STL file according to the slice axis. The facet grouping means is configured to produce a grouped facet file by grouping the facets of the sorted facet file according to facets having one of the following characteristics: common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis. The facet subgrouping means is configured to produce a subgrouped facet file by subgrouping the facets of the grouped facet file according to facets having the other of the following characteristics: common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis.
The key characteristic identifier (KCI) means is provided to ensure retention of the intended object's asymmetries and to deal more effectively with the complexities of the surface of the intended object. The key characteristic identifier means is configured to identify key characteristics of the profile represented by the STL file and to produce a key characteristic data file from the key characteristics of the profile represented by the STL file.
The thickness calculator means computes the thickness of the slice (a.k.a. layer) between adjacent slice planes according to a user defined criterion relating to the complexity in the shape of the boundary defining the model. Examples of the user defined criterion include the volumetric error per unit of perimeter, the maximum deviation, the chord length, and the cusp height. The thickness calculator means can include a simple back tracking means for determining the thickness of the next layer while retaining a single change in complexity level in the local profile. The thickness calculator means can include a repeated back tracking means for determining the thickness of the next layer while retaining more than a single change in the complexity level of the local profile.
The user supplies to the slicing engine the maximum and minimum values of the layer thickness. The thickness calculator means of the present invention is configured so that it automatically sets the layer thickness to this maximum value when warranted by the complexity of the shape of the boundary of the intended object. For a typical intended object, this results in a shorter build time with the same level of accuracy as would be obtained were the layer thickness set at the minimum value dictated by the application. The thickness calculator means of the present invention varies the layer thickness between the maximum and minimum values in relation to the complexity of the shape of the boundary of the intended object. For a typical intended object, this results in improved accuracy for the same build time as would be obtained were the layer thickness set at a uniform arbitrary value for the entire intended object.
The thickness calculator means employs the following technique to vary the thickness of each layer. The thickness calculator means chooses as the initial slice plane the slice plane that passes through the bottom-most point along the slice axis. To calculate the height along the slice axis of the next slice plane above the initial slice plane, the thickness calculator means pursues the following strategy. First, the thickness calculator means determines the controlling normal by considering all the facets that have at least one vertex through which the initial slice plane passes. The grouping of facets done earlier by the facet processor means greatly facilitates this process. However, all facets which lie in the initial slice plane are ignored. Once the controlling normal has been identified, the thickness calculator means calculates the layer thickness by applying a constraint equation that relates the geometrical error between the desired profile of the boundary of the intended object and the model profile of each layer having such layer thickness, and the height (Z.sub.tmp) of the next Possible Slice Plane is then computed by applying this constraint equation.
As noted above, the geometrical error can be expressed in the constraint equation using a number of alternative parameters, including: (1) cusp height, (2) maximum deviation, (3) chord length, and (4) volumetric error per unit of perimeter length. Expressing the geometrical error in terms of volumetric error per unit of perimeter length is particularly useful in applications concerned with minimizing the volume or mass of manufactured parts used in space vehicles. Other expressions of geometrical error can be devised to suit the particular needs of the application.
It is important to note that the Possible Slice Plane (at height Z.sub.tmp) need not become the next slice plane (at height Z.sub.next). The Possible Slice Plane becomes the next slice plane only in the absence of a new group (or groups) of facets between the current slice plane and the Possible Slice Plane. However, when the thickness calculator means encounters a new group of facets in this location, the thickness calculator means is configured to calculate the new slice thickness by obtaining the controlling normal from the new group of facets. If the "sum" of the Z-min of the new group and the layer thickness is less than the height of the Possible Slice Plane, then the thickness calculator means is configured to use the value of this "sum" as the height of the Possible Slice Plane. Then this process is repeated until either: (1) no new group of facets lies between the current slice plane and the Possible Slice Plane or (2) the "sum" is not less than the height of the Possible Slice Plane. At the end of this calculation loop performed by the thickness calculator means, the height of the Possible Slice Plane becomes the height of the Next Slice Plane, and the thickness calculator means is configured to place this information into a layer thickness data file to be used by the slicer means for computing the points of intersection for such Next Slice Plane.
Because of the configuration of the thickness calculator means of the present invention, two important observations can be made. First, if the controlling normal of a facet is perpendicular to the slice plane (indicating that the facet is parallel to the slice plane), then the thickness of the layer becomes zero and the slice plane is forced to passes through all three vertices of the facet. This result ensures that important features like horizontal walls (overhangs) are never missed. Moreover, this remains true regardless of the maximum layer thickness defined by the user. Second, if the controlling normal is parallel to the slice plane, then the thickness of the layer is set to the maximum value defined by user. This ensures that wherever there is a vertical wall, the thickness of the layer is set to a maximum value, thus reducing the number of layers. Moreover, multiple objects can be sliced together. In such cases, the thickness calculator means considers the most complex shape encountered for each slice plane when selecting the thickness of the next layer.
The slicer means generates two dimensional contours of the intended model. The slicer means is configured to operate on the sorted facet file, the grouped facet file, and the layer thickness data file to produce a slice plane intersection data file by calculating the intersection of each slice plane disposed at a height above the previous slice plane by the thickness calculated for each layer of the model of the intended profile represented by the STL file.
The slicer means desirably is configured to include a marching means that orders at least the grouped facet file (and desirably also any subgrouped facet file that is generated by the facet processor means) to produce a geometrically ordered facet file in the form of a continuous chain of facets wherein each facet in this continuous chain intersects one of the slice planes. Moreover, the marching means is configured to generate an adjacency list for each facet that is intersected by the one slice plane.
Once the thickness calculator means determines the height of the slice plane, the slicer means computes the points of intersection between the slice plane and the facets defining the model of the intended object. In accordance with the present invention, the slicer means optimizes the speed of this computation of the intersection. The slicer means is configured so that initially it forms a new group of facets by merging (if necessary) relevant existing groups. A relevant group is one which contains facets that potentially intersect the slice plane. The slicer means is configured so that the slicer means tests each facet in the new group for intersection with the slice plane as follows.
In the case involving the intersection of the slice plane through a vertex of a facet, any facet or vertex intersecting with the slice plane is deemed "active." The slicer means records the intersection point (which is the co-ordinates of the vertex), and the "active facet" containing this "active vertex" is identified and stored so that it will not be checked again on this layer. For the next intersection, the slicer means checks the first facet encountered that shares the "active vertex" for intersection. This first-encountered facet must have one other vertex or an edge through which the slice plane passes. Otherwise, this first-encountered vertex is recorded and discarded and will not be checked by the slicer means for an intersection with the current slice plane again. The slicer means is configured to repeat this process until the facet array is exhausted. This logic works for the tip of a cone or any other similar geometry where a number of facets share the same vertex.
In the case involving the intersection of the slice plane through an edge of a facet, the slicer means deems any facet or edge intersecting with the slice plane to be "active." If the slice plane is passing through an edge of a facet ("active edge" and "active facet"), the slicer means records the intersection point. The slicer means also records the "active facet" so that the active facet will not be checked again for an intersection with the current slice plane. There will be only one facet which shares the "active edge" and the slicer means identifies this facet by a simple comparison. This facet sharing the "active edge" will have at least one more edge or vertex through which the current slice plane passes. The slicer means computes the new intersection and repeates the process until all facets in the new group are exhausted. If there are no more facets which share the "active vertex" or "active edge" and still the facet array is not exhausted, this is a clear indication that there is a protrusion or a depression. In such case, the slicer means restarts this checking procedure for the facets left over in the facet array. This process is repeated until the facet array is exhausted.
The present invention also can include a model generating means for using the layer thickness data file and the slice plane intersection data file to generate at least a portion of a model of the intended object. This portion of the model of the intended object includes a plurality of successive layers. The cross-section of each layer in a plane of view is defined by the intersection with this plane of view, of a pair of parallel planes and a model profile connecting these parallel planes. For each layer, the minimum distance separating its pair of parallel planes defines the thickness of this layer.
The present invention also can include an interface means for providing the generated portion of the model of the intended object to operate the machine in successive steps with respect to at least a corresponding portion of the intended object. Each of the successive steps is based on a separate one of the layers of the generated portion of the model of the intended object.
In accordance with the method of the present invention, the steps of the method include the permitted sequences of functions performed by the programmable computer, the facet processor means, the facet sorting means, the facet grouping means, the facet subgrouping means, the key characteristic identifier means, the thickness calculator means, the simple back tracking means, the repeated back tracking means, the slicer means, the marching means, the model generating means, and the interface means.
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate several embodiments of the invention and, together with the description, serve to explain the principles of the invention.
Brief Description of the Drawings
FIGS. 1A-1C are an elevated perspective view of three prior art tesselated representations of a sphere, wherein FIG. 1(a) employs 67 facets, FIG. 1(b) employs 450 facets, and FIG. 1(c) employs 1450 facets;
FIG. 2 is an elevated perspective view of an E-shaped block wherein each face has been assigned a separate designation, and two intersecting slice planes A--A and B--B are shown perpendicular to the vertical direction indicated by the arrow labeled "Z";
FIG. 3 is an elevated perspective view of a preferred embodiment of apparatus in accordance with the present invention as applied to a rapid prototyping process using a stereolithography apparatus;
FIG. 4 is a schematic representation of the slicing engine of the present invention used in the apparatus of FIG. 3 for operating on an STL file to produce a sliced two-dimensional representation that provides both a data file and can be used for generating a visual display of the 2-D representation of the intended object;
FIG. 5 is an elevated perspective view of a block with a square hole wherein four outward normals (arrows 70) and four inward normals (arrows 71) are shown;
FIG. 6(a) is a schematic representation of facets failing to satisfy the vertex-to-vertex rule, and FIG. 6(b) is a schematic representation satisfying the vertex-to-vertex rule;
FIG. 7 is an elevated perspective view of a tesselated representation of a sphere intersected with a slice plane near the bottom of the sphere with the Z direction indicated as the vertical direction perpendicular to the slice plane;
FIG. 8 is a schematic representation illustrating the relationship between facet groups, facet sub-groups and individual facets constituting the STL file;
FIG. 9 is an elevated perspective view of a tesselated cube wherein the Z direction is indicated as the vertical direction and the coordinates of two corners are shown within parentheses adjacent to each corner;
FIG. 10 is a schematic representation of the grouping and subgrouping of the facets of the cube shown in FIG. 9;
FIG. 11 is a schematic representation of the facets of two different intended models, each with its own shape, as a progression is made from the initial state in the STL file to the state after grouping and subgrouping;
FIG. 12 is a schematic representation of a comparison between the error obtained at different layer thicknesses;
FIG. 13 is an elevated perspective view of a block with pointed edges;
FIG. 14(a) is an elevated perspective view of a rectilinear box having pointed ends, and FIG. 14(b) is a front plan view of a triangular face with pointed ends;
FIGS. 15(a)and 15(b) are a schematic representation of two different arrangements, (a) and (b), of multiple facets forming a horizontal face of an intended object;
FIG. 16(a) is a front plant view of a block with differently shaped holes shown in phantom (dashed line) intersecting with a plane A--A, which is shown from a top plan view in FIG. 16(b) to illustrate multiple empty contours within a larger contour;
FIG. 17(a) is a schematic cross-sectional view of a cylinder within a cylinder intersected by a slice plane A--A, which is shown from a top plan view in FIG. 17(b) to illustrate multiple contours surrounding each other;
FIG. 18(a) is a front plan view of a complex three-dimensional object intersected by a slice plane A--A, which is shown in FIG. 18(b) to illustrate unbounded contours;
FIG. 19(A) illustrates three examples of regular surfaces, while FIG. 19(B) illustrates a single example of a regular surface on a three dimensional block;
FIG. 20 schematically illustrates the effect of the slope and layer thickness on the error produced by stair stepping in a model of a tesselated intended object;
FIG. 21 is a graphical representation of the build time of an object created with an SLA as a function of the layer thickness wherein the layer thickness is "l" at the minimum build time "t.sub.min ";
FIG. 22 is a graphical illustration of the cusp height "c" as a surface attribute wherein the Z direction is the vertical direction along which the layers of thickness "l" are sliced;
FIG. 23 is a graphical representation of the relationship between the cusp height "c" and the angle .THETA.;
FIG. 24 is a schematic representation of the maximum deviation "d" as a surface attribute with the vertical direction indicated by the letter Z;
FIG. 25 is a schematic representation of the chord length "f" as a surface attribute with the vertical direction indicated by the letter Z;
FIG. 26 is a schematic representation of the error between the three surfaces: intended surface, tesselated model of intended surface, and layered model of tessalated surface;
FIG. 27 is a schematic representation illustrating the volummetric loss in a layered model with the slice axis direction indicated by Z;
FIG. 28(a) schematically illustrates the volumetric error due to Stair Step (SS) error for a single layer of a layered model having "n" surface complexities and shown in a top plan view;
FIG. 28(b) schematically illustrates the volumetric error due to SS error for a single layer of a layered model having "n" surface complexities and shown in a side plan view;
FIG. 28(c) schematically illustrates the volumetric error due to SS error for a single layer of a layered model having "n" surface complexities and shown in a front plan view;
FIG. 29 is an elevated perspective view of three-dimensional strip of a layer having a thickness "l" measured in the Z direction (the slice axis) and a perimeter length "p" measured in a linear direction that is normal to the Z direction;
FIG. 30 schematically illustrates three protrusions in an object being intersected by the slice planes A--A and B--B;
FIG. 31a-31care a schematic representation of how the back tracking means of the present invention provides the ability to retain a key characteristic in a layered model of an intended object;
FIG. 32 is a schematic representation of a three dimensional object wherein successive slice planes A--A and B--B pass through the same facet chain;
FIG. 33 is a schematic representation showing how the simple back tracking means of the present invention provides the ability to retain a change in the level of surface complexity in a layered model of an intended object;
FIG. 34 is a schematic representation showing the usefulness of repeated back-tracking in accordance with the present invention in cases of multiple changes in the level of complexity between Z.sub.curr and Z.sub.tmp ;
FIG. 35 is a schematic representation of a flow chart for adaptive slicing in accordance with the present invention;
FIG. 36 is a schematic representation of the intersection of a slice plane and a simple triangle;
FIG. 37 is a schematic representation of a slice plane passing through a common vertex of two triangles;
FIG. 38 is a schematic representation of a slice plane passing through a chain composed of six facets; and
FIG. 39a-39d are a schematic representation of how the direction of a contour is found in accordance with the marching means of the present invention.
Detailed Description of the Preferred Embodiments
Reference now will be made in detail to the presently preferred embodiments of the invention, one or more examples of which are illustrated in the accompanying drawings. Each example is provided by way of explanation of the invention, not limitation of the invention. In fact, it will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the scope or spirit of the invention. For instance, features illustrated or described as part of one embodiment, can be used on another embodiment to yield a still further embodiment. Thus, it is intended that the present invention cover such modifications and variations as come within the scope of the appended claims and their equivalents. The same numerals are assigned to the same components throughout the drawings and description.
The present invention operates with respect to the boundary of a model of an intended object, which can be two-dimensional or three-dimensional, as can the boundary of the model. As shown in FIG. 3, an example of an intended object could include an exemplar 40 of a three-dimensional object to be built by a stereolithography apparatus (SLA) 41 with a laser beam 42 incident on the surface of a bath composed of liquid polymer 43 under the control of a computer 44, which generates a layered model 45 of an STL file representative of the intended object that the SLA is building. Another example of an intended object in a scanning profile context such as shown in U.S. application Ser. No. 08/156,321, now U.S. Pat. No. 5,432,704, could include a patient's cancerous tumor that is to be subjected to irradiation. Yet another example of an intended object could include a part to be machined, such as shown in U.S. application Ser. No. 08/156,321, now U.S. Pat. No. 5,432,705. Thus, the present invention can be practiced in any application requiring a machine to be controlled with respect to a target, whether such target be expressed as a one-dimensional path, a two-dimensional perimeter or a three-dimensional surface of an intended object. Other examples of intended targets or objects are disclosed in U.S. application Ser. No. 08/156,321, now U.S. Pat. No. 5,432,704, which is incorporated herein in its entirety by this reference.
To facilitate understanding the present invention, the discussion herein will concern itself primarily with examples wherein the operation of the machine is controlled based on a two-dimensional sliced (a.k.a. layered) model of the profile of the cross-section of the intended object along a slice axis of the intended object. In particular, the present invention operates with respect to the model of the intended object wherein the model of the intended object is expressed in a computer file in the stereolithographic (STL) format.
In accordance with the present invention, one can begin with a three-dimensional model of the intended object. Such three-dimensional model of the intended object can be generated in a Computer-Aided Design (CAD) system such as the ARIES.RTM. CAD system available from ARIES Technology, Inc. As is the case with most available CAD systems, the ARIES.RTM. CAD system from ARIES Technology, Inc. converts the model of the intended object in the CAD system into a computer file in the stereolithographic (STL) format. The STL file defines the exterior surface of the intended object in a tesselated form as: (1) a plurality of flat planar sections known as facets and (2) the outward pointing normals to these planar sections. Before a process, including a process carried out by a machine operating under some form of automatic control in accordance with the present invention, can be performed with respect to the model of the intended object, this data in the STL file format must be transformed into an input that is useable by the process in question. At the heart of the present invention is a so-called slicing engine that is configured so as to effect this transformation by generating a two-dimensional sliced model of the profile of the cross-section of the intended object in a particular plane of view. The particular plane of view is chosen so as to take advantage of the symmetry of the intended object.
The basic function of the slicing engine of the present invention is to generate two-dimensional slices from a three-dimensional tesselated model, which typically is provided by CAD software in the form of an STL file. Thus, the output of a slicing engine is a two-dimensional representation of the STL model of the intended solid object. The slicing engine of the present invention differs from the existing similar software systems in two ways. First, there is a smaller number of slices in the output of the slicing engine of the present invention than in the output obtain from conventional slicing engine. Second, the quality of the two-dimensional representation comprising the output of the slicing engine of the present invention is superior to the quality of the output obtained from conventional slicing engines. In other words, the error between the desired profile of the cross-section of the intended object and the two-dimensional representation forming the output of the slicing engine is less in the output of the slicing engine of the present invention than in the output obtained from the conventional slicing engine. The slice data that is the output of the slicing engine of the present invention, when used to obtain the intended object, results in low build times because of the fewer number of slices and results in surfaces of the intended object that are high in quality because the slice representation is improved over that obtainable with a conventional slicing engine.
The slicing engine of the present invention includes four main constituents: (1) a facet processor means, (2) a key characteristic identifier (KCI) means, (3) a thickness calculator means, and (4) a slicer means. The basic function of each of the four means of the slicing engine of the present invention is briefly described as follows. The facet processor means is provided to improve the speed of the slicing operation. The key characteristic identifier (KCI) means is provided to ensure retention of the intended object's asymmetries and to deal more effectively with the complexities of the surface of the intended object. The thickness calculator means computes the thickness of the slice (a.k.a. layer) between adjacent slice planes according to a user defined criterion relating to the complexity in the shape of the boundary defining the model. The slicer means generates two dimensional contours of the intended model.
A preferred embodiment of the slicing engine in accordance with the present invention is schematically shown in the block diagram in FIG. 4 as the elements surrounded by the dashed-line rectangle and is represented generally by the numeral 50. Information flow between the various elements of the slicing engine moves in the direction indicated from the tail to the pointed head of each arrow in FIG. 4. Input to the slicing engine is an STL file 52, while the output from the slicing engine is two dimensional slice data 54. As schematically shown in FIG. 4, a facet processor means 56 receives the input data in the form of STL file 52, processes STL file 52 in the manner described below, and then feeds the processed STL file to the other three elements of slicing engine 50. The KCI means 58 operates on the processed STL file received from the facet processor means and sends the output of KCI 58 to a thickness calculator means 60. The thickness calculator means operates on the output received from KCI means 58 and on the processed STL file received from facet processor means 56 and sends the results of this operation to a slicer means 62. The slicer means operates on the output received from thickness calculator means 60 and on the processed STL file received from facet processor means 56 to generate two dimensional slice data 54 and presents this data 54 in the form of a data file 64, which must be converted by an interface means so that the information in data file 64 can be used by the machine being controlled (such as SLA 41). An example of an interface means is a machine controller 46 shown in FIG. 3.
The interface means desirably includes a computer that is programmed with software that converts the STL data file 64 output into a format compatible with the machine to be controlled. For example, to convert data file 64 so that it can be used by a SLA device such as SLA 41 in FIG. 3, a presently preferred embodiment of such component of the interface means program is contained in pages 96-175 of the Appendix which forms a part of this patent application. The data thus created can be used by any process employing a profile or cross-section that is decomposed by incremental parallel planes of intersection. For example, as shown in FIG. 3, the data 54 can be used by any layered manufacturing technique to realize the physical model 40.
As shown in FIGS. 3 and 4, two dimensional slice data 54 also can be outputted and presented in the form of a visual display 66. The interface means also can include a computer that is programmed with software that converts the STL data file 64 output into a format compatible with the output device generating the visual display 66. A presently preferred embodiment of such component of the interface means program is contained in pages 47-95 of the Appendix which forms a part of this patent application.
The properties of an STL file are taken into account in building an efficient slicing engine in accordance with the present invention. The configuration of a preferred embodiment of the slicing engine of the present invention assumes that the STL format conforms to the following criteria, which are typical of STL formats employed by the major computer-aided design (CAD) systems.
The first assumed criterion for the STL file is that each facet is represented by four coordinates. The first coordinate defines the unit normal to the plane of the facet. Each of the remaining three coordinates defines the location of a different one of the three vertices of the facet. The second assumed criterion for the STL file is that the facet unit normal information and the triangle vertex order are arranged so that they are used to properly distinguish the interior from the exterior of the intended object. Specifically, as shown in FIG. 5, each facet must have a corresponding unit normal vector 70, 71 pointing away from the solid object. In addition to the facet normal, the "right hand rule" is also used to specify the interior of the object. The right hand rule states that the vertices of each facet must be ordered so that when the fingers of the right hand pass from vertex one through vertex two to vertex three, with the thumb pointing in the direction of the normal, the normal will point away from the solid object.
The third assumed criterion for the STL file is that the actual data may be in any units, but negative and zero co-ordinates are not used in specifying this data. This makes it necessary to place the original model of the intended object in positive CAD space before the model of the intended object is tesselated. The fourth assumed criterion for the STL file is the so-called Vertex-to-vertex Rule, which states that each triangle must meet all adjacent triangles along a common edge. In no case may a triangle vertex intersect with the side of an adjacent triangle. FIG. 6(a) shows an example of a facet model which does not follow the Vertex-to-vertex Rule. FIG. 6(b) shows an example of a facet model that obeys the Vertex-to-vertex Rule. The reader should note that this Vertex-to-vertex Rule requires each edge to be shared by only two triangles and no more, and therefore each triangle has three and only three adjacent triangles. As explained hereafter, the Vertex-to-vertex Rule permits the specification and use of an adjacency list for each facet in the model of the intended object.
As shown in FIG. 7, about 90% or more of the facets will not intersect with the slice plane near the bottom of a tesselated sphere. Thus, considerable computation time can be saved merely by identifying and focusing on only those triangles which intersect with a proposed slice plane. The function of the facet processor means of the present invention is to bring order to the otherwise unordered collection of triangular facets of the STL file 52 that is inputted to the facet processor means 56.
In accordance with the present invention, the size and orientation of a facet are of less importance to the slicing engine than the facet's location relative to the slicing plane. Thus, sorting the facets according to their size or their unit normals was considered unavailing. Accordingly, a more coherent approach than a simple sorting was required. In accordance with the present invention, the objective of such a more coherent approach must be to arrange all of the facets used to specify the tesselated model of the intended object (the facets "thrown in the bucket" so to speak) according to the geometry of the object that the facets represent. Furthermore, in the general case, this "coherent" arrangement must be achieved without prior knowledge of the geometry of the intended object in question.
In accordance with the present invention, generally speaking the facets "in the bucket" are sorted, grouped, and subgrouped. In more particular accordance with the present invention, a facet processor means is provided for operating on the STL file input to produce a processed facet file. Desirably, a computer is programmed to function as the facet processor means. Suitable computers on which the facet processor means can be programmed include an IBM.RTM. personal computer, so-called IBM.RTM. clone brand personal computers, or an Apple.RTM. brand computer. Alternatively, the facet processor means can be programmed on an EPROM for example. As embodied herein, the facet processor means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such facet processor means program is contained in pages 15-18 and 30-31 of the Appendix which forms a part of this patent application. In the Appendix, the computer file containing the C language code was edited to fit the text into the format used to print the patent application.
In further accordance with the present invention, the facet processor means includes a facet sorting means for sorting the three vertices of each facet in the STL file and then sorting the facets themselves to produce a sorted facet file. The facet sorting means is configured to sort the facets of the STL file according to the slice axis to produce the sorted facet file. As noted above, a computer desirably is programmed to function as the facet sorting means. The facet sorting means desirably is programmed on the same computer on which the rest of the facet processor means is programmed.
The program embodying the facet sorting means is configured to sort the three vertices of each facet with reference to the slice axis, and such sorting usually is performed by the facet sorting means in ascending or descending order. For purposes of simplifying the example used to explain the operation of the facet sorting means, the slice axis has been chosen to be the Z-axis, which is assumed to be the vertical axis of the intended object in the three-dimensional Cartesian set of coordinates. Thus, for example, for each facet of the STL file of the tesselated model of the intended object, the facet sorting means orders the vertex with the smallest Z value into the first position, the vertex with the largest Z value into the third position, and the remaining vertex into the second position. After the facet sorting means has operated on the STL file, the output is a sorted STL file that is ordered as explained above. As embodied herein, the facet sorting means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such facet sorting means program is contained in pages 15-16 of the Appendix which forms a part of this patent application.
In still further accordance with the present invention, the facet processor means includes a facet grouping means for grouping the facets in the sorted facet file produced by operation of the facet sorting means. The facet grouping means is configured to group the facets of the sorted facet file into separate groups of facets to produce or compose a grouped facet file 72 shown schematically in FIG. 8. Desirably, a computer is programmed to function as the facet grouping means. The facet grouping means is programmed conveniently on the same computer on which the facet sorting means and the rest of the facet processor means is programmed. As embodied herein, the facet grouping means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such facet grouping means program is contained in pages 16-18 and 30-31 of the Appendix which forms a part of this patent application.
The strategy adopted by the facet grouping means is fairly simple. Each group composing the grouped facet file has either the same minimum Z-axis coordinate value (Z-min) or the same maximum Z-coordinate value (Z-max). In the example used herein, the program embodying the facet grouping means is configured to group the facets in the sorted facet file with reference to the Z-min values of the vertices of each facet. Since the vertices of each facet in the STL file provided as an input to the grouping means are already sorted by the facet sorting means, the first vertex will always contain the minimum Z-coordinate value for that facet. Thus, for example, each facet in one group of facets in the grouped facet file has the smallest Z-min value, while each facet in a second group of facets in the grouped facet file has the next smallest Z-min value, until in the last group of the grouped facet file each facet has a vertex with the largest Z-min.
As schematically shown in FIG. 8, all facets having exactly the same Z-min are stored as one group called a facet group 72. As embodied herein and schematically shown in FIG. 9 for example, triangles designated T.sub.l through T.sub.12 form the facets of a tesselated cube 78. Facets T.sub.1 and T.sub.2, which are not shown in FIG. 9, form the bottom face of cube 78. The top left corner of cube 78 has coordinates X=1.0, Y=1.0 and Z=2.0. The lower left corner of cube 78 has coordinates X=1.0, Y=1.0 and Z=1.0. As schematically shown in FIG. 10 for cube 78 of FIG. 9, facet group will include facets T.sub.1 through T.sub.10, since all these facets have the same Z-min value of 1.0. Facet group2 will have facets T.sub.11 and T.sub.12 forming the top face of cube 78, since these are the only facets with the same Z-min value of 2.0.
In yet further accordance with the present invention, the facet processor means can include a facet subgrouping means for subgrouping the facets in the grouped facet file, which has been produced by operation of the facet grouping means. In this way, as schematically shown in FIG. 8, the groups and subgroups have been arranged in a tree structure in which the facets in each group are further divided into subgroups called facet subgroups 74. The facet subgrouping means is configured to subgroup the facets of the grouped facet file into separate subgroups of facets to produce or compose a subgrouped facet file. Desirably, a computer is programmed to function as the facet subgrouping means. The facet subgrouping means is programmed conveniently on the same computer on which the facet grouping means and the rest of the facet processor means is programmed. As embodied herein, the facet subgrouping means is programmed in C language on a Sun Microsystem.RTM. work station.
Since grouping is based on the Z-min in the example given, subgrouping is done based on the maximum Z-coordinate value (Z-Max). Thus, each subgroup composing the subgrouped facet file has the same Z-max coordinate. The program embodying the facet subgrouping means is configured to subgroup the facets in the grouped facet file with reference to the Z-max values of the vertices of each facet in each group of facets. Thus, for example, each facet in one subgroup of facets in the subgrouped facet file has the smallest Z-max value, while each facet in a second subgroup of facets in the subgrouped facet file has the next smallest Z-max value, until in the last subgroup of the subgrouped facet file each facet has a vertex with the largest Z-max. While the program in the Appendix does not devote a separate section of code to a facet subgrouping means, such code would be the same as that shown for the facet grouping means on pages 16-18 and 30-31 of the Appendix, except that the facet subgrouping means would subgroup based on Z-max instead of the Z-min that the code in the Appendix uses for grouping.
The number of facet groups and the number of facet subgroups within a facet group depend solely on the complexity of the model of the intended object. However, the number of individual facets in each group or subgroup depends on the overall facet count.
As schematically shown in FIG. 10 for cube 78 of FIG. 9, facets in facet group can be further divided into two subgroups based on the Z-max values. Facet subgroup1 will contain facets T.sub.1 and T.sub.2, which have the same Z-max value of 1.0, while facets T.sub.3 through T.sub.10 having a Z-max value of 2.0 will be in facet subgroup2. However, there will be no subgroups in facet group2, because it has only two facets T.sub.11 and T.sub.12 and both of these facets have the same Z-min value as well as the same Z-max value. Accordingly, the facets constituting the bottom face of cube 78 are in a first grouping in the form of facet subgroup, the facets constituting the middle faces are in a second grouping in the form of facet subgroup2, and the facets constituting the top face are in a third grouping in the form of facet group2. Similarly, FIG. 11 schematically illustrates the operation of the facet processor means on two more facet models, which are taken from their disordered initial states in the STL file to their more ordered states after sorting, grouping and subgrouping.
In each of the above examples, each facet group or subgroup represents either a vertical or a horizontal face of the model. If a feature is defined as any topologically significant entity, each face of the above models can be considered to satisfy this definition of a feature. Accordingly, the grouping methodology that results from sorting vertices and grouping according to Z-min and subgrouping according to Z-max, allows for easy identification of facets that represent a feature. This important offshoot of grouping will be exploited in key characteristic identification, described below.
Unlike sorting, grouping and subgrouping are bi-directional, because both Z-min and Z-max are considered to arrange the facets. As grouping follows sorting, identifying facets with the same Z-min or Z-max for the purpose of grouping becomes relatively simplified. Consequently, the additional time spent on grouping and subgrouping becomes insignificant.
The technique used by a stereolithography apparatus to produce a prototype differs fundamentally from the technique used by a computerized numerically controlled (CNC) milling machine to produce the same prototype. A stereolithography apparatus such as an SLA-250 utilizes an additive process for producing a prototype. A CNC milling machine adopts layered machining, which is a hybrid process for producing a prototype. Because of these fundamental differences in their techniques, the stereolithography apparatus has different needs with regards to feature recognition than the needs of a CNC milling machine with regards to feature recognition.
For Rapid Prototyping processes such as employed by an SLA-250, the emphasis in the area of feature recognition must be on recognition of the nature of the feature, protrusion or depression, and on the base faces of each feature. Distinguishing depressions from protrusions enables identification of areas that must be left unfilled (additive processes) or where material must be removed (substractive processes) so as to form a hole.
In addition, recognition of base faces is essential to improve tolerances of layered models. For example, FIG. 12 schematically illustrates the importance of retaining horizontal faces. In FIG. 12, layered models of a stepped block at different values of layer thickness .increment.Z are schematically shown. At layer thicknesses of 0.4 and 0.6, models are considerably deviated from the intended object. In fact, at a layer thickness of 0.6, the layered model is out of shape, with one step completely eliminated. However, at a layer thickness of 0.5, which is a multiple of the step heights (0.5 and 1.0), there is no error. No error is shown at the top of the block, because it is assumed that all existing slicing engines can successfully retain both top and bottom surfaces of an object. This simple example demonstrates that the accuracy of the layered model of the intended object or target is a function of design features, dimensions, and layer thickness. Recognizing base faces eliminates this undesirable dependency. Thus, for Rapid Prototyping processes for example, it is adequate to identify the type of feature and the base of the feature. These two requirements can be met without a full scale feature recognition, and hence the term "key characteristics" has been adopted.
In contrast to Rapid Prototyping processes, layered machining performed by a CNC milling machine for example, demands a more comprehensive feature recognition than mere identification of the type of feature and the base faces of each feature of the intended object. Layered machining requires the dimensions, the geometric shape, and the orientation of each feature to be identified in addition to the type of feature and the base faces of each feature. This is mainly due to the fact that unlike other layered processes, the capabilities of layered machining are limited by the geometric complexity of the intended object. Features like threading, undercuts, grooves, and holes at an angle, are some of the constraints which demand retooling, refixturing or reorientation of the intended object. Identification of machining constraints is therefore an essential requirement for layered machining of a feature integrated design.
In accordance with the present invention, a Key Characteristic Indentifier (KCI) means is provided. The key characteristic indentifier means is configured to operate on a tesselated model with the result that the following key characteristics of the model of the intended object are identified: (1) the base faces of the model; (2) the type of features (protrusion or depression) of the model; (3) the geometric shape of the model's features; and (4) the orientation of the model's features. As schematically shown in FIG. 4, the key characteristic indentifier means 58 is configured to operate on the sorted facet file, the grouped facet file and the subgrouped facet file to produce or compose a key characteristic data file. Desirably, a computer is programmed to function as the KCI means. The KCI means is programmed conveniently on the same computer on which the facet sorting means and the rest of the facet processor means is programmed. As embodied herein, the KCI means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such KCI means program is contained in pages 9-10 of the Appendix which forms a part of this patent application.
In accordance with the present invention, the key characteristic indentifier means provides slice-based feature recognition in the profile of the intended object. The key characteristic indentifier means is configured to employ a simple knowledge base that is supported by a few rules. The rules are mostly based on simple geometric and engineering principles, and derived by observation and logical reasoning. In fact, some of the ideas discussed below are applicable for any solid model regardless of whether the representational scheme is a tesselated model.
Top and bottom faces of any feature in the model of the intended object, including the overall object itself, may be termed as base faces. Since a polygon consists of only two basic entities, a point and a line, only combinations of one or more points with one or more lines will result in any topologically significant characteristic. Consequently, as schematically shown in FIG. 13, any solid 80 represented by planar polygons can have only three types of base faces: a horizontal face 81, a pointed horizontal edge 82, and a pointed end 83. FIG. 14 shows additional examples of pointed ends 83 and pointed horizontal edges 82.
The key characteristic identifier means identifies a horizontal surface by applying the following test: A horizontal surface is represented by a single polygon disposed parallel to the slice plane or by multiple polygons disposed parallel to the slice plane. In the STL format that is of interest herein, the polygons will always be triangular facets. FIG. 15(a) shows a rectangular horizontal face composed of triangular facets and FIG. 15(b) shows an inverted T-shaped face composed of triangular facets.
The key characteristic identifier means identifies a pointed edge by applying the following test: When at least two non-coplanar polygons share an edge, the edge may be termed as a pointed edge. As schematically shown in FIG. 13, a block with v-grooves contains a number of pointed edges 82.
The key characteristic identifier means identifies a pointed end by applying the following test: When at least two non-coplanar polygons share a vertex, the vertex may be termed as a pointed end. As schematically shown in FIGS. 13 and 14, the tip of a cone and the corner of a block are examples of pointed ends 83.
For a layered manufacturing process, the existence of of each of the three types of base faces described above must be identified. Then the height of each base face must be determined so that the slice plane can be forced to assume that height. Identification of these three types of base faces from an STL model format is basically a searching operation. Horizontal surfaces can be identified by checking unit normals or ordinate values of facets. For example, if a unit normal of a facet describes 90.degree. (or 270.degree.) with the slice plane, there lies a horizontal surface with that facet contained in that surface. Alternately, if Z coordinate values of all the three vertices of a facet are equal, then that facet is part of a horizontal surface. However, if brute force approaches to searching horizontal facets are used, then computational costs will be tremendous as the facet count increases. Fortunately, the facet groupings performed in accordance with the present invention will greatly facilitate the searching process. In fact, from the cube example shown in FIGS. 9 and 10, it can be seen that during grouping in accordance with the present invention, all facets contained in a horizontal face are automatically stored in either a facet group by the facet grouping means or a facet subgroup by the facet subgrouping means.
With regards to implementation in a layered manufacturing process, the key characteristic identifier means applies further refinement of the above definition of an "edge" by following more elaborate rules. First, the key characteristic identifier means requires that none of the polygons sharing an edge should be horizontal. This is because horizontal faces are already identified, and the identification of a horizontal face means that the edges of such horizontal face have been determined. Second, the key characteristic identifier means requires that the common edge must be parallel to the slice plane, and then a separate effort is required for its retention in the sliced model. Thus, the key characteristic identifier means ignores an inclined edge, and only identifies horizontal edges. In STL format, any non-horizontal facet with two vertices having the same Z coordinate value is likely to form an edge provided the facet that shares these vertices (there will be only one such facet) is not in the same plane as the first facet and is non-horizontal. According to the grouping sequence implemented by the facet grouping means of the present invention, this facet will invariably lie in the current group or a group above or below the current group. Thus, the key characteristic identifier means limits the search to these three groups only, saving much computer time.
However, identifying two of a facet's vertices which have the same Z coordinate value, which is the very first step in identifying a pointed edge, demands searching virtually every vertex of all non-horizontal facets in the model. This again leads to high execution times. Finally, while the key characteristic identifier means can identify pointed ends from an STL file by looking for a single vertex shared by several non-coplanar facets, this calls for checking every vertex in the model for the number of facets that share it and their unit normals.
Clearly, identifying pointed ends and edges is a time intensive search operation. Isolation of the vertex that forms a pointed end or the edge that becomes a pointed edge demands a large scale search, and often that search must cover every facet of the model. This problem can be largely overcome by adopting a new strategy in accordance with the key characteristic identifier means of the present invention. This strategy is based on the fact that whenever there is a change in the number of contours in successive slices, there lies a base face of a feature.
A contour is a line that is continuous and defines an interior and an exterior. A continuous contour is a contour that does not have a break. An empty contour is a contour that lacks a second contour in the interior of the empty contour. For example, as shown in FIG. 16(b), the rectangular contour designated 84 is not an empty contour. This is because within the interior of contour 84, there resides an empty closed contour 85 in the form of a circle, an empty closed contour 86 in the form of a square, and an empty closed contour 87 in the form of a triangle.
In accordance with the configuration of the key characteristic identifier means of the present invention, the exact location of the base face of the feature can be anywhere between the two slice planes or in the slice plane which has the greater number of contours. A similar approach is suggested by Dolenc and Makela (1993) to identify peaks. Once a change in the number of contours is observed by the key characteristic identifier means of the present invention, the key characteristic identifier means finds the exact location of the base face by considering facet groups in that region and applying the methods suggested above.
Protrusions are any projections on the object, such as pads, bosses, and mounds, while depressions are voids of any shape. The easiest way of finding a feature's type (protrusion or depression) is by studying the slices and the contours within each slice. The following rules, which are true for any solid, are embodied in the KCI means of the present invention and enable the KCI means to distinguish between protrusions and depressions.
Rule 2.1: When a solid is sliced by an imaginary plane, the resulting slice may contain single or multiple, non-intersecting, closed contours. This is true, regardless of the angle and the position of the slice.
Rule 2.2: If a single empty contour is surrounded by a single larger contour or multiple empty contours are surrounded by a single larger contour, each inner contour, i.e., each surrounded contour, represents a depression feature. As shown in FIGS. 16(a) and (b), each of closed contours 85, 86 and 87 is a depression feature.
Rule 2.3: In general, if a contour is bounded by multiple non-intersecting contours, and each contour is numbered consecutively starting with the outermost contour as the first contour and each area between consecutive contours is also numbered consecutively starting with the outermost area, then the n'th area represents solid if "n" is odd and a void if "n" is even. Referring to FIG. 17 for example, a cross-sectional view of a cylindrical block 88 is shown in FIG. 17(a) in an axial plane, and a cross-sectional view of block 88 taken through transverse plane A--A is shown in FIG. 17(b). In FIG. 17, the first area 91 (the area between the two outermost contours 89 and 90) represents solid, the second area 92 (area between contours 90 and 93) represents a void, the third area 95 (area between contours 93 and 94) represents solid, and the fourth area 96 (area internal of contour 94) represents a void.
Rule 2.4: An unbounded contour is a contour that lacks a second contour surrounding the exterior of the unbounded contour. Any unbounded contour represents a protrusion feature. Referring to FIG. 18 for example, a cross-sectional view of a large solid block 100 is shown in FIG. 18(a) in an axial plane, and a cross-sectional view of block 100 taken through transverse plane A--A is shown in FIG. 18(b). As shown in FIG. 18, the first area 101 (the area within the empty contour 107) represents a solid protrusion, the second area 102 (area between empty contours 107 and 108) represents a void, the third area 103 (area within empty contour 108) represents a solid protrusion, the fourth area 105 (area between empty contours 108 and 109) represents a void, and the fifth area 105 (area within empty contour 109) represents a solid protrusion.
The KCI means is configured so that once it has determined the feature's type and location by employing the methods described above, the KCI means can obtain further details of the feature by analyzing the unit normals of the facets which compose the feature. The following example of an analysis of the unit normals of the facets which compose a feature of an intended model, assumes that the KCI means already has identified voids, projections, and base faces using the methods explained above.
Assuming that the KCI means has identified a particular feature as either a pad or a void using Rules 2.1 through 2.4, the KCI means applies the following tests to determine whether the pad or void is triangular in shape. If a feature comprising a pad or void is triangular, such feature is bounded by three faces where the sum of the included angles between the faces is exactly 180.degree. and all three faces are perpendicular to a common plane. Moreover, if the feature has a fourth face that connects all of these three commonly perpendicular faces, then the triangular feature is closed at one end. Similarly, if the feature has a fifth face that also connects the three commonly perpendicular faces, then the triangular feature is closed at both ends. The KCI means is configured to determine the type of triangle by testing the included angle between the three faces that are pendicular to the common plane. The KCI means is configured to determine the orientation of the feature by testing the angle between the slice plane and one of the bounding faces.
Assuming that the KCI means has identified a particular feature as either a pad or a void using Rules 2.1 through 2.4, the KCI means is configured to determine whether that feature is rectangularly shaped by testing whether such feature is bounded by four faces wherein each face is perpendicular to each adjacent face and all four faces are perpendicular to a common plane. Moreover, if there is a fifth face that connects all of the four faces, the rectangular feature is closed at one end. Similarly, if a sixth face connects the first four faces, then the rectangular feature is closed at both ends.
The KCI means is configured to apply the following tests to identify whether a feature, which has been determined to be a pad or void, is pentagonal. Each pentagonal pad or void is bounded by five faces, where the sum of the included angles between the faces is exactly 450.degree. and all five faces are perpendicular to a common plane. Moreover, if there is a sixth face that connects all of the five faces, the pentagonal feature is closed at one end. Similarly, if a seventh face also connects the first five faces, then the pentagonal feature is closed at both ends. The KCI means is configured to determine the type of pentagon (regular or irregular) by testing the included angle between adjacent faces. The KCI means is configured to determine the orientation of the pentagonal feature by testing the angle between the slice plane and one of the bounding faces. The foregoing logic can be extended to identify hexagonal or any higher order polygonal feature.
One limitation with slice-based feature recognition is that features parallel or at certain angles to the slice plane will be difficult to identify. In order for a comprehensive feature recognition, slicing may have to be done at different angles and the resulting slices studied independently and together. However, these difficult-to-identify features need not be identified for layered modeling techniques.
As noted in Chalasani and Bagchi (1992), one of the fundamental problems for better process planning and eventually for greater build efficiency (time versus quality) with all of the Rapid Prototyping technologies is the limitation of having to use a slice thickness that cannot be changed within the model. This aptly sums up the importance of adaptive slicing. Since the build time is directly proportional to the number of layers, the objective of any layered process must be to build superior quality parts with fewer layers.
As shown in FIG. 12, the stair stepping (SS) phenomenon introduces the concept of SS error, which is a function of the geometry of the object to be modeled and the thickness of the layers into which such object is resolved. As known, SS error is mainly responsible for the surface roughness of the model of the intended object and for the form error of the model from the intended object. Here a detailed analysis of SS error is done in order to arrive at methods for quantifying SS error. Then, the SS error so quantified is used by the thickness calculator means of the present invention as a criterion for determining how to vary the layer thickness to restrict the resulting model surface to a predetermined geometrical error that is acceptable for the application being used.
In accordance with the present invention, a thickness calculator means is provided for operating on the sorted facet file, the grouped facet file, the subgrouped facet file and the key characteristic data file, to calculate and select the layer thickness that achieves a predetermined surface roughness that is acceptable for the process being used. More particularly, the thickness calculator means is configured to calculate and select a thickness for each layer of the model of the desired profile represented by the STL file of the intended object such that the geometrical error between the desired profile of the intended object and the model profile of each layer having such layer thickness, remains no greater than a preselected geometrical error. The thickness calculator means is configured to produce a layer thickness data file containing such calculations. As schematically shown in FIG. 4, the thickness calclator means 60 is configured to operate on the sorted facet file, the grouped facet file, the subgrouped facet file, and the key characteristic data file to produce or compose a layer thickness data file. Desirably, a computer is programmed to function as the thickness calclator means. The thickness calclator means is programmed conveniently on the same computer on which the key characteristic identifier means and the facet processor means is programmed. As embodied herein, the thickness calclator means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such thickness calclator means program is contained in pages 25-29 of the Appendix which forms a part of this patent application.
With regards to layered manufacturing, the thickness calculator means is configured to classify each surface of the model of the intended object into one of two broad categories. The first broad category is Regular Surfaces, which are surfaces that are not susceptible to the SS phenomenon. Examples of regular surfaces are shown at (A) in FIG. 19. The thickness calculator means is configured to classify intended objects with multiple regular surfaces as a regular surface, if as shown at (B) in FIG. 19 each of the multiple regular surfaces is separated from the other by a sudden slope discontinuity.
The second broad category is Non-Regular Surfaces, which are surfaces that are subject to the SS phenomenon. Surfaces featuring slopes with respect to the slice axis are prone to SS error because of an inevitable loss of information during slicing. The amount of lost information is related to the slope of the surface of the intended object and to the layer thickness of the model. FIG. 20 shows two layered models at (b) and (c) of a single tessellated model at (a) having multiple slopes. Each of the two layered models at (a) and (b) employs a different value of layer thickness. As shown in FIG. 20, for a given layer thickness, the magnitude of the SS error (indicated by the area encompassed by the stippled region between the boundary of the intended tessellated object and the boundary of the layered model superimposed on the intended tessellated object) increases with slope, and for a given slope, SS error increases with layer thickness.
In further accordance with the present invention, the thickness calculator means is configured to quantify a parameter known as "Surface Complexity," which relates the layer thickness and the slope of the profile of the surface of the intended object (in this case a tessellated object) to the SS error. The "Surface Complexity" is defined in Equation 2.1 in terms of the slope of the profile of the surface of the intended object obtained in a two-dimensional cross-section of the STL model of the intended object.
The angle .THETA. defines the slope of the object's surface with respect to the slice axis along which the layers are "cut." According to Equation 2.1, the surface complexity of a surface that is perpendicular (.THETA. is 90 degrees) to the slice axis, a so-called "regular surface," is zero. For so-called "non-regular surfaces" in which the angle e between the surface and the slice axis varies between 90 degrees (perpendicularity) and zero degrees (parallel), the surface complexity increases with increasing slope (decreasing angle .THETA.) until the surface complexity becomes infinite when .THETA. equals 90.degree.. For example, if the slice axis is the Z-axis (vertical), then a horizontal surface (when .THETA. equals 90.degree.) has an infinite surface complexity.
Assuming ideal process conditions, surfaces with zero surface complexity can be precisely constructed by stacking layers appropriately. Theoretically, one layer should be enough to build all the intended objects shown at (A) in FIG. 19. However, in any practical application, the maximum layer thickness is limited by the process that is being carried out with respect to the intended object. For instance, the maximum layer thickness for a stereolithography apparatus is determined by the type of building material, the laser power, and other process parameters. Furthermore, as is known, a larger thickness for the slices or layers composing the model of the intended object does not automatically result in shorter process times. For example, in stereolithography, the laser scan velocity decreases exponentially with increased cure depth, which accounts for an increase in build time. FIG. 21 graphically depicts how the build time in a process employing a stereolithography apparatus varies as a function of the layer thickness. According to FIG. 21, at a time-optimized layer thickness of "l.sub.t ", build time "t.sub.min " is minimum. Thus, even for zero complexity surfaces, the thickness calculator means of the present invention must be configured so as to select a layer thickness in such a way that the fastest process time can be achieved. The determination of such an optimum layer thickness depends on the process to which the layered modeling is applied.
In accordance with the present invention, selection of layer thickness for non-regular surfaces mainly depends on the desired quality of the surface of the intended object. This quality parameter is related to SS error, which in turn is related to the surface complexity and to the layer thickness. In the thickness calculator means of the present invention, the relationship between SS error, layer thickness and slope can be represented generally as in Equation 2.2 where "e" is the SS error and is a function of "l" (the layer thickness) and the angle ".THETA." (defining the slope of the object's surface with respect to the slice axis along which the layers lie transversely, i.e., are "cut").
In accordance with the thickness calculator means of the present invention, a predetermined quality for the surface of the intended object is obtained by varying "l" for different "8" in order to control "e." This must be done by establishing a more specific relationship between "e", "l" and ".THETA.", which necessitates an investigation of various methods of quantifying the "SS error." Four methods of quantifying the "SS error" are discussed below. Before they are described, it may be noted that SS error can be either positive or negative. When each layer is slightly bigger than the intended cross-section of the object, the SS error is positive, otherwise it is negative. The following criteria can be applied for both positive and negative measures of the SS error.
One alternative embodiment of a method of quantifying the "SS error" in accordance with the thickness calculator means of the present invention, relies on a parameter known as the "Cusp Height." As shown in an example in FIG. 22, the layer thickness "l" is determined such that the "cusp height," which is denoted by the lower case letter "c," is within a user-defined value represented by the notation "C.sub.desired." Equation (2.3) relates the cusp height "c" to the layer thickness "l" and to the slope as follows.
In accordance with the thickness calculator means of the present invention, the following constraints are considered in determining a desired cusp height "c.sub.desired ". Referring to FIG. 23, which plots the relationship between "c" and the angle theta (.THETA.) for various values of "l", the following observations can be made. According to a first constraint applied by the thickness calculator means of the present invention, when a uniform layer thickness of l.sub.n is used, the cusp height of the intended object varies from c.sub.1 to c.sub.2, where c.sub.1 =l.sub.min Sin .THETA..sub.min and c.sub.2 =l.sub.min Sin .THETA..sub.max and is shown by the bottom-most curve in FIG. 23. Similarly, when l.sub.max is the layer thickness throughout the intended object, the cusp height varies from c.sub.3 to c.sub.4, where c.sub.3 =l.sub.max Sin .THETA..sub.min and c.sub.4 =1.sub.max Sin .THETA..sub.max and is shown by the top-most curve in FIG. 23. In the plot, c.sub.1 =c.sub.2 =0, because .THETA..sub.min is zero. The constraint equation for the cusp height will then be c.sub.1 .ltoreq.C.sub.desired .ltoreq.c.sub.4.
According to a second constraint applied by the thickness calculator means of the present invention, for all values of c.sub.desired, where c.sub.desired <c.sub.2, the cusp height, at best, can be varied from c.sub.desired to c.sub.2 by appropriately varying the layer thickness "l", then the minimum deviation that cannot be avoided is given by c.sub.2 -c.sub.desired.
According to a third constraint applied by the thickness calculator means of the present invention, a constant cusp height throughout the intended object can be maintained for all values of c.sub.desired if c.sub.2 .ltoreq.c.sub.desired .ltoreq.c.sub.3. This is possible only when c.sub.3 >c.sub.2, which then becomes the condition for maintaining a uniform cusp height. As shown in FIG. 23, since c.sub.2 <c.sub.2, a uniform cusp height cannot be achieved.
According to a fourth constraint applied by the thickness calculator means of the present invention, for all values of c.sub.desired, where c.sub.desired >c.sub.3, there will be at least one layer having a cusp height that is less than c.sub.desired and the deviation from the desired value can be, at best, controlled to c.sub.desired -c.sub.3.
According to a fifth constraint applied by the thickness calculator means of the present invention, when theta (.THETA.) is zero (corresponding to a situation lacking any surface complexity), the cusp height "c" is zero for any layer thickness "l". Moreover, when theta is zero, any layer thickness "l" can be used without increasing the cusp height "c". When theta is 90.degree. (corresponding to a situation of infinite surface complexity), the cusp height "c" equals the layer thickness "l".
A second alternative embodiment of a method of quantifying the "SS error" in accordance with the thickness calculator means of the present invention, relies on a parameter known as the Maximum Deviation. As shown in FIG. 24, in a stair stepping triangle, the side "d" opposite the side equal to the layer thickness "l" indicates the maximum deviation of the layer from the intended surface, which is represented by the hypotenuse of the triangle. Equation 2.4 relates the maximum deviation "d" to the layer thickness "l" and to the slope of the profile as follows.
In this case, the layer thickness "l" is determined such that the maximum deviation "d" is within a value that is defined by the user. According to Equation (2.4), as .THETA. approaches 0.degree., "l" approaches infinity. This indicates that at .THETA.=0.degree., the surface complexity is zero and the thickness calculator means can select virtually any layer thickness "l" without losing accuracy. However, as explained earlier, the maximum layer thickness selected by the thickness calculator means is limited by the characteristics of the process.
Yet another alternative embodiment of a method of quantifying the "SS error" in accordance with the present invention, relies on a so-called "surface attribute parameter" known as the "Chord Length." As shown in FIG. 25, the layer thickness "l" is varied such that the chord length "f" is within a user defined value f.sub.desired. Equation (2.5) relates the chord length "f" to the layer thickness "l" and to the slope as follows.
where,
The constraint equation applied by the thickness calculator means of the present invention will be
where
f.sub.1 is not equal to 1.sub.min /Cos.THETA..sub.min
and f.sub.2 is not equal to l.sub.min /Cos.THETA..sub.max
When an intended object with a complexity level from .THETA..sub.min to .THETA..sub.max is considered by the thickness calculator means of the present invention, the chord length "f" at best can be varied from f.sub.1 to f.sub.2 when f.sub.1 and f.sub.2 are set to their respective lower limits. As illustrated in FIG. 26, these criteria are useful in an application in which it becomes necessary to compute the deviation of the layered model from not only the tessellated model but also from the original model.
A uniform value of "f" throughout the intended object can be achieved if and only if Equation (2.6) is true: ##EQU1## If Equation (2.6) is true, then the thickness calculator means of the present invention keeps the "surface attribute" "f" uniform by setting the value of "f" such that ##EQU2## This is derived using a similar logic on which the condition for a uniform surface quality is based.
The fourth alternative embodiment of a method of quantifying the "SS error" in accordance with the thickness calculator means of the present invention, relies on a parameter known as the "Volumetric Error." However, the following volumetric analysis of layered and tessellated models is required before formulating the criteria for relating SS error to layer thickness "l" based on volumetric error.
Typically, a layered model can lose or gain volume over the tessellated model based on whether the SS error is positive or negative. The following analysis is done assuming a negative SS error. However, the results of this analysis can be applied for either positive or negative values of SS error. Loss of volume in layered models of an intended object with a complexity level more than 0.degree. is common and also inevitable (unless the process is modified). Volumetric loss is related not only to the layer thickness and to the surface complexity but also to the cross-sectional size of the layer. This is illustrated in FIG. 27 by considering two layers of the same thickness "l" but located at different heights h.sub.1 and h.sub.2 of the cone 104 shown in cross-section as the intended object. Due to the larger cross-section at h.sub.1, the loss of volume at h.sub.1 exceeds the loss of volume at h.sub.2. Thus, the volumetric error is a function of the size of the cross-section in addition to e and "l" and can be represented as:
Loss of volume per layer=f(.THETA., l, p) where "p" is the perimeter of the layer having thickness "l".
It will not be feasible to maintain a constant volumetric error from one layer to the next layer, because of large variations in the perimeter from layer to layer. Any attempt by the thickness calculator means of the present invention to maintain a constant volumetric error by varying the layer thickness is likely to lead to unacceptable form errors. One alternative to keeping the volumetric error constant from one layer to the next layer, could be the selection of an attribute "V" such as the volumetric loss per unit length and keeping such attribute constant. Equation (2.7a) presents this relationship in general.
For example, as shown in FIGS. 28(a), 28(b) and 28(c), consider a single layer that is bounded by "n" surfaces and has "n" levels of complexity. The stippled regions in FIGS. 28(b) and 28(c) schematically represent the volume that is lost due to SS error.
Where a.sub.1 -a.sub.n are cross-sectional areas and p.sub.1 -p.sub.n are perimeters. Thus, if V represents the volumetric loss per unit of length of the perimeter, then V can be expressed ##EQU3## which also can be expressed as ##EQU4##
Further, ##EQU5##
Therefore, ##EQU6## Assuming that "p" and .THETA. are constants for a given geometry, a suitable layer thickness "l" can be found for a V chosen by the user. For example, as shown in perspective in FIG. 29, a three-dimensional strip of a layer has a thickness "l" measured in the Z direction (the slice axis), and has a perimeter length "p" measured in a linear direction that is normal to the Z direction. The constraint equation for V can be obtained as follows: ##EQU7## V can be minimized by setting "l" to l.sub.min ##EQU8## For a part with a complexity from .THETA..sub.min to .THETA..sub.max, v varies from V.sub.1 to V.sub.2 such that
Where,
V.sub.1 is not less than ##EQU9## V.sub.2 is not less than ##EQU10##
A uniform volumetric error per unit length of the perimeter (V) throughout the model can be achieved only if ##EQU11## If this condition (2.09) is true, the thickness calculator means of the present invention achieves a uniform V by setting V such that: ##EQU12## Constraint equation (2.10) has been obtained using a logic similar to the logic underlying the constraint equation for a uniform surface quality. Expressing the geometrical error in terms of volumetric error per unit of perimeter length is particularly useful in applications concerned with manufacturing parts for components of space vehicles or any portion of the payload carried by space vehicles. In such applications, space and weight are major constraints, and the V parameter enables the user to control these characteristics of the part to be manufactured.
In accordance with the present invention, the thickness calculator means varies the slice thickness "l" according to a strategy that is easy to implement and independent of other criteria. That is, this same strategy for varying the slice thickness "l" can be used for any criteria that the user might come up with in the future. As embodied herein and explained below, the strategy by which the thickness calculator means selects the slice thickness is successfully tested for three different criteria. In connection with these explanations, a few simple rules, based on observation and logical reasoning, are defined and act as tools that facilitate this adaptive slicing process. Such rules are explained as and when necessary.
The first such rule (Rule 3.1) to be observed by the thickness calculator means states that: Triangles which are either contained in or lying completely below the current slice plane, which is designated by Z.sub.curr and defines the slice plane for which intersection points are being computed, will not, in any way, influence the location of the slice plane immediately above. For instance, if a cone is mounted on a cube and the current slice plane is passing through the top face of the cube (intersection plane of cube and cone), only the triangles of the cone but not the triangles of the cube will affect the location of the next slice plane.
Once again, for the sake of simplifying the following explanation of the strategy by which the thickness calculator means selects the intersection of the next slice plane in accordance with the present invention, the slice axis is arbitrarily chosen to be the Z-axis. The first slice plane must always pass through the lowest Z-value of the object to ensure that the bottom face is retained in the sliced model. This step in the strategy of selecting the triangles for calculating the intersection with the slice plane is a step that applies to any desired object and serves as a good starting point for adaptive slicing of a model of the desired tesselated object. As noted above, Z.sub.curr indicates the height of the current slice plane. Now, to calculate the height of the next slice plane, Z.sub.next, each of the triangles which satisfies the condition for intersection (which condition is explained below) with the current slice plane Z.sub.curr is taken into consideration by the thickness calculator means. This ensures that the triangles selected by the thickness calculator means are limited to those triangles which: (1) share either an edge or a vertex and lie above the current slice plane or (2) pass through the current slice plane. Any triangle which shares an edge or a vertex but lies below the current slice plane Z.sub.curr is eliminated from the group of triangles selected by the thickness calculator means. Of course, there will not be any triangles below the very first slice plane. This selection strategy employed by the thickness calculator means also eliminates from the group of selected triangles, horizontal triangles which lie in the current slice plane. Furthermore, all triangles which survive the elimination process employed by the thickness calculator means satisfy Rule 3.1. The intersection condition developed earlier, thus plays a dual role, (1) avoidance of redundancies and (2) selection of the group of triangles which satisfy Rule 2.1 for the purpose of calculating the thickness of the layer.
Because of the operation of the facet processor means, which has grouped the triangles according to Z-min and subgrouped the groups according to Z-max, the grouped data file and the subgrouped data file exist. Thus, the process employed by the thickness calculator means to select triangles to be used in determining the layer thickness does not require a triangle-by-triangle check. Instead, the thickness calculator means performs its checking by checking the grouped data file and the subgrouped data file. In accordance with the present invention, the checking occurs facet group-by-facet group rather than triangle-by-triangle, thus saving enormous computation time.
There might be one or more facet groups which will have an intersection with the slice plane. Again, there might be one or more facet subgroups in each such facet group which will have an intersection with the slice plane. All such facet groups and facet subgroups intersecting the slice plane are identified by the thickness calculator means and then are merged by the thickness calculator means to form a single new group. Interestingly, all the triangles in that merged group produced by the thickness calculator means form either: (1) a single, non-intersecting continuous chain representing a closed surface or (2) multiple, non-intersecting continuous chains, wherein each such chain represents a closed surface. For instance, as shown in FIG. 30, the slice plane at B--B passes through a single, non-intersecting continuous chain of triangles, while the slice plane at A--A passes through multiple (three), non-intersecting continuous chains of triangles. The three chains intersected by slice A--A represent the three closed surfaces, the hemisphere, the rectangular block, and the truncated cone. This leads to the following important axiom (Rule 3.2) that is employed by the thickness calculator means: For any closed surface, there can be one and only one continuous chain of triangular facets per slice plane.
As discussed earlier, the surface complexity can vary in each layer, and in each layer there can be any number of levels of surface complexity. However, in accordance with the present invention, the worst case, i.e., the most complex surface, in each layer is selected by the thickness calculator means when determining how to vary the layer thickness. As embodied herein, the thickness calculator means identifies from the current chain or chains of triangles the controlling normal that represents the most complex surface.
In theory, there can be as many controlling normals as there are chains, and the thickness calculator means can vary the slice thickness for each chain separately. The procedure essentially remains the same. However, this most meticulous procedure is not implemented in the thickness calculator means embodiment presented in the Appendix.
Next, depending on the user defined criteria, any of the four equations (2.3; 2.4; 2.5; 2.10) derived earlier (or any new relationship for other criteria developed by the user) can be used to calculate the layer thickness "l". Then, Z.sub.tmp , which is the height of the next probable (temporary, hence the "trap" subscript) slice plane, is simply the sum of Z.sub.curr and "l". In the thickness calculator means of the present invention, Z.sub.tmp still need not represent the height of the next slice plane Z.sub.next, for two reasons. First, the thickness calculator means checks to determine whether there could be a key characteristic (KC) between Z.sub.curr and Z.sub.tmp. If a KC does exist in this location, then the thickness calculator means employs a back tracking means to deal with this KC in order to retain this KC in the model. Second, the thickness calculator means checks to determine whether there could be a sudden change in the level of surface complexity between Z.sub.curr and Z.sub.tmp. If a sudden change in the level of surface complexity does exist in this location, then the thickness calculator means employs a back tracking means (that is either a simple back tracking means or a repeated back tracking means) to deal with this sudden change in the level of surface complexity in order to retain this sudden change in the level of surface complexity in the model.
First, in explaining how the thickness calculator means deals with a KC, consider the presence of a KC in the form of a horizontal face indicated by the designating numeral 110 in FIG. 31(a). In accordance with the present invention, if the KCI means identifies a KC 110 located between Z.sub.tmp and Z.sub.curr as shown in FIG. 31(b), such KC would be in the key characteristic data file in the form of Z.sub.KC, which represents the height of the KC and which is available to the thickness calculator means. As shown in FIG. 31(b), the "Error" is the distance along the Z-axis between Z.sub.temp and Z.sub.KC.
In accordance with the present invention, the thickness calculator means includes a means for back tracking the selection of the temporary slice plane at height Z.sub.tmp upon encountering a KC so that the temporary slice plane corresponds to the slice plane lying in the plane of the KC at height Z.sub.KC. Desirably, a computer is programmed to function as the back tracking means. The back tracking means is programmed conveniently on the same computer on which the thickness calculator means, the key characteristic identifier means and the facet processor means is programmed. As embodied herein, the back tracking means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such program for the back tracking means is contained in pages 25-29 of the Appendix which forms a part of this patent application. As shown in FIG. 31(c), the back tracking means assumes that Z.sub.tmp has the value of Z.sub.KC, and chooses this value as Z.sub.next.
Second, in accordance with the present invention, the thickness calculator means includes a means for accommodating the presence of a sudden change in the degree of surface complexity in the intended object. The means for accommodating the presence of a sudden change in the degree of surface complexity is complicated and thus merits the following detailed discussion. In accordance with the present invention, in order to ensure retention of a sudden change in the degree of surface complexity of the intended object, the thickness calculator means implements a back tracking strategy in the process of determining the intersection of the next layer in the model of the intended object. As the name suggests, the back tracking means adjusts the position (forward and backward) of the slice plane along the slice axis until a pre-selected attribute has been satisfied. Back tracking becomes desirable in an adaptive slicing process because the thickness of the next layer, and accordingly the height of next slice plane above the current slice plane, are not known beforehand.
When employing back tracking to ensure retention of a sudden change in the degree of surface complexity of the intended object, the thickness calculator means of the present invention ignores the notion that the thickness of the next layer depends on the current complexity level (or may be derived from the current chain of facets). Instead, when the thickness calculator of the present invention employs the back tracking to ensure retention of a sudden change in the degree of surface complexity of the intended object, the criteria used by the thickness calculator means to select the thickness (l.sub.next) of the next layer are: (1) the number of changes in complexity that occur within a distance of l.sub.max from the current slice plane and (2) the effect of each change in complexity that occurs within the same distance l.sub.max from the current slice plane. If the number of changes in complexity within that distance l.sub.max from the current slice plane is "n," then the effect of each one of them must be considered by the back tracking means in order to determine l.sub.next in accordance with the present invention.
The back tracking means used by the thickness calculator means to ensure retention of a sudden change in the degree of surface complexity of the intended object, can be classified into two types: (1) simple back tracking and (2) repeated back tracking. In accordance with the present invention, the thickness calculator means employs a simple back tracking means for cases having only one change of complexity and is explained first. Desirably, a computer is programmed to function as the simple back tracking means. The simple back tracking means is programmed conveniently on the same computer on which the thickness calculator means, the key characteristic identifier means and the facet processor means are programmed. As embodied herein, the simple back tracking means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such program for the simple back tracking means is contained in pages 25-29 of the Appendix which forms a part of this patent application.
To illustrate an example of simple back tracking, assume a surface with a one time slope discontinuity located at "P" immediately above the position Z.sub.curr of the current slice plane as shown in FIG. 33. The letter "P" designates the transition point between the two complexity levels, each complexity level being indicated by the respective slopes .THETA..sub.1 and .THETA..sub.2 of the cross-section of the desired object. In accordance with the present invention, whenever the complexity level of the profile increases when moving along the profile from a point just before P to a point just beyond P, the thickness calculator means uses the simple back tracking means to determine the position Z.sub.next of the next slice plane. In further accordance with the present invention, when the complexity level of the profile decreases when moving along the profile from a point just before P to a point just beyond P, the thickness calculator means forgoes employing simple back tracking to determine the position of the next slice plane. This can be proved mathematically as follows.
As shown in FIG. 33, the position of the current slice plane is denoted by Z.sub.curr, the starting assumption for the thickness of the next layer is denoted by l.sub.tmp1, and the slope of the cross-section of the desired object that intersects the position Z.sub.curr of the current slice plane is indicated by .THETA..sub.1. After reaching Z.sub.curr, the initial assumption for the thickness of the next layer thickness l.sub.tmp1, the "temporary" thickness of the next layer so to speak, is calculated by the thickness calculator means with respect to .THETA..sub.1. As can be seen, the location of P provides a convenient point of demarcation for dividing 1.sub.tmp1 into two components "a" and "b." Component "a" is the portion of l.sub.tmp1 before the change of complexity from .THETA..sub.1 to .THETA..sub.2, and component "b" is the portion of l.sub.tmp1 after the change of complexity from .THETA..sub.1 to .THETA..sub.2. A second temporary assumption for the thickness of the next layer thickness is denoted in FIG. 33 by l.sub.tmp2, and l.sub.tmp2 is then calculated considering the new complexity .THETA..sub.2. Note that l.sub.tmp2 is measured from P, the transition point between the two complexity levels .THETA..sub.1 and .THETA..sub.2.
In a first case, a decrease in the complexity level is indicated to the thickness calculator means if the slope of the cross-section of the surface of the intended object intersecting Z.sub.curr is greater than the slope of the cross-section of the surface of the intended object on the other side of P, i.e., .THETA..sub.1 >.THETA..sub.2. In this first case, since layer thickness increases as the complexity level decreases, l.sub.tmp2 will be more than l.sub.tmp1. Thus, the thickness calculator means need not consider l.sub.tmp2 for the simple reason that the surface attribute will exceed the allowable limit l.sub.max chosen by the user. In accordance with the present invention, the thickness calculator means does not employ the simple back tracking means to determine the location (Z.sub.next) of the next slice plane because the location (Z.sub.next) of the next slice plane is already calculated based on l.sub.tmp1. This corresponds to the case Z.sub.next =Z.sub.tmp1.
In a second case, no change in complexity level is indicated to the thickness calculator means if the slope of the cross-section of the surface of the intended object intersecting Z.sub.curr equals the slope of the cross-section of the surface of the intended object on the other side of P, i.e., .THETA..sub.1 =.THETA..sub.2. In this second case, l.sub.tmp2 will be equal to l.sub.tmp1, and in accordance with the present invention, the thickness calculator means is configured so that it does not employ the simple back tracking means to determine the position Z.sub.next of the next slice plane. This again corresponds to the case Z.sub.next =Z.sub.tmp1.
In a third case, an increase in complexity level is indicated to the thickness calculator means if the slope of the cross-section of the surface of the intended object intersecting Z.sub.curr is less than the slope of the cross-section of the surface of the intended object on the other side of P, i.e., .THETA..sub.1 <.THETA..sub.2. In this third case, which is shown in FIG. 33, since layer thickness decreases as the complexity level increases, l.sub.tmp2 will be less than l.sub.tmp1. Then if l.sub.tmp2 >b then Z.sub.next =Z.sub.tmp, and in accordance with the present invention, the thickness calculator means does not employ the simple back tracking means to determine the location of the next slice plane (and accordingly the thickness of the next layer in the model of the intended object). Similarly, if l.sub.tmp2 =b, then Z.sub.next =Z.sub.tmp, and in accordance with the present invention, the thickness calculator means does not employ the simple back tracking means to determine the location of the next slice plane (and accordingly the thickness of the next layer in the model of the intended object). Finally, if l.sub.tmp2 <b, then Z.sub.next =Z.sub.curr +a+l.sub.tmp2, and in accordance with the present invention, the thickness calculator means employs the simple back tracking means to back track the position Z.sub.tmp1 of the initial temporary slice plane to the position Z.sub.curr +a+l.sub.tmp2.
Two important observations can be made from the above analysis. First, the thickness calculator means is configured so that it does not employ the simple back tracking means when there is a sudden decrease in complexity level. Second, the thickness calculator means is configured to employ the simple back tracking means only in those certain cases in which the complexity level increases suddenly. In accordance with the present invention, the thickness calculator means is configured to decide to employ the simple back tracking means based not only on the value of the new complexity level relative to the current complexity level but also based on the location of the current slice plane Z.sub.curr with respect to the transition point (P) denoting the change from the current complexity level to the new complexity level.
However, as illustrated in FIG. 34 for example, a thickness calculator means configured with a simple back tracking means as explained above, fails to retain sudden changes in the complexity level when determining the position of the next slice plane if multiple levels of complexity are encountered between the current slice plane Z.sub.curr and the initial temporary slice plane Z.sub.tmp1. Therefore, in accordance with the present invention, the thickness calculator means is configured to employ a repeated back tracking means to handle this fourth case of multiple levels of complexity disposed between the current slice plane Z.sub.curr and the initial temporary slice plane Z.sub.tmp1. The configuration of the repeated back tracking means employed by the thickness calculator means for handling this fourth case is based on the fact that there will be as many transition points P.sub.1, P.sub.2, P.sub.3, . . . P.sub.n, as the number of complexity levels .THETA..sub.1, .THETA..sub.2, .THETA..sub.3, . . . .THETA..sub.n. Desirably, a computer is programmed to function as the repeated back tracking means. The repeated back tracking means is programmed conveniently on the same computer on which the thickness calculator means, the key characteristic identifier means and the facet processor means are programmed. As embodied herein, the repeated back tracking means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such program for the repeated back tracking means is contained in pages 26-27 of the Appendix which forms a part of this patent application.
In accordance with the present invention, after reaching Z.sub.tmp1, the thickness calculator means is configured to initiate a search to find any transition point P lying between the current slice plane Z.sub.curr and the initial temporary slice plane Z.sub.tmp1. If such search finds a transition point, the thickness calculator means is configured to initiate an additional procedure. According to this additional procedure, the thickness calculator means is configured to calculate the height of a new slice plane from the transition point P.sub.c that is closest to Z.sub.curr. The transition point P.sub.c may be referred to as the lowest of all of the transition points. If the height of the new slice plane is more than or equal to Z.sub.tmp1, the thickness calculator means ignores this transition point P.sub.c closest to Z.sub.curr and focuses on the transition point P.sub.c+1 immediately above P.sub.c. The transition point P.sub.c+1 is the second closest transition point to Z.sub.curr. If the height of the new slice plane is less than Z.sub.tmp1, then the thickness calculator means replaces Z.sub.tmp1 with the new height. Then, the thickness calculator means considers the group of transition points between the current transition point (P.sub.c or P.sub.c+1) and Z.sub.tmp1, and again employs the repeated back tracking means as described above until each transition point has been checked.
In accordance with the present invention, the thickness calculator means identifies transition points, which is the key to a successful implementation of this logic, by looking for new facet groups between Z.sub.curr and Z.sub.tmp1. Each new facet group corresponds to a possible change in the complexity level. When the thickness calculator means identifies facet groups between Z.sub.curr and Z.sub.tmp1, the thickness calculator means first considers the facet group that is closest to Z.sub.curr. The complexity level of the surface being represented by the new facet group is determined by the thickness calculator means from the new facet group's controlling normal, which is the normal that makes the greatest angle with the current slice plane Z.sub.curr and must belong to a facet through which the current slice plane Z.sub.curr is passing. If there is an increase in complexity, then the thickness calculator means employs the simple back tracking means to apply the procedure described above in the third case for .THETA..sub.1 <.THETA..sub.2. If back tracking is not necessary, then the thickness calculator means considers the facet group immediately above the earlier group and repeats the procedure.
However, when back tracking is performed by the thickness calculator means, then Z.sub.tmp1 assumes a new value. Then, all groups lying between the current group and the new Z.sub.tmp1 are considered by the thickness calculator means, which repeats the procedure until there are no more facet groups to be checked.
These steps are followed by the thickness calculator means until the top of the intended object is reached. Since the top face of the intended object is considered a key characteristic (KC), the last slice plane inevitably passes through the top face of the intended object. This completes adaptive slicing of the model of the intended object. The flow chart shown in FIG. 35 schematically illustrates the various steps involved in adaptive slicing as performed by the thickness calculator means of the present invention.
In accordance with the present invention, a slicer means is provided for operating on the sorted facet file, the grouped facet file, the subgrouped facet file and the layer thickness data file to compute the intersection between each slice plane and each facet of the STL file at any given height and in the way that is least expensive, i.e., uses computer time most efficiently. In accordance with the present invention as shown schematically in FIG. 4, the thickness calculator means 60 feeds the heights (Z.sub.1, Z.sub.2, Z.sub.3, etc.) in the layer thickness data file to the slicer means 62, and the facet processor means 56 feeds logically arranged facet groups and facet subgroups to the slicer means 62. The slicer means is configured to calculate the intersection of each slice plane disposed at a height above the previous slice plane by the thickness (l.sub.1, l.sub.2, l.sub.3, etc.) calculated by the thickness calculator means for each layer of the model of the intended profile represented by the STL file and to produce a slice plane intersection data file. The slicer means is configured to be programmed on a computer. As embodied herein, the slicer means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such slicer means program is contained in pages 31-41 of the Appendix which forms a part of this patent application. The working details of the slicer means are presented next.
In an exemplary case shown in FIG. 36 for example, a slice plane 120 at a height "h" will have an intersection with a triangle 121 denoting a facet, if the following Condition (2.11) is satisfied.
However, the imposition of Condition (2.11) leads to redundancies. To illustrate how these redundancies might arise, recall the cube example described in FIGS. 1.8 and 1.9. As shown therein, a total of ten triangles, T.sub.3 to T.sub.10 of facet group1 and T.sub.11 and T.sub.12 of facet group2, meet at the top face and share vertices. When the slice plane passes through the top face of the cube, technically each of these 10 triangles will have an intersection. However, this leads to unnecessary and duplicate intersection points. In accordance with the present invention, the slicer means is configured to employ the following Rule 4.1 to avoid such redundancies. Rule 4.1: When a slice plane passes through a common vertex shared by two triangles and the common vertex forms the first vertex (Z.sub.min) of the three vertices of one of the triangles and the last vertex of the three vertices of the other one of the triangles, for all computational purposes, the common vertex is considered to belong to the former triangle. Applying this Rule 4.1 shown in FIG. 37 for example, the common vertex 122 shared by triangle T.sub.1 and triangle T.sub.2 is the Z.sub.min vertex of triangle T.sub.2, and accordingly the computation is performed by the slicer means under the assumption that the common vertex shared by triangle T.sub.1 and triangle T.sub.2 belongs to triangle T.sub.2 only. Now the modified Condition (2.12) for an intersection will be
This Condition (2.12) can be applied to facet groups, to facet subgroups and to individual triangles.
Assume that "h" is the height at which the intended object is to be sliced by the thickness calculator means. There might be one or more facet groups which will have an intersection with a slice plane passing through the intended object at the height "h". Again, there might be one or more facet subgroups in each such facet group which will have an intersection with such slice plane. Applying the Condition (2.12) for intersection derived above, all such facet groups and facet subgroups are identified by the slicer means and merged to form a single new group. This new merged group defines a facet chain. All the facets in this merged group when properly ordered, will form a closed facet chain that is ordered according to a sequence determined by facets sharing common edges or common vertices.
According to the Slicing Axiom defined above, there will always be one continuous chain of triangles through which a slice plane passes. As explained previously, such chains are already obtained in accordance with the present invention by merging facet groups and facet subgroups. Now, the slicer means calculates intersections for each triangle in the chains by using basic analytic geometry. There are two types of intersections, a vertex intersection and an edge intersection, and each requires its own separate treatment by the slicer means in accordance with the present invention.
If the slice plane is passing through the vertex (active vertex) of a triangle (active triangle), the slicer means calculates the intersection point (which is nothing but the coordinates of the vertex) and stores the identity of the active triangle in the slice plane intersection data file in a manner so that the slicer means will not check this triangle for an intersection with the current slice plane a second time. When linked lists in C language are used, the active triangle can be dropped altogether, saving an enormous amount of computation time. For the next intersection, the first facet encountered that shares the active vertex is checked for intersection by the slicer means. This next facet to be checked by the slicer means must have at least one more vertex, other than the active vertex, or an edge through which the slice plane passes. (If not, it is a discontinuous chain, violating the slicing axiom. The STL file may be repaired, if necessary, before proceeding further). The intersection point is calculated by the slicer means and included in the slice plane intersection data file. This process is repeated until the chain of facets that intersect with the slice plane is complete.
If the slice plane is passing through the edge (active edge) of a triangle (active triangle), the slicer means calculates the intersection point and records the active triangle in the slice plane intersection data file in a manner so that the active triangle will not be checked a second time. There will be only one triangle in the chain that shares the active edge, and the slicer means identifies this one triangle by a simple comparison. The slicer means finds the triangle that shares an active edge by checking every triangle in the chain. This is necessary because the triangles are not in any order inside the chain. The slicer means terminates the search only when the slicer means finds the triangle which shares an active edge.
There still remains one more problem. Facets in each facet subgroup are not yet arranged in any geometrically coherent order. In other words, after the facets have been sorted, grouped and subgrouped, it still is not known whether facet T.sub.1 is next to facet T.sub.2 or next to facet T.sub.3 and/or next to facet T.sub.5. Referring to FIG. 10 for instance, it would be convenient to have facets T.sub.3 through T.sub.10 in "geometric order" so as to minimize the time needed to compute the intersection of each facet with the slicing plane (such as shown in FIG. 7).
Moreover, when there is a large number of triangles in the chain, the computation time used by the slicer means will be commensurately more than when the chain includes only a small number of triangles. However, the problems of lack of geometric order and of increasing computation time are addressed together. In accordance with the present invention, a marching means is provided to arrange in "geometric order," the facets which have been sorted, grouped and subgrouped. Desirably, the slicer means is provided with a marching means that generates an adjacency list for each facet (which is a triangle in the STL file) that is intersected by each slicing plane. Once the adjacency list is prepared, the facets of the model are said to be placed in geometric order. The marching means is configured to order the facets of the slice plane intersection data file so that the two neighboring facets of each facet in the slice plane intersection data file are identified as part of the adjacency list for that particular facet. The adjacency list becomes part of the slice plane intersection data file. Desirably, a computer is programmed to function as the marching means. The marching means is programmed conveniently on the same computer on which the slicer means is programmed. As embodied herein, the marching means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such marching means program is contained in pages 35-41 of the Appendix which forms a part of this patent application.
As embodied herein, the marching means examines the first facet in a subgroup of the slice plane intersection data file, and begins preparing an adjacency list for each active triangle. Such adjacency list is prepared by the marching means of the present invention by exploiting one of the properties (noted above) of STL files. According to that property, any edge can be shared by exactly two triangles and no more. Thus, each triangle will have exactly three adjacent triangles, which are listed in each triangle's completed adjacency list. As the intersection points are being calculated, the adjacency list is also prepared by the marching means of the slicer means and updated simultaneously. Every time the slicer means performs a search for a triangle which shares an active edge or an active vertex, the slicer means first consults the adjacency list of the active triangles. If the slicer means finds a matching triangle in one of the adjacency lists of those active triangles, then the slicer means calculates the intersection point. If the slicer means fails to find a matching triangle in the adjacency lists of one of those active triangles, then the next active triangle is identified from the chain of triangles by simple comparison and the intersection point is calculated. Then, the adjacency list of each of the active triangles as well as the active triangle's adjacent triangles, is updated by the marching means. The slicer means repeats the procedure until the chain of triangles in the intersection with the slice plane is complete.
In accordance with the present invention, the adjacency list generated and maintained by the marching means of the slicer means is particularly useful, in case of successive slice planes passing through chains which are similar or nearly similar in composition. For instance, consider two slice planes A--A and B--B as shown in FIG. 32. Both the slice planes are passing through exactly the same chain of triangles T.sub.1 through T.sub.8, which of course, are not in that order. While calculating intersections for slice plane A--A, an adjacency list is prepared simultaneously by the marching means. This adjacency list can be consulted by the slicer means while calculating intersections for slice plane B--B. Because the composition of chains at slice plane A--A and slice plane B--B is exactly the same, the need for checking every triangle until a match is found is eliminated in accordance with the procedure employed by the slicer means of the present invention. When the number of triangles in a chain runs into thousands, adjacency lists will be of immense help. In such cases, the computation time is often reduced by more than a factor of fifteen.
When triangles remain after the search employed by the slicer means returns to the first active edge or the first active vertex that triggered the search, it is an indication that there is at least one more chain, representing a protrusion or a depression. In such cases, the slicer means is configured as shown schematically in FIG. 35 to restart the search process for the remaining triangles. This is done until there are no more chains.
In accordance with the present invention, two important observations can be made from the above procedure adopted by the slicer means for calculating intersections. First, each triangle except the first triangle will have only one intersection per slice plane, and the first triangle will have two intersections. This can be seen from FIG. 38, where a chain with six (6) member triangles is broken and shown in extended form in relation to a slice plane 124. There are seven (7) intersections, for six (6) triangles, with the first triangle T.sub.1 having two intersections. Thus, the slicer means of the present invention completely eliminates duplicate intersection points. Second, the slicer means of the present invention uses the marching means to march from one triangle to its adjacent triangle in calculating intersections between the triangle and the slice plane. Typically, the march starts at the first triangle of the chain, encompasses all the member triangles and ends again at the first triangle where it began. This marching attribute for this procedure justifies the name marching means for this aspect of the procedure employed by the slicer means.
The greatest advantage of the slicer means that employs a marching means, lies in the fact that intersection points are always calculated in the correct sequence and shape of the contour at any slice plane and can be obtained by joining the points of intersection in that order. Accordingly, the marching means enables the slicer means of the present invention to eliminate any need to determine the direction of vectors, etc. Moreover, because the intersections are calculated chain-after-chain for each slice plane, identifying protrusion and depression features becomes much easier. In addition, the marching means enables the slicer means to operate with a significant reduction in overall computation time.
Another interesting problem exists because the intersection points or vectors obtained using the slicer means with the marching means in accordance with the present invention, may be ordered in a clockwise (CW) or counterclockwise (CCW) direction and this direction may vary from contour to contour within a single slice. Accordingly, before the slice plane intersection data file is utilized to build physical models, the present invention employs a directional ordering means that imposes a uniform direction that is maintained. Desirably, a computer is programmed to function as the directional ordering means. The directional ordering means is programmed conveniently on the same computer on which the slicer means and the marching means is programmed. As embodied herein, the directional ordering means is programmed in C language on a Sun Microsystem.RTM. work station. A presently preferred embodiment of such directional ordering means program is contained in pages 42-46 of the Appendix which forms a part of this patent application.
As a first step, the directional ordering means obtains Y.sub.min (or X.sub.min) of the data points of any contour. If the i.sup.th point represents Y.sub.min there will be at least two more points, one point (i-1) preceding the i.sup.th point and one point (i+1) following the i.sup.th point. This condition will never fail because a polygon does not exist without at least three non-co-linear points. As shown in FIGS. 39(a)-39d, if .THETA..sub.1 is the angle subscribed by lines 126, 128 joining the i-1.sub.th point and the i.sub.th point, and .THETA..sub.2 is the angle subscribed by lines 127, 128 joining the i+1.sub.th point and the i.sub.th point, then the conditions for determining the direction of the contour in accordance with the directional ordering means of the present invention are:
When .THETA..sub.1 >.THETA..sub.2 then the direction of the contour is CCW
when .THETA..sub.1 <.THETA..sub.2 then the direction of the contour is CW
when .THETA..sub.1 =.eta..sub.2 then point i-1, point i, and point i+1 are co-linear, meaning that there is an error in the data.
Once the directional ordering means employed by the slicer means of the present invention finds the direction of each contour by applying the above conditions, then the directional ordering means of the present invention makes whatever adjustment, if any, is necessary so that the direction of each contour remains the same, either CW or CCW. In other words, the directional ordering means finds the direction of each contour defining each intersection of each slice plane in the slice plane intersection data file and ensures uniformity of such direction (CW or CCW) with the direction of each other contour defining each other intersection of each other slice plane in the slice plane intersection data file. In the embodiment of the directional ordering means shown in pages 42-46 of the Appendix, the directional ordering means imposes the CW direction.
A few words about the software in the Appendix are appropriate. First, it is necessary that two programs "slice.c" and "reverse.c" be compiled. Both programs were written in C language. The "slice.c" program takes an STL file in ASCII format as input and stores the slice data as output in a file called "int.sub.-- pts" file by default. The "reverse.c" program functions as the directional ordering means and so ensures that the line segments which form each contour are in CW order.
The commands to compile the programs are
cc -g slice.c -lm -o slice
cc -g reverse.c -lm -o reverse
It is important to note that the name of the object code of "reverse.c" must be "reverse" as shown above. The reason being that "reverse" is invoked from inside the object code of "slice c" and no other name will be recognized. However, the name of the object code of slice.c, called "slice" in this document, can be anything.
Before the slicing program can be executed, the STL file of the object being sliced must be in the current directory. This is a command driven program and the input provided by the user, including the STL file, is assumed to be correct and is not counterchecked by the program. Therefore any error in input either aborts the program or outputs garbage. A step by step tutorial on how to use the program is presented here. Underlined words indicate program commands which will be displayed on the console. Words/numerals in bold and within "<>" are examples.
At the command prompt type <slice name of the STL file> The STL file is an argument to the "slice" program and must be typed along with it. The two names should be separated by a white space.
Enter 1 to read sorted STL file, else enter any other number. <enter 2>
A sorted STL file will be created upon the use of "slice" and the file will be named "sorted.stl". This file contains exactly the same information as the original STL file with the difference being that the facets are sorted. Incidentlly, "sorted.stl" file should be renamed since it is written over each time the slicing program is used.
Some information, like number of facets in the STL file, the dimensions of the object, etc., will appear on the console. For uniform slicing enter 1, for adaptive slicing enter any other number. <enter 2>
Enter the maximum slice thickness that you would prefer. <enter 0.005>
Choose a criteria by entering the appropriate number
1. Cusp height
2. Maximum deviation
3. Chord length <enter 1>
Four values c1, c2, c3, c4 of cusp height will be displayed on the screen. Enter a value for limiting cusp height based on these four values. You may take the help of cusp height Vs. theta plot. If you need help enter 911. <enter the value of c2>
The number of slices will be displayed and the program ends. A file named "int.sub.-- pts", which contains the slice data, will appear in the current directory. The "int.sub.-- pts" file, which is short for intersection points, contains intersections between slice plane and triangular facets in an STL file. The format of the "int.sub.-- pts" file is as follows. The first line of the file will have two numbers which are Z-min and Z-max of the intended object. In each of the second line and subsequent lines of the "int.sub.-- pts" file, there will be exactly five numbers in each line. The first two numbers are integers and the first number represents the slice number while the second number represents the island number. The next three numbers are x, y and z coordinates of the particular intersection point. An example "int.sub.-- pts" file is shown below. Beneath each row of numbers, an explanation of such numbers is written within the parentheses. ##EQU13##
The slice data generated and stored in the form of "int.sub.-- pts" file can be used to produce the desired object using any layered process. However, typically the interface means is required to postprocess the slice data, depending on the process chosen. For example, postprocessing of the slice data will be shown for CNC milling and stereolithography processes.
The slice data obtained above can be postprocessed using the CNC machining software in the appendix of U.S. application Ser. No. 08/156,321, now U.S. Pat. No. 5,432,704, to obtain the CNC code which can be used to produce the part on any 3-axis CNC milling machine. But, first it is necessary to convert the slice data stored in "int.sub.-- pts" into a format compatible with the CNC machining software. An embodiment of the interface means capable of converting the slice data stored in "int.sub.-- pts" into a format compatible with the CNC machining software is the program called "tata.sub.-- to.sub.-- vouze.c" (in Appendix pages 176-179). This "tata.sub.-- to.sub.-- vouze.c" program assumes that "int.sub.-- pts" file is in the current directory and looks for it.
Compile "tata.sub.-- to.sub.-- vouze.c" as shown below.
cc -g tata.sub.-- to.sub.-- vouze.c -o tata.sub.-- to.sub.-- vouze
At the command prompt enter <tata.sub.-- to.sub.-- vouze >
A file named "vouze" will be created in the current directory, which contains the slice data in binary format and is the input to the CNC code generator in the appendix of U.S. application Ser. No. 08/156,321, now U.S. Pat. No. 5,432,704.
As noted above, the interface means desirably includes a computer that is programmed with software that converts the STL data file 64 output into a format compatible with the machine to be controlled. If the machine to be controlled is a SLA-250, a presently preferred embodiment of such component of the interface means program is contained in pages 96-175 of the Appendix which forms a part of this patent application. The slice data needs to be postprocessed to obtain a .sli file, which is necessary to produce the part on an SLA-250 . This includes hatching all of the closed contours. In order to obtain the .sli file, the following files are required to be in the current directory.
main.c (Appendix pages 96-97)
io.c (Appendix pages 98-116)
cides.h (Appendix pages 117-123)
3d.sub.-- io.h (Appendix pages 124-126)
seg.h (Appendix page 127)
file.h (Appendix pages 128-129)
read.sub.-- brdrpts.c (Appendix pages 130-132)
user.sub.-- input.c (Appendix pages 133-135)
hatch.sub.-- cal.c (Appendix pages 136-168)
new.sub.-- sli.c (Appendix pages 169-174)
header (Appendix page 175)
The following command "cc -g main.c io.c -lm -o create.sub.-- sli" can be used to compile the program. Next, at the command prompt, type in <create.sub.-- sli >. The program assumes that the "int.sub.-- pts" file is in the current directory. In the following step-by-step tutorial, underlined words indicate program commands that will be displayed on the computer console. Words/numerals in bold and within "<>" are examples.
For Clemson hatching style enter 1, else any other number E.g., <1>
Enter hatch Spacing for bottom and top most layers E.g., <0.05>
Enter number of hatch angles followed by their values E.g., <3 090 120>
Enter the number of different zones (Z heights) you would like to have for hatching. If you type in 1, there will be only one zone which includes all layers other than bottom and top layers. E.g., <2>
Enter hatch spacing for zone 1 which starts from second layer from bottom E.g., <0.1>
Enter number of hatch angles followed by their values for zone 1 E.g., <2 45 135>
Enter starting z-height for zone 2 E.g., <0.7> (Note: z.sub.-- min and z.sub.-- max of this part are 0.5 and 1.5. Compute z.sub.-- height accordingly)
Enter hatch spacing for zone 2 E.g., <0.03>
Enter number of hatch angles followed by their values for zone 2 E.g., <3 10 40 70>This ends the execution of create sli file, and a file named "Clemson.sli" appears in the current directory. Clemson.sli can be used to produce the part on any SLA-250 .
In order to visualize sliced models, a separate program "visual.sub.-- slice.c" was developed using HOOPS graphic routines. The following command compiles the program.
hcc -g visual.sub.-- slice.c -o visual.sub.-- slice
Note the usage of "hcc" for compiling HOOPS routines. Usage of "visual.sub.-- slice" is the same as "slice" which is explained above. ##SPC1##
Claims
What is claimed is:
1. A device for effecting automatic operation of a stereolithography apparatus (SLA) with respect to a part to be built by the SLA, the part having a desired profile characterized by at least one key characteristic and at least two local complexity levels, wherein the operation of the SLA is controlled based on a two-dimensional sliced model of the local profile of the cross-section of the part to be built with each slice plane disposed transversely with respect to and along a slice axis of the part, which is represented by a tesselated model of the part and provided as an input to the device in the form of an STL file specifying triangular facets obeying the vertex-to-vertex rule, the device comprising: a programmable computer; a facet processor means for operating on the STL file input to produce a processed facet file, said facet processor means being programmed on said computer, said facet processor means including a facet sorting means, said facet sorting means being configured to sort the facets of the STL file according to the slice axis to produce a sorted facet file; said facet processor means including a facet grouping means for operating on said sorted facet file to produce a grouped facet file, said facet grouping means being configured to group the facets of said sorted facet file according to facets having common minimum vertex values with respect to the slice axis and to produce a grouped facet file; said facet processor means including a facet subgrouping means for operating on said grouped facet file to produce a subgrouped facet file, said facet subgrouping means being configured to subgroup the facets of said grouped facet file according to facets having common maximum vertex values with respect to the slice axis and to produce a subgrouped facet file; a key characteristic identifier means for operating on said sorted facet file, said grouped facet file and said subgrouped facet file, said key characteristic identifier means being programmed on said computer, said key characteristic identifier means being configured to identify key characteristics of the profile represented by the STL file and to produce a key characteristic data file from said key characteristics of the profile represented by the STL file; a thickness calculator means for operating on said sorted facet file, said grouped facet file, said subgrouped facet file and said key characteristic data file, said thickness calculator means being programmed on said computer, said thickness calculator means being configured to calculate a thickness for each layer of the model of the desired profile represented by the STL file of the part such that the geometrical error between the desired profile of the part and said model profile of each said layer having said layer thickness, remains no greater than a preselected geometrical error and to produce a layer thickness data file containing the results of such calculations; a slicer means for operating on said sorted facet file, said grouped facet file, said subgrouped facet file and said layer thickness data file, said slicer means being programmed on said computer, said slicer means being configured to calculate the intersection of each slice plane disposed at a height above the previous slice plane by said thickness calculated by said thickness calculator means for each layer of the model of the intended profile represented by the STL file and to produce a slice plane intersection data file; a directional ordering means for finding the direction of each contour defining each intersection of each slice plane and ensuring uniformity of such direction with the direction of each other contour defining each other intersection of each other slice plane; a model generating means for using said layer thickness data file and said slice plane intersection data file to generate at least a portion of a model of the part wherein said portion of said model of the part includes a plurality of successive layers wherein the cross-section of each layer in a plane of view is defined by the intersection with said plane of view, of a pair of parallel planes and a model profile connecting said parallel planes, and wherein for each layer the minimum distance separating its pair of parallel planes defines the thickness of said layer; and an interface means for controlling the operation of the machine based on said portion of a model of the part.
2. An apparatus for effecting automatic operation of a machine with respect to an intended object having a desired profile characterized by at least one key characteristic and at least two local complexity levels, wherein the intended object is represented by a tesselated model of the intended object oriented with respect to a slice axis and provided as an input to the apparatus in the form of an STL file specifying triangular facets obeying the vertex-to-vertex rule, the apparatus comprising: a programmable computer; a facet processor means for operating on the STL file input to produce a processed facet file, said facet processor means being programmed on said computer, said facet processor means including a facet sorting means, said facet sorting means being configured to produce said sorted facet file by sorting the facets of the STL file according to the slice axis; and said facet processor means including a facet grouping means for operating on said sorted facet file to produce a grouped facet file, said facet grouping means being configured to produce said grouped facet file by grouping the facets of said sorted facet file according to facets having one of the following characteristics: common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis.
3. An apparatus as in claim 2, further comprising: a facet subgrouping means for operating on said grouped facet file to produce a subgrouped facet file, said facet subgrouping means being programmed on said computer and configured to subgroup the facets of said grouped facet file according to facets having the other of: common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis.
4. An apparatus as in claim 2, further comprising: a thickness calculator means for operating on at least said sorted facet file and said grouped facet file to produce a layer thickness data file defining the profile of the tesselated model in terms of at least two stacked layers, said thickness calculator means being programmed on said computer.
5. An apparatus as in claim 4, wherein: said thickness calculator means being configured to produce said layer thickness data file by calculating a thickness for each layer defining the tesselated model of the desired profile represented by the STL file of the intended object such that the geometrical error between the desired profile of the intended object and said model profile of each said layer having said layer thickness, remains no greater than a preselected geometrical error.
6. An apparatus as in claim 4, further comprising: a slicer means for operating on at least said sorted facet file, said grouped facet file, and said layer thickness data file to produce a slice plane intersection data file, said slicer means being programmed on said computer, said slicer means being configured to produce a slice plane intersection data file by calculating the intersection of each slice plane disposed at a height above the previous slice plane by said thickness calculated for each layer defining the tesselated model of the intended profile represented by the STL file.
7. An apparatus as in claim 6, further comprising: an interface means for controlling the operation of the machine based on a two-dimensional sliced model of the local profile of the cross-section of the intended object with each slice plane disposed transversely with respect to and along the slice axis.
8. An apparatus as in claim 5, wherein said thickness calculator means includes a simple back tracking means for determining the thickness of the next layer while retaining a single change in complexity level in the local profile.
9. An apparatus as in claim 5, wherein said thickness calculator means includes a repeated back tracking means for determining the thickness of the next layer while retaining more than a single change in complexity level in the local profile.
10. An apparatus as in claim 4, wherein said thickness calculator means includes a means for selecting the thickness of the next layer based on a predetermined geometrical parameter relating the selected thickness of the next layer to the profile of the intended object in the vicinity of the next layer, said predetermined geometrical parameter being the volumetric error per unit of perimeter.
11. An apparatus as in claim 6, further comprising: a marching means for ordering at least said grouped facet file to produce a geometrically ordered facet file, said marching means being programmed on said computer and configured to geometrically order the facets of at least said grouped facet file to produce a continuous chain of facets wherein each facet in said continuous chain intersects one of the slice planes.
12. An apparatus as in claim 11, wherein said marching means is configured to generate an adjacency list for each facet that is intersected by the one slice plane.
13. An apparatus as in claim 2, further comprising: a key characteristic identifier means for operating on at least said sorted facet file and said grouped facet file, said key characteristic identifier means being programmed on said computer, said key characteristic identifier means being configured to identify key characteristics of the profile represented by the STL file and to produce a key characteristic data file from said key characteristics of the profile represented by the STL file.
14. An apparatus as in claim 4, wherein said thickness calculator means includes a back tracking means for determining the thickness of the next layer while retaining at least one key characteristic in the local profile.
15. An apparatus as in claim 6, further comprising: a model generating means for using said layer thickness data file and said slice plane intersection data file to generate at least a portion of a model of the intended object wherein said portion of said model of the intended object includes a plurality of successive layers wherein the cross-section of each layer in a plane of view is defined by the intersection with said plane of view, of a pair of parallel planes and a model profile connecting said parallel planes, and wherein for each layer the minimum distance separating its pair of parallel planes defines the thickness of said layer.
16. An apparatus as in claim 18, further comprising: an interface means for converting said generated portion of said model of the intended object in a format for operating the machine with respect to at least a corresponding portion of the intended object in successive steps with each said step based on a separate one of said layers of said generated portion of said model of the intended object.
17. A method involving automatic operation of a machine with respect to an intended three-dimensional object having a desired profile characterized by at least two local complexity levels, wherein the operation of the machine is controlled based on a two-dimensional sliced model of the profile of the cross-section of the intended object along a slice axis of the intended object, which is representated by a three-dimensional tesselated model of the three-dimensional intended object and provided as an input in the form of an STL file specifying triangular facets obeying the vertex-to-vertex rule, the method comprising: using a computer to sort the facets of the STL file according to the slice axis; using a computer to group said sorted facets according to facets having one of common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis; generating at least a portion of a model of the object, said portion of said model including a plurality of successive layers wherein the cross-section of each layer in a plane of view is defined by the intersection with said plane of view, of a pair of parallel planes and a model profile connecting said parallel planes, and wherein for each layer the distance separating its pair of parallel planes defines the thickness of said layer; and operating the machine with respect to at least a portion of the object in successive steps with each said step based on a separate one of said layers.
18. A method as in claim 17, further comprising: using a computer to subgroup said grouped facets according to facets having the other of common minimum vertex values with respect to the slice axis and common maximum vertex values with respect to the slice axis.
19. A method as in claim 17, wherein said step of generating at least said portion of said model of said object includes selecting each layer thickness such that the geometrical error between the desired profile of said portion of said object and said model profile of said layer with said layer thickness, remains no greater than a preselected geometrical error; wherein more than one layer thickness is selected during said step of generating at least said portion of said object.
20. A method as in claim 17, further comprising: before the step of generating at least said portion of the model of the object, identifying each key characteristic in the intended object and using a back tracking means for retaining in said portion of said model each said identified key characteristic of said portion of said model.
21. A method as in claim 17, further comprising: before the step of generating at least said portion of the model of the object, identifying each change in complexity level in the local profile in the intended object and using a back tracking means for retaining in said portion of said model each said identified change in complexity level of said portion of said model.
22. A method involving automatic operation of a machine with respect to an intended three-dimensional object having a desired profile characterized by at least two local complexity levels, wherein the operation of the machine is controlled based on a two-dimensional sliced model of the profile of the cross-section of the intended object along a slice axis of the intended object, which is representated by a three-dimensional tesselated model of the three-dimensional intended object and provided as an input in the form of an STL file specifying triangular facets obeying the vertex-to-vertex rule, the method comprising: producing from the STL file input a sorted facet file that is sorted according to the slice axis; producing from said sorted facet file a grouped facet file that is grouped according to facets having one of the following: common minimum vertex values with respect to said slice axis and common maximum vertex values with respect to said slice axis; producing from said grouped facet file a subgrouped facet file that is subgrouped according to facets having the other of the following: common minimum vertex values with respect to said slice axis and common maximum vertex values with respect to said slice axis; identifying from said grouped facet file and said subgrouped facet file at least one key characteristic of the profile represented by the STL file and producing a key characteristic data file containing each key characteristic so identified; using said grouped facet file, said subgrouped facet file and said key characteristic data file to calculate a thickness for each layer of the model of the desired profile represented by the STL file of the intended object such that the geometrical error between the desired profile of the intended object and said model profile of each said layer having said layer thickness, remains no greater than a preselected geometrical error and producing a layer thickness data file containing the result of each such calculation; using said grouped facet file, said subgrouped facet file and said layer thickness data file to calculate the intersection of each slice plane disposed at a height above the previous slice plane by said thickness calculated for each layer of the model of the intended profile represented by the STL file and producing a slice plane intersection data file; using said layer thickness data file and said slice plane intersection data file to generate at least a portion of a model of the intended object, said portion of said model including a plurality of successive layers wherein the cross-section of each layer in a plane of view is defined by the intersection with said plane of view, of a pair of parallel planes and a model profile connecting said parallel planes, and wherein for each layer the minimum distance separating its pair of parallel planes defines the thickness of said layer; wherein more than one layer thickness is selected during said step of generating at least said portion of said model of the intended object; and operating the machine with respect to at least a portion of the intended object in successive steps with each said step based on a separate one of said layers of said model of said intended object.
Patent Citations (15)
| Patent | Date | Inventor | Cited By |
|---|---|---|---|
| US4618924 | 1986-10-01 | Hinds | |
| US4752352 | 1988-06-01 | Feygin | |
| US4789931 | 1988-12-01 | Kuragano et al. | |
| US4837703 | 1989-06-01 | Kakazu et al. | |
| US4863538 | 1989-09-01 | Deckard | |
| US4866631 | 1989-09-01 | Kuragano et al. | |
| US4907164 | 1990-03-01 | Guyder | |
| US4938816 | 1990-07-01 | Beaman et al. | |
| US4944817 | 1990-07-01 | Bourell et al. | |
| US4949270 | 1990-08-01 | Shima et al. | |
| US4961041 | 1990-10-01 | Seki et al. | |
| US4996010 | 1991-02-01 | Modrek | |
| US5017753 | 1991-05-01 | Deckard | |
| US5053090 | 1991-10-01 | Beaman et al. | |
| US5432704 | 1995-07-01 | Vouzelano et al. |
Non-Patent Literature (2)
- Vouzelaud et al, "Solid Freeform Fabrication Proceedings", Dec. 1992 pp. 291-300.
- Dolenc et al, "Slicing Procedures for Layered Manufacturing Techniques," Mar. 1993, pp. 1-13.