Automatically Optimized NC Tool Path Generation for Machining

This invention relates to a method based on solid modeling for automatically generating tool paths to achieve more efficient milling.

Efficient 21/2-axis milling is the key to fast production and cost reduction in the milling of molds and dies. In this type of machining the z depth is set and all tool motion is in the x and y planes, another z depth is set and so on. For any one set of tool paths, z is constant. Efficiency requires accurate tool paths, a minimized number of tool retractions, and automation with flexibility. Solid modeling systems can define parts to be milled, and milling machines can follow programmed paths faithfully, but the step from a solid model to an optimized process for automatically generating the numerical control (NC) code has been lacking and is addressed by this invention.

A solid model of a part to be machined may be constructed in a computer using the principles of constructive solid geometry by adding and subtracting a basic set of primitive shapes; this is explained in U.S. Pat. No. 4,618,924 - J. K. Hinds. After the model is defined by means of boolean operations on solids, the stock from which the model is to be milled, and avoidance and containment regions must be modeled. A typical avoidance region is a clamp to hold the workpiece, which can push against the sides of a workpiece or can clamp over the top of it. In either case, the centerline of a milling tool cannot move along the edge of a workpiece in the region of the clamp without milling the clamp. Containment regions may be needed when milling complex surfaces where it is necessary to generate one setup, mill a portion of the part, change the setup and mill a portion, etc. For each setup only a portion of the part is milled. Such an approach cannot be accomplished unless the system allows the user to define containment regions for each setup. Having the conglomerate model, part plus stock plus avoidance and containment regions, maximum and minimum z levels are identified from the conglomerate, and planar sections are passed through the workpiece from the maximum to the minimum level as discussed by Hinds. Tool paths are defined for each planar section, and the aggregate of these tool paths define the motion for the tool in the milling process. Thus the central problem to be solved involves the definition of optimal tool paths and retraction requirements on a single plane.

It is a simple matter to accumulate the optimized results as they are obtained on successive planes. The conglomerate defines the total milling process, which can be used with any milling machine. Machine code for a specific milling machine is generated in a post processing phase since each machine requires its own code format.

Tool paths are curves that describe the route for the centerline of the tool. Since the tool has a diameter, at all times these centerlines must maintain a minimum distance of the tool radius from the boundaries. In addition to the tool radius (HRAD), the user must be able to specify the distance between tool paths (STEPOVER) and extra thickness desired on the finished boundaries (THICK). All tool paths generated must honor these distance requirements within a fine tolerance.

The NC optimization process being described defines the generation of gouge-free tool paths on a single plane, optimizes the ordering of tool paths, and specifies required retractions. It optimizes the process of milling cavities which define the mold or die for forging parts. All tool paths are contained within the cavity. The workpiece surrounds the cavity and is of size and shape suitable for the purposes of the forging operation, but is not of concern for this invention. Hence the workpiece is shown only once and as a rectangular slab in the figures.

A consideration is that the cavity to be milled is of arbitrary shape and may have deep, concave portions and/or narrow strictures. As the tool paths are generated, major challenges arise with maintaining data concerning the position of each tool path relative to the others to determine accessibility of the tool from one path to another without gouging. The technique developed involves establishing algorithmic rules to nest, or group, certain sets of tool paths and then deriving another set of algorithmic processes to mill these nests in an expedicious manner while honoring milling guidelines and minimizing the number of required retractions. These algorithmic processes are the essence of the invention.

Constant interference checking is mandatory to prevent gouging of the boundaries. A problem in the development of this process is the handling of avoidance and containment regions. Another consideration is that pocketing tool paths, which are closed loop tool paths to clear material inside cavity boundaries, are generated for flat-end cutters only. Paths are also appropriate for ball-end cutters when the curve is a section cut of an offset surface. Thus the optimized system needs a mechanism for knowing when the part is an offset surface.

Finishing curves are tool paths closest to cavity boundaries. Quality molds require smooth boundaries, which puts constraints on the manner in which a cutting tool approaches and leaves finishing paths. Predrilling and straight line motion from neighboring arcs can cause dwell marks where the tool is deflected and seats in a location before traversing a boundary. To prevent this, the process should augment all finishing curves with tangential approach and exit arcs, and determine that there is sufficient room for the cutter tool diameter at all points of the augmented curve.

An object of the invention is to automate optimal milling especially for clearing material within cavities for molds and dies.

Another object is to define the tool motion for milling a cavity on a plane by automatically generating the geometric representation of all tool paths and of tool motion between tool paths, provide the ordering of tool paths to accomplish optimal milling, limit retractions of the milling tool, and specify predrill locations for required tool plunges.

Yet another object is to provide for containment regions to which milling is restricted and avoidance regions where milling is prohibited.

Still another object is to isolate finishing tool paths from roughing tool paths, and augment the former with tangential approach and exit routes.

According to one aspect of the invention, a method of generating optimized tool paths for milling mold and die parts with a cutting tool comprises providing a three-dimensional solid model of a part cavity to be milled, which optionally has one or more islands that are not milled, and taking a planar section through the model and obtaining a closed curve which describes boundaries of the cavity and has a plurality of loops oriented such that material to be cut is on the left. A first inset of the section curve is obtained, each loop of which is a finishing tool path across which the cutting tool cannot move without gouging and each starting a nest of tool paths. Other inset curves are generated repeatedly that are a constant distance from the previous inset; each loop is a roughing tool path and is a member of one nest or starts a new nest. This process stops when a last inset curve has no arcs. The roughing tool paths are ordered in an optimized sequence to mill the cavity so as to follow a set of ordering guidelines while minimizing retractions of the tool to go to another roughing tool path, and tool motion paths to move between roughing tool paths are generated. The finishing tool paths are then augmented with tangential approach and finish arcs.

Other features of the method are that all of the tool paths are numbered in sequence so that innermost tool paths have the highest numbers, and a nest is milled from innermost to outermost tool paths unless an ordering guideline is violated. A list of numbered tool paths is outputted in the order to be milled separated by a retraction symbol, and a list of predrill locations at starting points of the tool paths where the cutting tool plunges from a retract position.

Additional features of the step of ordering tool paths in an optimized sequence are that the tool motion paths between roughing tool paths are straight lines. Frequent interference checks are made to determine that there is no intersection with the first inset curve, with an unmilled nest of tool paths, and with an about-to-be-milled tool path except at a starting point. The ordering guidelines were generated as part of this invention and are: (a) gouge free milling is required, (b) prevent milling both sides of one tool path before milling the tool path itself, (c) prevent cutting all or part of one tool path more than once, (d) mill the tool paths that are innermost first, (e) finishing tool paths are milled separately from roughing paths and last, (f) the tool intersects the path-to-be-milled only at the starting point, (g) the tool intersects no unmilled nest of tool paths, and (h) the tool may cross milled nests of tool paths.

Another aspect of the invention is that the tool path optimization procedure makes provision for containment and/or avoidance regions. A 3D solid model of the part cavity is constructed and of the containment and avoidance regions, and a milling region is formed by intersecting the cavity and any containment region and obtaining the Volume common to both, and subtracting therefrom any avoidance region. The method after this initial step is the same; there is no change in the nesting and optimization algorithms.

FIG. 1 is a flowchart giving an overview of the tool path optimization procedure.

FIG. 2 shows a closed curve comprising four loops.

FIG. 2A shows a closed curve comprising two loops. FIGS. 2 and 2A show one closed curve comprising six loops.

FIG. 3 is a simplified cross section through a mold illustrating the cavity to be milled.

FIG. 4 shows the tool paths on a single z-level plane during the milling of a cavity from a rectangular slab.

FIG. 5 shows the roughing tool paths to mill the cavity at the section in FIG. 4, ordered for optimized milling. The ordered list of tool paths, predrill coordinates, and nesting of tool paths are printed.

FIG. 6 shows the finishing tool paths for the section in FIG. 4 augmented with tangential approach and exit arcs;

FIG. 7 shows a containment region imposed on the milling process of the cavity illustrated in FIGS. 4-6, and the reordered tool paths.

FIG. 8 shows the same section and the optimization of tool paths to mill those portions of the cavity outside the boundary of an avoidance region.

FIG. 9 is an example of optimized tool path organization for a cavity having deeply concave recesses and a narrow stricture.

FIG. 10 shows tool paths during the milling of one plane of a cavity having a more complicated configuration.

The major steps of the optimizing process are given in boxes 100-103 in FIG. 1. The user designs a computerized solid model of the part and any containment and avoidance regions. The construction of a solid model using a solid modeling system is explained in U.S. Pat. No. 4,618,924, the disclosure of which is incorporated herein by reference. Although other solid modelers may be used, the preferred solid modeling system is General Electric's TRUCE (Tri-Dimensional Rational Unified Cubic Engine), which is described in the TRUCE User's Manual, July 1984. A milling region is formed as the boolean intersection of the part and the sum of any containment regions minus the avoidance regions. The intersection operation provides the volume common to both the part and the containment regions. The user defines the cutter radius, the STEPOVER distance between roughing tool paths, and a THICK dimension if a prescribed thickness is to be left on the surface of the part.

Tool paths to cover the milling region are generated and nested, and finishing tool paths are identified. For 21/2 axis milling, this is done at a series of planar section cuts through the model at different heights. An optimal ordering of the tool paths is generated following a set of ordering guidelines. Retractions of the cutting tool are reduced to the minimum required, and predrill locations at which the cutter plunges to begin milling are produced. The finishing tool paths are augmented with tangential approach and exit arcs to prevent dwell marks. The conglomerate of tool paths from the consecutive section levels define the milling process.

When it is stated that the geometric modeling system generates a model, the modeler generates a mathematical representation of that model which is stored in the memory of the computer. All operations on the model, such as will be described later, are mathematical operations on this internal representation. Plots and pictures can be generated, but algorithms make no use of the physical rendering of the model, and perform mathematical and logical operations on the internal representations. The TRUCE modeler used to implement the optimization process utilizes the well known Coons geometric representation for curves, sculptured surfaces and solids; D.V. Ahuja and S.A. Coons, "Interactive Graphics in Data Processing--Geometry for Construction and Display", IBM Systems Journal, Nov. 3 and 4, 1968. The behavior of x, y and z along arcs of a curve and paths of a surface is expressed in terms of rational cubic polynominals, parameterized on the interval [0.0, 1.0]. A z-level section through a model, i.e. the intersection of a plane z=constant with a solid, is a 2D curve with only x and y varying.

The terms used in describing the invention are defined in Appendix A, and a few of them are as follows. A curve is a collection of arcs. A loop is a continuous. non-self intersecting, closed chain of arcs which defines an area. Disjoint loops are loops with no area in common. In FIG. 2, loops A, B, C and D are disjoint from loops E and F. Nested loops are a set of loops that for each pair one is completely inside the other. Considering the pair A-D, loop D is nested inside loop A and loop A encloses D. Loop F is nested inside loop E. A closed curve is a set of nested and/or disjoint loops; it may have loops inside loops to any depth. FIG. 2 shows one closed curve comprising six loops. The orientation of a loop is the direction, clockwise or counterclockwise, traveled along that loop as consecutive arcs are traversed. A positive or counterclockwise nest is a list of counterclockwise loops such that each loop (except the first) is entirely enclosed by its predecessor. A negative or clockwise nest is a list of clockwise loops such that each loop (except the first) entirely encloses its predecessor. Either of these nests may consist of a single loop.

FIG. 3 is a simplified cross section through a two part mold 104,105. Mold half 104 has a cavity 106 that is milled out, leaving projections 107 on the molds which are hereafter called "islands" because on any given plane the cutter mills around them. The islands are shown schematically but are essential parts of the mold.

Geometric modeling systems provide the designer with the capability to model a mold or die and its cavity, or the cavity regions for molds and dies with optional islands, features, and avoidance and containment regions. In order for a design to be suitable for 21/2 axis milling, on each z-level the tool must be constrained to mill only those regions interior to, or on the boundary of, the regions milled on higher z-levels. All molds and dies satisfy this requirement, which is why 21/2 axis milling is used for most molds and dies. Maximum and minimum z-levels for section cuts are identified from the conglomerate of the regions designed, and appropriate z-levels for sectioning are calculated based on characteristics of the part and desired scallop height. If all sides of the cavity are vertical, a rapid dissent is possible. If smooth, sloped side surfaces are required, a gradual dissent is required.

If the model is correctly designed, each planar section through the solid model forms a closed curve with a complexity of nested and disjoint loops. The tool path optimization system being described imposes proper orientation on the aggregate of loops so the material to be removed by milling is encased between counterclockwise loops and clockwise loops. This implies that loops forming an outer boundary of the cavity must be counterclockwise, loops defining islands must have clockwise orientation, loops defining a "hole" in an island must have counterclockwise orientation, and so forth. Since the optimization method imposes the correct orientation on loops, there is no orientation requirement on solids as defined by the geometric modeler.

FIG. 4 shows the tool paths on a single z-level plane during the milling of a cavity from a rectangular slab 108. The cavity boundary is indicated at 109 and there are three islands 110-112 that are not milled. The planar section cut through the solid model yields a closed curve comprising a cavity boundary loop 109 and three loops at the peripheries of islands 110-112. There are two finishing tool paths 113 and 114 both having tangential approach and exit arcs. According to the curve orientation convention that material to be removed is on the left, finishing path 113 is traversed in a counterclockwise direction and finishing tool path 114 in a clockwise direction. All of the interior loops are roughing tool paths 115. The tool paths are formed by successive insets, which are offsets with negative offset size, of the closed curve and are processed through nesting and ordering optimization algorithms.

The following guidelines for ordering tool paths may be used by both algorithms, and were generated in consultation with milling machine operators who have programmed milling operations manually. They are given in order of priority, from the most important to less important.

1. Gouge free milling is required.

2. Prevent milling both sides of a path before milling the path itself.

3. Prevent over-milling, i.e. cutting part or all of a path more than once.

4. Mill paths that are innermost first unless previous guidelines affect this rule.

5. Paths closest to a boundary of a part (finishing paths) should:

be milled separately from the roughing paths

be augmented by tangential approach and exit

be milled under user control for potential

change in cutter diameter

6. Tool intersects the path to-be-milled only at its starting point. Tool paths can be reorganized to have a different starting point, if by so doing, a retraction is avoided.

7. Tool intersects no UNMILLED nest of tool paths.

8. Tool may cross MILLED nests of tool paths.

A summary of the nesting algorithm is given in paragraphs A, B, C and D and a discussion of how the. algorithm relates to FIG. 5 follows.

A. Validate the curve and impose proper loop orientation. Test to ascertain the section cut through the solid model has generated a closed curve, i.e. a set of loops that are either disjoint or properly nested to any depth. The loops may not touch nor intersect. Using the internal, cannonical form of the loops, identify the nesting and reorient individual loops to impose the convention of the material on the left.

B. Generate the milling region by using the offset radius of -(HRAD+THICK), as defined by the user, to obtain the first inset of the closed curve. Each of the n loops of this first inset are the finishing tool paths (113 and 114) and serve as a boundary curve across which the tool cannot move without gouging into the mold surface. The cannonical forms for each loop are stored, using an identifying number from 1 to n, so as to be available for intersection checks throughout the process.

C. Initialize the nest generated by the boundary loops, the finishing tool paths. Each counterclockwise (CCW) loop starts a positive nest of tool paths, and each clockwise (CW) loop starts a negative nest. Thus, loop 113 starts a positive nest and loop 114 starts a negative nest.

D. Repeatedly generate an inset curve at STEPOVER distance from the previous inset. Each loop of the inset is stored on a disk using a consecutive number starting at (n+1) and is either assigned to an existing nest or is used to establish a new nest based on the milling and ordering guidelines provided above and the nesting logical criteria elaborated in Appendix B. CCW loops and CW loops are nested separately from one another.

FIG. 5 shows the roughing tool paths, numbered as paths 3-16, to mill a cavity around three islands at the same z-level as in FIG. 4, ordered for optimized milling. FIG. 6 shows separately the augmented finishing tool paths at the same level, here numbered paths 1 and 2. The generation and nesting of the roughing tool paths is done as is explained in Appendix B where the first three pages are an elaboration of the nesting algorithm. A simplified explanation of how the algorithm applies to FIG. 5 is given. Reference is made to the positive and negative nesting of tool paths listed at the side of the figure.

The first inset curve at STEPOVER distance from tool path 1, the finishing path, is loop 3 which follows the contour of cavity boundary 109 except where it follows the finishing path around islands 110 and 111. It is a counterclockwise loop and starts a positive nest of tool paths. The loops are numbered in sequence, so adding 1 to the previous loop count of 2 gives 3. The second inset curve yields loops 4 and 5 at the same time, but the first is a clockwise loop and is nested with the finishing path 2, and the second is a counterclockwise loop. A new positive, CCW nest is opened for loop 5 because it is found that loop 3 encloses both of loops 4 and 5. The third inset at STEPOVER distance yields three loops at the same time, loops 6, 7 and 8. Loop 6 is entirely inside loop 5 and does not enclose loop 4; therefore loop 6 is nested with loop 5. Loops 7 and 8, however, meet the criteria for starting new positive or CCW nests. The fourth inset produces loops 9 and 10; the first is added to the nest having loops 5 and 6 and the second to the nest with loop S. The fifth inset gives loops 11 and 12 which are respectively handled the same. The sixth and seventh insets yield loops 13 and 14 which are both added to the nest of loops 5, 6, 9 and 11. The eight inset produces loops 15 and 16; loop 15 is added to the same nest as loops 13 and 14. Loop 16 becomes a new positive nest because loop 14 encloses both of the new loops. In light of the loop numbering convention and the manner of generating tool paths by successive insets, the innermost tool paths are assigned the highest identification numbers. The process stops when the last inset curve has no arcs and is a null curve.

A summary of the algorithm to produce optimization of tool paths and dictate ordering throughout the process is given as paragraphs E, F, G, H and I. A discussion of how the algorithm relates to FIG. 5 follows. Throughout the algorithm, intersection checks are performed repeatedly. That functionality is implemented using computer subroutines. The correctness of these routines bear the responsibility of assuring gouge-free milling for surfaces. The interference logic generates the geometrical representation of the straight line tool motion from one roughing path to the next and determines that there is no intersection of it with the first inset curve, which has been identified as the non-augmented finishing paths, with an unmilled nest of tool paths, or with an about-to-be milled tool path except at a starting point. If these interference checks are not satisfied the algorithm attempts to reorganize the starting point of the tool path about to be milled.

E. The roughing loops in a nest are always used in the order specified in the nest, but can be used either top/down (outermost to innermost) or bottom/up (innermost to outermost). The direction is determined by the ordering guidelines specified above. The current position of the tool is known at all times. At the start of the process, the tool is in retract position so a plunge to the first point of the first loop chosen is required. Each time the tool is in retract position, the beginning point of the next tool path is retained as a predrill location.

F. The decision process must decide the next tool path to be used, either the next loop in the nest, if there is one, or the starting loop of another nest, and then whether or not linear motion from the current tool position to new start position in the next loop is safe without retraction of the cutting tool. The latter decision is made by the intersection subroutines, which test for intersections with the about-to-be milled tool path and with all the boundary loops (the finishing tool paths) and attempts reorganization of the about-to-be milled path when intersections occur. The effect of these subroutines, given on the first page of the optimization algorithms in Appendix B, is to determine tool clearance.

G. As these decisions are made, an ordered list is generated containing the identifying numbers of the loops in the order they are to be milled interspersed by -1, or another retraction symbol, indicating when a retraction of the cutting tool is necessary to prevent gouging.

H. Generate the geometric representation for all tool motion between roughing paths.

I. None of the boundary loops, i.e. the finishing tool paths, appear in the ordered list. They are stored on a disk during the nesting process and are finally augmented with tangential approach and exit arcs. The user has an opportunity to assign a different tool diameter for the augmentation of the finishing paths.

Referring to FIG. 5 and to the recommended ordering of the tool paths set forth below and to the optimization algorithm in Appendix B, the innermost tool path in an unprocessed nest is taken first, namely roughing tool path 16.

The next innermost tool path is loop 15 which is in another positive nest. The intersection subroutines determine that there is no clear path from loop 16 to the beginning point of loop 15 and a retraction is necessary. Predrill locations are needed to plunge the cutting tool to mill loop 16 and to mill loop 15; the X and Y coordinates of the loop start points are stored. The straight line tool path 116 from the end point of loop 16 to the start of loop 15 for the tool in retracted position is generated and stored. The identifying numbers of the tool paths in the order they are to be milled is outputted and stored. The nest with loops 15-5 is milled bottom/up, innermost loop to outmost loop, following the ordering guidelines. The linear tool paths 117 and 118 to move the cutting tool from one roughing path to the next in the nest are generated in sequence and stored. The unprocessed nest with the innermost tool path, 12, is identified and it is determined that positive nest 12-B will be milled top/down, outermost to innermost, because a milled tool path encloses this nest and the guideline as to not milling both sides of a path before milling the path itself would be violated. An interference check shows that there is no clear path from loop 5 to loop 8, and a retraction is unavoidable. The value -1 is added to the ordered list of tool paths and the straight line tool motion path 119 is produced and stored for the tool in retracted position.

The positive nest 12-B is ordered such that tool paths 8, 10 and 12 are to be milled in that sequence and the linear motion paths 120 from roughing path 8 to 10 and 10 to 12 are generated on the milling plane. The unprocessed nest with the innermost tool path is now loop 7. A retraction is essential to go from loop 12 to loop 7 because the tool would intersect a boundary loop. The value -1 is added to the output list and linear retracted motion 121 is generated. Tool path 4 becomes the innermost tool path in an unprocessed nest; this is a negative nest which is always taken bottom/up to assure that loops away from the finishing tool paths are milled first. Loop 2 is a finishing tool path and is not taken after loop 4. A retraction is necessary to go from loop 7 to loop 4 and straight line motion path 122 is generated. Loop 3 is now the innermost unmilled tool path. The optimization algorithm reorganizes loop 3 and its starting point so as to allow safe tool motion between negative loop 4 and positive loop 3 without retraction. Straight line path 123 is produced.

The procedure for augmenting finishing tool paths 1 and 2, FIG. 6, is given on the last two pages of Appendix B. The first has tangential approach and exit arcs 124,125 and the second arcs 126,127. Fairly long arcs are easier to fit; if these tool paths cannot be augmented a smaller tool could be used or the tool could plunge to a start point. The approach and exit arcs are preferably quarter circles with a large radius. The geometric representation of the finishing tool paths augmented by these two arcs is reconstructed. Their starting and end points are such that the approach and exit paths are always internal to the roughing region, and predrilling is not required. If cavity boundaries do not allow sufficient clearance for the augmentation of the finishing path, the path is not augmented.

FIG. 7 shows a containment region of arbitrary shape imposed on the milling process of the mold cavity shown in FIGS. 5 and 6. The boundary of the containment region is indicated at 12B. Finishing tool path 1 and roughing tool paths 2-10 are confined to the milling region which is that area common to both the cavity and the containment region. There are two positive nests of tool paths, and the recommended order for optimized milling of the tool paths is printed below.

After milling one positive nest there is a mandatory retraction of the cutting tool before milling loop 5. The straight line tool motion paths between the roughing tool paths are indicated at 129 and 130. There are two predrill locations, at the start points of roughing tool paths 10 and 5.

FIG. 8 shows the optimization of tool paths to mill those portions of a cavity outside the boundary of an avoidance region 131. The milling region in this case is the mold cavity having boundary 109 from which is subtracted the avoidance region. Following the same nesting and optimization methods already discussed, the optimized order for milling the roughing tool paths, which is set forth below, is the nest comprising paths 12, 10, 8 and 6, after which there is a required retraction before milling paths 11, 9, 7, 5 and 3. A retraction is made before proceeding to path 4.

Finishing paths 1 and 2 are augmented with approach and exit arcs and are milled at the end. Straight line tool motion paths 132-134 are generated to provide for tool movement from a roughing path to the next. Despite the vastly different milling regions in FIGS. 5-8 and the radical change in the arrangement of the tool paths, the ordering optimization method is successful in providing for optimized machining.

FIG. 9 provides an example of optimized tool path organization, as set forth below, for a cavity having deeply concave recesses and a narrow stricture.

FIG. 10 shows tool motion during the milling of a more complicated configuration of the milling cavity having a boundary 141 and island 142, and two holes 143 and 144 within the island. There are four finishing tool paths 1-4 and twenty-three roughing paths 5-27. The recommended ordering for the loops is printed below; there are only five required retractions.

The guidelines and algorithms to generate an optimized ordering of 2-D tool paths for 21/2-axis milling of cavities for molds or dies have been presented. The optimized tool path generation method dictates all motion for the milling tool, and reduces the number of retractions to the minimum number possible within the guidelines. It specifies predrill locations for tool plunges. In addition, it incorporates mechanisms for handling optional avoidance and containment regions of arbitrary shape.

These optimization principles can be extended to accomplish highly efficient profile milling. In principle, the techniques can be extended to three and five axis milling.

While the invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention as defined in the appended claims.