This invention relates to filters. More specifically, a programmable logic controller is utilized to release bodyfeed into fluid to be filtered so that the filter obtains the maximum in service life between required cleanings or the timely depletion of the bodyfeed tank.
Background of the Invention
Introduction
When filtering certain types of materials a dense deposit can form on the filter bed and cause the filter to become clogged very quickly. This requires that the filter be cleaned frequently. The life of the filter bed can be extended by mixing a filter aid (commonly referred to as "bodyfeed") such as diatomaceous earth with the liquid being filtered and/or as a pre-coat for the filter media. The enhanced efficiency of a "bodyfeed" filter system of this type more than pays for the added cost of the system and the filter aid.
In many cases the load on the filter is not constant over time. In these cases, the filter aid should be added at a rate consistent with the level of contamination of the liquid being filtered. When this approach is taken, the operating time of the time of the filter can be increased even more.
The present method of bodyfeeding a pressure or vacuum filter requires operator input to control the rate of bodyfeed injection. The most common procedure of optimizing this injection rate is to adjust the "bodyfeed injection rate time" to a rate high enough to accommodate the maximum foreseeable dirt load. This results in a shorter cycle time that would be otherwise witnessed if the injection rate was adjusted to the load at any time. Due to the tedious as well as time consuming task this method would take, it is not the normal practice in an industrial setting. Some examples of inefficient timer settings would be the process equipment that generates the dirt load being at idle for some period of time; if the timer were set for the anticipation of a given dirt load and there was none generated, this would result in an inappreciable pressure increase and a depletion of the bodyfeed tank. When the dirt load resumed, the filter would immediately go to pressure. Another example would be if a low injection rate was set in anticipation of a light dirt load; if the load increased a short filter cycle would result.
A previous solution to this problem was the use of an electro-mechanical device that changed the bodyfeed rate of a pressure diatomaceous earth water filter. This system was used to filter suspended solids from river water. Due to variables in solids concentrations as a result of seasonal changes as well as vessel traffic, this device was needed. The sensing device was a turbidimeter which measured the loss of light across a fixed gap of incoming water. This signal was then used to select from a group of preset times that would vary the bodyfeed injection rate. The higher the turbidity, the higher the injection rate. Due to the inherent problems associated with turbidimeters used in industrial applications such as the fouling of the sensing cell and the resulting need of maintenance and recalibration, this system proved less than desirable.
This system did not feature the integration of a "bodyfeed level signal" which supplies another signal to better optimize the length of filter cycle run. As the dirt load increased the bodyfeed rate increased the resulting pressure would increase closer to the termination pressure. This would sometimes result in the bodyfeed tank being less than empty. Since the bodyfeed tank has the desired amount of filter aid needed to form the optimum cake thickness, any filter termination occurring with less than an empty bodyfeed tank would result in an inefficient use of the disposable filter aid support septum.
Summary of the Invention
An automated computer protocol is disclosed for the dispensing rate of a filter aid "bodyfeed" slurry into a pressure or vacuum filter in order to efficiently use filter aid and disposable filter septum. The control system accepts three input variables; these variables are total time of filter use, differential pressure across the filter media and the volume (level) in the (slurry or filter aid) bodyfeed tank for augmenting the filter performance. In the disclosed protocol, these variables are processed in turn to vary the rate in which bodyfeed will be introduced into the filter. Specifically, and on an adjustable time incremental basis, the pressure differential is monitored across the filter. This differential is compared to that differential which would exist if the filter had reached the end of service life "set point"--which requires septum replacement and cleaning. The total quantity of bodyfeed available and the rate at which bodyfeed is being added to the filter are monitored. An extrapolation of successive filter differential pressure against time is used to analyze the rate of the pressure differential curve increase. A predicted operating time to the filter set point is computed based on discrete selected operating intervals including a given number of sample points. This is in turn used to control the bodyfeed rate to the filter. These disclosed controls result in the filter reaching its terminal set point pressure at over the greatest possible time interval. The system enables variances in solids concentration being filtered to be accommodated while preserving the maximum time interval between required filter replacements. Provision is made to vary the time sample for slope determination depending upon consistency of the material being filtered as well as the pressure differential response of the filter.
Description of the Drawings
FIG. 1 is a schematic of the bodyfeed system of this invention applied to an unfiltered fluid from a process illustrating schematically the respective fluid and signal routings to obtain maximum system operating time in accordance with the protocol of this invention; and,
FIG. 2 is a time based plot of filter inlet (differential) pressure plotted against time with three windows of slope analysis being shown for three corresponding estimates of filter end point operation illustrating three computed end points.
Description of the Preferred Embodiment
The program and control strategy described are used to control a bodyfeed filter system such as that shown in FIG. 1 below. The filter system described below is used to filter the oil used in a rolling mill which is a typical application of this type of filter system. As shown in the attached schematic, oil from process is inlet to pump P at inlet 14. Discharge occurs at outlet 16 to filter F. After passing through filter F, oil returns to process at discharge 18.
A pressure transducer 20 measure the pressure differential occurring across filter F. It will be understood that although only one pressure sampling point is here shown, what must be measured is the usually increasing filter differential pressure across filter F. Accordingly, and dependent upon the type of filter utilized, more than one pressure transducer may be utilized.
The control system for this filter is implemented in PLC (programmable logic controller) 30 that contains a programmable co-processor. The co-processor shares memory and I/O (input/output) with PLC 30 and is programmed in a dialect of the BASIC programming language, an intelligible listing of which is provided below. The PLC provides the primary control of the elements of the filter and operates the filter independently when it is run in manual mode. The program in the co-processor provides control of the bodyfeed pump 40 when the filter is operated in the automatic modes.
Completing the summary of FIG. 1, bodyfeed tank 50 outputs through respective check valves 52, 54. Air driven bodyfeed pump 40 through air valve 42 monitors bodyfeed outflow by first drawing in a positively displace amount of bodyfeed from conduit 51 through check valve 52 and then discharging the bodyfeed through check valve 54 into conduit 16 before the unfiltered fluid reaches filter F. Thus the bodyfeed is introduce to the filter F with the fluid to be filtered.
Level transmitter 60 outputs the (level) volume of bodyfeed remaining in tank 50. Since it is common practice to introduce sufficient bodyfeed to fully charge filter F when it reaches its setpoint, typically bodyfeed tank 50 level transmitter will output a signal indicative of the total remain volume of bodyfeed available.
During operation, the filter F cycles through three operating phases. First the filter media is precoated with the filter aid. Second, the filter is then used to filter the oil from the mill. Bodyfeed is continually added to obtain optimum filter performance. Thereafter, the filter reaches the setpoint. In common parlance, it becomes unacceptably clogged with both the filtered material and the bodyfeed and operation of the filter must stop. Third, the filter F is then cleaned and made ready for precoating.
The filter can be operated in three modes set at switch 80; manual 81, automatic 82 and timed automatic 83.
Manual mode 81 is operator controlled. The operator sets the rate of the bodyfeed pump from the control panel and it remains at the set rate until the operator changes it.
Automatic mode 82 uses the control strategy programmed in the co-processor to set the rate of the bodyfeed pump. In this mode the strategy is to extend the time the filter can operate as long as possible. This is accomplished by using the inlet pressure of the filter as an indication of how clogged filter F is. A pressure setpoint is established which is used to indicate the point at which the filter must be cleaned. The inlet pressure verses time relationship is measured and used to continuously adjust the rate of the bodyfeed pump.
The objective of this strategy is to exhaust the supply of bodyfeed at the same time the inlet pressure reaches its setpoint. This assumes that the amount of bodyfeed required is determined by the total contamination load. This is equivalent to saying that when a fixed amount of bodyfeed has been added to the system that the inlet pressure will have reached its setpoint and the filter will have to be cleaned. The operating time of the filter can therefore be maximized by adding bodyfeed at a rate proportional to the contamination load of the fluid being filtered.
Timed Automatic mode uses the co-processor to set the bodyfeed pump rate so that the supply of filter aid is exhausted after a set time interval. This mode monitors the pressure verses time relationship in the same manner as during operation in Automatic mode. If this relationship indicates that the pressure setpoint will be reached before the set time, then the control program attempts to extend the time by reverting to the control strategy used during Automatic mode operation. The predicted time remaining before a clean cycle is required is always displayed for the operators information.
Principles of Operation
During operation the control program periodically samples the bodyfeed tank level and the filter inlet pressure. The bodyfeed pump is a diaphragm type and pumps a fixed amount of fluid on each stroke. The count of the pump strokes is also monitored in order to continuously calibrate the pump volume per stroke. The pump stroke period is set by the control program. These inputs are shown 90, 92, and 94 respectively. It will be realized that output bodyfeed signal is shown at 94; the reader will understand that so long as this signal is indicative of the volume of bodyfeed being pumped, the signal will suffice for the disclosed logic.
Referring to FIG. 2, a plot of filter inlet pressure versus time is set forth. Pressure setpoint is set forth at 100. The filter inlet pressure readings, sampled over a specified period of time are plotted. In actual fact, these respective values are stored in the control program in the co-processor.
The program is subdivided to operate in discrete time intervals required for analysis. Consequently, windows W.sub.1, W.sub.2, and W.sub.3 are shown. Each of these respective windows constitutes a window moving along the pressure verses time history of the filter operation and is used to predict the time at which the inlet pressure will reach its setpoint. FIG. 2 shows a pressure verses time curve for a typical filter. Three windows W.sub.1, W.sub.2, and W.sub.3 are shown that depict snapshots of this curve as they would exist stored in the control program at the times T.sub.1, T.sub.2 and T.sub.3.
The data points stored in the control program are updated each tune the control loop is run. At that time the oldest point is deleted and the newest point is added. In this manner the data window moves along in tune but always contains the same amount of history. The exception to this is at the time when the filter is first started running. Until a sufficient time has elapsed the number of points contained in the window is increasing.
The pressure verses time data contained in the control program is used to predict the time at which the inlet pressure will reach a value equal to the pressure setpoint. This is accomplished by fitting a straight line to the data by using the Least Squares method. This provides values for the intercept, P.sub.0, and the Slope, in the following equation.
The time at which the inlet pressure will reach the setpoint, T.sub.sp, can then be calculated from the following equation.
where: P.sub.sp is the pressure setpoint.
This prediction is made for each cycle of the control loop regardless of the operating mode of the filter. FIG. 2 shows the results of the least squares fit (LSF) for the data points in each of the three windows. They are shown as the dotted lines originating in each of the windows and extending until they intersect with the pressure setpoint line. The times predicted for the inlet pressure to reach the setpoint are shown as T.sub.sp1, T.sub.sp2 and T.sub.sp3 (the intersection of the LSF lines With the pressure setpoint line).
In manual mode the predicted time is displayed for the operators information but no control action is taken based on its value. In timed automatic mode action is taken only if the predicted inlet pressure at the time set by the operator is greater than the pressure setpoint P.sub.sp. In this case timed automatic mode operate in the same manner as if the fiber were in automatic mode.
The amount of bodyfeed remaining in the tank, V.sub.bf, is continuously monitored. Knowing this and the T.sub.sp predicted from the LSF, the required bodyfeed rate can be calculated as:
Where: T is the elapsed operating time.
The bodyfeed pump cycle time .THETA..sub.bf, is then calculated as:
Where V.sub.p is the volume per stroke.
Control Algorithm Tuning
The automatic control algorithm attempts to prolong the operating time of a bodyfeed filtering system by monitoring the load on the filter and adjusting the bodyfeed flow rate. There are several parameters that are implemented in the control program that provide a means of tuning this algorithm. Tuning is necessary in order to match the sensitivity and responsiveness of the algorithm to the various types of applications where a bodyfeed filter system can be used.
The two main parameters are the frequency at which points are sampled from the filter and the size of the prediction windows.
Sampling Frequency--The control program is a control loop that runs periodically. Each time it "wakes up" it read all of its inputs, calculates the required outputs which it then sets, and then goes back to "sleep" until it is time to run again. This control loop cycle time is set in the program and determines the frequency with which data is sampled.
It is best to sample frequently when the data being read is subject to noise. This results in a form of data averaging. When the data being read is stable, the sampling frequency can be relatively infrequent, since the values read are more likely to be accurate.
Sample Size--The sample size is also set in the program and defines the size of the window used in the LSF calculations. Since samples are taken each time the control loop runs, the width of the window, in time, is determined by the product of the sample size and the control loop cycle time. For example, a sample size of 40 and a control loop cycle time of 15 seconds would correspond to a 10 minute moving window.
The size of the moving window determines how quickly the control algorithm will respond to changes in the filter inlet pressure. Smaller window sizes are best for applications where the contamination load changes over a large range and/or quick response time is desired. Larger windows are less responsive and are best where changes fluctuate rather rapidly over a smaller range.
Program Structure and Flow Logic
The following is a description of the structure and flow logic of the control program. It is presented in pseudo code which is a descriptive form of program documentation that uses English language commentary within logic programming constructs. It uses indentation and square brackets, [. . . ], to group descriptions of equal hierarchy and execution order. Programming operators and constructs are shown in bold type.
The main control program consists of two primary execution phases. The first phase initializes all system constants, variables and state vectors. The second phase is the control loop which is effectively a never ending DO loop. The control loop is driven by an interrupt that is generated by an interval timer that is set to the control loop cycle time. In addition to the main program, several subroutines exist that perform functions such as the LSF calculations, converting values to BCD representation for display in the PLC, and resetting state vectors and variables during transition of a clean cycle.
The pseudo code descriptions of these programs follows. The main program is described first, followed by descriptions of the subroutines.