GB2224369A

GB2224369A - "Updating output parameters for controlling a process"

Info

Publication number: GB2224369A
Application number: GB8822447A
Authority: GB
Inventors: Carl Arthur William Aylen
Original assignee: MANAGEMENT FIRST Ltd
Current assignee: MANAGEMENT FIRST Ltd
Priority date: 1988-09-23
Filing date: 1988-09-23
Publication date: 1990-05-02
Also published as: GB8822447D0

Abstract

Values are assigned to a set of input parameters and these are supplied to a processor 1 which compares these values with corresponding values in an up-datable reference 4 representing a library of previous input parameter values and derived output parameter values. One or more of the stored parameter sets most resembling the input parameter values is selected and, from the or each selected set, the processor 1 derives values for the output parameters which are output at 5 to control a process, e.g. operation of a lathe, represented by the delay 6. The performance of the process is subsequently assessed at 7 and, on the basis of this, the values of the parameter sets in the reference 4 are undated. Subsequent input of parameter values then gives rise to derived output parameter values which benefit from the previous experience gained by the system. <IMAGE>

Description

Method of and APparatus for Optimising output Parameters for Controlling a Process The present invention relates to a method of optimising a set of output parameters for controlling a process, and to an apparatus for performing a method of this type.

According to a first aspect of the invention. there is provided a method as defined in the appended claim 1.

According to second aspect of the invention, there is provided an apparatus as defined in the appended claim 22.

Preferred embodiments of the invention are defined in the other appended claims.

The invention will be further described. by way of example, with reference to the accompanying drawings. in which: Figure 1 is a schematic block diagram of a first embodiment of the invention; Figure 2 is a schematic block diagram of a second embodiment of the invention; Figure 3 is a flow chart illustrating operation of the embodiment shown in figure 2; Figure 4 is a schematic block diagram of a third embodiment of the invention; and Figures 5 to 7 are functional diagrams illustrating three possible modes of operation of the embodiment of Figure 4.

Figure 1 illustrates a system for optimising a set of output parameters for controlling a process. The system comprises a processor 1, such as a computer, which receives a set of input parameters at a parameter input 2. These input parameters represent user selectable values which a user selects or defines in order to influence the operation of a process. Specific examples of these will be given hereinafter with reference to the other drawings.

The processor 1 also receives reference values at an input 3 from an updatable reference 4. The reference values represent previously used sets of input parameters which have been previously specified and used as the basis of controlling the process, together with the corresponding output parameters which were applied to the process for direct control or which have subsequently been updated.

The processor 1 compares the values of the input parameters supplied to the input 2 with the input parameter values of the reference parameter sets from the updatable reference 4 supplied to the input 3 and selects one or more of these parameter sets whose input parameter values resemble or are closest to the assigned values selected by a user. The corresponding output parameter values are then processed by the processor 1 and are output at 5 for controlling the process. In some cases the output 5 supplies the output parameter values directly to means for controlling the process imiediately. In other cases, the output parameter values are presented. for instance, as printed output, and may then optionally be applied to control a process, for instance, at a later time.

The process which is controlled by the output parameters is not explicitly shown as such in figure 1.

However, there is an intrinsic delay between the supply of the output parameter values and the performance or conclusion of performance of the process using these values, and this delay is represented at 6 in figure 1.

The delay is such that the result of the process could not be used to control the process itself, for instance as a conventional negative feedback loop. Essentially, the performance of the process, or at least the performance which is relevant to the system of figure 1, can only be assessed once the process has been completed.

Upon completion of the process and after the delay represented at 6, the performance of the process is assessed at 7. This assessment of the performance determines whether the reference 4 is updated and possibly also how this updating is performed, for instance depending upon the improvement in performance achieved. If updating is to be performed, this will generally be represented either by an entry of a new reference point in the form of a new set of parameters based on the input parameter value selected by the user and the output parameter values which controlled the process, or by modification of existing parameter sets, for instance by modifying the output parameter values of the parameters sets selected by the processor 1 in accordance with the output parameter values supplied to the output 5.

Thus, the system can be thought of as having a degree of artificial intelligence such that it learns from the previous experience of actual processes. The subsequent selection of input parameter values by a user will then result in the supply of output parameter values for controlling the process in accordance with the experience gained from the performance of previous processes including the immediately preceding one. The system thus optimizes the output parameters essentially automatically and with little or no human intervention, depending on the specific application to which the system is put.

Figure 2 illustrates a specific application of a system of the type shown in Figure 1 to the control of a machine tool, in this case an automatic lathe. The system comprises a processor 10 connected by a bi-directional bus 11 to a library 12, for instance in the form of an electronic memory. The processor 10 has an input bus 13 connected to an input interface 14. The processor 10 also has an output bus 15 connected to an output interface 16. The output interface 16 has output ports connected to control inputs for speed and depth of cut of the automatic lathe 17.

The lathe 17 is provided with a sensor 18 which monitors the temperature at the tip of the cutting tool and supplies an output signal to an input port of the input interface 14. A timer 19 which measures the total time of the machining operation is also provided. In figure 2, the timer is shown as a device associated with the lathe 17 and having an output connected to another port of the input interface 14. However, it is also possible for the timer to be implemented by the processor lo, in which case the input interface 14 need only receive a signal representing the operational state of the lathe.

The input interface 14 has a further port which receives input parameter values selected by the user.

These input parameters differ from the control parameters used to control the operation of the lathe 17. For instance, the input parameters selected by the user may include the total depth of material to be machined during the machining operation, an indication of the type of material and its hardness, and an indication of the maximum or optimum temperature of the cutting tool tip. The output parameters specified by the processor 10 and used to effect direct control of the lathe 17 are speed of the lathe and depth of cut, as shown in figure 2.

The library 12 contains sets of parameter values pertaining to previous machining operations. In particular, each set of parameters comprises the values of the input parameters specified for the particular operation (i.e. depth of material to be machined, material hardness, temperature data), the associated values of speed and depth of cut supplied by the processor 10 for controlling the lathe 17 (possibly modified by experience gained during subsequent machining operations), and the time taken for the machining operation.

Operation of the system shown in Figure 2 will now be described with reference to the flow chart in figure 3. In the first step 20, the values of the input job parameters are supplied, for instance by a machinist or lathe operator, to the input interface 14. These parameter values are received by the processor which then interrogates the library 12 and selects the parameter sets whose input parameter values are nearest to the job parameter values entered by the user (step 21).

In step 22, the output parameters values of the selected parameter sets are processed by the processor 10 in order to define the optimum lathe speed and depth of cut. The processor 10 may use one or more of various techniques to determine the optimum output parameter values, such as weighted averaging, interpolation, and extrapolation. The optimum lathe speed and depth of cut are then output at step 23 to the lathe 17, which begins the machining operation.

In step 24, the performance of the machining operation is assessed. In particular, the total time of the cutting operation is noted and the temperature of the tool tip is analysed, for instance by noting the maximum temperature during the cutting operation. This analysis is performed by the processor 10, for instance by comparing the performance with previous performances resulting from the selected parameter sets in the library 12.

In step 25. the library is. up-dated depending on the result of the assessment of performance by the processor 10. This may comprise forming a new library entry of a parameter set comprising the input parameter values supplied by the user and the speed and depth of cut output by the processor 10, possibly in combination with details of the machining time and maximum tool tip temperature. Alternatively or additionally, one or more of the existing parameter sets in library may be modified, for instance by modifying the output parameter values of the parameter sets which were used by the processor 10 to derive the speed and depth of cut values used in the machining operation. The system is then ready to receive new input parameter values specifying the next machining operation to be performed by the lathe 17.

The next, and all subsequent, machining operations benefit from the experience gained by the just-completed machining operation. For instance, it may be desired to minimise machining time while preventing the temperature of the tool tip from exceeding a threshold value. By making use of data gathered during all previous machining operations, the system can, for instance, increase speed and depth of cut such that the tool tip temperature remains just below the threshold. This has the effect of reducing the machining time. The processor 10 may additionally be programmed to monitor the tool tip temperature continuously so that, if the temperature show signs of exceeding the threshold, the speed and/or depth of cut are reduced accordingly.

However, this is an auxiliary instantanous feedback system which is not an essential part of the "learning system" represented in Figures 2 and 3. Of course, any danger of exceeding the threshold of the tool tip temperature can be noted in the library so that subsequent machining operations are performed with lower speed and/or depth of cut. Thus, as the system builds up experience of machining operations with a variety of material hardnesses and depth of machining, so that machining operations are controlled more accurately by optimising the output parameter values for speed and depth of cut so as to minimise machining time while preventing the maximum cool tip temperature from exceeding the threshold.

In order to select the parameter sets from the library 12 to be used in deriving the output parameter values, the processor 10 preferably compares each of the library parameter sets in turn with the input parameters values. For instance, for each of the library parameter sets, the processor forms the differences between the input parameter values specifying depth of material to be machined and the corresponding value of the parameter from the library 12, and the difference between the requested and stored material hardnesses. These differences are then weighted with suitable weighting factors and added together to form a weighted sum which indicates how "close the parameter set from the library is to the selected input parameter values.Once all of the stored sets of parameter values have been compared in this way, selection of suitable sets may be made.

For instance, it may be sufficient to select the parameter set who weighted sum of differences is the smallest. In this case, the output parameter values of the stored set could be used to control the lathe 17.

However, a more effective approach is to derive suitable speed and depth of cut values from the selected parameter set by extrapolating from the stored values, the extrapolation depending on the weighted sum of differences or on the individual differences.

Alternatively, a number of stored parameter sets having the smallest weighted sums, such as the four parameter sets closest to the selected input parameter values, may be used to derive the speed and depth of cut. In this case, the value for speed and depth of cut may be derived from forming a weighted sum of the individual output parameter values of the selected parameter sets.

Another possibility is that the stored parameter set having the least positive weighted sum of differences and the parameter set having the least negative weighted sum of differences are selected, and the speed and depth of cut are derived by interpolating between the output parameter values of the two selected parameter sets.

The interpolation may be weighted in accordance with the sizes of the weighted sums of differences.

As previously mentioned, the library 12 may be up dated by entering new parameter sets. However, when the existing parameter sets in the library are up dated, preferably the or each parameter set selected for use in deriving the output of parameter values is modified in accordance with the derived values. For instance, in the case where interpolation between two parameter sets or weighted sums of several parameter sets was used, the output values of these sets may be modified according to the derived output parameter values with a weighting equal or equivelant to that used in the deriving step.

It is of course, possible to modify the value of the input parameters of the stored parameter sets without modifying the output parameter values, or to modify both the input and output parameter values in the stored sets, when up dating the library. However, some caution must be exercised in order to ensure that the "learning" of the system is stable; because the system may be thought of as a form of feedback with intrinsic delay.

over-zealous modification of the parameter sets could result in substantial instability with the result that erratic parameter values tending away from optimised values would be produced. This can be avoided effectively by up dating the library in such a way that the loop gain is made sufficiently low or a sufficient amount of damping is provided.

Another interesting application of the system of the type shown in Figure 1 is the controlling of placement of personnel in an organisation, for instance placing executives within a management structure of a large industrial firm on the basis of leadership skills.

Figure 4 illustrates the hardware for a system for performing such a role. The system is based on a processor 30, generally a computer, connected by a bi-directional bus 31 to a library 32 formed by computer storage, for instance electronic memory, disc memory, and tape memory. The processor 30 has an input connected to a document reader 33 and an output connected to a printer 34. The document reader is used for automatically reading questionnaires 35 and may comprise a conventional optical character reading device. However, it is preferable for the questionnaires to be of the multiple-choice type in which a candidate makes suitable marks for selecting answers to the questions. In this case, a much simpler document reader may be used since it is only necessary to detect the marks made by the candidate.

The system of Figure 4 may be used in different ways, each being effectively defined by the software controlling the processor 30. Three possible modes of operation will now be described with reference to Figures 5 to 7 of the drawings.

Figure 5 shows, in the form of a functional diagram, the basic steps performed by the system in order to predict executive leadership skills. In step 41, a standardised questionnaire with multiple-choice answer-type questions is completed by an executive and the answers are fed to the processor automatically by the document reader 33. In step 42, the processor compares the answers, which represent input parameters, with answers to questionnaires which have previously been read and stored in a library, and selects those answers, each of which essentially comprises input and corresponding output parameters, which most closely resemble the answers read in the step 41.The selected answers from the library each contain output parameters which essentially predict the leadership skills exhibited by previous executives, and these form the basis of an executive profile including predicted leadership skills (step 43). Any of the techniques disclosed hereinbefore with reference to controlling a machine tool can be used to predict the leadership skills. The executive profile is then printed by the printer 34.

The predictions contained in the executive profile are used to select a suitable position in the organisation for the executive. Step 44 which refers to performance by the executive indicates the actual performance of a job by the executive. At a suitable time after the executive has begun in the position, his performance is assessed by a manager at step 45, which manager completes a standardised multiple-choice questionnaire recording perfomance data and particularly the leadership skills actually exhibited by the executive. This questionnaire is read by the document reader 33 and the data supplied to the processor 30.

The processor then processes the data to assess whether its previous prediction of the executive leadership skills was accurate. In general, there will be room for improvement, and the library is up-dated at step 46, for instance by adding a new entry or by modifying the previous entries such as those selected by the processor for deriving the prediction of leadership skills.

Again, the techniques disclosed hereinbefore for up-dating the memory may be applied to the present system.

The questionnaires are designed in conjunction with the analysis preformed by the processor 30 so as to provide data which is as consistant as possible. The use of multiple-choice questionnaires effectively eliminates human participation in the procedure for predicting leadership skills. Thus, unpredictable elements are largely eliminated from the process.

Because the system effectively learns from its previous predictions and applies the result of this learning to each new prediction, the accuracy with which leadership skills can be predicted increases and can track any variations caused by changes in the importance of input parameters and changes in the organisation which call for different skills as the organisation itself evolves. The system therefore predicts with far greater accuracy and consistancy the leadership skills and other output parameters of the executive profile.

A similar type of system can be used to predict the demands of particular jobs within the organisation as shown in Figure 6. In a first step 51, standardised multiple-choice questionnaires are read by the document reader 33 and the resulting data supplied to the processor 30. The questionnaires are completed by managers in the organisation and provide input parameters relating to various specific requirements of the job being analysed. At 52, these parameters are used to select from a library correlating previous input parameters with corresponding job analysis profiles, including predicted job demands. An analysis profile for the specific job is then supplied at 53 in essentially the same way as the executive profile was derived at 43 in Figure 5.

At 54, the job is actually performed by one or more executives and, after a suitable time period, the job demands are analysed by a manager by means of a standardised multiple-choice appraisal questionnaire at 55. The result of this appraisal is then used as the basis for up-dating the library at 56. This up-dating may adopt any of the techniques described hereinbefore.

This system provides an up-to-date data base of the requirements of various executive jobs in an organisation. Thus, the job demands predicted by the system can track changes in the actual jobs and their demands within an organisation as the organisation itself changes and develops. The job demands predicted by the system therefore represent an accurate and up-to-date profile of the demands and can be used in selecting suitable personnel as and when vacancies arise.

The system shown in Figure 7 is used for matching specific executives to specific positions within an organisation. The executive profiles, for instance obtained by the system of Figure 5, are supplied to the processor 30 in step 61 and a job analysis profile from the system of Figure 6 is supplied at 62. The processor then predicts the performance of executives, based on the executive profiles, in the job, based on the job analysis profile, by the same comparison and selection procedure as described hereinbefore in conjuction with a library containing previous profiles and selections of executives for jobs (step 63). At step 64, a suitable executive is selected for the specific job and the executive performs the job at 65. The performance of the executive in the job is assessed by an appraisal questionnaire answered by a manager at 66, and the result of this appraisal is used to up-date the library at 67 using any of the techniques previously described.

The system thus learns from previous experience how to make accurate matchings of executives and positions within an orgnaisation. and is capable of tracking any changes caused by development of the organisation. Each new selection of an executive for a specific job benefits from the experience gained with all previous selections.

It has been found in practice that the systems illustrated in Figure 5 to 7 provide very accurate optimisation of predictions based on executive analysis and job requirement analysis. Such systems have achieved remarkable accuracy and reliability in making predictions and have replaced previously used selection precedures, which rely very heavily on human intervention, with a substantially automatic system achieving far higher accuracy and reliability. The process has thus been transformed from an art into a science with substantial improvements in the running of organisations.

Claims

1. A method of optimising a set of output parameters for controlling a process, which output parameters are influenced by a set of input parameters, the method comprising the steps of: assigning values to the set of input parameters; comparing the values of the set of input parameters with a plurality of stored parameter set values, each stored set of values comprising a set of input parameter values which has previously been supplied and a corresponding set of output parameter values; selecting from the plurality of stored parameter set values at least one selected set of parameter values; deriving from the correponding set of output parameter values of the or each selected set of parameter values a derived set of output parameter values; supplying the derived set ofvoutput parameter values to control the process;; assessing the performance of the process when controlled by the derived set of output parameter values; and updating the stored parameter set values, depending on the performance assessment, in accordance with the assigned values of the set of input parameters and the derived set of output parameter values.

2. A method as claimed in claim 1, in which the selecting step comprises selecting the or each selected set of parameter values comprising a set of input parameter values resembling the assigned values of the set of input parameters.

3. A method as claimed in claim 1 or 2, in which the comparing step comprises forming the difference between the assigned value of each input parameter and the stored value of the corresponding parameter for each of the stored parameter set values and forming a weighted sum of the differences corresponding to the stored parameter set values.

4. A method as claimed in claim 3, in which the selecting step comprises selecting the set of stored parameter values corresponding to the smallest of the weighted sums.

5. A method as claimed in claim 4, in which the deriving step comprises extrapolating from the output parameter values of the selected set to form the derived set of output values.

6. A method as claimed in claim 3, in which the selecting step comprises selecting the x sets, where x is an integer greater than 1, of stored parameter values corresponding to the x smallest weighted sums.

7. A method as claimed in claim 6, in which the deriving step comprises forming weighted averages of the stored output parameter values of the selected x sets to form the derived set of output values.

8. A method as claimed in claim 3, in which the selecting step comprises selecting the two sets of stored parameter values corresponding to the smallest positive and negative weighted sums, respectively.

9. A method as claimed in claim 8, in which the deriving step comprises interpolating between the output parameter values of the two selected sets to form the derived set of output values.

10. A method as claimed in any one of the preceeding claims, in which the updating step comprises storing a new set of parameter values for use in subsequent comparing and selecting steps, the new set of parameter values comprising the assigned set of input parameter values and the derived set of output parameter values.

11. A method as claimed in any one of claims 1 to 9, in which the updating step comprises modifying at least one of the plurality of stored parameter set values in accordance with the assigned set of input -parameter values and the derived set of output parameter values.

12. A method as claimed in claim 11, in which the or each modified parameter set value is the or each selected parameter set value.

13. A method as claimed in claim 11 or 12, in which the values of the or each modified parameter set which are modified are the outout parameter values of the or each selected set.

14. A method as claimed in claim 13, in which each of the values of the or each modified parameter set is formed as a weighted average of the values of the corresponding parameters in the respective selected set and in the input set.

15. A method as claimed in any one of the preceding claims for controlling a machine tool.

16. A method as claimed in claim 15, in which the input parameters include depth of material to be machined, data relating to physical properties of the machine, and permissible tool tip temperature, the output parameters control rate of machining. and the performance assessed in the assessing step is machining time and tool tip temperature.

17. A method as claimed in anyone of the preceeding claims for controlling placement of personnal in an organisation.

18. A method as claimed in claim 17, in which the input parameters include data relating to the abilities of an executive extracted from a first standardised questionnaire, the output parameters include predicted performance in a specific job within the organisation, and the assessing step includes extraction of data relating to actual performance of the executive in the specific job from a second standardised questionnaire assessed by a manager of the organisation.

19. A method as claimed in claim 17, in which the input parameters include data relating to the requirements a job within the organisation extracted from a third standardised questionnaire, the output parameters include predicted performance of the job by a specific executive and the assessing step includes extraction of data relating to actual performance of the job from a forth standardised questionnaire assessed by a manager of the organisation.

20. A method as claimed in claim 17, in which the input parameters include data relating to the abilities of executive extracted from a fifth standardised questionnaire and data relating to the requirement of a job within the organisation extracted from a sixth standardised questionnaire, the output parameters include predicted performances of a selected executive in the job, and the assessing step includes extracting data relating to actual performances by the selected executive of the job from a seventh standardised questionnaire assessed by a manager of the organisation.

21. A method of optimising a setsof output parameters for controlling a process, substantially as hereinbefore described with reference to the accompanying drawings.

22. An apparatus for optimising a set of output parameters for controlling a process, which output parameters are influenced by a set of input parameters, the apparatus comprising.

means for receiving assigned values of the set of input parameters; a memory containing a plurality of stored parameter set values, each stored set of values comprising a set of previously received input parameter values'and a corresponding set of output parameter values: means for selecting from the memory at least one selected set of parameter values; means for deriving from the corresponding set of output parameter values of the or each selected set of parameter values a derived set of output parameter values; means for supplying the derived set of output parameter values to control the process; means for receiving an assessment of the performance of the process controlled by the derived set of parameter values; and means for updating the memory, depending on the perforaance assessment, in accordance with the assigned values of the set of input parameters and the derived set of output parameter values.

23. An apparatus as claimed in claim 22 for controlling a machine tool, in which the selecting means, the deriving means, and the updating means comprise a data processor, the assigned value receiving means and the performance assessment receiving means comprise an input interface for the data processor, and the supplying means comprise an output interface for the data processor.

24. An apparatus as claimed in claim 23, in which the data processor is arranged to supply via the output interface signals for controlling machining rate of the machine tool and the input interface is connected to a temperature sensor for sensing the tool tip temperature of the machine tool, the apparatus being connected to or including a timer for timing machining of a workpiece.

25. An apparatus as claimed in claim 22 for controlling placement of personnel in an organisation, in which the selecting means, the deriving means, and the updating means comprise a data processor, the assigned value receiving means and the performance assessment receiving means comprise automatic document reading means, and the supplying means comprise an output device of the data processor.

26. An apparatus for optimising a set of output parameters for controlling a process, substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.