GB2241097A - Supervisory device - Google Patents

Abstract

A device for monitoring a parameter in a manufacturing process is set up in a learning mode to establish upper and lower limits for the parameter at particular times in the process cycle, and thereafter generates an output signal (preferably a simple alarm) if the parameter falls outside the band defined by the tolerance values. In the learning mode, a number of repetitions, assumed to be fault free, are monitored, and the tolerance limits are derived by first applying a "time tolerance" and then applying a magnitude tolerance. The monitored parameter may be pressure, temperature, force, acceleration or vibration. <IMAGE>

Description

Supervisory Device This invention relates to manufacturing industry and more especially to the supervision of manufacturing processes that entail repetition of at least one operation. The repetition may be total, as for example the cyclic operation of an automatic lathe producing a series of identical products; or it may be ancillary to a 'continuous1 process, as for example the revolutions of an extruder screw when producing a product of indefinite length.

Currently there are two approaches to supervising such processes: either the process is observed by a human "operator" using appropriate senses and instruments, which is inefficient in labour utilisation since the operator has little to do unless something goes wrong; or control is entrusted to a computer which monitors appropriate process parameters and shuts down the operation if any of the parameters moves beyond its capacity to effect adjustment, which is expensive both in terms of capital cost and in terms of down-time since the process may be stopped needlessly and personnel may not be immediately available to re-start it.

The present invention provides a new and useful compromise in which human supervision is assisted by a device that is very much simpler and cheaper than a process-control computer and will usually be set up to call for human intervention when repetition deviates from its normal pattern, though it may stop the process if desired.

In accordance with the invention, a device for assisting in the operation of a manufacturing process that entails repetition of at least one operation comprises (a) means for receiving and storing data representing current values of at least one parameter that may change during the said operation; (b) means for analysing in a learning mode data so received and stored (i) to determine the highest and lowest values of the said parameter observed at a plurality of corresponding times in a number of repetitions of the said operation; and; (ii) to derive from the said highest and lowest values upper and lower tolerance limits for each of the said times, each upper tolerance limit exceeding the highest value observed within a first predetermined interval including that time by an amount that increases with the difference between the highest and lowest values observed in a second predetermined interval including that time and each lower tolerance limit being lower than the lowest value observed within a third predetermined interval including that time by an amount that increases with the difference between the highest and lowest values observed in a fourth predetermined interval including that time; and (c) means for comparing in a running mode the data values received at specific times in the current repetition of said operation with the derived upper and lower tolerance limits for those specific times and for generating an output signal if any said data value exceeds the respective upper tolerance limit or falls below the respective lower tolerance limit.

The parameter, or usually parameters, selected for observation may be any of those that are susceptible to instrumental measurement and appropriate to the process being monitored, and need not necessarily be the same at all stages of the process repetition. Temperatures and pressures at particular places, forces, accelerations, displacements and vibrations (including sound volume and/or pitch) will often be suitable parameters.

The rules for deriving the upper and lower tolerance limits may be different for different parameters, and if required may differ from one another for the same parameter. In simple, accurately repetitive processes, a multiplier of the order of 0.5 may be all that is required; in other cases the exercise of skill and experience may lead to the use of a rule in which proportionality applies to the logaritham or exponent (or some other increasing function) of the parameter in question, or of a step function rule.

The four predetermined time intervals are needed to ensure that the device does not generate alarm signals merely because a steep change that normally occurs in some parameter happens a little earlier or a little later than it did during any of the repetitions observed in the learning mode. In most cases all four of these intervals can be equal and evenly distributed before and after the time for which the limits are being derived; but unequal or unsymmetrically distributed times could be used if desired (for instance, if there is reason to expect that tool wear will result in a gradual slowing down of the operation).

Preferably, in order to keep down costs and maximise the utilisation (and job satisfaction) of the human operator, the device produces an output signal representing no more than that a deviation from the expected range of repetition has occurred, leaving the operator to determine whether, and if so what, adjustment or other remedial action is called for; for example it may merely illuminate a signal or sound an audible alarm. If the process is complex, it may be desirable to provide a signal indicating which of the parameters being observed has given rise to the signal; and if the process is hazardous or the product very expensive, the process may be stopped (or slowed down) automatically in response to the output signal until re-set after inspection by the operator.

The invention will be further described, by way of example, with reference to the accompanying drawings in which Figure 1 is a graph showing observed values of variables for a number of repetitions of the cycle; Figure 2 is an enlargement of the part of Figure 1 relating to the time interval indicated by the Roman numeral II in Figure 1 and; Figure 3 is a further graph, corresponding to Figure 1, illustrating the derivation of tolerance limits by the device in accordance with the invention.

Referring first to Figure 1, the graph plots the magnitude of some convenient variable against time, and repeats for each separate cycle of the process. This produces (in this particular case) four traces 1 which are similar in shape but differ in detail because the repetition of the process is not absolutely precise.

The upper and lower traces 2 represent (as nearly as may be done without obscuring the original traces 1) the envelope within which all of the observed traces 4 lie.

The device in accordance with the invention operates in a series of cycles of predetermined length which are each initiated by a "start of cycle" signal from the process being monitored. In the initial "learning" phase of operation, it first collects data corresponding to the first of the traces 1 by sampling at regular intervals; during the second cycle, incoming magnitude values are compared with the stored value for the corresponding sample in the previous cycle to determine at each instant whether the incoming value or the stored value is higher, and the higher value is stored into one record and the lower value into another.

In the third and each subsequent cycle of the learning phase, the incoming magnitude is compared with both the higher and lower stored value for the corresponding sample, substituted for the higher if it falls above it, substituted for the lower if it falls below it, or discarded if it falls between the upper and the lower values. This generates data record arrays corresponding to the envelope curves 2 in Figure 1.

It will be clear that these envelopes enclose only the traces for the particular cycles encountered in the learning phase, and just as the individual traces 1 differ from one another, so may subsequent observed traces even though the process is being performed normally. Not only will the magnitudes observed at a particular point in the cycle vary, but variations will also occur in the time at which particular events are observed. This is illustrated by Figure 2, in which the thick curves reproduce on a larger scale the parts of the curves of Figure 1 falling within the time interval II. In the cycles that occur during the learning phase, a sudden increase in the magnitude of the observed parameter has occurred within the approximate time interval from T1 to T2.If, instead, the rise had begun just a little earlier, at TO (as illustrated by the dotted trace 3), then the observed value would have gone outside the envelope by a considerable margin, in terms of magnitude, though only slightly in terms of time.

Satisfactory results cannot, in our experience, be obtained by merely applying a tolerance to the magnitude observed at each time in the cycle, since the size of the tolerance required to avoid false alarms in rapidly changing circumstances, such as those illustrated in Figure 2, are so large that significant changes at times when rapid change is not occurring would be overlooked.

In order, therefore, to obtain adequately close monitoring without unduly frequent false alarms, the envelope data corresponding to curves 2 are first subjected to a process which may be called "time tolerancing" in which the upper stored value for each time is compared with all the upper stored values recorded for all the samples within a time interval AT before and after the time in question, and the largest of those values substituted for the value in question; and correspondingly for the lower limit values. (Note that these comparisons may be done in any convenient order; for example each stored value could be taken in turn and compared with all the stored values within the required time interval to determine the time-toleranced value corresponding to it before proceeding to the next stored value; or each stored value could first be compared with the first stored value within its appropriate time interval [with appropriate conventional adjustments at the ends of the stored array) to generate a new, temporary array, then each stored value in the temporary array compared with the second value within the appropriate time interval for the original array and so on until all the values within the time intervals have been compared).

This time tolerancing process is illustrated in Figure 3, in which 1 is a typical one of the traces 1 from Figure 1, and 2,2 are the upper and lower envelope curves, as in Figure 1, so that the cross-hatched area 6 represents the range of values actually observed during the learning phase. Upper and lower curves 4 show the effects of the time tolerancing operation just described; it will be noted that the area 7 added to the graph by this process is considerable in regions where rapid change is occurring and much smaller elsewhere.

Preferably data corresponding to the original envelope curves 2 is retained in store, so that it can be re-used, if required, with different computational rules without the need to re-run the learning phase.

One further derivational step is required before the apparatus is ready to begin its normal monitoring functions; this is to apply a magnitude tolerance to the values representing the curves 4, this tolerance being related, in the example proportionally, to the instantaneous difference between the upper and lower values 4. This generates new data corresponding to the upper and lower curves 5,5, which are stored and used as the actual basis for comparison during the monitoring phase.

In the monitoring phase, the incoming data magnitude is simply compared with the upper and lower values on the curves 5 for the corresponding time in the cycle, and an output signal (alarm) is produced if the observed value is above the upper stored value or below the lower one.

Claims (5)

1. A device for assisting in the operation of a manufacturing process that entails the repetition of at least one operation comprising (a) means for receiving and storing data representing current values of at least one parameter that may change during the said operation; (b) means for analysing in a learning mode data so received and stored (i) to determine the highest and lowest values of the said parameter observed at a plurality of corresponding times in a number of repetitions of the said operation; and; (ii) to derive from the said highest and lowest values upper and lower tolerance limits for each of the said times, each upper tolerance limit exceeding the highest value observed within a first predetermined interval including that time by an amount that increases with the difference between the highest and lowest values observed in a second predetermined interval including that time and each lower tolerance limit being lower than the lowest value observed within a third predetermined interval including that time by an amount that increases with the difference between the highest and lowest values observed in a fourth predetermined interval including that time; and (c) means for comparing in a running mode the data values received at specific times in the current repetition of said operation with the derived upper and lower tolerance limits for those specific times and for generating an output signal if any said data value exceeds the respective upper tolerance limit or falls below the respective lower tolerance limit.

2. A device as claimed in Claim 1 including means for monitoring at least one said parameter selected from temperatures, pressures, forces, accelerations and vibrations.

3. A device as claimed in Claim 1 or Claim 2 in which the said upper and lower tolerance limits are derived by applying a multiplier to the said highest and lowest values observed within the second and fourth predetermined intervals respectively.

4. A device as claimed in one of the preceeding claims in which the said output signal is a simple indicator that a deviation from the expected range of repetition has occurred

5. A device for assisting in the operation of the manufacturing process that entails repetition of at least one operation substantially as described with reference to the accompanying drawings.