JP2008226006A - Facility equipment diagnostic device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a facility equipment diagnostic device, capable of providing information for performing diagnosis of failure (abnormality) of facility equipment or the presence/absence thereof even if a feature quantity extracted in a normal state varies. <P>SOLUTION: The device includes a feature extraction part 2 which extracts a feature quantity based on a state change of input signal or output signal of a programmable controller, acquired by a measurement part 1; a statistic quantity calculation part 3 which determines a statistic quantity for a fixed period for the extracted feature quantity; a comparison part 5 which determines the degree of slippage of a current statistic quantity from the calculated past statistic quantity; and a variable information collation part 6 which forms diagnostic result obtained by associating the degree of slippage determined by the comparison part with facility equipment, and outputs it to a display part 8 or stores it in an information accumulation part 9. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an equipment diagnosis apparatus and program for making a diagnosis on equipment controlled by a programmable controller.

  2. Description of the Related Art Conventionally, there is an equipment failure diagnosis device that uses a PLC input / output signal to diagnose whether there is an abnormality in an equipment whose operation is controlled by the PLC. As an example of this equipment failure diagnosis apparatus, there is an invention disclosed in Patent Document 1. The present invention is premised on a production system in which various equipments start operation at independent timings and repeat a series of operations in various production processes. In such a production system, the equipment failure diagnosis apparatus records state changes of sensors and contacts in time series, and performs abnormality diagnosis (failure diagnosis) based on this. That is, in the conventional failure diagnosis, a reference pattern of time-series state changes at normal time is created in advance, and normality / abnormality of the equipment is determined by comparison with this.

Japanese Patent No. 3277247

  The conventional failure determination disclosed in Patent Literature 1 and the like is performed by generating a reference pattern of time-series state change by performing feature extraction processing when the production system is operating normally, and at the time of actual abnormality diagnosis, A feature extraction process similar to that at the time of pattern creation is performed, and it is determined that there is a failure (abnormality) when a preset threshold value with respect to the reference pattern is exceeded. As described above, since normality and abnormality are clearly defined and separated, notification is made when an abnormality occurs, and the production system is stopped, the following problems occur.

  In other words, there are cases where state changes occur regularly at regular intervals, and cases where state change timing varies even during normal operation. Therefore, when the normal / abnormal determination is performed, if the threshold value processing is uniformly performed, there is a risk of erroneous determination. In addition, when the normal / abnormal determination is made based on the one having the small variation in the timing of the state change, the abnormality diagnosis cannot be performed on the equipment related only to the one having the large variation. In addition, even if threshold values are individually set according to the degree of variation, it is possible to determine whether a deviation from the reference pattern is due to variations in the normal range or due to a failure or the like. Absent. Therefore, accurate failure (abnormality) diagnosis cannot be performed for items that cause variations even if they are normal.

  The present invention relates to a facility equipment diagnosis apparatus and program capable of providing information for diagnosing whether there is a failure (abnormality) of equipment equipment or the possibility of the presence or absence of the equipment equipment even when there are variations in the feature values extracted in the normal state The purpose is to provide.

  A facility equipment diagnosis apparatus according to the present invention is (1) a facility equipment diagnosis apparatus controlled by a programmable controller, and a feature extraction means for extracting a feature quantity based on a state change of an input signal and / or an output signal of the programmable controller; , A statistic calculating means for obtaining a statistic for a certain period of the feature amount extracted by the feature extracting means, and a comparison comparing the past statistic calculated by the statistic calculating means with the current statistic Means and means for outputting the obtained diagnosis result information by comparing with the comparing means.

  Input signals and output signals may be acquired by monitoring signals input / output to the programmable controller, or may be acquired by accessing the IO memory of the programmable controller, and the acquisition method is arbitrary. In short, any device that can detect a change in the state of a signal may be used. In addition, the diagnosis result is not limited to that for determining the presence or absence of a specific abnormality, but also includes information that provides information for determination.

  In the embodiment, the output means creates diagnostic result information obtained by associating the degree of detachment obtained by the comparison unit with the equipment and outputs it to the display unit 8 or stores it in the information storage unit 9. Corresponds to the “variable information matching unit 6”. The “output” can be variously displayed, such as “displayed”, “stored”, and “printed out”.

  In the present invention, instead of performing the threshold processing on the extracted feature value as in the prior art and clearly determining normal / abnormal, the feature value statistics (average, variance, etc.) are used. Since the amount of deviation from a certain measurement time (the change is large) was obtained and information based on it was presented, even if the feature value obtained in the normal state varies within a certain range It can be determined with high accuracy by comparing the statistics.

  (2) The feature amount can include at least one of a time from a state change of one signal to a state change of another signal and a time calculated based on a state change of each signal. The former corresponds to the feature A in the embodiment, and the latter corresponds to the feature B in the embodiment.

  (3) The diagnosis result information can be information (device name) that identifies a facility diagnosis device related to a statistic having a large change (deviation) from a past statistic.

  (4) Obtaining the distribution of past feature values used in calculating the past statistics and the distribution of current feature values used in calculating the current statistics, and calculating each distribution It is preferable to provide means for outputting in contrast.

  In addition, the program according to the present invention includes a feature extraction unit that extracts a feature amount based on a change in state of an input signal and / or an output signal of a programmable controller, and a fixed period of time for the feature amount extracted by the feature extraction unit. Statistics calculation means for obtaining statistics, past statistics calculated by the statistics calculation means, comparison means for comparing the current statistics, and comparison by the comparison means, and output of diagnostic result information obtained To function as a means to

  In the present invention, since the statistical quantities of the feature quantities are compared with each other, information for diagnosing whether there is a failure (abnormality) of the equipment or whether there is a possibility of the malfunction or the like even when the feature quantities extracted in the normal state vary. Can be provided.

  FIG. 1 shows a preferred embodiment of a facility equipment diagnostic apparatus. In a network system in FA (Factory Automation), a PLC (programmable controller) that controls production facilities and a device whose operation is controlled by the PLC are connected to a network of a control system. These PLCs and devices communicate with each other cyclically via the network of the control system, thereby transmitting / receiving IO data and controlling production facilities.

  The PLC connects a CPU unit that executes calculations based on a control program, an input device such as a sensor or a switch and inputs an on / off signal as an input signal, and an output device such as an actuator or a relay. It is composed by combining multiple units such as an output unit that sends output signals to them, a communication unit that sends and receives data to and from other devices connected to the network, and a power supply unit that supplies power to each unit. Yes. The IO data for these input devices and output devices is stored in the IO memory of the CPU unit. The IO data stored in the IO memory is updated every cycle.

  The equipment diagnosis apparatus of the present embodiment diagnoses equipment (equipment) constituting the network system based on IO data stored in the CPU unit of the PLC. This equipment diagnostic apparatus includes a measuring unit 1, a feature extracting unit 2, a statistic calculating unit 3, a past value storage unit 4, a comparing unit 5, a variable information collating unit 6, and an equipment setting information database 7. The display unit 8 and the information storage unit 9 are provided.

  The measurement part 1 is implement | achieved by the data collection unit which comprises PLC. The data collection unit is one unit constituting the PLC, and is connected to the PLC bus together with the other units, and collects and stores IO data transmitted through the PLC bus. The collection period is set to one cycle of the CPU unit that operates cyclically. Thus, the history of IO data for each cycle is stored in the unit for data collection.

  2 and 3 show the concept of data collection processing in the measurement unit 1. The PLC 10 of the present embodiment is configured by connecting a CPU 11, a data collection unit (measurement unit 1), and various other units. A control command (OUT data) based on the result of the CPU unit 11 calculating and executing the user program is given to the actuator 12 which is an output device. The operation of the actuator 12 is detected by an input device such as a sensor / switch and is taken into the CPU unit 11 as a control result (IN data). Since the IO data composed of the OUT data and the IN data is stored in the IO memory of the CPU unit 11, the IO data is taken in by the data collection unit 1, and as shown in FIG. Recorded as historical data. In the present embodiment, the history information of the IO data is recorded using the data collection unit 1. However, as another method, for example, a lot of IO data is transmitted through the field network. Any data may be used as long as it is measured by a measuring instrument and recorded by a data logger, such as using a data collection device that monitors data transmitted and acquires necessary information.

  The feature extraction unit 2 reads out the history information of each IO data recorded by the measurement unit 1 and uses each as a time-series pulse signal (ON / OFF pulse signal of each IO data), and performs feature extraction from a plurality of pulse signals. . Features extracted in this embodiment are prepared in two types: feature A used for viewing the state of a signal group and feature B used for viewing the state of each signal, using the relationship between a plurality of signals. .

  As a premise, since the equipment to be controlled by the PLC repeatedly performs the same movement, the signal to be subjected to feature extraction is repeatedly performed with the ON and OFF switching timings being substantially constant.

  That is, focusing on one signal, the time when the m-th ON from the reference point (for example, operation start) is TON (m), and the time when the m-th OFF is from the reference point (for example, operation start). Is TOFF (m), TON (m) is substantially constant for each value from m = 1 to M, and TON (m) is substantially constant for each value from m = 1 to M. Indicates the value. Of course, TON (m) and TOFF (m) may indicate the same value (duty ratio: 1/2) or different values. Furthermore, the time of one cycle of ON / OFF also becomes a substantially constant value. The “variation range” that takes the “substantially constant value” varies depending on each signal.

  Since each signal is a device control signal, a device operation detection signal, or the like, the generation timing between the signals is also fixed. Therefore, the order in which each signal is turned on (rises) or the order in which each signal is turned off (falls) is the same in each cycle. In other words, if the signal does not repeat ON / OFF at a predetermined timing or the appearance order is different, it is not a target for feature extraction of this embodiment.

  As an example of the feature A, there is a time difference between two signals in which the generation order is continuous in a signal group in which the generation order of each signal is the same and repeated. That is, as shown in FIG. 4, in the logic signals Sj, j = 1, 2,..., J, the time difference (difference between the state change times) between two logic signals in which the rise of each signal continues is represented by Dn (m , M = 1, 2,..., M, n = 1, 2,. Here, m means the m-th cycle after starting the feature extraction process. One cycle is a period in which the reference signal (here, S1) is turned ON / OFF once. Further, one cycle period of each signal and one cycle of the reference signal are not necessarily equal. That is, for example, there may be a signal that repeats ON / OFF a plurality of times while the reference signal is turned ON / OFF once. When all the logic signals Sj are turned ON / OFF once during one cycle of the reference signal, N = J.

  As another example of the feature A, each time from when the reference signal is turned on to when each signal is turned on may be used. That is, as shown in FIG. 5, the time difference between the rising edge of the reference signal S1 and the rising edge of each signal (difference between state change times) is expressed as Dn (m), m = 1, 2,. n = 1, 2,... Here, M means the Mth cycle after the feature extraction process is started. One cycle is a period in which the reference signal (here, S1) is turned ON / OFF once. In each of the above examples, the time difference between the rising edges of the two signals is characterized. However, the time difference between the falling edges of each signal may be characterized.

  As an example of the feature B, as shown in FIG. 6, the time when each signal rises, the time when each signal does not rise as shown in FIG. 7, the period of each signal as shown in FIG. There is. In these cases, if all the logic signals Sj are turned ON / OFF once during one cycle of the reference signal, N = J.

  The actual feature extraction processing in the feature extraction unit 2 executes the processing steps from S1 to S11 in the flowcharts shown in FIGS. Specifically, the feature extraction unit 2 determines whether or not the process is to extract a feature that looks at the relationship between a plurality of signals and (feature A) (S1). If the branch determination is Yes, signal setting is performed. A logic signal for determining information is sampled (S2). Next, the feature extraction unit 2 extracts the time of the ON state for each sampled signal (S3). In the extraction of the ON state, for example, focusing on the rise of the signal, the ON generation interval (time of one cycle) for an arbitrary period is calculated. Of course, the falling edge of the signal may be used as a reference.

  The feature extraction unit 2 calculates the variation (variance) in each signal based on the obtained ON-state time for a plurality of cycles of each signal (S4). An arbitrary period in the processing step S3 is set to a value sufficient to calculate this variance.

  A signal having the smallest variation obtained in the processing step S4 is used as a reference signal, and the rise of the signal is used as a trigger. Then, with reference to the rising edge of the trigger signal, the signals are labeled in the order of their rising speed.

  If each signal is labeled, the feature extraction unit 2 stores the information of the trigger signal and the information that associates each signal with the label as setting information in the storage means (S6). . The relation between the trigger and the label and the signal is a table structure in which the address and the label in which the value of each signal is stored are associated (see FIG. 9). The storage means may be either a volatile memory or a nonvolatile memory. The setting information stored in this storage means is then used for the feature A extraction process (S11) from the signal sampled by the execution of the processing step S8. That is, as a relationship between a plurality of signals, processing such as obtaining a time between signals of two preset labels (from the rise of one signal to the rise of the other signal, etc.) is performed. A combination of signals for checking the relationship is set in advance.

  Further, when extracting the feature amount of each individual signal like the feature B, since the branch determination processing step is No, the processing jumps to the processing step S7 and the signal is sampled (S7).

  Then, the feature extraction unit 2 calculates the operation time based on the sampled data. That is, the feature extraction unit 2 extracts the features A and B (S10, S11). Specifically, as illustrated in FIGS. 4 to 8, the relationship (time difference) between the plurality of signals and the individual signals. The operation time (ON time, etc.) is extracted. As shown in FIG. 10, the data of the operation time (after feature extraction) of each signal (single and mutual) is accumulated in time series for each period (m). This time-series accumulation may be performed using various storage means not shown in FIG. 1 or may be temporarily stored in a buffer memory or the like provided in the feature amount calculation unit 3. good. Of course, other storage methods can be used.

  Each feature extracted by the feature extraction unit 2 is given to the statistic calculation unit 3 at the next stage. The statistic calculator 3 calculates a past value, which is a temporal data change for each feature, as a statistic (S12, S13, S14 in FIG. 20). Each calculated statistic is stored in the past value storage unit 4 (S15 in FIG. 20). The statistics calculated here are the mean, variance, and distribution for each feature.

  The comparison unit 5 acquires the current statistical amount calculated by the feature amount calculation unit 3 and the past statistical amount stored in the past storage unit 4 (S16 and S20 in FIG. 20), and the degree of deviation for each variable. (S17 in FIG. 20) and the degree of deviation from the past value is compared between the variables (S18 in FIG. 20).

In particular,
μ: Average obtained from newly measured data μ ′: Average obtained from previously measured data σ2: Variance obtained from newly measured data σ′2: Variance obtained from previously measured data if you did this,
μ / μ ′
σ2 / σ'2
Thus, the degree of deviation from the past value can be obtained. In this case, if the newly measured data is almost the same as the statistical amount of the data measured in the past (the degree of deviation is small), these values are approximately 1.

Also,
(Μ−μ ′) / μ ′
(Σ−σ ′) 2 / σ′2
The degree of deviation from the past value can also be obtained. In this case, if the newly measured data is almost the same as the statistical amount of the data measured in the past (the degree of deviation is small), these values are approximately zero.

  The statistic calculation cloth section 3 measures and records, for example, every 5 minutes as a fixed period, calculates a statistic (average, variance, distribution, etc.) and stores it in the database (past value storage section 4). The comparison unit 5 can extract arbitrary past data from the past value storage unit 4 and compare the degree of deviation. These are obtained and stored for every period such as every hour, every day, every week, every month, etc. (see FIGS. 11, 12, etc.).

  The variable information matching unit 6 sequentially arranges the equipment devices having a large degree of deviation from the past values with respect to the average, variance, and other statistics obtained by the comparing unit 5 and displays the result on the display unit 8 or information Or stored in the storage unit 9. Which variable is associated with which equipment is read out from the equipment setting information database 7 in advance and used.

  As an example, when the degree of deviation from the past value is calculated by the comparison unit 5, for example, as shown in FIG. 13A, when a graph is drawn with the horizontal axis as a variable and the vertical axis as the degree of deviation, The graph goes up and down around a certain one. When the variable information matching unit 6 rearranges this graph in the order of the degree of deviation, the graph is replaced as shown in FIG. Then, a variable located near both ends (less than 1 / greater than 1) deviating from the standard level “1” has a large degree of deviation, and a device related to the variable becomes a maintenance candidate device. The variable information matching unit 6 outputs the graph to the display unit 8 or stores it in the information storage unit 9. In this graph, the degree of deviation is obtained by an arithmetic expression of μ / μ ′, but the same applies to those using other arithmetic expressions.

  The value (length of time) of each variable has a small variation such as a reference signal and a large variation originally. Therefore, if attention is paid only to the value of a variable at a certain moment, even if the value is largely deviated from the standard value of the variable, the variable may be within the normal range when the variable is originally large. Conversely, even when the value of a variable at a certain moment does not deviate so much from the standard value of that variable, it may be abnormal if the variable is originally small in variation.

  Therefore, in the present embodiment, as described above, the statistic is obtained for both the reference past value and the value to be judged, and the degree of deviation between the statistics is obtained, so that it is possible to correctly determine whether or not the product is acceptable. I made it.

  In other words, for example, when an average value is used as a statistic, even if the value of a variable at a certain moment is significantly different from a standard value (average value of past values), it is within the range of normal (normal) variation. In the case of, it becomes close to the standard value by taking an average over a certain period. Therefore, the deviation amount becomes a value close to the standard level (1 or 0) and can be determined to be normal.

  Conversely, even if there is not much difference between the value of the variable at a moment and the standard value (average value of past values), if it is outside the range of normal (normal) variation, the average over a certain period By taking, it will not be close to the standard value. Therefore, the deviation amount becomes a value far from the standard level (1 or 0), and it can be determined that the abnormality is approaching or is abnormal.

  This is not limited to the average that is the first-order statistic, and the same applies to the variance that is the second-order statistic, and also to other statistics such as the third-order and fourth-order statistics. In addition, the certain period suitable for the pass / fail judgment varies depending on each variable and the statistic to be used. Therefore, the user sets a suitable period in advance, and obtains statistics for the set period, obtains statistics for various periods, and obtains and determines deviations for necessary periods. Can be.

  Then, as shown in FIG. 14, by comparing with the past statistics in a time series, it is possible to extract a variable whose amount of deviation is large, and before a complete abnormality occurs, that is, normal It is possible to detect a variable that is within the range but becomes abnormal in the near future if left as it is. Further, in FIG. 14, the comparison is made in units of one week, but this comparison period (cycle) can be arbitrarily set. Note that the statistic obtained for the pass / fail judgment is also used as a past value for comparison with a statistic calculated thereafter.

  Also, as shown in FIG. 15, those that are particularly worrisome among the monitoring targets can be displayed by displaying the distribution in the period for which the statistics are calculated, and confirming how far the distribution is from the reference time distribution. . Such processing is also performed by the variable information matching unit 6 based on the designation of the variable (facility equipment) for displaying the distribution from the user (S21 in FIG. 20).

  Further, the variable information collating unit 6 acquires the equipment name related to the equipment with the large amount of deviation calculated for each variable (the rate of change is greater than or equal to the set value) from the equipment information setting information database 7 (S19 in FIG. 20). ), It is displayed on the display unit 8 in association with the address of the variable (S22 in FIG. 20), or stored in the information storage unit 9 (S23 in FIG. 20). The information to be displayed and / or stored (maintenance candidate device list) is as shown in FIG. 16, for example. In particular, when the information is displayed in ascending order from the one with the largest change rate as shown in FIG. Can easily recognize equipment to be urgently maintained.

  17 and 18 show an example of an actual display layout for the display unit 8. In this example, an example is shown in which evaluation is performed using the average and variance as statistics, and each is displayed on the same screen in a separate window. As shown in FIG. 17, in the state where the window for dispersion is displayed in front, by selecting the rear window, the window for average is switched to be displayed in front as shown in FIG. Show.

  The display layout of each display window W is as follows. First, an area W1 for selecting and specifying the type of statistic used from the pull-down menu is provided at the upper left of the display window W. In addition, an area W2 for designating the feature quantity to be extracted and a comparison time designation area W3 for specifying the past value to be compared are provided at the lower right of the display window W. The designation method in both these areas W2 and W3 is also a pull-down menu method.

  In accordance with these input conditions, the statistic calculation unit 3 obtains the statistic based on the feature extracted by the feature extraction unit 2, and the comparison unit 5 deviates from the past statistic stored in the past value storage unit 4. The degree is obtained, and the variable information collation unit 6 sorts out the degree of deviation of each variable, and outputs the sorted graph in the degree-of-offset result display area W4.

  Further, the variable information matching unit 6 outputs information about the top X equipment names having a large change rate to the maintenance candidate equipment list display area W6. This X is the number (pull-down menu method) designated in the change rate upper display number designation area W5.

  Further, the variable information collating unit 6 outputs a distribution graph of current and past values for the equipment specified by the distribution display number designation area W7 and the distribution display number designation area W8 to the departure display area W9.

It is a block diagram which shows suitable one Embodiment of the equipment diagnostic apparatus of this invention. It is a figure explaining an effect | action. It is a figure which shows an example of the data structure of the time series data of the state change collected and memorize | stored in a data collection unit. It is a figure explaining the feature B. FIG. It is a figure explaining the feature B. FIG. It is a figure explaining the feature A. It is a figure explaining the feature A. It is a figure explaining the feature A. It is a figure which shows an example of setting information (The relationship between the signal and the address in which the value of the signal is stored). It is a figure which shows an example of the data structure in the memory | storage means which memorize | stores the data by which the feature extraction was carried out. It is a figure explaining an effect | action (function). It is a figure explaining an effect | action (function). It is a figure explaining an effect | action (function). It is a figure explaining an effect | action (function). It is a figure explaining an effect | action (function). It is a figure explaining an effect | action (function). It is a figure which shows an example of the display layout which a display part displays. It is a figure which shows an example of the display layout which a display part displays. It is a flowchart which shows the function of an equipment diagnostic apparatus. It is a flowchart which shows the function of an equipment diagnostic apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Measurement part 2 Feature extraction part 3 Statistics calculation part 4 Past value storage part 5 Comparison part 6 Variable information collation part 7 Equipment apparatus setting information database 8 Display part 9 Information storage part

Claims (5)

A diagnostic device for equipment controlled by a programmable controller,
Feature extraction means for extracting a feature quantity based on a change in state of an input signal and / or an output signal of the programmable controller;
Statistic calculation means for obtaining a statistic for a certain period of the feature quantity extracted by the feature extraction means;
A comparison means for comparing a past statistic calculated by the statistic calculation means with a current statistic;
Means for comparing by the comparison means and outputting the obtained diagnosis result information;
A facility equipment diagnostic apparatus comprising:   The feature amount includes at least one of a time from a state change of one signal to a state change of another signal and a time calculated based on a state change of each signal. Equipment equipment diagnostic device according to.   The equipment diagnosis apparatus according to claim 1, wherein the diagnosis result information is information for identifying equipment diagnosis equipment related to a statistic having a large change from a past statistic. Obtaining a distribution of past feature values used in calculating the past statistics and a distribution of current feature values used in calculating the current statistics;
The equipment diagnosis apparatus according to any one of claims 1 to 3, further comprising means for comparing and outputting the respective distributions. Computer
Feature extraction means for extracting a feature quantity based on a change in state of an input signal and / or an output signal of a programmable controller;
Statistic calculation means for obtaining a statistic for a certain period of the feature quantity extracted by the feature extraction means;
A comparison means for comparing a past statistic calculated by the statistic calculation means with a current statistic;
Means for comparing with the comparison means and outputting the obtained diagnosis result information;
Program to function as.

Images