Abnormal modulation cause determination device, abnormal modulation cause determination method, and abnormal modulation cause determination program

Abstract

Improved performance in determining the cause of abnormal modulation in production equipment. The abnormal modulation cause determining device comprises: a process data acquisition unit that reads process data from a storage device that stores process data continuously output by multiple sensors in the production equipment and associated with a management number of a processing target; an abnormality determination unit that continuously calculates an abnormality degree representing the degree of modulation of the process data read by the process data acquisition unit for the multiple sensors; and a cause diagnosis unit that uses causal relationship information to determine whether the abnormality degree calculated by the abnormality determination unit meets a predetermined benchmark for the process data corresponding to the management number of the processing target and output by the multiple sensors, wherein the causal relationship information defines a combination of a cause and the modulation of the process data output by the multiple sensors that occurs as an effect of the cause.

Description

Technical Field

[0001] The present disclosure relates to an abnormal modulation cause determination device, an abnormal modulation cause determination method, and an abnormal modulation cause determination program. Background Art

[0002] In the past, a technology for estimating the cause of an abnormality using operating data received from a unit has been proposed. For example, a technology for estimating the cause of an abnormality precursor by weighting the probability of a first abnormal event that occurred in the past relative to the probability of a second abnormal event that did not occur has been proposed (Patent Document 1). In addition, a technology has also been proposed that, when an abnormality is detected in a process to be diagnosed, estimates a contribution rate representing the proportion of the contribution of a process variable, wherein the process variable represents the state of the process to be diagnosed, and estimates an event that may be the cause of the abnormality from predefined registered events based on the contribution rate (Patent Document 2).

[0003] In addition, a technique has been proposed that uses multiple sub-models for predicting process states to calculate deviation indicators from normal states of the process, and infers the cause of abnormal states occurring in the process based on a set of deviation indicator patterns formed by the deviation indicators calculated for each sub-model (Patent Document 3). Furthermore, a technique has been proposed that calculates a monitoring contribution relative to the monitoring abnormality for each of the multiple operating data constituting the monitoring operating data, extracts a diagnostic target data group consisting of the top N operating data with the largest monitoring contributions, calculates a conformity index between the diagnostic target data group and a diagnostic reference data group consisting of the top M operating data with the largest reference contributions included in the reference operating data, and predicts abnormalities in the unit based on the reference operating data for which the conformity index is above a conformity determination threshold (Patent Document 4).

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2018-109851

[0007] Patent Document 2: Japanese Patent Application Publication No. 2018-120343

[0008] Patent Document 3: Japanese Patent Application Publication No. 2019-16039

[0009] Patent Document 4: Japanese Patent Application Publication No. 2019-57164 Summary of the Invention

[0010] Problems to be solved by the invention

[0011] Generally speaking, in production facilities, it is desirable to prevent abnormal modulation and suppress the effects on safety, stability, product quality, cost, etc. The present technology aims to improve the ability to identify the cause of abnormal modulation in production facilities.

[0012] Technical Solution

[0013] The device for determining the cause of abnormal modulation comprises: a process data acquisition unit, which reads out process data from a storage device that stores process data continuously output by multiple sensors possessed by production equipment and associated with the management number of a processing object; an abnormality determination unit, which continuously calculates, for multiple sensors, an abnormality degree representing the degree of modulation of the process data read by the process data acquisition unit; and a cause diagnosis unit, which uses causal relationship information to determine whether the abnormality degree calculated by the abnormality determination unit meets a prescribed benchmark for the process data corresponding to the management number of the processing object output by multiple sensors, wherein the causal relationship information defines a cause and a combination of the modulation of the process data output by multiple sensors that occurs as an influence due to the cause.

[0014] In production equipment, for example, where a process object is processed across multiple steps, it's impossible to correlate process data obtained from multiple sensors measuring the same process object from the moment the process data was measured. By associating process data based on management numbers as described above, it becomes easier and more accurate to calculate the degree of abnormality based on process data obtained from multiple sensors measuring the same process object. This improves the ability to identify the cause of abnormal processing in production equipment.

[0015] Furthermore, the cause diagnosis unit may determine the accuracy of the cause of modulation based on the proportion of process data whose abnormality, as calculated by the abnormality determination unit, meets a predetermined standard, among the process data corresponding to the management number of the processing target output by the multiple sensors. This allows the detection of signs of abnormality due to a certain cause, even if only a small amount of process data meets the predetermined standard among the process data output by the multiple sensors affected by modulation. This improves the ability to identify the cause of abnormal modulation in production equipment.

[0016] Alternatively, the system may further include a pre-processing unit that calculates average time-series data using process data corresponding to processes performed in the production equipment and having different management numbers, synchronizes the process data based on a degree of similarity with the calculated average time-series data, and an abnormality determination unit that calculates an abnormality degree based on a degree of deviation from a predetermined reference for the synchronized process data corresponding to the processes having different management numbers. Alternatively, the process data of the same process subject to processing but having different management numbers may be synchronized, for example, as described above, and the abnormality degree calculated.

[0017] Alternatively, the process data may be stored in a storage device in association with steps representing stages of processing within a process performed in the production facility, and the abnormal preparation cause identification device may further include an output control unit that causes the process data corresponding to the steps, associated with different management numbers, to be displayed superimposed on the output device. Furthermore, by synchronously displaying process data for the same process, a user can easily compare multiple sets of process data.

[0018] It should be noted that the contents of the technical solutions can be combined as much as possible without departing from the scope of the problems and technical concepts of the present disclosure. In addition, the contents of the technical solutions can be provided as a system including a computer or multiple devices, a method executed by a computer, or a program for causing a computer to execute the method. It should be noted that a recording medium storing the program can also be provided.

[0019] Effects of the Invention

[0020] According to the disclosed technology, it is possible to improve the accuracy of determining the cause of abnormal modulation in a production facility. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a diagram showing an example of a system according to this embodiment.

[0022] Figure 2 This is a schematic diagram showing an example of a process performed by the equipment included in the plant.

[0023] Figure 3 This is a diagram for explaining an example of process data in a batch process.

[0024] Figure 4 This is a diagram showing an example of a preset process line definition table.

[0025] Figure 5 This is a diagram showing an example of a preset label definition table.

[0026] Figure 6 This is a diagram for explaining an example of process data in a continuous process.

[0027] Figure 7 This is a diagram showing an example of traceability information.

[0028] Figure 8 This diagram illustrates the correspondence between process data in a continuous process and production numbers in a batch process.

[0029] Figure 9 This is a diagram showing an example of information pre-registered in the knowledge base.

[0030] Figure 10This is a diagram showing an example of a logic tree that represents the relationship between modulation and its cause.

[0031] Figure 11 This is a diagram for explaining the synchronization processing of process data.

[0032] Figure 12 This is a diagram for explaining an example of calculating the degree of abnormality based on the distance from a reference for time series data.

[0033] Figure 13 This is a diagram for explaining an example of calculating the abnormality degree based on the distance from a reference in consideration of positive and negative directions for time series data.

[0034] Figure 14 This is a diagram for explaining abnormality detection using an autoencoder.

[0035] Figure 15 This is a block diagram showing an example of the configuration of an abnormal modulation cause identifying device.

[0036] Figure 16 This is a process flow chart showing an example of a learning process executed by the abnormal modulation cause identification device.

[0037] Figure 17 This is a diagram showing an example of an action table.

[0038] Figure 18 This is a process flow chart showing an example of abnormality sensing processing executed by the abnormal modulation cause identification device.

[0039] Figure 19 This is a diagram showing an example of a screen output to an input / output device.

[0040] Figure 20 It is a diagram showing another example of a screen output to the input / output device. DETAILED DESCRIPTION

[0041] Hereinafter, embodiments of the abnormal modulation cause identification device will be described with reference to the accompanying drawings.

[0042] <Implementation Method>

[0043] Figure 1This diagram illustrates an example of a system according to the present embodiment. System 100 includes an abnormal modulation cause determination device 1, a control station 2, and a unit 3. System 100 is, for example, a distributed control system (DCS) and includes multiple control stations 2. Specifically, the control system of unit 3 is divided into multiple partitions, each of which is controlled in a distributed manner by a control station 2. Control station 2 is a conventional device in a DCS that receives status signals from sensors and other components of unit 3 and outputs control signals to unit 3. Based on the control signals, control stations 2 control actuators such as valves and other components of unit 3, as well as other equipment.

[0044] The abnormal modulation cause determination device 1 obtains the status signal (process data) of the unit 3 via the control station 2. The process data includes the temperature, pressure, flow rate, etc. of the processing object as raw materials and intermediate products, or the set values ​​that determine the operating conditions of the equipment possessed by the unit 3. In addition, the abnormal modulation cause determination device 1 creates an abnormality sensing model based on a knowledge base, which stores the correspondence between assumed causes and effects that appear as abnormalities, for example. For example, the following model is created: based on the knowledge base, it is used to determine abnormal modulation, its precursors, and its causes based on a method of sensing changes in process data that deviate from the allowable range. Then, the abnormal modulation cause determination device 1 can use the model and process data to sense the occurrence of abnormal modulation or its precursors. In addition, it is also possible that the abnormal modulation cause determination device 1 obtains candidates for operating conditions, such as those for suppressing abnormal modulation, based on a table storing the causes of abnormal modulation and actions for coping with them, as well as the determined causes, and presents them to the user.

[0045] Figure 2 This is a schematic diagram illustrating an example of a process performed by the machines in a unit. In this embodiment, the process may include a batch process 31 and a continuous process 32. In the batch process 31, the processing target is processed sequentially according to each specified processing unit, for example, the raw materials are received, held, and discharged in sequence by each machine. In the continuous process 32, the continuously introduced processing target is processed continuously, for example, the raw materials are received, held, and discharged in parallel. Furthermore, the process may include multiple series 33 performing the same process in parallel.

[0046] The machines that perform each process include, for example, reactors, distillation devices, heat exchangers, compressors, pumps, tanks, etc., and these machines are connected via piping. In addition, sensors, valves, etc. are provided at specified locations on the machines and piping. The sensors may include thermometers, flow meters, pressure gauges, level meters, concentration meters, etc. In addition, the sensors monitor the operating status of each machine and output status signals. In addition, the sensors provided by the unit 3 are configured as sensors with "tags" attached as identification information for determining each sensor. That is, the type of process data can be determined based on the tags. Then, the abnormal modulation cause determination device 1 and the control station 2 manage the input and output signals to each machine based on the tags.

[0047] <Batch process>

[0048] Figure 3 This is a diagram for explaining an example of process data in a batch process. Figure 3 The left column represents Figure 2 The batch process 31 is shown as part of the process. Specifically, the process includes a pulverizer 301, a cyclone 302, a pretreatment 303, a precooler 304, and a reactor 305. In addition, these processes are classified into a pretreatment step, a precooling step, and a reaction step. Figure 3 The column on the right side of represents an example of process data obtained in each process. In the pretreatment process, time series data is obtained from sensors labeled 001 and 002. In the precooling process, time series data is obtained from sensors labeled 003 and 004. In the reaction process, time series data is obtained from sensors labeled 005, 006, and 007. In addition, in the batch process, the processing objects corresponding to the production numbers (also called "product numbers", "batch numbers", "management numbers") are processed intermittently. That is, the production number is identification information for identifying the processing objects that are centrally processed in the batch process. As Figure 3 As shown, as time passes, data regarding the timing of subsequent processing objects associated with the production numbers is obtained. In this embodiment, the control station 2 manages production numbers and steps indicating the stages of processing within the subdivided processes that constitute the batch process. It should be noted that, when a step is reset by a PLC (Programmable Logic Controller) within a unit 3 connected to the control station 2, the production number of the process data output from the control station 2 may be appropriately adopted (for example, after the step is switched in the PLC or after a set time has passed) based on the timing of communication between the control station 2 and the unit 3. Furthermore, the set time may be set for each manufacturing line or for each subdivided process.

[0049] Figure 4This figure shows an example of a pre-set process line definition table. The process line definition table registers the production number for each series and process, step definitions indicating the processing stages within each process, and the product types processed within each process. The process line definition table can be a table in a so-called database or a file in a specified format such as a CSV file. Furthermore, the process line definition table is also created in advance by the user and read by the abnormal preparation cause identification device 1.

[0050] The process line definition table includes attributes such as series, process, product number, step, and variety. The series column registers identification information for identifying the process series. The process column registers identification information for the subdivided process within the batch process. The product number column registers the production number, which is used to identify the processing objects processed collectively within the batch process. The step column registers the timing definitions for multiple steps that represent the processing stages within the process. The variety column registers the category of the processing object.

[0051] Figure 5 1 is a diagram showing an example of a preset tag definition table. The tag definition table defines the acquisition timing of process data obtained from the sensor corresponding to each tag.

[0052] It should be noted that the label definition table may be a table in a so-called database or a file in a predetermined format such as CSV (Comma Separated Values).

[0053] The label definition table contains attributes such as label, series, process, and collection interval. The label column registers the sensor's identification information. The series column registers the series' identification information used to identify the process. The process column registers the identification information indicating the subdivided process within a batch process. The collection interval column registers the interval for acquiring sensor output values.

[0054] <Continuous process>

[0055] Figure 6 This is a diagram for explaining an example of process data in a continuous process. Figure 6 The left column represents Figure 2 A portion of a process is shown as a continuous process 32. Specifically, the process includes a reservoir 311 and a pump 312. Figure 6The column to the right of shows an example of process data acquired during each process. In continuous process 32, time-series data associated with tags, but not with production numbers, is continuously acquired from sensors. In the continuous process, time-series data is acquired from sensors labeled 102 and 103. In the continuous process, the machine continuously receives and processes objects.

[0056] When a continuous process is performed after a batch process, in order to associate the processing targets in the batch process with the processing targets in the continuous process, in this embodiment, traceability information set in advance by the user is used. Figure 7 This diagram shows an example of traceability information. Traceability information includes attributes such as sampling interval and residence time. The sampling interval column records the interval at which samples are taken for process inspection in a continuous process, for example, using the batch method. The residence time column records the time the processed object remains in the process from the completion of the batch process until it reaches the continuous process.

[0057] Figure 8 This diagram illustrates the association between process data in a continuous process and production numbers in a batch process. For example, process data is acquired at intervals specified in traceability information. Furthermore, when a continuous process is performed after a batch process, products from the batch process completed within a specified period are introduced into a storage device, etc., as processing targets for the continuous process. Therefore, for process data in a continuous process, the residence time of the processing target from the completion of the batch process to the time of sensor measurement can be traced, and the completion time of the batch process can be associated with the group of production numbers included in the specified period. This association improves the accuracy of identifying the cause of anomalies using process data in the batch process when batch and continuous processes are performed continuously.

[0058] As described above, by associating the production numbers in the batch process with the measurement timings in the continuous process, it is possible to improve the accuracy of identifying the cause of the abnormality.

[0059] Figure 9 : is a diagram showing an example of information pre-registered in the knowledge base. The knowledge base is pre-stored in the storage device of the abnormal modulation cause identification device 1. Figure 9 The table includes an "Influence" column corresponding to each sensor (label) and a row indicating the "Assumed Cause" of the modulation. Specifically, the direction of value fluctuation is registered in the column corresponding to the sensor affected by the cause (e.g., "Cause 1" or "Cause 2") shown in each row. In the knowledge base, the direction of fluctuation is indicated by "Up" indicating an increase (rise) in the sensor output value or "Down" indicating a decrease (fall) in the sensor output value.

[0060] It should be noted that if Figure 9 As shown, the combination of cause and effect is not limited to one-to-one. In addition, the operation method of process data, extraction timing, threshold value for abnormality judgment, etc. are determined in correspondence with each sensor. Information representing the operation performed on the output value of each sensor is registered in the row of the operation method. It should be noted that, in this embodiment, for example, it is assumed that machine learning methods such as Hotelling's method, k-nearest neighbor algorithm, DTW Barycenter Averaging, Autoencoder, Graphical Lasso, etc. are used to perform the operation. Information representing the timing of extracting the value used for abnormality judgment from the output value of each sensor is registered in the row of the extraction timing. As for timing, for example, in batch processing, it can be defined by steps representing the stage of processing in each process, a certain period, a time point, etc. In addition, in continuous processing, it can be defined by, for example, Figure 7 The threshold row registers the threshold value used as the basis for determining an abnormality in each abnormality determination method. For example, the threshold value includes both an upper limit and a lower limit. As described above, the knowledge base defines the causal relationship between the event that serves as a cause and the influence of the modulation of process data caused by the event. Furthermore, the causal relationship can be represented in a tree format, with the modulation that serves as the influence as the root and its assumed cause as the leaf, connecting the events that occur in the process from the cause to the modulation in a hierarchical manner along time.

[0061] The knowledge base is created in advance by the user, for example, based on a Hazard and Operability Study (HAZOP). A HAZOP, for example, is a method for comprehensively listing and associating sensing units at monitoring points of equipment and devices constituting a plant, a management range (threshold values ​​of upper and lower limits and alarm setpoints), deviations from the management range (anomalies, modulations), a list of hypothetical causes of deviations from the management range, logic (sensing units) for determining which hypothetical cause caused the deviation, the impact of the deviation, measures taken when the deviation occurred, and actions taken in response to the measures. It should be noted that the knowledge base is not limited to HAZOP and can also be created based on FTA (Fault Tree Analysis), FMEA (Failure Mode and Effect Analysis), ETA (Event Tree Analysis), or methods applying these, similar methods, content extracted from operator interviews, or content extracted from work standards and technical standards. In this embodiment, abnormality detection is performed based on parameters that are set to have a causal relationship in the knowledge base.

[0062] Based on the information set in the above table, the abnormal modulation cause identification device 1 extracts data at a predetermined timing from the process data acquired from the unit 3 and performs abnormality determination using a predetermined method. Figure 10 is a diagram showing an example of a logic tree representing the relationship between modulation and its cause. Figure 9 The knowledge base shown is used to create a logical tree. In addition, Figure 10 The logic tree arranges upstream and early events in the production process on the left and downstream and subsequent events in the production process on the right, connecting them in layers with arrows in a manner from the assumed cause toward the modulation that appears as an influence. In addition, in the logic tree, when there are multiple assumed causes for one modulation in the knowledge base table, they are connected in branches, and events that appear in common in the process from the assumed cause to the modulation are displayed in a bundle. The thick solid rectangle at the end of the upstream side of each branch corresponds to the assumed cause in the knowledge base table. Figure 9 and Figure 10The numbers in parentheses correspond to each other. Furthermore, the thin solid rectangles correspond to the influences in the knowledge base table and represent events that can be observed through process data. For each of these influences, calculations corresponding to the calculation methods defined in the knowledge base table are performed. Furthermore, for each hypothetical cause, a model including formulas for performing these calculations can be defined. This model can be used to detect anomalies or their precursors, or to assist in identifying their causes.

[0063] <Calculation method>

[0064] The above-mentioned calculations may include, for example, the following methods: Furthermore, the abnormal modulation cause identification device 1 may display the calculation results of these methods.

[0065] Hotelling method 2 Law)

[0066] For example, assuming that a plurality of process data obtained from a sensor follow a prescribed probability density function, the mean and standard deviation of the population are estimated based on the sample mean and sample standard deviation calculated using the process data. The prescribed probability density function is, for example, a normal distribution. Then, the degree of abnormality is calculated based on the distance from the mean of the population to the process data of the verification object. For example, the degree of abnormality is determined based on the square of the Mahalanobis distance. It should be noted that the degree of abnormality based on the Hotelling theory can be calculated using the instantaneous value of the process data itself, or the maximum value, minimum value, cumulative value, standard deviation or differential coefficient (slope) of the process data during a prescribed period. According to the Hotelling method, outliers from the prescribed benchmark can be sensed.

[0067] k-nearest neighbor algorithm

[0068] For example, time-series process data acquired from one or more sensors is vectorized or matrixed, and the distances between the data are calculated. Distances can be Euclidean, Mahalanobis, or Manhattan distances. The degree of anomaly is then determined based on the distance to the k-th closest data point to the target data. The k-nearest neighbor algorithm makes its determination based on its relationship to other data points. Therefore, for example, if normal values ​​can be categorized into multiple clusters, it is possible to detect outliers that are far from any of these clusters.

[0069] DTW (Dynamic Time Wrapping) Barycenter Averaging

[0070] The average time series data is calculated based on a plurality of time series data such as process data in different batch processes. For example, the distance from the average time series data to process data of different production numbers in corresponding intervals in the batch process can be calculated. Figure 11This is a diagram for explaining the synchronous processing of process data. For each value of the elements included in the time series data of batch processing with different production numbers, the shortest distance between the values ​​included in the different time series data is cyclically calculated, and the time series data is aligning by sliding along the time axis in a manner that the cumulative value of the shortest distance is minimized. That is, based on the similarity of the time series data, multiple time series data are synchronized. In this way, multiple process data can be displayed in an overlapping manner so that the steps in the process implemented in unit 3 correspond to each other in time sequence. Then, based on the cumulative value of the distance between the synchronized time series data, the abnormality is calculated by the k-nearest neighbor algorithm and the Hotelling theory. According to DTW Barycenter Averaging, abnormalities can be sensed based on the similarity between the time series data.

[0071] Alternatively, the abnormality level with a positive or negative sign may be calculated for the deviation from a reference such as the average value. Figure 12 This is a diagram for explaining an example of calculating the degree of abnormality based on the size of the distance from a reference for time series data. Figure 13 This is a diagram for explaining an example of calculating the abnormality degree based on the distance from a reference in consideration of positive and negative directions for the same time series data. Figure 12 and Figure 13 In the example, the vertical axis represents the degree of deviation from the average value. In the portion indicated by the dotted rectangle, modulation is actually generated, but only based on Figure 12 The value shown in the example is difficult to sense. On the other hand, Figure 13 In the example of , there is a tendency for the positive and negative directions to deviate in opposite directions, and therefore, modulation sensing becomes easy.

[0072] For example, in the Hotelling method described above, the degree of deviation from the reference is obtained without squaring the distance and the value is given a positive or negative sign. Figure 13 In DTW BarycenterAveraging and the like, for characteristic points such as maximum values ​​in time series data, the positive or negative sign is determined by, for example, the following formula, and the calculated value is multiplied by the distance.

[0073] Sign determination formula = (μ-x) / |μ-x|

[0074] It should be noted that μ is the average value (reference value) of the training data, and x is the process data of the verification object. In this way, according to the sign determination formula, the sign indicating the direction of deviation from the reference at the time point can be determined according to the size relationship between the reference value of the time series data at a specified time point and the process data of the verification object at the corresponding time point. In addition, by using a value with a sign indicating the degree of deviation from the reference, the image can be obtained. Figure 13Such abnormality can suppress false sensing. In addition, as a characteristic point in the time series data, in addition to using the maximum value, the minimum value, the difference between the process data at a certain time point and the process data at another time point, etc. can also be used. · Autoencoder (autoencoder, self-encoder)

[0075] Figure 14 This is a diagram for explaining abnormality sensing using an autoencoder. In this method, abnormality determination is performed based on the characteristics of the relationship between process data from multiple sensors. Specifically, a neural network is used, and the process data itself, such as continuous processing or batch processing as input data, is set as a teaching value to create a model that can compress (encode) and restore (decode) the input data. In a neural network, for example, the number of nodes in the input layer and the output layer corresponds to the number of sensors, and the number of nodes in the intermediate layer is less than the number of sensors. The information input to the input layer is compressed in the intermediate layer and restored in the output layer. It should be noted that there can be multiple intermediate layers, and the connection structure between the layers is not limited to full connection. Then, the process data under normal conditions is used as training data for learning processing, and a model is created in which the parameters are adjusted in a manner that reduces the difference between the value of the input layer and the value of the output layer. In addition, in the abnormality determination process, the process data of the verification object is input, and the abnormality degree corresponding to the difference between the value of the input layer and the value of the output layer is calculated. Specifically, when abnormal process data is input, the information compressed in the intermediate layer cannot be properly restored in the output layer, increasing the difference between the input and output layer values. Therefore, abnormality detection can be performed based on this difference. The autoencoder can detect abnormalities based on the characteristics of the relationship between the output values ​​of multiple sensors.

[0076] Graphic Lasso

[0077] For example, the inter-variable dependency relationship is quantified based on the covariance matrix of process data from multiple sensors in continuous or batch processing, and represented as a sparse graph serving as a benchmark. Under normal circumstances, the inter-variable dependency relationship can be determined to be one that does not significantly deviate from the benchmark. Then, during abnormality determination, the inter-variable dependency relationship is determined using the process data being verified, and the degree of abnormality is calculated based on the magnitude of the deviation from the benchmark. Using the graphical lasso, the correlation between process data can be quantified, and the degree of abnormality can be detected based on any deviation from the relationship.

[0078] In addition, it is also possible to use a general abnormality detection method or a method that applies these. In addition, as for the threshold value used for abnormality detection in each method, it is possible to use the process data actually obtained during the operation of the unit 3 to explore a value that can minimize misjudgment in normal times and quickly detect the occurrence of abnormalities and their signs in abnormal times, and pre-register it in the system. Figure 9 The knowledge base shown.

[0079] <Device Configuration>

[0080] Figure 15 This is a block diagram showing an example of the configuration of an abnormal modulation cause determination device 1. The abnormal modulation cause determination device 1 is a general computer and includes a communication interface (I / F) 11, a storage device 12, an input / output device 13, and a processor 14. The communication I / F 11 may be, for example, a network card or a communication module, and communicates with other computers based on a predetermined protocol. The storage device 12 may also be a primary storage device such as RAM (Random Access Memory) or ROM (Read Only Memory), or an auxiliary storage device (secondary storage device) such as an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory. The primary storage device temporarily stores programs read by the processor 14, information exchanged between other computers, and a work area for the processor 14. The auxiliary storage device stores programs executed by the processor 14, information exchanged between other computers, and the like. The input / output device 13 may be, for example, an input device such as a keyboard or mouse, an output device such as a monitor, or a user interface such as an input / output device such as a touch panel. The processor 14 is a CPU (Central Processing Unit) or other arithmetic processing device, and performs various processes of this embodiment by executing programs. Figure 15 In the example of FIG, functional blocks are shown within the processor 14. That is, the processor 14 functions as a process data acquisition unit 141, a preprocessing unit 142, a learning processing unit 143, an abnormality determination unit 144, a cause diagnosis unit 145, and an output control unit 146 by executing a predetermined program.

[0081] The process data acquisition unit 141 acquires process data from sensors included in the unit 3, for example, via the communication I / F 11 and the control station 2, and stores the acquired data in the storage device 12. As described above, the process data is associated with the sensors via tags.

[0082] The pre-processing unit 142 processes the process data when creating the abnormality sensing model. For example, the pre-processing unit 142 associates the process data with the production number. That is, based on the above-mentioned traceability information pre-stored in the storage device 12, the process data corresponding to the specified label, system and production number in the batch process is associated with the process data corresponding to the specified label in the continuous process and output at the specified time. In addition, based on the set values ​​of the table of the knowledge base, etc., the data of the specified period for abnormality judgment is extracted, and the feature quantities corresponding to each method are calculated. It should be noted that it can also be set that, in the learning process, the pre-processing unit 142 performs data cleaning to exclude outliers such as data during non-steady-state operation, data when abnormalities occur, and noise to extract training data.

[0083] The learning processing unit 143 creates an abnormality detection model including one or more operations based on a knowledge base, for example, and stores it in the storage device 12. In this case, the learning processing unit 143 determines parameters after learning the features of the training data. It should be noted that when learning is performed using the output values ​​of multiple sensors, appropriate normalization may be performed.

[0084] The abnormality determination unit 144 calculates the degree of abnormality using the process data and the abnormality sensing model.

[0085] That is, the abnormality determination unit 144 calculates the abnormality degree using the test data for cross-validation and the abnormality sensing model in the learning process. In addition, the abnormality degree is calculated using the process data acquired from the unit 3 in the abnormality determination process.

[0086] The cause diagnosis unit 145 uses the calculated abnormality degree to calculate the degree of validity (accuracy) for each of the multiple hypothetical causes. The degree of validity is calculated, for example, based on the abnormality degree calculated by the abnormality determination unit and the proportion and degree of influences associated with each hypothetical cause in the knowledge base and appearing in the process data. Alternatively, actions indicating the countermeasures to be taken against the cause may be pre-associated with each hypothetical cause and stored in the storage device 12, thereby prompting the user with the action.

[0087] The output control unit 146, for example, via the input / output device 13, issues an alarm or outputs the validity of each assumed cause when an abnormality is detected. The output control unit 146 appropriately connects the above-mentioned components via the bus 15 according to the user's operation. It should be noted that, for convenience, Figure 15 The illustrated apparatus includes a process data acquisition unit 141 , a preprocessing unit 142 , a learning processing unit 143 , an abnormality determination unit 144 , a cause diagnosis unit 145 , and an output control unit 146 . However, at least some of the functions may be dispersed across different apparatuses.

[0088] <Learning Process>

[0089] Figure 16 is a processing flow chart showing an example of a learning process performed by the abnormal modulation cause determination device 1. The processor 14 of the abnormal modulation cause determination device 1 performs the image by executing a predetermined program. Figure 16 As for the learning process, the process data obtained from the past operation of the unit 3 is used and executed at an arbitrary timing. In addition, the learning process mainly includes pre-processing ( Figure 16 :S1), model building process (S2) and verification process (S3). That is, it can also be set to use part of the process data as training data and the rest as test data for cross-validation. It should be noted that the above-mentioned tables are set as tables created by the user and pre-stored in the storage device 12. For convenience, Figure 16 The illustrated single processing flow describes preprocessing, learning processing, and verification processing. However, at least a portion of the preprocessing, verification processing, and the like may be distributed and executed by different devices.

[0090] The process data acquisition unit 141 of the abnormal preparation cause determination device 1 acquires the process data ( Figure 16 :S11). In this step, the image is extracted Figure 3 、 Figure 6 The process data shown is used for the abnormality detection model. The process data is stored in the storage device 12 in the form of OPC (Object Linking and Embedding for Process Control) data, a database table, or a file in a specified format such as CSV. Furthermore, the process data includes attributes such as date and time, and tags. In particular, for batch-processed process data, it may also include attributes such as product number and step.

[0091] Furthermore, the pre-processing unit 142 of the abnormal preparation cause determination device 1 associates the continuously processed process data with the production number ( Figure 16 :S12). In this step, as Figure 8 As shown in FIG, the process data obtained in the continuous process is associated with the production number group of the process data obtained in the batch process, and the process data used in the calculation of the abnormality is associated. Figure 9 The knowledge base shown, Figure 11 In the logic tree shown, when a certain factor affects both the process data of the batch process and the process data of the continuous process, the abnormality degree and the validity degree are calculated based on the data associated in this step.

[0092] Then, the pre-processing unit 142 extracts and processes the data used in the abnormality determination model ( Figure 16 : S13) In this step, the pre-processing unit 142 extracts data for a predetermined period used for abnormality determination based on the set values ​​of a table such as a knowledge base, and calculates feature quantities corresponding to each method.

[0093] For example, when calculating anomaly levels using the Hotelling method, the preprocessing unit 142 extracts process data at a specified time and period, calculates the instantaneous value of the process data itself, the maximum value, minimum value, cumulative value, or difference of the process data, cumulative value of the reaction speed, differential coefficients at specified time points, and stores the data in the storage device 12. Furthermore, when calculating anomaly levels using the k-approximation method, the time series process data is vectorized or matrixed. Furthermore, when calculating anomaly levels using DTW Barycenter Averaging, multiple process data are synchronized to determine the average time series data. Furthermore, when calculating anomaly levels using an autoencoder or graphical lasso, multiple process data are synchronized.

[0094] It should be noted that the pre-processing unit 142 may also perform a predetermined data cleanup on the process data. Data cleanup is a process for excluding outliers, and various methods can be used. For example, the most recent data can be used to calculate a moving average. In addition, the difference between the moving average and the measured value is taken to obtain the standard deviation σ representing the unevenness of the difference. Then, for example, values ​​that do not fall within a predetermined reliable interval such as the interval from the mean value of the probability distribution -3σ to the mean value of the probability distribution +3σ (also called the 3σ interval) may be excluded. Similarly, for the difference between the measured values ​​before and after, values ​​that do not fall within the 3σ interval may be excluded.

[0095] After that, the learning processing unit 143 of the abnormal modulation cause determination device 1 performs abnormality sensing model construction processing ( Figure 16 :S2). In this step, based on Figure 9 The knowledge base shown in the figure creates an abnormality sensing model including the calculation of abnormality degree. Figure 9For each "hypothetical cause", one or more corresponding "influences" are established, and the abnormality degree obtained by the method registered in the "calculation method" is calculated respectively, and an abnormality detection model represented by a combination of abnormality degrees is created. In addition, the learning processing unit 143 adjusts the parameters of the model using the training data according to the abnormality detection method. For example, when calculating the abnormality degree obtained by the autoencoder, the weight coefficients between layers are adjusted in a manner that can restore the information of the input process data after compression. When calculating the abnormality degree obtained by the graphic lasso, the dependency relationship between variables is digitized based on the covariance matrix of the process data from multiple sensors. Then, the learning processing unit 143 stores the created abnormality detection model in the storage device 12.

[0096] The abnormality determination unit 144 of the abnormal modulation cause determination device 1 calculates the abnormality degree using the created abnormality sensing model and test data ( Figure 16 : S31). In this step, the abnormality determination unit 144 calculates the abnormality according to the abnormality calculation method. For example, when calculating the abnormality obtained by the Hotelling method, the process data is used to estimate the sample mean and sample standard deviation of the population, and the abnormality is calculated based on the distance from the population mean to the process data of the verification object. When calculating the abnormality obtained by the k-nearest neighbor algorithm, the distance between the data is calculated, and the abnormality corresponding to the distance to the data closest to the verification object is calculated. When calculating the abnormality obtained by DTW Barycenter Averaging, the abnormality is calculated based on the cumulative value of the distance between the time series data synchronized in the preprocessing using the k-nearest neighbor algorithm and the Hotelling theory. When calculating the abnormality obtained by the autoencoder, the process data of the verification object is input to the autoencoder, and the abnormality corresponding to the difference between the value of the input layer and the value of the output layer is calculated. When calculating the abnormality obtained by the graphic lasso, the dependency relationship between the variables is calculated using the process data of the verification object, and the abnormality corresponding to the magnitude of the difference from the dependency relationship serving as the benchmark is calculated.

[0097] The cause diagnosis unit 145 of the abnormal modulation cause identification device 1 uses the calculated abnormality degree to obtain the validity degree of the assumed cause ( Figure 16 :S32) In this step, for each knowledge base's assumed cause, the establishment degree is calculated based on the proportion of the corresponding modulation occurrence established as an influence. For example, Figure 9 The reason (2) corresponds to the three effects of the increase in the moisture of tag 002, the increase in the temperature 1 of tag 004, and the decrease in the temperature 2 of tag 005. Figure 16For the abnormality calculated for each influence in S31, the proportion of the three influences whose abnormality exceeds the threshold is set as the validity. Assuming that the abnormality of two of the three influences exceeds the threshold, the validity can be set to 66.7%, for example. Furthermore, in the calculation of the validity, weighting can be applied based on the type of influence (label) or the magnitude of the abnormality. For example, the validity can be calculated by multiplying the weights of each influence and finding the sum.

[0098] Furthermore, the output control unit 146 outputs the abnormality calculated in S31 and the validity calculated in S32 for the user to evaluate the created model. Figure 16 : S33). In this step, cross-validation is performed using test data different from the training data used to build the model, among the process data collected during the past operation of the unit 3. In addition, in this step, process data at the time when an abnormality occurred in the past is also used to appropriately detect the abnormality and verify whether an alarm and an action to respond to it are output. In addition, the learning processing unit 143 determines whether the abnormality is detected with sufficient accuracy ( Figure 16 : S4). If the accuracy is determined to be insufficient (S4: No), the threshold value registered in the knowledge base (in other words, the normal range of the process data) is corrected so that the abnormality can be appropriately detected, and the process from S31 onwards is repeated. If it is determined in S4 that the abnormality can be detected with sufficient accuracy (S4: Yes), the abnormality detection model and threshold value created in S2 are applied. It should be noted that at least part of the determination in S4 can also be performed by the user.

[0099] Note that, regarding the actions, for example, actions that the operator of the unit 3 should perform to deal with the assumed causes are stored in advance in the storage device 12 in association with the assumed causes. Figure 17 This is a diagram showing an example of an action table. Figure 17 The table includes attributes such as cause, action 1, and action 2. The cause column contains the cause corresponding to the assumed cause in the knowledge base. The action 1 and action 2 columns contain information indicating the measures that the operator of unit 3 should take to eliminate the corresponding cause.

[0100] <Abnormality Sensing Process>

[0101] Figure 18 FIG. 1 is a process flow chart showing an example of abnormality sensing processing performed by the abnormal modulation cause determination device 1. The processor 14 of the abnormal modulation cause determination device 1 executes a predetermined program to perform image Figure 18 As for the abnormality sensing process, it is executed in real time using the process data obtained from the operation of the unit 3. The abnormality sensing process mainly includes pre-processing ( Figure 18 :S10)、Model reading process (S20) and abnormality determination process (S30). Figure 18 In, with Figure 16 Corresponding steps in the learning process are designated by the same reference numerals, and the following description focuses on the differences from the learning process. For convenience, the description assumes that the process is implemented by the same device as the device performing the learning process. However, the device performing the abnormality sensing process may be a different device from the device performing the learning process. Furthermore, it is assumed that the tables of abnormality sensing models, thresholds, knowledge bases, and the like created during the learning process are pre-stored in the storage device 12.

[0102] The process data acquisition unit 141 of the abnormal preparation cause determination device 1 acquires the process data ( Figure 18 :S11). The process data is stored in the storage device 12 in the form of OPC data, a table of a so-called database, or a file in a prescribed format such as CSV. Figure 16 The S11 of the apparatus is substantially the same, but data related to the process in operation is obtained in the unit 3. In addition, the pre-processing unit 142 of the abnormal modulation cause determination device 1 associates the continuously processed process data with the production number ( Figure 18 :S12). This step is the same as Figure 16 Then, the pre-processing unit 142 extracts and processes the data used in the abnormality determination model ( Figure 18 :S13). This step is the same as Figure 16 The S13 is much the same, but without the data cleanup.

[0103] After that, the abnormality determination unit 144 of the abnormal modulation cause determination device 1 reads the abnormality sensing model ( Figure 18 :S20). In addition, the abnormality determination unit 144 uses the created abnormality sensing model and the process data obtained from the operation of the unit 3 to calculate the abnormality degree ( Figure 18 : S31). This step is the same as Figure 16 In addition, the cause diagnosis unit 145 of the abnormal modulation cause determination device 1 uses the calculated abnormality degree to obtain the validity degree of the assumed cause ( Figure 18 : S32). This step is the same as Figure 16 The same is true for S32.

[0104] Furthermore, the output control unit 146 outputs the abnormality calculated in S31 and the degree of satisfaction calculated in S32, and issues an alarm when either abnormality exceeds a predetermined threshold value ( Figure 18In this step, process data indicating the operating status of the unit 3 , the degree of abnormality, and the validity of the assumed cause are presented to the user via the input / output device 13 .

[0105] Figure 19 This is a diagram showing an example of a screen output to the input / output device 13 . Figure 19 As an example of the main management chart, the progress of individual process data is shown in a broken line chart. The area 131 displayed on the input / output device 13 shows a combination of identification information of multiple process data obtained from the unit 3 and the latest value. In the management chart of area 132, the progress of the value of the specific process data is shown in a broken line chart. It should be noted that the vertical axis represents the value of the process data and the horizontal axis represents the time axis. In addition, Figure 19 In the example, the solid line represents the true value and the dotted line represents the estimated value. It should be noted that the true value is the process data of the object for calculating the abnormality, and the estimated value can be the estimated value obtained by regression analysis of the process data of the object for calculating the abnormality. The thin dotted line is set to represent the upper and lower limits of the normal range (in other words, the threshold value for abnormality sensing). It should be noted that it can also be set as Figure 19 As indicated by the circle in the figure, when a user operates an input / output device 13, such as a pointing device, to move the pointer on the graph, the numerical value of the process data at the time indicated by the pointer is displayed. The factor-effect graph in area 133 displays the cause of the process data modulation displayed in area 132 or a label that identifies it on the horizontal axis, and the vertical axis shows the validity of the cause using a bar graph. The higher the validity, the more likely it is that the cause of the process data modulation. Furthermore, the factor diagnosis unit 145 calculates the validity of the event based on the abnormality calculated by the abnormality determination unit 144 for the event considered as the presumed cause of the process data modulation. The user can identify the candidate cause of the modulation and its accuracy based on the validity, making it easier to identify the cause of the modulation. Furthermore, the factor-effect graph is a graph that is displayed by the output control unit 146 when the "Diagnosis" button in area 134 is pressed, and the abnormality at a specified time or the current time is calculated by the abnormality determination unit 144.

[0106] Then, when the user operates the input / output device 13 such as a pointing device and selects any of the bar graphs in the cause-effect diagram, the modulation cause corresponding to the bar graph is highlighted in the logic tree.

[0107] Figure 20 This is a diagram showing another example of the screen output by the output control unit 146 to the input / output device 13 . Figure 20 An example of a tree diagram is shown with Figure 10 The logic tree is as shown in Figure 19When the bar graph of tag 004 is selected, the influence of the process data of tag 004 is highlighted on the logic tree. The highlighted display can be achieved by changing the display form, such as changing the color or line type. Figure 20 In the example, the corresponding rectangles are shaded. In addition, the rectangles with thick lines connected to the upstream side of the logic tree represent the assumed causes of the influence. Figure 20 The reasons can be displayed as circled in the figure, or the influence of other factors on process data can be displayed. Figure 18 The degree of validity of each cause calculated in S32 of FIG. 1 may be displayed, or an action may be displayed. In addition, the cause may be displayed when the user moves the pointer on each rectangle.

[0108] It should be noted that, after pressing Figure 19 、 Figure 20 The "Trend" button shown in the figure can display the process trends of the labels listed in the cause-effect diagram, or specifically display the process trends of labels that can be identified as the cause of the modulation. The process trends are calculated using the process data stored in the storage device 12, for example, at intervals of a specified time, a specified number of days, a specified number of months, or each season, and are plotted on a graph.

[0109] Furthermore, the output control unit 146 may output a log of the abnormality level when, for example, the abnormality level calculated by each calculation method exceeds a predetermined threshold. Alternatively, a log of the assumed cause and the degree of validity may be output. By outputting each log in association with the date and time, production number, calculation method, abnormality detection model, etc., analysis of abnormal modulation can be facilitated.

[0110] <Modification>

[0111] The various configurations and combinations thereof in the various embodiments are merely examples, and additions, omissions, substitutions, and other modifications may be made as appropriate without departing from the spirit of the present invention. The present disclosure is not limited by the embodiments but solely by the claims. Furthermore, each of the various aspects disclosed in this specification may be combined with any other feature disclosed in this specification.

[0112] Furthermore, while the above embodiments illustrate a chemical plant as an example, the invention can be applied to manufacturing processes in general production facilities. For example, instead of using the production numbers for the batch processes described in the embodiments, the batch number can be used as the processing unit and the processing according to the batch process described in the embodiments can be applied.

[0113] Alternatively, at least a portion of the functions of the abnormal modulation cause determination device 1 may be implemented in a distributed manner across multiple devices, or multiple devices may provide the same functions in parallel. Furthermore, at least a portion of the functions of the abnormal modulation cause determination device 1 may be implemented in the so-called cloud.

[0114] Furthermore, the present disclosure includes a method for executing the above-mentioned processing, a computer program, and a computer-readable recording medium having the program recorded thereon. The recording medium having the program recorded thereon can execute the above-mentioned processing by causing a computer to execute the program.

[0115] Here, a computer-readable recording medium refers to a recording medium that can store information such as data and programs and then be read from a computer through electrical, magnetic, optical, mechanical, or chemical action. Examples of such recording media that can be removed from a computer include floppy disks, magneto-optical disks, optical disks, magnetic tapes, and memory cards. Furthermore, examples of recording media that are fixed to a computer include HDDs, SSDs (Solid State Drives), and ROMs.

[0116] Description of Reference Numerals

[0117] 1: Abnormal modulation cause determination device;

[0118] 11: Communication I / F;

[0119] 12: storage device;

[0120] 13: Input and output devices;

[0121] 14: Processor;

[0122] 141: process data acquisition department;

[0123] 142: pre-processing unit;

[0124] 143: Learning Processing Department;

[0125] 144: Abnormality determination unit;

[0126] 145: Cause diagnosis department;

[0127] 146: output control unit;

[0128] 2: Control station;

[0129] 3: Crew.

Claims

1. An abnormal modulation cause determination device, comprising: a process data acquisition unit that reads the process data from a storage device that stores process data continuously output by a plurality of sensors provided in the production equipment and associated with management numbers of processing objects; an abnormality determination unit that continuously calculates, for the plurality of sensors, an abnormality degree indicating a degree of modulation of the process data read by the process data acquisition unit; as well as A cause diagnosis unit determines whether the abnormality degree calculated by the abnormality determination unit satisfies a predetermined standard for the process data corresponding to the management number of the processing object and output by the plurality of sensors using causal relationship information, wherein the causal relationship information defines a combination of a cause and modulation of the process data output by the plurality of sensors that occurs as an effect of the cause. The abnormal modulation cause determination device further comprises: a pre-processing unit that calculates average time series data using process data corresponding to processes implemented in the production equipment and having different management numbers, and synchronizes the process data based on similarity with the calculated average time series data; The abnormality determination unit calculates the abnormality degree based on a degree of deviation from a predetermined reference for the synchronized process data corresponding to the steps and having different management numbers.

2. The abnormal modulation cause determination device according to claim 1, wherein: The cause diagnosis unit determines the accuracy of the cause of the modulation based on a ratio of process data whose abnormality degree calculated by the abnormality determination unit satisfies a predetermined reference, among the process data corresponding to the management number of the processing target outputted by the plurality of sensors.

3. The abnormal modulation cause determination device according to claim 1 or 2, wherein: The process data is stored in a storage device in association with steps representing stages of processing in a process performed in the production facility. The abnormal modulation cause determination device further comprises: The output control unit causes the process data corresponding to the steps, which are associated with the different management numbers, to be displayed superimposed on an output device.

4. A method for determining the cause of abnormal modulation, wherein a computer performs the following processing: The process data is read from a storage device storing process data continuously output by a plurality of sensors provided in the production equipment and associated with the management number of the processing object. continuously calculating an abnormality degree indicating a degree of modulation of the read process data for the plurality of sensors, Using the causal relationship information, it is determined whether the calculated abnormality degree satisfies a predetermined standard for the process data corresponding to the management number of the processing object output by the plurality of sensors. The causal relationship information defines a combination of a cause and modulation of the process data output by the plurality of sensors that occurs as an effect of the cause. The abnormal modulation cause determination method further comprises: Preprocessing: using process data corresponding to the process performed in the production equipment and having different management numbers to calculate average time series data, and synchronizing the process data based on similarity with the calculated average time series data, The abnormality determination unit calculates the abnormality degree based on a degree of deviation from a predetermined reference for the synchronized process data corresponding to the steps and having different management numbers.

5. A storage medium of a program for determining the cause of abnormal modulation, causing a computer to execute the following processing: The process data is read from a storage device storing process data continuously output by a plurality of sensors provided in the production equipment and associated with the management number of the processing object. continuously calculating an abnormality degree indicating a degree of modulation of the read process data for the plurality of sensors, Using the causal relationship information, it is determined whether the calculated abnormality degree satisfies a predetermined standard for the process data corresponding to the management number of the processing object output by the plurality of sensors. The causal relationship information defines a combination of a cause and modulation of the process data output by the plurality of sensors that occurs as an effect of the cause. The abnormal modulation cause determination program further executes: Preprocessing: using process data corresponding to the process performed in the production equipment and having different management numbers to calculate average time series data, and synchronizing the process data based on similarity with the calculated average time series data, The abnormality determination unit calculates the abnormality degree based on a degree of deviation from a predetermined reference for the synchronized process data corresponding to the steps and having different management numbers.

Images