Driving support device and driving support method

Abstract

To achieve both suppression of a false notification of an abnormality sign detection and estimation of an unexperienced malfunction mode in the diagnosis of the operation state of an apparatus.SOLUTION: An operation assisting device 200 includes diagnosing means 300, a diagnosis result database 810, a malfunction knowledge database 820, and malfunction mode estimating means 400. The diagnosing means 300 classifies, by clustering technology, measurement signal data of multiple sensors into respective clusters, detects an abnormality sign when the classified cluster differs from a past normal cluster, and calculates an abnormality contribution level for each sensor measurement item of each sensor. The malfunction mode estimating means 400 refers to the malfunction knowledge database 820, extracts a malfunction mode associated with the sensor measurement item that has the abnormality contribution level satisfying a predetermined condition among the abnormality contribution levels for each sensor measurement item calculated when the diagnosing means 300 detects the abnormality sign, and outputs as a malfunction mode estimation result.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a driver assistance device and a driver assistance method, and is particularly suitable for application to a driver assistance device and a driver assistance method that diagnose the operating status of a target device and provide information necessary for responding when an abnormality is predicted to occur. [Background technology]

[0002] Conventionally, techniques for diagnosing anomalies using measurement data from equipment, and techniques for estimating failure modes when equipment malfunctions occur, have been proposed.

[0003] For example, Patent Document 1 discloses a plant anomaly diagnosis system that can determine anomalies in the entire plant at an early stage. The plant anomaly diagnosis system of Patent Document 1 includes an anomaly degree calculation unit that calculates the degree of anomaly of the entire plant to be diagnosed based on the difference in spatial distance between the data belonging to clusters identified by normal data and the above-mentioned multiple measurement data, using Adaptive Resonance Theory (ART) on multiple measurement data from various sensors installed in the plant to be diagnosed.

[0004] Furthermore, Patent Document 2 discloses an anomaly detection and diagnosis method for detecting abnormalities early and with high sensitivity in equipment such as plants. In the anomaly detection and diagnosis method disclosed in Patent Document 2, maintenance history information consisting of past cases such as work history and replacement parts information is correlated on a keyword basis, and anomalies are detected based on anomaly detection targeting the output signals of multidimensional sensors attached to the equipment. By linking the detected anomaly with the maintenance history information associated with it, the diagnosis or action to be taken for the occurring anomaly is clarified.

[0005] Furthermore, Patent Document 3 discloses a maintenance support system that improves the accuracy of identifying equipment failure modes. The maintenance support system of Patent Document 3 includes a failure causality model database which stores a failure causality model created based on a manual for the equipment to be maintained and which describes the causal relationship between the state of the equipment to be maintained and inspection items; a linking database unit which stores a linking database which describes the correspondence between the manual and the failure causality model; a manual update part extraction unit which extracts the updated parts when the manual is updated; and a model update part identification unit which identifies parts that should be updated in the failure causality model by searching the linking database based on the updated parts. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] International Publication No. 2018 / 051568 [Patent Document 2] Japanese Patent Publication No. 2011-227706 [Patent Document 3] Japanese Patent Publication No. 2021-68073 [Overview of the project] [Problems that the invention aims to solve]

[0007] The anomaly diagnosis system described in Patent Document 1 above is capable of detecting anomaly precursors by considering the correlation of data measured by multiple sensors, and can estimate failure modes by associating cluster numbers with experienced failure events. However, the anomaly diagnosis system in Patent Document 1 has the problem that it is difficult to estimate unprecedented failure modes that have not been experienced in the past.

[0008] Furthermore, in the system anomaly detection and diagnosis method described in Patent Document 2, the diagnosis and actions to be taken for an anomaly that occurs are clarified using maintenance history information consisting of past cases. Therefore, it was difficult to clarify the actions to be taken for an unprecedented failure mode that had not been experienced in the past.

[0009] On the other hand, since the maintenance support system of Patent Document 3 estimates a failure mode based on knowledge about the equipment to be maintained, it is possible to estimate an inexperienced failure mode that has not been experienced in the past. However, since the maintenance support system of Patent Document 3 estimates a failure mode using only a single piece of data indicating whether each inspection item has been checked as a judgment criterion, it is difficult to make a judgment (abnormality detection) considering the correlation of a plurality of measurement data, and as a result, there was a possibility of false negatives occurring.

[0010] The present invention has been made in consideration of the above points, and proposes an operation support device and an operation support method capable of achieving both suppression of false negatives in detecting abnormal signs and estimation of inexperienced failure modes in diagnosing the operating state of equipment.

Means for Solving the Problems

[0011] In order to solve such problems, in the present invention, based on operation data of a plant obtained from measurement signal data of the equipment measured by a plurality of sensors, diagnostic means for diagnosing the operating state of the plant, a diagnostic result database for storing the diagnostic results obtained by the diagnostic means, a sensor measurement item of each of the sensors, and a failure knowledge database in which a failure mode that is assumed to occur when the measured value deviates from the normal state is associated, and failure mode estimation means for estimating the failure mode that is assumed to occur when the diagnostic means detects an abnormal sign. Image display information generation means for generating image display information for displaying the failure mode estimation results output from the failure mode estimation means; interface for acquiring evaluation results regarding whether the failure mode estimation results displayed by the image display information generation means are correct or incorrect from an external source; and database update means for updating information stored in at least one of the diagnostic result database or the failure knowledge database based on the evaluation results. The diagnostic means classifies the measurement signal data of the plurality of sensors into clusters by a clustering technique, and when the classified cluster is different from the normal cluster classified during normal times in the past, detects an abnormal sign, and calculates an abnormality contribution degree, which is the degree to which the measured values of the plurality of sensors deviate from the normal state, for each sensor measurement item. The failure mode estimation means refers to the failure knowledge database and extracts a failure mode associated with a sensor measurement item having an abnormality contribution degree that satisfies a predetermined condition among the abnormality contribution degrees for each sensor measurement item calculated when the diagnostic means detects an abnormal sign, and outputs it as a failure mode estimation result. Furthermore, if the evaluation result indicates that the failure mode estimated in the failure mode estimation result was correct, the database update means updates the diagnostic result database by associating the cluster number indicating the relationship of the operating data with the failure mode; if the evaluation result indicates that the failure mode estimated in the failure mode estimation result was incorrect, the database update means registers the correct failure mode information in the failure knowledge database and then updates the diagnostic result database by associating the cluster number with the failure mode.There is provided a driving support device characterized by the following.

[0012] Also, in order to solve such problems, in the present invention, based on operation data of a plant from measurement signal data obtained by measuring a device with a plurality of sensors, diagnostic means for diagnosing an operating state of the plant, a diagnostic result database for storing a diagnostic result by the diagnostic means, sensor measurement items of each of the sensors, and a failure knowledge database in which a failure mode that is assumed to occur when a measured value thereof deviates from a normal state is associated, and failure mode estimation means for estimating the failure mode that is assumed to occur when the diagnostic means detects an abnormal sign, Image display information generation means for generating image display information for displaying the failure mode estimation results output from the failure mode estimation means; interface for acquiring evaluation results regarding whether the failure mode estimation results displayed by the image display information generation means are correct or incorrect from an external source; and database update means for updating information stored in at least one of the diagnostic result database or the failure knowledge database based on the evaluation results. are provided. The diagnostic means classifies measurement signal data of a plurality of sensors into clusters by a clustering technique, and when the classified clusters are different from normal clusters classified during normal times in the past, detects an abnormal sign, and calculates an abnormality contribution degree, which is a degree to which measured values of the plurality of sensors deviate from a normal state, for each sensor measurement item of each of the sensors. The failure mode estimation means refers to the failure knowledge database and extracts a failure mode associated with a sensor measurement item having an abnormality contribution degree that satisfies a predetermined condition among the abnormality contribution degrees for each sensor measurement item calculated when the diagnostic means detects an abnormal sign, and outputs it as a failure mode estimation result Furthermore, if the evaluation result indicates that the failure mode estimated in the failure mode estimation result was correct, the database update means updates the diagnostic result database by associating the cluster number indicating the relationship of the operating data with the failure mode, and if the evaluation result indicates that the failure mode estimated in the failure mode estimation result was incorrect, it registers the correct failure mode information in the failure knowledge database and then updates the diagnostic result database by associating the cluster number with the failure mode. There is provided a driving support method characterized by the following.

Effect of the Invention

[0013] According to the present invention, in diagnosing an operating state of a device, it is possible to achieve both suppression of false negatives in detecting an abnormal sign and estimation of an unexperienced failure mode.

Brief Description of the Drawings

[0014] [Figure 1] It is a block diagram showing a configuration example of a driving support device 200 according to an embodiment of the present invention. [Figure 2] It is a diagram showing a configuration example of a compression refrigerator 111 which is an example of a device 110. [Figure 3] It is a diagram showing an example of measurement signal data. [Figure 4] This figure shows an example of diagnostic result data. [Figure 5] This figure shows an example of cluster data. [Figure 6] This figure shows an example of failure mode data. [Figure 7] This figure shows an example of chain failure mode data. [Figure 8] This figure shows an example of a graphical representation of failure knowledge data. [Figure 9] This flowchart shows an example of the processing procedure for anomaly prediction diagnosis. [Figure 10] This is a block diagram showing an example configuration of the clustering unit 310 using Adaptive Resonance Theory (ART). [Figure 11] This figure shows an example of the classification results obtained by classifying measurement data into clusters. [Figure 12] This diagram illustrates the relationship between measured driving data and the results of classifying them into clusters. [Figure 13] This diagram explains how the degree of abnormality is calculated. [Figure 14] This diagram explains how to calculate the anomaly contribution. [Figure 15] This figure shows an example of the relationship between inspection items (sensor measurement items) and failure modes. [Figure 16] This figure shows an example of the relationship between the anomaly contribution and the failure mode estimation result. [Figure 17] This figure shows an example of a diagnostic results screen. [Figure 18] This figure shows an example of how abnormal signs are detected in this embodiment. [Figure 19] This flowchart shows an example of the processing steps for a database update. [Figure 20] This figure shows an example of an evaluation result input screen. [Figure 21] This figure shows an example of failure mode data. [Figure 22] This is a diagram illustrating the adjustment of resolution. [Figure 23]This figure shows an example of failure mode data. [Modes for carrying out the invention]

[0015] Embodiments of the present invention will be described in detail below with reference to the drawings.

[0016] The following descriptions and drawings are illustrative examples for explaining the present invention, and have been omitted and simplified as appropriate for clarity of explanation. Furthermore, not all combinations of features described in the embodiments are necessarily essential to the solution of the invention. The present invention is not limited to the embodiments, and any application that aligns with the spirit of the invention falls within its technical scope. Those skilled in the art can make various additions and modifications within the scope of the present invention. The present invention can also be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

[0017] In the following explanation, various types of information may be described using terms such as "table," "list," and "queue," but these types of information may also be represented using other data structures. To indicate independence from data structure, "XX table," "XX list," etc., may be referred to as "XX information." When describing the content of each type of information, terms such as "identification information," "identifier," "name," "ID," and "number" will be used, but these terms are interchangeable.

[0018] Furthermore, in the following explanations, when describing similar elements without distinction, a reference code or a common number in the reference code will be used. When describing similar elements with distinction, the reference code of that element will be used, or the ID assigned to that element will be used instead of the reference code.

[0019] Furthermore, while the following description may include explanations of processes performed by executing a program, the processor may be the primary entity performing the processing, as a program is executed by at least one processor (e.g., a CPU) and performs defined processes using appropriate memory resources (e.g., memory) and / or interface devices (e.g., communication ports). Similarly, the primary entity performing the processing by executing a program may be a controller, device, system, computer, node, storage system, storage device, server, management computer, client, or host having a processor. The primary entity performing the processing by executing a program (e.g., a processor) may include hardware circuits that perform some or all of the processing. For example, the primary entity performing the processing by executing a program may include hardware circuits that perform encryption and decryption, or compression and decompression. The processor operates as a functional unit that realizes predetermined functions by operating according to the program. Devices and systems including a processor are devices and systems including these functional units.

[0020] A program may be installed from its program source into a device such as a computer. The program source may be, for example, a program distribution server or a non-temporary storage medium readable by a computer. If the program source is a program distribution server, the program distribution server includes a processor (e.g., a CPU) and non-temporary storage resources, which may further store the distribution program and the program to be distributed. The processor of the program distribution server may then execute the distribution program, thereby distributing the program to other computers. Furthermore, in the following description, two or more programs may be implemented as a single program, or one program may be implemented as two or more programs.

[0021] (1) Composition Figure 1 is a block diagram showing an example configuration of an operation support device 200 according to one embodiment of the present invention. As shown in Figure 1, the operation support device 200 is connected to the plant 100, which is the target of operation support, and to an external device 900.

[0022] Plant 100 is comprised of equipment 110 and a control device 120. Equipment 110 is the equipment whose operating condition is to be diagnosed. The control device 120 is a device that controls equipment 110 by transmitting control signals, and receives measurement signals (measurement data) such as temperature, flow rate, or pressure measured by multiple sensors (not shown) installed on equipment 110. A specific example of equipment 110 is a compression chiller (see Figure 2).

[0023] (1-1) Configuration of the driver assistance system 200 The driver assistance system 200 is a computer that performs various functions by executing arithmetic processing, input / output processing, etc., by having a processor such as a CPU read a program stored in a memory device into the main memory (memory) and execute it.

[0024] The driver assistance system 200 includes a diagnostic means 300, a failure mode estimation means 400, an image display information generation means 500, and a database update means 600 as a computing device, and the operation of these computing means is controlled by the driver assistance system operation control means 700.

[0025] The driver assistance system 200 also includes a database (DB) consisting of a measurement signal database 800, a diagnostic result database 810, and a fault knowledge database 820. As will be described in detail later, the measurement signal database 800 stores the measurement signals of the plant 100 (measurement signal data 801 shown in Figure 3). The diagnostic result database 810 stores information indicating the results of operating the diagnostic means 300 (diagnosis result data 811 and cluster data 812 shown in Figures 4 and 5). The fault knowledge database 820 stores information that associates sensor measurement items with failure modes that are expected when their measurement values ​​deviate from the normal state (failure mode data 821 and chain failure mode data 822 shown in Figures 6 and 7). Note that in Figure 1, database is abbreviated as DB. The database stores digitized information, and the information is usually stored in a form called an electronic file (electronic data).

[0026] Furthermore, the driver assistance device 200 is equipped with an external input interface 210 and an external output interface 220 as interfaces with the outside world. The driver assistance device 200 receives measurement signals collected at the plant 100 and external input signals generated by the external input device 910 (keyboard 920 and mouse 930) of the external device 900 in response to user operations via the external input interface 210. The driver assistance device 200 stores the measurement signals received from the plant 100 in the measurement signal database 800. The driver assistance device 200 also outputs image display information generated by the image display information generation means 500 to the image display device 940 via the external output interface 220.

[0027] In the driver assistance device 200, the diagnostic means 300 includes a clustering unit 310, an abnormality determination unit 320, and an abnormality contribution calculation unit 330, and the database update means 600 includes a diagnostic result update unit 610 and a fault knowledge update unit 620.

[0028] (1-1-1) Configuration of diagnostic means 300 The diagnostic means 300 diagnoses the operating state of the plant 100 using the measurement signals stored in the measurement signal database 800 and the diagnostic result database signals stored in the diagnostic result database 810. The diagnostic means 300 performs calculations to diagnose the operating state of the plant 100 and stores the resulting diagnostic results in the diagnostic result database 810.

[0029] More specifically, the diagnostic means 300 diagnoses the operating state of the plant 100 by, for example, the method disclosed in Patent Document 1.

[0030] In the diagnostic means 300, first, the clustering unit 310 classifies the normal operating data (composed of multiple measurement data) into multiple clusters (normal clusters). Next, the clustering unit 310 classifies the current operating data into clusters. If this time-series data cannot be classified into a normal cluster, the clustering unit 310 generates a new cluster (new cluster). The generation of a new cluster means that the state of the subject being diagnosed has changed to a new state (new state).

[0031] Therefore, the anomaly detection unit 320 determines the occurrence of an anomaly from the occurrence of a new cluster. If the anomaly detection unit 320 detects the occurrence of an anomaly, the anomaly contribution calculation unit 330 calculates the anomaly contribution, which is the degree to which the measured values ​​of the multiple sensors deviate from the normal state.

[0032] Furthermore, the diagnostic means 300 may be configured to accept not only the measurement signal data, but also data that has been arbitrarily processed from the measurement signal data, such as the rate of change of the measurement signal data. By inputting the rate of change of the measurement signal data into the diagnostic means 300, it becomes possible to detect abnormalities while taking into account the rate of change of the data.

[0033] The diagnostic means 300 then outputs the diagnostic results of the operating status of the plant 100 to the diagnostic result database 810 and the failure mode estimation means 400. These diagnostic results also include the aforementioned abnormality contribution.

[0034] (1-1-2) Configuration of failure mode estimation means 400 The failure mode estimation means 400 estimates the failure mode by referring to the failure knowledge database 820 based on the abnormality contribution calculated when the diagnostic means 300 detects an abnormality precursor. Specifically, in estimating the failure mode, the failure mode estimation means 400 extracts failure modes associated with sensor measurement items whose calculated abnormality contribution exceeds a predetermined threshold from among the failure modes stored in the failure knowledge database 820 in association with sensor measurement items, or it selects sensor measurement items in descending order of their calculated abnormality contribution and extracts the failure modes associated with the selected sensor measurement items. The failure mode estimation means 400 then outputs the failure modes extracted by the above method as the failure mode estimation result. In addition, if there are failure modes associated with cluster numbers assigned when the diagnostic means 300 classifies the data, the failure mode estimation means 400 also extracts those failure modes.

[0035] The failure mode estimation means 400 stores the failure mode estimation result obtained from the above calculation in the failure knowledge database 820, together with the diagnosis result from the diagnosis means 300. The diagnosis result and the failure mode estimation result are also output to the image display information generation means 500 for display output.

[0036] (1-1-3) Configuration of the image display information generation means 500 The image display information generation means 500 generates image display information using the diagnostic results and failure mode estimation results, and transmits it to the image display device 940 of the external device 900 via the external output interface 220. The image display information includes at least one of the following: information for displaying the failure mode estimation results, information for displaying the relationship between sensor measurement items and associated failure modes in a tree structure, or information for displaying the relationship between sensor measurement items and associated failure modes in a directed graph structure. In addition, failure modes associated with cluster numbers may be highlighted. The highlighted display can be any method that notifies the operator that the failure mode is one that deserves particular attention, such as coloring the failure mode.

[0037] Furthermore, the information stored in the measurement signal database 800, the diagnostic result database 810, and the fault knowledge database 820 may also be visualized by any means and displayed on the image display device 940.

[0038] In the external device 900, when image display information including the failure mode estimation results is displayed on the image display device 940, the operator confirms the display content and evaluates the display content, then creates the result by operating the external input device 910 (keyboard 920, mouse 930). The evaluation result of the failure mode estimation results thus created is taken into the driving support device 200 via the external input interface 210 as an external input signal output from the external device 900. Note that the evaluation result of the failure mode estimation results is not limited to being created by input by the operator, but may also be created by obtaining information from other sources, such as reports showing the results of countermeasures against malfunctions.

[0039] (1-1-4) Configuration of the database update means 600 The database update means 600 updates the information stored in the diagnostic result database 810 and the failure knowledge database 820 based on the evaluation results of the failure mode estimation results acquired from the external device 900 via the external input interface 210.

[0040] The database update means 600, in the evaluation result of the failure mode estimation result, associates the cluster number indicating the relationship of the operating data with the failure mode if the estimated failure mode is correct, and updates the failure knowledge database 820 to register the correct failure mode information in the failure knowledge database and associate the cluster number with the failure mode if the estimated failure mode is incorrect.

[0041] Furthermore, if the failure mode associated with a cluster number is incorrect in the evaluation results of the failure mode estimation, the database update means 600 adjusts the resolution used by the diagnostic means 300 to classify the data so that there is only one failure mode corresponding to each cluster number, and updates the diagnostic result database 810.

[0042] In this embodiment, the driving support device 200 is equipped with a computing device and a database inside the driving support device 200. However, at least a portion of these may be located outside the driving support device 200, and configured to communicate only data between the devices.

[0043] Furthermore, all the signals stored in each database can be displayed on the image display device 940 via the external output interface 220, and this information can be modified based on external input signals generated by operating the external input device 910.

[0044] Furthermore, in this embodiment, the external input device 910 is configured with a keyboard 920 and a mouse 930, but the devices that make up the external input device 910 are not limited to these, and any device for inputting data may be used, such as a microphone for voice input or a touch panel for manual input.

[0045] Furthermore, in this embodiment, the application of the operation support system 200 to the plant 100 is used as an example, but it goes without saying that the operation support system 200 can also be applied to facilities other than plants.

[0046] (1-2) Configuration of equipment 110 Figure 2 shows an example configuration of a compression chiller 111, which is an example of equipment 110. In this embodiment, the compression chiller 111 shown in Figure 2 is used as an example for explanation, but the equipment 110 of the plant 100 that is the target of operation support by the operation support device 200 is not limited to this.

[0047] As shown in Figure 2, the compression chiller 111 mainly consists of a compressor 112, a condenser 113, an expansion valve 114, and an evaporator 115. In the compression chiller 111, the refrigerant is repeatedly compressed, condensed, expanded, and evaporated to produce chilled water for the air conditioner 116 to supply cold air. This series of cycles is called the refrigeration cycle.

[0048] The compressor 112 compresses the low-temperature, low-pressure gaseous refrigerant, transforming it into a high-temperature, high-pressure gas. The condenser 113 condenses the high-temperature, high-pressure refrigerant sent from the compressor 112, exchanging heat with the cooling water. As the heat from the refrigerant is transferred to the cooling water, the refrigerant changes from a high-temperature, high-pressure gas to a medium-temperature, high-pressure liquid. This cooling water is sent to the cooling tower 118 by the cooling water pump 119, and the heat is dissipated from the cooling tower 118.

[0049] The expansion valve 114 reduces the pressure of the medium-temperature, high-pressure liquid refrigerant sent from the condenser 113. As the pressure is reduced, the refrigerant expands, its temperature decreases, and it changes into a low-temperature, low-pressure liquid. The evaporator 115 evaporates the low-temperature, low-pressure liquid sent from the expansion valve 114, exchanging heat with the chilled water. The heat from the chilled water is transferred to the refrigerant, and the cooled chilled water is sent to the air conditioner 116 by the chilled water pump 117, making the air blown from the air conditioner 116 cool air. Meanwhile, the refrigerant that has absorbed the heat from the chilled water changes into a low-temperature, low-pressure gas. The low-temperature, low-pressure gas is sent back to the compressor 112, and the refrigeration cycle is repeated.

[0050] As described above, in the compression chiller 111, the refrigerant circulates through multiple components, and if a malfunction occurs in one component and the refrigerant does not change to the desired state, the effect will cascade throughout the entire refrigeration cycle. Therefore, the operation support device 200 according to this embodiment diagnoses the compression chiller 111 (component 110) and supports its operation.

[0051] (1-3) Database The nature of the data stored in each database of the driver assistance system 200 will be described.

[0052] Figure 3 shows an example of measurement signal data. The measurement signal data 801 shown in Figure 3 is data stored in the measurement signal database 800, which stores multiple measurement data (operation data) measured by sensors installed on the equipment 110 for each sampling period.

[0053] The measurement signal data 801 has data items for time 8011 and measurement item 8012. Time 8011 indicates the measurement time of the operating data to be recorded in the record. Measurement item 8012 is a code that indicates the measurement item of multiple measurement data that constitute the operating data. The codes X1, X2, X3, etc. shown in Figure 3 specifically mean measurement items such as compressor temperature and compressor pressure.

[0054] Figure 4 shows an example of diagnostic result data. The diagnostic result data 811 shown in Figure 4 is data stored in the diagnostic result database 810 and represents the diagnostic result of abnormality prediction obtained when the diagnostic means 300 is in operation.

[0055] The diagnostic result data 811 contains data items for time 8111, cluster number 8112, anomaly degree 8113, and anomaly contribution degree 8114. Time 8111 indicates the measurement time of the measurement signal that was the subject of the diagnostic result for the record. Cluster number 8112 indicates the identifier (cluster number) of the cluster to which the measurement signal was classified. Detailed information of the cluster identified from cluster number 8112 is shown in the cluster data 812 of Figure 5, which will be described later, in the record for the corresponding cluster number 8131. Anomaly degree 8113 indicates the anomaly degree of the measurement signal. Anomaly contribution degree 8114 indicates the anomaly contribution of the measurement signal. Anomaly degree and anomaly contribution degree will be described later with reference to Figures 13 and 14.

[0056] Figure 5 shows an example of cluster data. The cluster data 812 shown in Figure 5 is data stored in the diagnostic results database 810 and holds attributes for each cluster. The configuration information held by the cluster data 812 is set in advance by an administrator, etc., but may also be updated when the database update process described later is executed.

[0057] Cluster data 812 contains data items for cluster number 8121, attribute 8122, resolution 8123, and weight coefficient 8124. Cluster number 8121 indicates an identifier (cluster number) assigned individually to each cluster. Attribute 8122 indicates whether the cluster is normal or abnormal. Resolution 8123 indicates the size of the cluster, and specifically corresponds to the parameter ρ (vigilance parameter) described later in Figure 10. Weight coefficient 8124 indicates coefficient information representing the center coordinates of the cluster.

[0058] Figure 6 shows an example of failure mode data. The failure mode data 821 shown in Figure 6 is data stored in the failure knowledge database 820, and it holds information for predicting the failure mode for each component.

[0059] The failure mode data 821 has data items for part 8211, function 8212, failure mode 8213, functional failure 8214, impact 8215, and inspection item 8216. Part 8211 indicates the part in question, and function 8212 indicates the function of the part in question. Failure mode 8213 indicates the name of a failure mode that may occur in the part in question, and functional failure 8214 indicates the nature of the functional failure caused by the failure mode shown in failure mode 8213. Impact 8215 indicates the impact on the part caused by the failure mode shown in failure mode 8213. Inspection item 8216 corresponds to a sensor measurement item and is associated with a failure mode 8213 that is expected if the measured value deviates from the normal state.

[0060] To illustrate a specific example of the failure mode estimation method using Figure 6, if the compressor pressure deviates from the normal state, the relevant record is narrowed down from the combination of part 8211 and inspection item 8216, and from the failure mode 8213 in the relevant record, the following failure modes can be assumed: "insufficient liquid refrigerant," "abnormal pressure due to compression valve failure," or "refrigerant inflow due to expansion valve failure."

[0061] Figure 7 shows an example of chain failure mode data. The chain failure mode data 822 shown in Figure 7 is data stored in the failure knowledge database 820, and it holds information about failure events (failure modes to be chained) that may occur in a chain reaction due to the occurrence of a certain failure event (the failure mode of the chain source), using the combination of the component of the chain source and its effect as the key.

[0062] The chain failure mode data 822 includes data items for the source component 8221, the source influence 8222, the cause 8223, the target component 8224, and the target failure mode 8225. The source component 8221 and source influence 8222 indicate the component (corresponding to component 8211) and the influence (corresponding to influence 8215) at the source. The cause 8223 indicates the cause of the failure event that occurred at the source. The target component 8224 and target failure mode 8225 indicate the component and failure mode for the chain failure caused by the failure event that occurred at the source.

[0063] To give a specific example, if a refrigerant leak occurs in the low-pressure piping and the amount of refrigerant decreases, this will lead to a decrease in the amount of refrigerant at the compressor inlet, which in turn will result in a shortage of liquid refrigerant in the compressor. In this way, information is stored that a failure mode occurring in the low-pressure piping can chain into a failure mode in the compressor. The first row of records in Figure 7 shows information about this chain of failures.

[0064] Figure 8 shows an example of a graph representation of failure knowledge data. Here, failure knowledge data refers to the data stored in the failure knowledge database 820, i.e., the failure mode data 821 shown in Figure 6 and the chain failure mode data 822 shown in Figure 7. The image display information generation means 500 can also display the failure knowledge data stored in the failure knowledge database 820 on the image display device 940 using a directed graph representation as shown in Figure 8.

[0065] The operation of the driver assistance device 200, which has the configuration described above, will be explained below.

[0066] The operation of the driver assistance device 200 according to this embodiment is broadly divided into two parts: an abnormality prediction diagnosis process that diagnoses the driving state to be diagnosed and, if an abnormality is detected, estimates the expected failure mode and outputs the failure mode estimation result; and a database update process that updates the information stored in the database as feedback based on the driver's (user's) evaluation of the failure mode estimation result. The execution timing of the abnormality prediction diagnosis process and the database update process does not need to be synchronized; it is sufficient that the database update process is executed at least once after the abnormality prediction diagnosis process has been executed.

[0067] Both the abnormality prediction diagnosis process and the database update process are realized by the driver assistance system operation control means 700 operating each arithmetic unit. However, for the sake of simplicity, the description of the control by the driver assistance system operation control means 700 may be omitted in the following detailed explanation.

[0068] (2) Abnormality prediction diagnostic processing Figure 9 is a flowchart showing an example of the processing procedure for abnormality prediction diagnosis. Of the steps shown in Figure 9, step S101 is a calculation step for setting data under normal conditions and only needs to be executed at least once after the equipment 110 in plant 100 starts operating under normal conditions. Steps S102 to S108 are calculation steps for diagnosis and are repeatedly executed at predetermined sample cycles until the operation of equipment 110 is stopped. The sample cycle is, for example, the cycle in which the measurement signal measured by equipment 110 in plant 100 is received from the control device 120.

[0069] As shown in Figure 9, first, the diagnostic means 300 classifies past measurement signals stored in the measurement signal database 800 using the clustering unit 310 and generates clusters for normal operation (normal clusters) (step S101).

[0070] Next, the diagnostic means 300 acquires measurement signals from the plant 100 in real time and stores the acquired measurement signals in the measurement signal database 800 (step S102).

[0071] Next, the diagnostic means 300 classifies the measurement signals acquired in step S102 into clusters using the clustering unit 310 (step S103).

[0072] Next, the diagnostic means 300 operates the abnormality determination unit 320 to diagnose an abnormality indication from the classification result of the measurement signal in step S103 and confirm whether or not an abnormality indication has been detected (step S104). The abnormality determination unit 320 diagnoses "abnormal (abnormality indication exists)" if the cluster classified in step S103 is different from the cluster in the normal state, and diagnoses "normal (no abnormality indication)" if it matches the cluster in the normal state. In step S104, if the diagnosis result by the abnormality determination unit 320 is "normal" (NO in step S104), the system returns to step S102, and if the diagnosis result is "abnormal" (YES in step S104), the system proceeds to step S105.

[0073] In step S105, the diagnostic means 300 operates the abnormality contribution calculation unit 330 to calculate an abnormality contribution, which represents the degree to which the values ​​of the measurement signals from the multiple sensors acquired in step S102 deviate from the normal state. The diagnostic means 300 then outputs a diagnostic result that includes the calculated abnormality contribution.

[0074] Next, the failure mode estimation means 400 estimates the failure modes that are expected to occur based on the abnormality contribution calculated in step S105 (step S106). The failure mode estimation means 400 then outputs the failure mode estimation result, which shows the result of the estimation.

[0075] Next, the image display information generation means 500 generates image display information using the diagnostic results output in step S105 and the failure mode estimation results output in step S106, and outputs it to the image display device 940 of the external device 900 via the external output interface 220 (step S107).

[0076] Finally, the driver assistance device operation control means 700 performs a determination to determine the end of the abnormality prediction diagnosis process (step S108). If the driver instructs the driver assistance device 200 to stop (YES in step S108), the driver assistance device operation control means 700 terminates the abnormality prediction diagnosis process. On the other hand, if the driver assistance device 200 is not instructed to stop (NO in step S108), the process returns to step S102 and continues.

[0077] (2-1) Operation of diagnostic means 300 The operation of the diagnostic means 300 (steps S101 to S105) in the abnormality prediction diagnostic process shown in Figure 9 will be explained in detail below.

[0078] First, we will explain the clustering unit 310, which classifies the operation data (measurement data) recorded in the measurement signal data 801 and generates clusters.

[0079] FIG. 10 is a block diagram showing a configuration example of the clustering unit 310 using the Adaptive Resonance Theory (ART). The ART module 340 shown in FIG. 10 is an embodiment of the clustering unit 310 using ART.

[0080] Data Nx i normalized to the range from 0 to 1 based on the normalization range set for the operation data and the raw material information data, and the complement CNx i of the normalized data (i.e., "1 - Nx i (n)") are included in the input data I i and input as (n).

[0081] The ART module 340 includes an F0 layer 341, an F1 layer 342, an F2 layer 343, a memory 344, and a selection subsystem 345, which are interconnected. The F1 layer 342 and the F2 layer 343 are connected via weight coefficients. The weight coefficients represent the prototypes of the clusters into which the input data is classified. Here, the prototype represents the representative value of the cluster.

[0082] The F0 layer 341 normalizes the input vector and removes noise, and creates u i 0 to be input to the F1 layer 342 and the selection subsystem 345.

[0083] The F1 layer 342 holds u i 0 output from the F0 layer 341 as short-term memory, and calculates P i to be input to the F2 layer 343.

[0084] The F2 layer 343 calculates the fitness using the weight coefficients defined for each cluster and P i and selects the cluster that best matches P i .

[0085] The selection subsystem 345 receives the output from the F0 layer 341. i 0 The degree of agreement between the selected cluster and the F2 layer 343 is calculated, and if the degree of agreement is greater than the parameter ρ, the selected cluster is adopted as the output cluster.

[0086] Next, we will explain the algorithm of the ART module 340. The general outline of the algorithm when input data is input to the ART module 340 is as follows: Processes 1 to 5.

[0087] (Process 1) The F0 layer 341 normalizes the input vector and removes noise.

[0088] (Process 2) The input data and weight coefficients entered into the F1 layer 342 are compared to select a suitable cluster candidate.

[0089] (Process 3) The validity of the cluster selected in the selection subsystem 345 is evaluated by its ratio to the parameter ρ. If it is deemed valid, the input data is classified into that cluster and the process proceeds to step 4. On the other hand, if it is not deemed valid, the cluster is reset, and a suitable candidate cluster is selected from other clusters (step 2 is repeated). Increasing the value of parameter ρ results in finer cluster classification (smaller cluster size). Conversely, decreasing the value of parameter ρ results in coarser classification (larger cluster size). This parameter ρ is called the vigilance parameter (resolution parameter).

[0090] (Process 4) If all existing clusters selected in process 2 are reset in process 3, the input data is determined to belong to a new cluster. In this case, new weight coefficients are generated that represent the prototype of the new cluster.

[0091] (Process 5) When the input data is classified into cluster J, the weight coefficient WJ(new) corresponding to cluster J is updated by the following equation (1) using the past weight coefficient WJ(old) and the input data p (or data derived from the input data).

Equation

[0092] The feature of the data classification algorithm of the ART module 340 lies in the above-mentioned process 4. In process 4, when input data different from the pattern at the time of learning is input, a new pattern can be recorded without changing the recorded pattern. Therefore, it is possible to record a new pattern while recording the patterns learned in the past.

[0093] Thus, when the previously given operation data is given as input data, the ART module 340 learns the given pattern. Therefore, when new input data is input to the learned ART module 340, it is possible to determine which past pattern it is close to by the above algorithm. Also, if it is a pattern that has not been experienced in the past, it is classified into a new cluster.

[0094] The enlarged view shown at the lower left in FIG. 10 is a block diagram showing a configuration example of the F0 layer 341. The F0 layer 341 i [[ID=2i]]re-normalizes the input data I i 0 at each time and creates a normalized input vector u

[0095] to be input to the F1 layer 342 and the selection subsystem 345. i 0The calculation method will be explained in detail. First, layer F0 341 receives input data I i Therefore, according to equation 2 below, w i 0 Calculate the following, where a is a constant.

number

[0096] Next, layer 341 of F0 is w i 0 normalized x i 0 This is calculated using the following equation 3. Here, "||w 0 ||」 is w 0 It represents the norm.

number

[0097] Next, layer F0 341 uses the following equation 4 to calculate x i 0 V with noise removed i 0 We calculate the following. Here, θ is a constant for removing noise. As a result of the calculation in Equation 4, very small values ​​become 0, so the noise in the input data is removed.

number

[0098] Finally, the F0 layer 341 uses the following equation 5 to normalize the input vector u i 0 The u calculated in this way i 0 However, this becomes the input to layer 342 of F1.

number

[0099] The enlarged view shown in the lower right of Figure 10 is a block diagram showing an example of the configuration of F1 layer 342. F1 layer 342 is obtained using equation 5 u i 0 The data is held in short-term memory and input to the F2 layer 343. i Calculate.

[0100] Below, F1 layer 342 is P i The formulas necessary to calculate are shown in Equations 6 to 12. Note that a and b are constants, and f() is the function shown in Equation 4, and T j This is the goodness of fit calculated using F2 layer 343.

number

number

number

number

number

number

number

[0101] Next, we will explain the determination of anomaly signs by the anomaly determination unit 320 based on the data classification results by the clustering unit 310 (step S104 in Figure 9).

[0102] Figure 11 shows an example of the classification results obtained by classifying measurement data into clusters. As an example, Figure 11 displays two items of the measurement data from the operation data (X1 and X2, shown as measurement item 8012 in Figure 3), and is represented as a two-dimensional graph. The vertical and horizontal axes of Figure 11 also show the normalized operation data for each item.

[0103] The operating data is divided into multiple clusters 350, indicated by circles in Figure 11, by the processing of the ART module 340 described above. Each circle represents one cluster.

[0104] In Figure 11, the operational data is classified into four clusters of 350, numbered 1 to 4. Cluster 1 is the group with large X1 values ​​and small X2 values. Cluster 2 is the group with small values ​​for both X1 and X2. Cluster 3 is the group with small X1 values ​​and large X2 values. Cluster 4 is the group with large values ​​for both X1 and X2.

[0105] Figure 12 illustrates the relationship between the measured operational data and the results of its classification into clusters. In Figure 12, the horizontal axis represents time, and the vertical axis represents the strength of the measured signal or the cluster number. The upper part of Figure 12 shows the time-series changes of the measured signals for each item X1 and X2, while the lower part of Figure 12 shows the cluster numbers 1 to 4 to which the operational data containing X1 and X2 is classified. In Figure 12, the section labeled "Learning" refers to the learning period of the ART340, and the section labeled "Diagnostic" refers to the period during which the operational plant 100 (equipment 110) was diagnosed.

[0106] As shown in Figure 12, during the learning period, which is a normal state, the operating data was classified into one of cluster numbers 1 to 3. However, during the diagnostic period, it was classified into cluster number 4, which was not experienced during normal operation. Since cluster number 4 is a "new state," in step S104 of the abnormality prediction diagnostic process shown in Figure 9, the abnormality determination unit 320 determines that an abnormality has occurred.

[0107] Next, the operation of the anomaly contribution calculation unit 330 will be explained.

[0108] Figure 13 illustrates the method for calculating the anomaly score. To make the explanation easier to understand, Figure 13 shows the case where clusters are created in a two-dimensional space using two signals (signal X1, signal X2) as inputs, similar to Figures 11 and 12.

[0109] Figure 13 assumes that normal cluster 1 and normal cluster 2, representing normal conditions, are created by the process of step S101 of the abnormality prediction diagnostic process shown in Figure 9.

[0110] Next, let's assume that a measurement signal is acquired in step S102 of Figure 9, and a new cluster is generated as a result of operating the clustering unit 310 in step S103. As shown in Figure 13, the newly acquired measurement signal 351 does not fall within the range of either of the existing normal clusters 1 and 2, so a new cluster containing the measurement signal 351 is generated.

[0111] In this case, an abnormality indicator is detected in step S104, and in step S105, the abnormality contribution calculation unit 330 calculates the abnormality contribution and the degree of abnormality.

[0112] The "abnormality score" is defined as the distance from the centroid of the nearest cluster. As can be seen in Figure 13, normal cluster 1 is closer to the measurement signal 351 mapped in two-dimensional space than normal cluster 2. Therefore, the abnormality contribution calculation unit 330 calculates the distance to the measurement signal 351 with respect to the centroid of normal cluster 1 as the abnormality score for the entire plant 100 that is being diagnosed.

[0113] Furthermore, the anomaly contribution calculation unit 330 calculates the anomaly contribution of each signal (signal X1, signal X2) based on the positional relationship between the measurement signal 351 and normal clusters 1 and 2 in a two-dimensional space.

[0114] Figure 14 illustrates the method for calculating the anomaly contribution. Figure 14 shows an enlarged view of the centroid X of normal cluster 1, which is the nearest cluster in Figure 13, and the plotted points of the measurement signal 351. In this case, the distance difference in each axial direction, i.e., the axial direction of signal X1 and the axial direction of signal X2, represents the difference between each signal (signal X1, signal X2) and the normal state.

[0115] Therefore, in the case of Figure 14, the anomaly contribution calculation unit 330 calculates the anomaly contribution Rn for each signal (signal X1, signal X2) using the following equation 13, based on the distance difference ΔX1 between the centroid X of the nearest normal cluster 1 and the axial component of the signal X1 of the measurement signal 351, and the distance difference ΔX2 between the centroid X of the nearest normal cluster 1 and the axial component of the signal X2 of the measurement signal 351. Here, ΣΔX i This represents the sum of the values ​​obtained by adding ΔX1 to ΔXn.

number

[0116] In the above-described explanation of the operation of the anomaly contribution calculation unit 330, the normal state referenced when calculating the anomaly degree and anomaly contribution was the centroid of the normal cluster closest in distance to the measured signal data. However, for example, the centroid of the normal cluster with the closest time from among the normal clusters that occurred in the past may also be used as the referenced normal state.

[0117] To explain this in detail using Figures 11 and 12 as examples, in Figure 11, cluster numbers 1 to 3 are normal clusters, and cluster number 4 is a new (abnormal) cluster. In this case, as shown in Figure 12, the normal cluster immediately before cluster number 4 occurred is cluster number 2, so the abnormality contribution calculation unit 330 can use this cluster number 2 as the cluster to refer to when calculating the abnormality contribution.

[0118] (2-2) Operation of the failure mode estimation means 400 The operation by the failure mode estimation means 400 (step S106) of the abnormality prediction diagnostic process shown in Figure 9 will be explained in detail.

[0119] The failure mode estimation means 400 refers to the failure knowledge database 820 (failure mode data 821) and extracts failure modes (failure modes 8213) stored in association with sensor measurement items (inspection items 8216) from among the failure modes associated with sensor measurement items (failure modes 8213), specifically those associated with sensor measurement items whose abnormality contribution, calculated when the diagnostic means 300 detects an abnormality precursor, exceeds a predetermined threshold, or those associated with sensor measurement items selected in descending order of abnormality contribution, and outputs the extracted failure modes as failure mode estimation results.

[0120] Figure 15 shows an example of the relationship between inspection items (sensor measurement items) and failure modes. As mentioned above, inspection items and failure modes are stored in association with each other in the failure mode data 821 of the failure knowledge database 820 (see Figure 6).

[0121] Figure 15 shows the relationship between component 8211, failure mode 8213, and inspection item 8216, extracted from the record where component 8211 is "compressor" in the failure mode data 821 of Figure 6, and represented graphically. According to the graph in Figure 15, the inspection item "temperature measurement" is associated with one type of failure mode, "compressor overheating," and the inspection item "pressure measurement" is associated with three types of failure modes: "insufficient liquid refrigerant," "abnormal pressure due to compression valve failure," and "refrigerant inflow due to expansion valve failure."

[0122] Referring to the graph in Figure 15, we will now explain the case in which the failure mode estimation means 400 extracts failure modes associated with sensor measurement items (inspection items) whose abnormality contribution exceeds a predetermined threshold, and outputs them as failure mode estimation results.

[0123] If only the abnormal contribution to the compressor's temperature exceeds a predetermined threshold, one failure mode associated with "temperature measurement" is output as the failure mode estimation result. If only the abnormal contribution to the compressor's pressure exceeds a predetermined threshold, three failure modes associated with "pressure measurement" are output as the failure mode estimation result. If the abnormal contributions to both temperature and pressure exceed predetermined thresholds, four failure modes associated with "temperature measurement" and "pressure measurement" are output as the failure mode estimation result.

[0124] Furthermore, as described above, the failure mode estimation means 400 may select sensor measurement items in order of their degree of abnormality contribution, extract the failure modes associated with the selected sensor measurement items, and output the extracted failure modes as failure mode estimation results. When this method is adopted, the number of sensor measurement items (inspection items) to be selected in order of their degree of abnormality contribution is set by the system or the user.

[0125] Figure 16 shows an example of the relationship between the degree of anomaly contribution and the failure mode estimation results. Figure 16 shows that, in the relationship between inspection items and failure modes shown in Figure 15, the extracted failure modes differ depending on whether one sensor measurement item (inspection item) is selected in order of the degree of anomaly contribution or two items are selected.

[0126] Specifically, when the order of highest anomaly contribution is "pressure measurement" followed by "temperature measurement," selecting one item results in the output of three failure modes associated with "pressure measurement" as the failure mode estimation result. On the other hand, selecting two items results in the output of four failure modes associated with "pressure measurement" and "temperature measurement" as the failure mode estimation result.

[0127] In the above explanation, it was stated that only failure modes directly associated with the inspection item are output as failure mode estimation results. However, if the failure knowledge database 820 stores not only the failure mode data 821 exemplified in Figure 6 but also the chain failure mode data 822 exemplified in Figure 7, the failure mode estimation means 400 may consider that failures are chained and output all related chain failure modes up to the related chain failure mode 8225 in the failure mode estimation results.

[0128] (2-3) Result display Figure 9 describes the abnormality prediction diagnostic process, specifically the operation by the image display information generation means 500 (step S107), and the diagnostic result screen displayed on the image display device 940 as a result of this operation.

[0129] Figure 17 shows an example of a diagnostic results screen. The diagnostic results screen 941 shown in Figure 17 is an example of a screen displayed on the image display device 940 of the external device 900 using image display information generated by the image display information generation means 500, and can be considered as one embodiment of image display information.

[0130] In the abnormality prediction diagnosis process, the image display information generated by the image display information generation means 500 in step S107 in response to the operation of the failure mode estimation means 400 includes at least one of the following: information for displaying the failure mode estimation result; information for displaying the relationship between sensor measurement items (inspection items) and associated failure modes in a tree structure; and information for displaying the relationship between sensor measurement items and associated failure modes in a directed graph structure.

[0131] In the diagnostic results screen 941 shown in Figure 17, areas 942, 943, and 944 are arranged as areas for displaying the diagnostic results (failure mode estimation results). Area 942 displays the estimated failure mode, area 943 displays the relationship between sensor measurement items and associated failure modes in a tree structure, and area 944 displays the relationship between sensor measurement items and associated failure modes in a directed graph structure.

[0132] When such a diagnostic results screen 941 is displayed on the image display device 940, the plant operator can be notified not only of the estimated failure mode but also of the events that formed the basis of the estimation.

[0133] By performing the abnormality prediction diagnostic process in the manner described above, the operation support device 200 according to this embodiment can diagnose the operating state of the plant 100 (equipment 110) and, upon detecting an abnormality, quickly estimate and output the failure modes that are expected to occur, including previously unexplored failure modes. The effects of such abnormality prediction diagnostic processing will be explained in more detail below.

[0134] Conventionally, plant diagnostic systems utilizing clustering technologies such as ART were capable of detecting anomaly precursors by considering the correlation of data measured by multiple sensors, but it was difficult to estimate or identify previously unexplored failure modes.

[0135] On the other hand, while failure knowledge databases can estimate anticipated failure modes through methods such as Failure Mode and Effects Analysis (FMEA) and Fault Tree Analysis (FTA), they detect anomalies based on the presence or absence of inspection items, making it difficult to detect anomalies while considering data correlation, which could lead to false positives.

[0136] Thus, while conventional technology had the potential to produce false alarms, the driver assistance device 200 according to this embodiment can suppress false alarms. The reason for this will be explained using Figure 18.

[0137] Figure 18 shows an example of how abnormality prediction is detected in this embodiment. In Figure 18, it is assumed that normal data is processed by the diagnostic means 300, resulting in the generation of three normal clusters 360 to 362, and that three measurement signals 352 to 354 are then input as diagnostic data.

[0138] If the diagnostic means 300 attempts to detect an anomaly based solely on the information in the fault knowledge database 820 for the input measurement signals 352 to 354, the timing of detection will be when the value of the inspection item exceeds a threshold (threshold judgment). For example, in Figure 18, measurement signals 352 and 353 are higher than the threshold, so they can be detected by threshold judgment. On the other hand, measurement signal 354 is lower than the threshold, so an anomaly cannot be detected by threshold judgment, resulting in a false alarm.

[0139] However, in the driver assistance device 200 according to this embodiment, the diagnostic means 300 detects abnormal signs based on whether or not the measurement signal can be classified into a normal cluster (i.e., whether or not the measurement signal is within the range of a normal cluster), thus enabling more accurate abnormality detection than threshold determination. Specifically, the measurement signal 354 shown in Figure 18 is outside the range of normal clusters 360 to 362, so a new cluster is generated and an abnormal sign can be detected.

[0140] In other words, the driver assistance device 200 according to this embodiment utilizes both clustering technology (ART) and a fault knowledge database to achieve both suppression of missed abnormality prediction detections and estimation of previously unexplored failure modes in diagnosing the operating state of the equipment 110.

[0141] (3) Database update process Figure 19 is a flowchart illustrating an example of the processing procedure for a database update. The database update process shown in Figure 19 is mainly performed by the database update means 600. As a prerequisite for the database update process to be executed, when the failure mode estimation result from the abnormality prediction diagnosis process is displayed on the image display device 940 (diagnosis result screen 941 in Figure 17), the operator or other personnel evaluate it. In this evaluation, the operator inputs whether the estimated failure mode was correct or incorrect, and if incorrect, what the correct failure mode was. These evaluation results (evaluation results of the failure mode estimation result) are then sent from the external device 900 to the driver assistance device 200.

[0142] According to Figure 19, first, the database update means 600 acquires an external input signal that includes the evaluation result of the failure mode estimation result described above (step S201).

[0143] Next, the database update means 600 refers to the evaluation result obtained in step S201 to confirm whether the estimated failure mode was correct or not (step S202). If the estimated failure mode was correct (YES in step S202), the process proceeds to step S203; otherwise, it proceeds to step S205.

[0144] In step S203, the database update means 600 checks whether the cluster number and the failure mode are associated. If they are associated (YES in step S203), there is no need to update the database, and the database update process is terminated. If they are not associated (NO in step S203), the process proceeds to step S204.

[0145] In step S204, the database update means 600 operates the failure knowledge update unit 620 to update the failure knowledge database 820 (failure mode data 821 or chain failure mode data 822) to associate cluster numbers with failure modes, and then terminates the database update process.

[0146] On the other hand, in step S205, the database update means 600 checks whether information regarding the correct failure mode is stored in the failure knowledge database 820. If the correct failure mode exists in the failure knowledge database 820 (YES in step S205), the process proceeds to step S206. If the correct failure mode does not exist in the failure knowledge database 820 (NO in step S205), the process proceeds to step S207.

[0147] In step S206, the database update means 600 operates the diagnostic result update unit 610 to adjust the resolution used by the diagnostic means 300 when classifying data, so that there is only one failure mode corresponding to each cluster number, and updates the cluster data 812 (particularly resolution 8123) of the diagnostic result database 810. After that, the process proceeds to step S204, where the same processing as described above is performed, and the database update process is terminated.

[0148] Furthermore, in step S207, the database update means 600 operates the failure knowledge update unit 620 to add information about the correct failure mode to the failure knowledge database 820 (failure mode data 821 or chain failure mode data 822), and updates the failure knowledge database 820 to associate the cluster number with the failure mode. After that, the database update process is terminated.

[0149] The following sections will explain in more detail the processing details in steps S204, S206, and S207.

[0150] In the database update process, the database update means 600 updates the information stored in the diagnostic results database 810 or the failure knowledge database 820 based on an external input signal that includes the evaluation result of the failure mode estimation result. The evaluation result of the failure mode estimation result is entered, for example, manually by the operator of the plant 100 or obtained from a report of the results of the malfunction countermeasures.

[0151] Figure 20 shows an example of an evaluation result input screen. The evaluation result input screen 951 shown in Figure 20 is an example of a screen provided for the operator of plant 100 to manually input the evaluation results of the failure mode estimation results.

[0152] Based on the information entered into the evaluation result input screen 951, the driver assistance device 200 (database update means 600) obtains information on whether the failure mode estimation result was correct or incorrect, and, if incorrect, the name of the correct failure mode. In addition, the report on the results of the malfunction countermeasures is a report that describes the details of the malfunction and the countermeasures taken, and it is also possible to obtain information on the malfunction mode that actually occurred from such a report.

[0153] (3-1) Processing of step S204 Figure 21 shows an example of failure mode data. The failure mode data 824 shown in Figure 21 is data stored in the failure knowledge database 820, and its data structure is the same as the failure mode data 821 shown in Figure 6, so a detailed explanation is omitted.

[0154] As mentioned above, if the estimated failure mode is correct and this failure mode is not associated with a cluster number, the failure knowledge update unit 620 associates the cluster number, which indicates the relationship with the operating data, with the failure mode (step S204 in Figure 19).

[0155] Specifically, Figure 21 shows an example in failure mode data 824 where cluster number 4 is associated with the failure mode "refrigerant inflow due to expansion valve failure" in data 825.

[0156] If the estimated failure mode is incorrect, in step S204, the failure knowledge update unit 620 registers the correct failure mode information in the failure knowledge database 820 and then similarly associates the cluster number with the failure mode.

[0157] By associating the cluster number with the failure mode in the database update process in this way, the driver assistance device 200 can estimate the failure mode for the next time the same cluster occurs, based on past performance.

[0158] On the other hand, if the failure mode estimation result is incorrect, possible reasons include a failure to properly associate the cluster number with the failure mode, or the knowledge of the correct failure mode not being stored in the failure knowledge database 820. Step S206 in Figure 19 is a process to solve the former problem, and step S207 in Figure 19 is a process to solve the latter problem.

[0159] (3-2) Processing of step S206 As mentioned above, if the failure mode associated with a cluster number is incorrect and the correct failure mode is stored in the failure knowledge database 820, the diagnostic result update unit 610 adjusts the resolution used by the diagnostic means 300 to classify the data so that there is only one failure mode corresponding to one cluster number (step S206 in Figure 19). Here, resolution refers to the parameter ρ (vigilance parameter) described in Figure 10, which is a parameter that determines the size of each cluster.

[0160] Figure 22 illustrates the adjustment of resolution. Figure 22(A) shows the resolution before adjustment, and Figure 22(B) shows the resolution after adjustment.

[0161] In Figure 22(A) before adjusting the resolution, cluster 370 is a cluster associated with failure mode A that occurred in the past. Here, the measured signal was classified as cluster 370 (i.e., estimated to be failure mode A), but according to the evaluation of the failure mode estimation results, the correct failure mode was failure mode B. In such a case, it is thought that the failure mode could not be correctly estimated because the association between the cluster number and the failure mode was not properly established.

[0162] Therefore, the diagnostic result update unit 610 adjusts the resolution of the clusters so as to generate clusters 371 and 372 in a manner that separates failure mode A and failure mode B, as shown in Figure 22(B). After this adjustment, in step S204 in Figure 19, the failure knowledge update unit 620 associates cluster 371 with failure mode A and cluster 372 with failure mode B.

[0163] (3-3) Processing of step S207 Figure 23 shows an example of failure mode data. The failure mode data 826 shown in Figure 23 is data stored in the failure knowledge database 820, and its data structure is the same as the failure mode data 821 shown in Figure 6, so a detailed explanation is omitted.

[0164] As mentioned above, if the failure mode associated with the cluster number is incorrect and the correct failure mode is not stored in the failure knowledge database 820, the failure knowledge update unit 620 adds the information of the correct failure mode to the failure knowledge database 820 and updates the failure knowledge database 820 to associate the cluster number with the failure mode (step S207 in Figure 19).

[0165] Specifically, Figure 23 shows an example where the failure mode data 826 has the data for the correct failure mode (827) added to the failure mode, and the cluster number data (828) added to the inspection items.

[0166] As described above, the driver assistance device 200 according to this embodiment can update two databases, the diagnostic result database 810 and the failure knowledge database 820, by performing a database update process using the database update means 600, and by feeding back the evaluation results of the failure mode estimation results. This allows for more efficient data updates with simpler processing than updating each database independently. [Explanation of symbols]

[0167] 100 plants 110 Equipment 111 Compression Refrigeration Unit 112 Compressor 113 Condenser 114 Expansion valve 115 Evaporator 116 Air conditioner 117 Chilled water pump 118 Cooling Tower 119 Cooling water pump 120 Control device 200 Driving assistance devices 210 External Input Interface 220 External Output Interface 300 diagnostic methods 310 Clustering section 320 Abnormality judgment section 330 Anomaly Contribution Calculation Unit 340 ART Module 400 Failure Mode Estimation Method 500 Image display information generation means 600 Database update methods 610 Diagnostic Result Update Section 620 Fault knowledge update section 700 Operation control means for driving assistance devices 800 Measurement Signal Database 801 Measurement signal data 810 Diagnostic Result Database 811 Diagnostic Result Data 812 cluster data 820 Failure Knowledge Database 821,824,826 Failure Mode Data 822 Chain failure mode data 900 External device 910 External Input Device 920 Keyboard 930 Mouse 940 Image Display Device 941 Diagnosis Results Screen 951 Evaluation Result Input Screen

Claims

1. A diagnostic means for diagnosing the operating state of the plant based on plant operating data obtained from measurement signal data measured by multiple sensors on the equipment, A diagnostic results database for storing the diagnostic results obtained by the aforementioned diagnostic means, A failure knowledge database is provided, which associates the sensor measurement items of each of the aforementioned sensors with failure modes that are expected to occur when the measured values ​​deviate from the normal state. When the diagnostic means detects an abnormality, a failure mode estimation means estimates the failure mode that is expected to occur, Image display information generation means for generating image display information for displaying the failure mode estimation results output from the failure mode estimation means, An interface for acquiring external evaluation results regarding whether the failure mode estimation result displayed by the image display information generation means is correct or incorrect, A database update means that updates the information stored in at least one of the diagnostic results database or the fault knowledge database based on the evaluation results, Equipped with, The diagnostic means is Clustering technology is used to classify the measurement signal data from multiple sensors into clusters. If the classified cluster differs from the normal cluster classified during past normal conditions, an anomaly is detected, and the degree of anomaly contribution, which is the degree to which the measured values ​​of the multiple sensors deviate from the normal state, is calculated for each sensor measurement item. The failure mode estimation means is By referring to the aforementioned failure knowledge database, the system extracts failure modes associated with sensor measurement items that have an abnormality contribution that satisfies predetermined conditions from among the abnormality contributions for each sensor measurement item calculated when an abnormality precursor is detected by the diagnostic means, and outputs them as failure mode estimation results. The aforementioned database update means is If the evaluation result indicates that the failure mode estimated in the failure mode estimation result was correct, the diagnostic result database is updated by associating the cluster number indicating the relationship of the operating data with the failure mode. If the evaluation results indicate that the failure mode estimated in the failure mode estimation results was incorrect, the correct failure mode information is registered in the failure knowledge database, and the cluster number and failure mode are updated in the diagnostic result database. A driving assistance device characterized by the following features.

2. The failure mode estimation means is By referring to the aforementioned failure knowledge database, the system extracts failure modes associated with sensor measurement items whose abnormality contribution, calculated when an abnormality precursor is detected by the diagnostic means, exceeds a predetermined threshold, or failure modes associated with a predetermined number of sensor measurement items selected in descending order of abnormality contribution, and outputs them as failure mode estimation results. The driving support device according to feature 1.

3. The input data to the diagnostic means consists of the measurement signal data and the rate of change of the measurement signal data. The driving support device according to feature 1.

4. The normal state referenced when calculating the anomaly contribution is the centroid of the nearest normal cluster that is close to the measurement signal data in which the anomaly precursor was detected, or the centroid of the most recent past normal cluster from the measurement signal data in which the anomaly precursor was detected. The driving support device according to feature 1.

5. The image display information includes at least one of the following: information for displaying failure mode estimation results, information for displaying the relationship between sensor measurement items and associated failure modes in a tree structure, or information for displaying the relationship between sensor measurement items and associated failure modes in a directed graph structure. The driving support device according to feature 1.

6. The aforementioned evaluation results are obtained from the results manually entered by the plant operator or from reports of the results of troubleshooting measures. The driving support device according to feature 1.

7. The database update means adjusts the resolution at which the diagnostic means classifies the data so that, if the evaluation result indicates that the failure mode associated with the cluster number was incorrect, there is only one failure mode corresponding to each cluster number. The driving support device according to feature 1.

8. A diagnostic means for diagnosing the operating state of the plant based on plant operating data obtained from measurement signal data measured by multiple sensors on the equipment, A diagnostic results database for storing the diagnostic results obtained by the aforementioned diagnostic means, A failure knowledge database is provided, which associates the sensor measurement items of each of the aforementioned sensors with failure modes that are expected to occur when the measured values ​​deviate from the normal state. When the diagnostic means detects an abnormality, a failure mode estimation means estimates the failure mode that is expected to occur, Image display information generation means for generating image display information for displaying the failure mode estimation results output from the failure mode estimation means, An interface for acquiring external evaluation results regarding whether the failure mode estimation result displayed by the image display information generation means is correct or incorrect, A database update means that updates the information stored in at least one of the diagnostic results database or the fault knowledge database based on the evaluation results, Equipped with, The diagnostic means, Clustering technology classifies measurement signal data from multiple sensors into clusters. If the classified cluster differs from the normal cluster classified during past normal conditions, an anomaly is detected, and the degree of anomaly contribution, which is the degree to which the measured values ​​of the multiple sensors deviate from the normal state, is calculated for each sensor measurement item. The failure mode estimation means, By referring to the aforementioned failure knowledge database, the system extracts failure modes associated with sensor measurement items that have an abnormality contribution that satisfies predetermined conditions from among the abnormality contributions for each sensor measurement item calculated when an abnormality precursor is detected by the diagnostic means, and outputs them as failure mode estimation results. The aforementioned database update means If the evaluation result indicates that the failure mode estimated in the failure mode estimation result was correct, the diagnostic result database is updated by associating the cluster number indicating the relationship of the operating data with the failure mode. If the evaluation result indicates that the failure mode estimated in the failure mode estimation result was incorrect, the correct failure mode information is registered in the failure knowledge database, and the cluster number and failure mode are linked and updated in the diagnostic result database. A driving assistance method characterized by the following features.

Images