JP2005165375A - Diagnostic device and method for facility - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a diagnostic device and method for narrowing down an abnormal part even when any abnormal event other than a preliminarily estimated abnormal event occurs. <P>SOLUTION: This diagnostic device is provided with a database (2) in which the operation data of each of measurement items set for each of equipment configuring a facility (1) are recorded, a display means (7) for displaying a measurement item relevant figure (5) representing a causal relation between the operation data for each of the measurement items, a designating means (4) for designating measurement items set for each of equipment based on the measurement item relevant figure (5), a retrieving means (6) for retrieving a similar event similar to the operation data in the measurement item set by the inputting means by referring to the database based on the measurement items set by the designating means (4), and the display means (7) for displaying the retrieval results. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a facility diagnosis apparatus and diagnosis method, and more particularly, to a facility diagnosis apparatus and diagnosis method including a plurality of devices.

For example, in Patent Document 1, a time series pattern of each event is stored in a database, and a pattern having a high similarity to a time series pattern measured online is searched from a pattern stored in the database to diagnose a current state. A diagnostic device is shown. Further, in this apparatus, measurement items to be noted for each event are stored in the database. Therefore, since the similarity of the pattern can be obtained only with the item of interest, the similar pattern can be searched with high accuracy without being affected by the value of the measurement item unrelated to the state determination. For this reason, it is possible to efficiently specify a similar pattern for an assumed event.
JP 2002-99319 A

  However, in the technique shown in the above document, when a plant is operated in various operation patterns, it is necessary to store an abnormal event pattern in a database for each operation pattern. Further, when an unexpected abnormal event occurs, a similar pattern cannot be searched and it is difficult to specify an abnormal part.

  The present invention has been made in view of these problems, and provides a diagnostic device and a diagnostic method capable of narrowing down an abnormal part even when an abnormal event other than an abnormal event assumed in advance occurs.

  In order to solve the above problems, the present invention employs the following means.

  A database that records operation data for each measurement item set for each device constituting the equipment, a display means for displaying a measurement item relation diagram representing a causal relationship between operation data for each measurement item, and the measurement item Measurement item designation means for designating measurement items to be set for each device on the basis of the related figure, and measurement set by the designation means with reference to the database based on the measurement items designated by the designation means Search means for searching for similar cases similar to the operation data in the item, and display means for displaying the search results are provided.

  Since the present invention has the above-described configuration, it is possible to provide a diagnostic device and a diagnostic method that can narrow down an abnormal portion even when an abnormal event other than an abnormal event assumed in advance occurs.

  Hereinafter, the best embodiment will be described with reference to the accompanying drawings. FIG. 1 is a diagram for explaining an equipment abnormality diagnosis apparatus according to a first embodiment of the present invention. In the figure, reference numeral 1 denotes equipment to be diagnosed by the abnormality diagnosis apparatus 10. Reference numeral 2 denotes a database that acquires and stores operation data of the facility 1 in time series. A data classification device 3 classifies the obtained operation data into a plurality of categories including the degree of abnormality. Reference numeral 4 denotes a measurement item designation means for designating measurement items necessary for specifying an abnormal part (abnormal device). 5 is a measurement item relation diagram showing the causal relationship between operation data for each measurement item, and 6 is a database 2 based on the measurement items set in advance, and similar cases similar to the operation data in the measurement items. Search means for searching, 7 is a display means for displaying the search results.

  The facility 1 is a diagnosis target and is composed of a plurality of devices. A plurality of devices are provided with measuring devices, respectively, and the measured operation data is sent to the database 2 and the data classification device 3.

  In the database 2, the operation data sent from the facility 1 is recorded. The data sampling interval is 1 minute. In addition to the operation data, the database 2 also records the result of the operation data classified by the data classification device 3.

  The data classification device 3 is a data classification device using adaptive resonance theory. This device determines whether or not newly input data has the same tendency as before by learning normal and abnormal operation data in advance and learning causal relationships between variables. can do.

  The measurement item relation FIG. 2 is a system diagram shown in FIG. 2 to be described later, for example, and represents the relation between the devices (A to E) and the measurement items (x1 to x7) constituting the facility. This system diagram is displayed on the display means 7. Thereby, the operator can know the causal relationship between the measurement items. Further, by referring to the measurement item relation diagram 2, it is possible to more reliably specify the measurement items necessary for specifying the abnormal part of the facility.

  FIG. 2 is a system diagram of the equipment. As shown in FIG. 2, the facility is composed of five devices A to E. In addition, a measuring device is installed at the input / output unit of each device, and seven state quantities x1 to x7 are measured. When the target facility is viewed as one system, the inputs are x1 and x2, and the output is x7. The output of the device D is the input of the device C and constitutes a feedback system. The database 2 records seven measurement values x1 to x7, which are operation data of the facility 1.

  FIG. 3 is a diagram for explaining details of the data classification device, and FIG. 4 is a diagram for explaining processing (algorithm) of the data preprocessing device 31 shown in FIG. The data classification device 3 is a data classification device using adaptive resonance theory (ART), and includes a data preprocessing device 31 and an ART module 32 as shown in FIG.

  The data preprocessing device 31 converts the operation data into the input data of the ART module 32 according to the algorithm shown in FIG. The steps will be described below.

  First, in step 311, the maximum value and the minimum value are calculated for each measurement item. In step 311, the calculated maximum value and minimum value are recorded in the database 2. In step 312, the data is normalized using the maximum value and the minimum value obtained in step 311.

  The normalization method will be described by taking xi as an example. The number of data in xiDATA1 is N, and the nth measurement value is xi (n). Further, when the maximum value and the minimum value in N pieces of data are Max_i and Min_i, normalized data Nxi (n) is calculated by the following equation.

Nxi (n) = α + (1−α) × (xi (n) −Min_i) / (Max_i−Min_i) (Expression 1)
Here, it is a constant of α (0 ≦ α <0.5), and the data is normalized to the range of [α, 1−α] by Equation 1. In this embodiment, α = 0.2.

  In step 313, the complement of the data normalized in step 312 is calculated and added to the input data. The complement CNxi (n) of the normalized data Nxi (n) is calculated by the following equation.

CNxi (n) = 1-Nxi (n) (Formula 2)
Next, in step 314, 14 items of data consisting of the data Nxi (n) and CNxi (n) created in step 312 and step 313 are input to the ART module 32 as input data.

  Next, the details of the ART module 32 will be described with reference to FIG. The ART module 32 classifies input data into a plurality of categories as described above. The ART module 32 includes an F0 layer, an F1 layer, an F2 layer, and an Orienting Subsystem, which are coupled to each other. The F1 layer and the F2 layer are connected through a weighting factor, and the weighting factor represents a prototype (prototype) of a category into which input data is classified.

  Next, the algorithm of the ART module will be described. The outline of the algorithm when input data is input to the ART module is as shown in the following steps 1 to 5.

  Step 1: The F0 layer is normalized so that the size of the input vector becomes 1, and noise is removed.

Step 2: A suitable category candidate is selected by comparing the input data input to the F1 layer with a weighting factor.

Step 3: The validity of the category selected in the Orienting Subsystem is evaluated by comparison with the parameter ρ. If it is determined to be valid, the input data is classified into the category, and the process proceeds to step 4. If it is not judged to be valid, the category is reset, and an appropriate category candidate is selected from the other categories (the process of step 1 is repeated). Increasing the value of parameter ρ makes the category classification finer, and decreasing ρ makes the classification coarser.

Step 4: When all the existing categories are reset in Step 2, it is determined as a new category, and a new weighting factor representing a prototype of the new category is generated.

Step 5: When the input data is classified into the category J, the weighting factor WJ ^(new) corresponding to the category J is the next from the past weighting factor WJ ^(old) and the input data p (or data derived from the input data). Updated with expression.

WJ ^(new) = Kw * p + (1-Kw) * WJ ^(old) Formula (3)
Here, Kw is a learning rate parameter and determines the degree to which the input vector is reflected in the new weighting factor.

  The feature of the data classification algorithm of the ART module is the processing in step 4. When input data different from the stored pattern is input by the process of step 3, the stored pattern is not changed and a new pattern can be stored. For this reason, it is possible to store a new pattern while storing patterns learned in the past.

  As described above, when the operation data given in advance is given as input data, the ART module 32 learns the given pattern. Therefore, when new input data is input to the learned ART module 32, it is possible to determine which pattern is near by the above algorithm. If the pattern has never been experienced before, it is classified into a new category.

  In the present embodiment, the normal data is learned in advance, and the degree of whether the data included when the new category is generated is normal operation data or abnormal data is determined.

  FIG. 5 is a diagram for explaining a procedure for designating measurement items with reference to the measurement item relation diagram (FIG. 2). As shown in FIG. 2, the measurement item relation diagram schematically shows the configuration of the device and the name of the measurement data, and the measurement data is written at the measurement position. For this reason, it becomes possible to intuitively understand the relationship between measurement items by referring to the measurement item relation FIG. In this embodiment, measurement data inside the device is not included. However, when there are a plurality of measurement items in one device, a plurality of devices are used so that these relationships can be intuitively understood. It is better to use a diagram that divides the modules into the modules and shows the relationship between the measurement items.

  The measurement item designating means 4 selects a measurement item necessary for specifying an abnormal part of the facility using the measurement item relation diagram 2. In the case of this example, for example, the measurement item relation diagram (FIG. 2) is displayed on the display means 7, and the measurement item is selected on the screen of the display device 7 by operating a pointing device such as a mouse. FIG. 5 shows an example of the selection screen. On this selection screen, there are a measurement item related diagram 2, an upstream selection button 51, a decision button 52 and a cancel button 53. When selecting a measurement item, the pointer 54 on the screen is moved by operating the pointing device, and necessary data is clicked or an area is selected with the pointer 54. The selected measurement item changes its item name color to indicate that it has been selected. In the example of FIG. 5, items from x5 to x7 are selected. When the determination button 51 is pressed, the selected data is determined as an index for similarity calculation. If there is an error in the selected data, pressing the cancel button 53 resets the selected item. In addition, when the upstream selection button 51 is pressed, an item located upstream of the item selected by the pointing device operation is selected. When the upstream selection button 51 is pressed in the state of FIG. 5, x1 to x4 are selected.

  In the present embodiment, by selecting a component device, a variable corresponding to the output can be selected. For example, if device A is specified, x3 can be selected, and if device C is specified, x5 can be selected. When the devices C, D, and E are selected, items x5 to x7 are selected.

  Moreover, the similar case search means 6 pays attention to the item specified by the measurement item specifying unit 5 and searches for similar data from the database. In the present embodiment, the dissimilarity (distance) D (n) is calculated by the following equation, and a case where the distance D (n) is smaller than the threshold D_std is selected as a similar case.

D (n) = Σai (Nxi (n) −Nxi ^* ) Formula (4)
However, Nxi ^* : reference data ai = 1 (when item xi is selected), 0 (when item xi is not selected) In this embodiment, the distance is Euclidean as shown by the equation (4). Calculated by distance. This is because in the present embodiment, each measurement item is normalized and used. When such normalization processing is not performed, a weighted distance may be calculated with ai = wi (when item xi is selected) in Expression (4).

  In this embodiment, the distance is calculated based on the data Nxi (n) input to the ART module 32. However, the distance may be calculated based on the weight coefficient of the ART module 32. Since the weighting coefficient is a vector representing a plurality of input data included in each category, the number of weighting coefficients is significantly smaller than the number of data n. Therefore, when the distance is calculated based on the weight coefficient, the calculation load can be significantly reduced as compared with the case where the distance is calculated based on the input data. In the present embodiment, the similar case search range is all input data, but the search range may be limited.

  FIG. 6 is a diagram illustrating an example of a display screen that displays search results. The figure shows the result of searching for similar cases using x1 to x4 as reference data. In the figure, reference numeral 71 is a reference data display screen, 72 is a similar case display screen, and 73 is a detail display button. The reference data display screen 71 shows data used as a reference when calculating the distance. The similar case display screen 72 displays the number of similar cases and the average value of data selected as the similar cases. By comparing the data on the reference data display screen 71 and the similar case display screen 72, the location of the abnormality can be identified. The determination method will be described below.

  In this example, since the data classification means has detected the occurrence of a new category, the measurement item designation means 4 extracts data having similar reference data portions (x1 to x4) as similar cases using x1 to x4 as reference data. As a result, as shown in the similar case display screen 72, 18 pieces of similar data were extracted. The meaning of this result is that “an event that has never been experienced before has occurred when comprehensively judging from the data x1 to x7, but the data from x1 to x4 has been experienced in the past” That's what it means. That is, it is estimated that any of the characteristics of the devices C, D, and E shown in FIG. 5 has changed and items x5 to x7 have changed. Next, comparing the average values of the reference data and the similar data for the items x5 to x7, it can be seen that only the item x7 has a large deviation. Therefore, it is understood that there is a high possibility that the characteristic of the device E has changed.

  As described above, according to the present embodiment, by searching for similar cases based on the data upstream of the subsystem focused by the operator, it is possible to specify the location even if it is an abnormal event experienced for the first time. It becomes possible. If necessary, all the similar data selected by the detail display button 73 can be displayed on the screen as the detail display screen 74. In the present embodiment, an example in which data in which a new category has occurred is used as reference data has been described. However, any data included in the database can be used as reference data.

  FIG. 7 is a diagram for explaining an equipment abnormality diagnosis apparatus according to the second embodiment of the present invention. In the figure, reference numeral 8 denotes state diagnosis means for automatically identifying an abnormal part. In the figure, parts that function in the same way as the part shown in FIG.

  When the operation diagnosis acquired by the data classification unit 3 is classified into a new category, the state diagnosis unit 8 designates a measurement item by the measurement item designation unit 4, and uses the similar case retrieval unit based on the designated measurement item to make a similarity. Search for cases. The state diagnosis means 8 can identify an abnormal device by changing the measurement item of the similar case search from the upstream side to the downstream side and repeating the process.

  FIG. 8 is a diagram for explaining the processing of the state diagnosis means 8. First, in step 811, a similar case is searched based on the facility input condition. Here, the input condition is an item corresponding to “input” when a facility composed of a plurality of devices is regarded as one system. In the case of the equipment shown in FIG. 2, x1 and x2 correspond to the input conditions. In step 812, similar cases are searched under this input condition, and the presence or absence of similar cases is determined. If there is no similar case, it means that the input condition is different from the previous data. That is, the operation data is not classified into a new category because a failure has occurred in the equipment in the facility, and it is determined that the facility has been operated under an unprecedented operation condition (new operation condition).

  On the other hand, if there are similar cases, the operating conditions of the equipment are the same as those experienced so far, so it is considered that the characteristics of one of the devices have changed. Therefore, in step 813, the similarity is calculated based on the data downstream of the device X. At this time, the device X is selected in order from the upstream side. In the example of the measurement item relation diagram shown in FIG. 2, device A is selected, and x3 is selected as the measurement item. Further, when searching for similar cases based on x3, the search is made only from the similar cases searched in step 812.

  Next, in step 814, it is determined whether there is a similar case. As a result of the determination, if there is no similar case, it can be seen that the item x3 is different from the previous data. That is, the item x1 is a value that has been experienced so far, but x3 is a value that has never been experienced, and from these, it can be seen that the characteristics of the device A have changed.

  However, in this state, since it is unclear whether or not the characteristic of the device B has changed, in step 816, the device A is removed and the device E is considered as a new facility from the device B, and the same processing is repeated. Thus, it is possible to identify a device whose device characteristics have changed.

  If there is a similar case in step 814, the presence / absence of remaining data is determined in step 815. If there is remaining data, the process returns to step 813, and the determination is gradually performed based on downstream data. . The above processing is repeated until a device whose characteristics have changed is identified or data is exhausted. Through the above processing, it is possible to identify a device whose characteristics have changed even if it is an event experienced for the first time.

  FIG. 9 is a diagram illustrating a display example that displays a process of identifying a device whose characteristics have changed. Here, the devices whose characteristics have changed, the devices that have not changed, and the devices that do not know whether they have changed are color-coded. Then, the display color is changed in accordance with the progress of the process shown in FIG.

  FIG. 9A shows a situation in which a new category has occurred and the device cannot be specified. For example, all devices are colored yellow. FIG. 9B shows a situation where it is determined that there is a change in the characteristics of the device A and there is no change in the characteristics of the device B. The devices A and B are displayed in different colors. Since the specific process is displayed for each device whose device characteristics have changed in this way, the operator can check the inference process of the abnormality diagnosis apparatus.

  In addition, although the example which designates by a pointing device operation was demonstrated when specifying a measurement item, it is not limited to the thing by pointing device operations, such as a mouse | mouth, A measurement item can be designated by keyboard operation and touch panel operation. it can. In addition, although the example in which the abnormality diagnosis device is provided in the facility has been described, a remote diagnosis device that receives a measurement signal of the facility via the Internet or the like and diagnoses at a remote place may be used.

  As described above, according to the present embodiment, it is possible to identify a device in which a device abnormality whose characteristics have changed even when a new abnormal event occurs in the diagnosis of a facility including a plurality of devices.

It is a figure explaining the equipment abnormality diagnostic device concerning a 1st embodiment of the present invention. It is a systematic diagram of equipment. It is a figure explaining the detail of a data classification device. It is a figure explaining the process of the data pre-processing apparatus. It is a figure explaining the procedure which designates a measurement item with reference to a measurement item related figure. It is a figure explaining the example of the display screen which displays a search result. It is a figure explaining the equipment abnormality diagnostic apparatus concerning 2nd Embodiment. It is a figure explaining the process of the state diagnostic means. It is a figure explaining the example of a display which displays the process which specifies the apparatus from which the characteristic changed.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Equipment 2 Database 3 Data classification means 4 Measurement item designation means 5 Relationship between measurement items 6 Similar case search means 7 Display means 8 State diagnosis means 3 Data classification means 10 Abnormality diagnosis device 31 Data pre-processing device 32 ART module

Claims (6)

A database that records operation data for each measurement item set for each device constituting the facility;
Display means for displaying a measurement item relation diagram representing a causal relationship between the operation data for each measurement item;
A measurement item specifying means for specifying a measurement item to be set for each device based on the measurement item relation diagram;
Based on the measurement item designated by the designation means, a search means for referring to the database and searching for similar cases similar to the operation data in the measurement item set by the designation means, and a display means for displaying the search result A facility diagnostic apparatus characterized by comprising: In the equipment diagnostic apparatus according to claim 1,
The operation data classification means for classifying the operation data into a plurality of categories based on the causal relationship between the operation data for each measurement item is provided, and when the operation data belonging to the new category is acquired by the means, the operation data is similar to the operation data A facility diagnostic device characterized by searching for similar cases. In the equipment diagnostic apparatus according to claim 1,
The facility diagnostic apparatus characterized in that the designation means designates a measurement item by instructing each device of the measurement item relation diagram by a pointing device. A database that records operation data for each measurement item set for each device constituting the facility;
Driving data classification means for classifying driving data into a plurality of categories based on the causal relationship between driving data for each measurement item,
The measurement item relation diagram showing the causal relationship between the operation data for each measurement item,
A measurement related item specifying means for specifying a measurement item to be set for each device based on the measurement item related diagram;
Based on the measurement item set by the specifying means, referring to the database, similar example search means for searching similar cases similar to the operation data in the measurement item specified by the specifying means,
A state diagnosis unit is provided, and when the state diagnosis unit acquires operation data of a new category different from the data stored as normal and abnormal operation data, the measurement unit is specified by the specifying unit, and a similar case search is performed. A facility diagnosis apparatus characterized in that a similar case search by means is repeatedly executed by changing the measurement item. In the equipment diagnostic apparatus according to claim 4,
The facility diagnosis apparatus, wherein the designating unit sequentially designates measurement items from the upstream side in the signal flow of the facility. Record the operation data for each measurement item set for each device that makes up the equipment,
Display a measurement item relation diagram representing a causal relationship between operation data for each measurement item,
Specify the measurement item to be set for each device based on the measurement item relation diagram,
Based on the set measurement item, refer to the database, search for similar cases similar to the operation data in the set measurement item, and display the search result. .

Images