JP5292477B2 - Diagnostic device and diagnostic method - Google Patents

Description

  The present invention relates to a diagnostic apparatus and a diagnostic method.

  When an abnormal transient or accident occurs in the plant, the plant diagnosis device detects the occurrence of the abnormality or accident based on the measurement data from the plant.

  Patent Document 1 discloses a diagnostic apparatus that uses adaptive resonance theory (ART).

  A diagnostic apparatus using ART has a function of classifying multidimensional data into categories according to the similarity.

  In the technique of Patent Document 1, first, normal measurement data is classified into a plurality of categories (normal categories) using ART. Next, current measurement data is classified into categories by ART. When this measurement data cannot be classified into normal categories, a new category (new category) is generated. Finally, an abnormality is diagnosed when the incidence of a new category exceeds a threshold.

JP 2005-165375 A

  When an abnormality occurs in a plant or the like (diagnosis target), an operator (operator) estimates the cause based on the measurement data in order to deal with the problem accurately. For example, since thermal power plants measure hundreds to thousands of data, it takes time to extract data related to abnormal events.

  An object of the present invention is to automatically extract data items related to an abnormal event to be diagnosed.

  The diagnostic apparatus of the present invention includes a measurement signal database that stores a measurement signal to be diagnosed, a processing data extraction unit that extracts a diagnosis signal used for diagnosing the state of the diagnosis target from the measurement signal database, and the diagnosis A reference signal database for storing signals, a classification unit for classifying data stored in the reference signal database into categories, a classification result database for storing the categories as normal categories, and the latest extracted by the processing data extraction unit A diagnostic unit for generating and classifying a new category different from the category when the diagnostic signal of the classification does not belong to the normal category stored in the classification result database, and a diagnosis for storing the classification result classified by the diagnostic unit Results database. And it has a diagnostic result visualization unit for creating diagnostic result display information for visualizing the cause of the occurrence of the new category, the diagnostic result visualization unit includes a cause item extraction unit for extracting the data item that is the cause It is characterized by including.

  According to the present invention, data items related to an abnormal event such as a plant can be automatically extracted, and the time for estimating the cause of the abnormal event can be shortened.

  Further, according to the present invention, even an unskilled operator can immediately determine data items to be confirmed when an abnormality occurs in a plant or the like.

It is a block diagram which shows the diagnostic apparatus of an Example. It is a flowchart which shows the basic operation in the normal state learning mode of the diagnostic apparatus of an Example. It is a flowchart which shows the basic operation | movement in the diagnosis mode of the diagnostic apparatus of an Example. It is a block diagram which shows the timing which performs the flowchart of the normal state learning mode and diagnostic mode of the diagnostic apparatus of an Example. It is a block diagram which shows the timing which performs the flowchart of the normal state learning mode and diagnostic mode of the diagnostic apparatus of an Example. It is a block diagram which shows the timing which performs the flowchart of the normal state learning mode and diagnostic mode of the diagnostic apparatus of an Example. It is a block diagram which shows the Example of the classification | category part 500 and the diagnostic part 600 of FIG. It is a graph which shows the classification result by the example of Drawing 4A. It is a display screen which shows the aspect of the data preserve | saved at a measurement signal database. It is a display screen which shows the aspect of the data preserve | saved at a reference signal database. It is a display screen which shows the mode of the data preserve | saved at a classification result database. It is a display screen which shows the mode of the data preserve | saved at a classification result database. It is a block diagram which shows the thermal power plant of an Example. It is a graph which shows the case where the measurement signal acquired from the thermal power plant of FIG. 6A changes by generation | occurrence | production of abnormality. It is a flowchart which shows the basic operation | movement of the diagnostic result visualization of an Example. It is a graph which shows the relationship between the normal category and new category in a cause item extraction part. It is a graph which shows the relationship between the measurement position and measured value in an Example. It is a graph which shows an example of the relationship between a measurement position and a correction coefficient. It is the graph which showed the relationship between the fuel consumption which is a data item, and the electric power generation amount of a thermal power generator. It is the graph which showed the relationship between the temperature of the combustor which is a data item, and the electric power generation amount of a thermal power generator. It is a block diagram which shows the example of the control logic between items. It is a systematic diagram which shows the example of arrangement | positioning of each item. It is a display screen of the image display apparatus which shows the example of a contribution. It is a display screen of the image display apparatus which shows the example of a contribution. It is a display screen of the image display apparatus which shows the example of the extracted related item.

  Hereinafter, a diagnostic apparatus and a diagnostic method for detecting an abnormality of equipment such as a plant according to an embodiment of the present invention, a program for causing a computer to execute a diagnostic procedure, and a computer-readable recording medium storing the program are disclosed. .

  The diagnostic apparatus includes a measurement signal database for storing a measurement signal to be diagnosed, a processing data extraction unit for extracting a diagnosis signal used for diagnosing the diagnosis target state from the measurement signal database, and a reference for storing the diagnosis signal A classification database that classifies the data stored in the signal database, the data stored in the reference signal database into categories, a classification result database that stores the categories as normal categories, and a latest diagnostic signal extracted by the processing data extraction unit A diagnostic unit that generates and classifies a new category different from the above category when it does not belong to a normal category stored in the database, a diagnostic result database that stores classification results classified by the diagnostic unit, a diagnostic result database, and a normal category The frequency of new category generation using information Generating an alarm when it exceeds and an alarm generator. And it has a diagnostic result visualization part which creates the diagnostic result display information for visualizing the cause which the new category generate | occur | produced, and a diagnostic result visualization part contains the cause item extraction part which extracts the data item which is the said cause.

  The diagnostic apparatus further includes a control logic database in which control logic information for controlling a diagnosis target is stored, and a design information database in which design information of the diagnosis target is stored, and the diagnosis result visualization unit includes: A cause item extraction unit, and a related item extraction unit that extracts a related data item closely related to the data item from a control logic database or a design information database.

  In the diagnostic apparatus, the cause item extraction unit calculates the similarity between the normal category and the data classified into the new category, extracts the normal category having the maximum similarity as the maximum similarity category, and determines the maximum similarity category. And the data classified into the new category, the distance between the data items is calculated as the contribution to the occurrence of the new category, and the data item with the contribution higher than the threshold is extracted as the cause data item.

  In the diagnostic apparatus, the cause item extraction unit has a function of performing correction to increase the contribution degree of the data item or the related data item at the upstream measurement position when the generated abnormal event propagates due to the movement phenomenon.

  The diagnostic apparatus further includes an image display unit, and the image display unit can display a relationship between the data item and the contribution degree.

  In the diagnostic apparatus, the image display unit can display a result obtained by adding the contributions of the data items for each element device constituting the diagnosis target.

  The diagnostic method includes a first step of extracting a diagnostic signal used for diagnosing a state of the diagnostic target from a measurement signal database that stores the measurement signal of the diagnostic target, and a second step of classifying the diagnostic signal into a category. And a third step of generating and classifying a new category different from the category when the latest diagnostic signal does not belong to the category, and a fourth step of extracting a data item causing the new category to be generated And a fifth step of extracting related data items closely related to the data items.

  The diagnostic method repeats from the first step to the fifth step.

  The program includes a first procedure for extracting a diagnostic signal used for diagnosing a state of a diagnostic target from a measurement signal database that stores the diagnostic signal for a diagnostic target in a computer, and a second procedure for classifying the diagnostic signal into a category. A procedure, a third procedure for generating and classifying a new category different from the category when the latest diagnostic signal does not belong to the category, and a fourth procedure for extracting the data item that caused the new category to be generated. And a fifth procedure for extracting a related data item closely related to the data item.

  The said program is for making a computer perform the procedure which repeats from a 1st procedure to a 5th procedure.

  The recording medium includes a first procedure for extracting a diagnostic signal to be used for diagnosing a state of a diagnostic target from a measurement signal database that stores the measurement signal of the diagnostic target in a computer, and a second procedure for classifying the diagnostic signal into a category. A third procedure for generating and classifying a new category different from the category when the latest diagnostic signal does not belong to the category, and a fourth procedure for extracting the data item that caused the new category to be generated. And a computer-readable recording medium storing a program for executing the fifth procedure for extracting related data items closely related to the data item.

  The recording medium is a computer-readable recording medium storing a program for causing a computer to execute a procedure for repeating the first procedure to the fifth procedure.

  Hereinafter, specific examples in the present embodiment will be described with reference to the drawings.

  FIG. 1 is a block diagram illustrating a diagnostic apparatus according to an embodiment.

  In this figure, the state of the plant 100 is diagnosed by the plant diagnostic apparatus 200.

  The diagnosis apparatus 200 includes a processing data extraction unit 400, a classification unit 500, a diagnosis unit 600, an alarm generation unit 700, and a diagnosis result visualization unit 800 as arithmetic units. The diagnostic apparatus 200 includes a measurement signal database 310, a reference signal database 320, a classification result database 330, a diagnosis result database 340, a control logic database 350, and a design information database 360 as databases. In this figure, the database is abbreviated as DB. The database here is a component of the diagnostic apparatus 200, but information recorded in each database is digitized and is usually called an electronic file (electronic data).

  The diagnostic apparatus 200 includes an external input interface 210 and an external output interface 220 as interfaces with the outside. Then, the measurement signal 1 including values obtained by measuring various state quantities of the plant 100 via the external input interface 210 and the external input signal 2 created by operating the external input device 900 including the keyboard 910 and the mouse 920 are diagnosed. Input to the device 200. Further, the image display information 17 is output from the diagnostic device 200 to the image display device 950 (image display unit) via the external output interface 220.

  The measurement signal 3 input via the external input interface 210 is stored in the measurement signal database 310.

  In the processing data extraction unit 400, the diagnostic signal 6 used for diagnosis is extracted from the measurement signal 5 stored in the measurement signal database 310 and stored in the reference signal database 320. The reference signal database 320 stores measurement signals for a period determined by the operator as normal.

  The classification unit 500 classifies the reference signal 7 into categories. The classification result 9 is stored in the classification result database 330. The processing content of the classification unit 500 will be described later with reference to FIGS. 4A and 4B.

  In the diagnosis unit 600, when the latest diagnosis signal 6 extracted by the processing data extraction unit 400 belongs to the classification result database 330, the diagnosis signal 6 is classified into the category. On the other hand, when the latest diagnostic signal 6 extracted by the processing data extraction unit 400 is compared with the classification result 10 stored in the classification result database 330, when the category does not belong to the category stored in the classification result database 330. A new category (hereinafter referred to as a new category) is generated. The diagnosis result 12 that is the classification result created by the diagnosis unit 600 is stored in the diagnosis result database 340. The processing content of the diagnosis unit 600 will be described later with reference to FIGS. 4A and 4B.

  In the alarm generation unit 700, the diagnosis result 13 stored in the diagnosis result database 340, the classification result 14 stored in the classification result database 330, and the measurement signal 4 at the latest time stored in the measurement signal database 310 are stored. Is used to determine whether or not to generate an alarm.

The alarm generation unit 700 has a determination criterion for generating the following two types of alarms, and determines whether to generate an alarm by arbitrarily combining them. (For example, an alarm is generated when both of the following condition 1 and the following condition 2 are satisfied, and an alarm is generated when either of the following condition 1 and the following condition 2 is satisfied)
Condition 1: The measurement signal 4 at the latest time deviates from a predetermined range (threshold value).

  Condition 2: The generation frequency of a new category within a predetermined period exceeds a certain value (threshold value).

  When the alarm generation unit 700 determines to generate an alarm, the alarm generation unit 700 transmits the alarm signal 15 to the external output interface 220. The alarm signal 15 is converted into the image display information 17 by the external output interface 220 and displayed on the image display device 950.

  In this embodiment, an alarm is communicated to the operator using the image display device 950, but the present invention is not limited to this. First, an alarm is generated by generating an alarm sound. You may make it notice. Alternatively, the image display information 17 may be set to be displayed on the image display device 950 after the alarm sound is generated or simultaneously with the generation of the alarm sound.

  The diagnosis result visualization unit 800 includes a cause item extraction unit 810 and a related item extraction unit 820. In the present embodiment, the diagnosis result visualization unit 800 includes the cause item extraction unit 810 and the related item extraction unit 820, but only the cause item extraction unit 810 may be included.

  The cause item extraction unit 810 extracts a data item that causes a new category using the classification result 11 stored in the classification result database 330 and the diagnosis result 13 stored in the diagnosis result database 340. To do. The processing content of the cause item extraction unit 810 will be described later with reference to FIGS. 8A, 8B, and 8C.

  In the related item extraction unit 820, data extracted by the cause item extraction unit 810 using at least the control logic information 18 stored in the control logic database 350 or the design information 19 stored in the design information database 360. Extract data items related to the item. The processing content of the related item extraction unit 820 will be described later with reference to FIGS. 9A and 9B.

  The data items extracted by the cause item extraction unit 810 and the related item extraction unit 820 are sent to the processing data extraction unit 400 as the extracted data item information 20.

  The processing data extraction unit 400 also has a function of extracting the measurement signal 5 of the data item included in the extracted data item information 20 from the information stored in the measurement signal database 310.

  In addition, the data items extracted by the cause item extraction unit 810 and the related item extraction unit 820, the reference signal 8 that is a measurement value of these data items, and the result of processing the information when extracting the data items are displayed as diagnosis results. Information 16 is transmitted to the external output interface 220. The diagnosis result display information 16 is converted into the image display information 17 by the external output interface 220 and displayed on the image display device 950.

  The diagnostic device information 50 stored in the measurement signal database 310, the reference signal database 320, the classification result database 330, the diagnostic result database 340, the control logic database 350, and the design information database 360 can be displayed on the image display device 950. It has become. These pieces of information can be corrected using the external input device 900 as necessary.

  In this embodiment, the processing data extraction unit 400, the classification unit 500, the diagnosis unit 600, the alarm generation unit 700, the diagnosis result visualization unit 800, the measurement signal database 310, the reference signal database 320, the classification result database 330, the diagnosis result Although the database 340, the control logic database 350, and the design information database 360 are all inside the diagnostic apparatus 200, some of them may be arranged outside the diagnostic apparatus 200 to communicate only the data.

  In addition, although there is one plant 100 to be diagnosed in the present embodiment, a plurality of plants 100 can be diagnosed by one diagnostic device 200.

  2A and 2B are flowcharts showing the basic operation of the diagnostic apparatus 200 of FIG. FIG. 2A shows a normal state learning mode, and FIG. 2B shows a diagnostic mode.

  In the following description, the components described in FIG. 1 are also used.

  The diagnostic apparatus 200 has two basic operations: a normal state learning mode for classifying normal data into categories based on information stored in the reference signal database 320, and a diagnostic mode for diagnosing the state of the plant 100.

  In FIG. 2A, the normal state learning mode is executed by sequentially performing steps 1000 and 1010.

  First, in step 1000, the processing data extraction unit 400 is operated to extract the diagnostic signal 6 from the measurement signal 5 in the measurement signal database 310. The diagnostic signal 6 is stored in the reference signal database 320. The data stored in the reference signal database 320 is data for a period when the operator (operator) determines that the operation state of the plant 100 is normal.

  Next, in step 1010, the classification unit 500 is operated to classify the reference signal 7 stored in the reference signal database 320 and store the classification result 9 in the classification result database 330.

  In the diagnostic mode shown in FIG. 2B, steps 1100, 1110, 1120, 1130, 1140, and 1150 are executed in order.

  First, in step 1100, the measurement signal 1 from the plant 100 is taken into the diagnostic device 200 via the external input interface 210 and the measurement signal 3 is stored in the measurement signal database 310.

  Next, the processing data extraction unit 400 is operated, the measurement signal 5 is extracted from the measurement signal database 310, and the diagnosis signal 6 with the latest time is transmitted to the diagnosis unit 600.

  In step 1110, the diagnosis unit 600 is operated and the diagnosis result 12 is transmitted to the diagnosis result database 340.

  In step 1120, at least the diagnostic result visualization unit 800 is operated, and when a new category is generated in step 1110, a data item that causes the new category is extracted. Details of step 1120 will be described later with reference to FIG.

  In step 1130, the diagnosis result display information 16 output from the diagnosis result visualization unit 800 is converted into image display information 17 by the external output interface 220 and output to the image display device 950.

  In step 1140, alarm generation unit 700 is operated to determine whether an alarm can be generated. If it is determined in step 1140 that an alarm can be generated, the process proceeds to step 1150. On the other hand, if the alarm is not generated, the process ends.

  In step 1150, the alarm signal 15 output from the alarm generation unit 700 is converted into image display information 17 by the external output interface 220 and output to the image display device 950. Thereby, an alarm is notified to the operator (operator) of the plant 100.

  3A, 3B, and 3C are diagrams illustrating timings for executing the flowchart of the normal state learning mode (FIG. 2A) and the flowchart of the diagnostic mode (FIG. 2B) of the diagnostic apparatus 200. FIG.

  The diagnostic apparatus 200 acquires the measurement signal 1 from the plant 100 for each sampling period.

  In FIG. 3A, diagnosis is performed by operating both the normal state learning mode and the diagnostic mode for each sampling period.

  Further, in FIG. 3B, the normal state learning mode is operated every predetermined set period, and only the diagnosis mode is operated every sampling period for diagnosis.

  Further, in FIG. 3C, the operator performs an operation of setting a learning period and a diagnosis period, and the normal state learning mode and the diagnosis mode are operated at this timing.

  FIG. 4A is a block diagram illustrating an example of the classification unit 500 and the diagnosis unit 600 of FIG.

  Hereinafter, a case where adaptive resonance theory (ART) is applied to the classification unit 500 and the diagnosis unit 600 will be described, but other clustering methods such as vector quantization may be used.

  In this figure, the classification unit 500 and the diagnosis unit 600 are composed of a data preprocessing device 610 and an ART module 620. The data preprocessing device 610 converts the operation data into input data for the ART module 620.

  Hereinafter, the procedure (process) will be described.

  First, the maximum value and the minimum value are calculated for each measurement item. Data is normalized using the calculated maximum and minimum values.

  Here, the normalization method will be described by taking the process amount xi of the plant as an example.

  The number of data of xi is N and the nth measurement value is xi (n). Further, if the maximum value and the minimum value in N pieces of data are Max_i and Min_i, respectively, normalized data Nxi (n) is expressed by the following formula (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 the above equation (1).

  Next, the complement of the normalized data is calculated and added to the input data.

  The complement CNxi (n) of the normalized data Nxi (n) is calculated by the following equation (2).

  Next, data including data Nxi (n) and CNxi (n) is input to the ART module 620 as input data.

  The above procedure is included in the input data conversion processing of the operation data to the ART module 620 performed in the data preprocessing device 610.

  The ART module 620 classifies input data into a plurality of categories.

  The ART module 620 includes an F0 layer 621, an F1 layer 622, an F2 layer 623, a memory 624, and a selection subsystem 625, which are coupled to each other. The F1 layer 622 and the F2 layer 623 are coupled via a weighting factor, and the weighting factor represents a prototype (prototype) of a category into which input data is classified. Here, the prototype represents a representative value of the category.

  Next, the algorithm of the ART module 620 will be described.

  The outline of the algorithm when input data is input to the ART module 620 is as shown in the following processing 1 to processing 5.

  Process 1: The input vector is normalized by the F0 layer 621, and noise is removed.

  Process 2: A suitable category candidate is selected by comparing the input data input to the F1 layer 622 with a weighting factor.

  Process 3: The validity of the category selected by the selection subsystem 625 is evaluated by the ratio with the parameter ρ. If it is determined to be valid, the input data is classified into the category, and the process proceeds to process 4. On the other hand, if it is not judged to be valid, the category is reset, and an appropriate category candidate is selected from the other categories (repeat processing 2). Increasing the value of parameter ρ makes the category classification finer, and decreasing the value of ρ makes the classification coarse. This parameter ρ is referred to as a vigilance parameter.

  Process 4: When all the existing categories are reset in Process 2, it is determined that the input data belongs to the new category, and a new weighting factor representing the prototype of the new category is generated.

  Process 5: When input data is classified into category J, weight coefficient WJ (new) corresponding to category J uses past weight coefficient WJ (old) and input data p (or data derived from input data). And updated by the following equation (3).

  Here, Kw is a learning rate parameter (0 <Kw <1), and is a value that 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 620 is the processing 4 described above.

  In process 4, when input data different from the pattern recorded (saved) in the classification result database 330 of FIG. 1 is input, a new pattern can be recorded without changing the recorded pattern. Therefore, it is possible to record a new pattern while recording a pattern learned in the past.

  As described above, when the operation data given in advance is given as input data, the ART module 620 learns the given pattern. Therefore, when new input data is input to the learned ART module 620, it is possible to determine which pattern in the past is close by the above algorithm. If the pattern has never been experienced before, it is classified into a new category.

  FIG. 4B is a graph illustrating an example of the classification result.

  This figure displays two items of measurement data as an example, and is represented by a two-dimensional graph. The vertical axis and the horizontal axis indicate standardized measurement data for each item.

  The measurement data 625 is divided into a plurality of categories 630 (circles shown in FIG. 4B) by the ART module 620 of FIG. 4A.

  In this figure, the two-dimensional measurement data is shown as a two-dimensional graph. However, the present invention is not limited to this, and a category may be created using multi-dimensional coordinates for three or more measurement data. Good.

  5A, 5B, 5C, and 5D show aspects of data stored in the measurement signal database 310, the reference signal database 320, and the classification result database 330 (two cases). These drawings may be considered as display screens of the image display device 950 of FIG.

  For example, in the case of the measurement signal database in FIG. 5A, a wide range of data can be scroll-displayed by using scroll boxes 56 a and 56 b that can be moved vertically and horizontally on the display screen 55.

  As shown in this figure, the measurement signal database 310 stores values of a plurality of data items (items A, B, C, etc.) measured by the plant 100 for each sampling period (vertical time).

  Further, in the case of the reference signal database of FIG. 5B, by selecting the tabs 57a, 57b and 57c indicating the data sheets of the references 1 to 3, only the items classified for each reference can be displayed together.

  In addition, the aspect of the data preserve | saved in the classification result database 330 and the diagnostic result database 340 is the same.

  1 extracts a data group used for diagnosis of the plant 100 from the measurement signal database 310.

  For example, in FIG. 5B, there are three reference signal data groups ("reference 1", "reference 2", and "reference 3"), and "reference 1" is selected as the reference tabs 57a, 57b, and 57c. A state in which the data group is composed of item A, item C, and item D is displayed.

  As described above, the measurement values of all data items are stored in the measurement signal database 310 in time series as one data group, whereas the reference signal database 320 is selected and limited according to the reference. The measured values of the data items are stored in time series as a plurality of data groups.

  5C and 5D are display screens showing modes of data stored in the classification result database 330 of FIG.

  In FIG. 5C, the relationship between the time and the category number into which the data at that time is classified is displayed. On the other hand, in FIG. 5D, the relationship between the category number and the weighting factor is displayed.

  The classification result database 330 stores the classification results for each data group stored in the reference signal database 320.

  FIG. 6A is a block diagram illustrating a thermal power plant according to an embodiment.

  In this figure, the thermal power plant 100 includes a gas turbine generator 110, a control device 120, and a data transmission device 130. The gas turbine generator 110 includes a generator 111, a compressor 112, a combustor 113, and a turbine 114.

  At the time of power generation, the air sucked by the compressor 112 is compressed into compressed air, and this compressed air is sent to the combustor 113 and mixed with fuel and burned. The turbine 114 is rotated using the high-pressure gas generated by the combustion, and the generator 111 generates power.

  In the control device 120, the output of the gas turbine generator 110 is controlled according to the power demand. The control device 120 uses the operation data 102 measured by a sensor (not shown) installed in the gas turbine generator 110 as input data. The operation data 102 is state quantities such as intake air temperature, fuel input amount, turbine exhaust gas temperature, turbine rotation speed, generator power generation amount, turbine shaft vibration, and the like, and is measured at each sampling period. It also measures weather information such as atmospheric temperature.

  In the control device 120, the control signal 101 for controlling the gas turbine generator 110 is calculated using these operation data 102.

  The signal data transmission device 130 transmits the measurement signal 1 including the operation data 102 measured by the control device 120 and the control signal 101 calculated by the control device 120 to the diagnosis device 200.

  FIG. 6B shows an example of a change over time when the measurement signal 1 acquired from the plant 100 of FIG. 6A changes due to the occurrence of an abnormality.

  The horizontal axis represents time, and the vertical axis represents measurement signals, category numbers, and the occurrence rate (generation frequency) of new categories. Data I and II correspond to items A and B, respectively.

  In this example, item A is a generator output, and item B is an atmospheric temperature measurement signal.

  In this figure, items A and B are initially stable at a substantially constant value, but data I (item A) decreases immediately before time t1, and then data II (item B) immediately after time t1. increased. Thereafter, data II (item B) decreases, and finally both data I (item A) and data II (item B) increase.

  Further, before reaching time t1, the category numbers of items A and B are 1 to 4, which is a reference time category, that is, a normal category. On the other hand, after the time t1, the category numbers of the items A and B are 5 to 7, which is a new category indicating the occurrence of abnormality.

  Along with this, the occurrence rate (occurrence frequency) of the new category increases immediately after the time t1, and the abnormality is diagnosed exceeding the threshold. Here, the occurrence ratio of new categories is calculated using a moving average of the number of new categories that occur in a predetermined period.

  When the occurrence rate of the new category exceeds a preset threshold value, the alarm generation unit 700 in FIG. 1 issues an alarm.

  When an abnormality occurs due to a change in the state of the plant, the operator needs to estimate the cause based on the measurement data in order to deal with the problem accurately.

  However, since a thermal power plant measures hundreds to thousands of data, it takes time to extract data related to state changes.

  Therefore, in the present invention, by operating the diagnostic result visualization unit 800 provided in the diagnostic apparatus 200, a data item (cause data item) causing a new category can be extracted. Yes.

  Thereby, the data items to be confirmed when the state of the plant changes can be narrowed down, and the operator's work can be greatly reduced. Steps 1120 and 1130 in FIG. 2B correspond to this function.

  FIG. 7 is a flowchart showing the basic operation for visualizing the diagnosis result, and shows the details of the operation in step 1120 of FIG. 2B.

  In this figure, the visualization of the diagnosis result is executed by sequentially performing steps 1200, 1210, 1220, 1230, 1240, 1250, 1260, and 1270.

  First, in step 1200, the cause item extraction unit 810 in FIG. 1 is operated to extract the data item that caused the new category to be generated. The cause item extraction unit 810 uses the classification result 11 stored in the classification result database 330 and the diagnosis result 13 stored in the diagnosis result database 340. The processing content of the cause item extraction unit 810 will be described later with reference to FIGS. 8A, 8B, and 8C.

  Next, in step 1210, it is determined whether or not related items can be extracted. If a related item is to be extracted, the process proceeds to step 1220. If not, the process proceeds to the end. Whether or not the related item can be extracted is set in advance by the operator.

  In step 1220, the related item extraction unit 820 is operated to extract data items related to the data items extracted by the cause item extraction unit 810. In the related item extraction unit 820, the data item extracted by the cause item extraction unit 810 using at least the control logic information 18 stored in the control logic database 350 or the design information 19 stored in the design information database 360. Extract data items related to. The processing content of the related item extraction unit will be described later with reference to FIGS. 9A and 9B.

  In step 1230, it is determined whether or not the data item extracted by the related item extraction unit 820 can be re-evaluated. Here, the reevaluation means that diagnosis is performed by operating the classification unit 500 and the diagnosis unit 600 using a signal including the data item extracted by the related item extraction unit 820. If re-evaluation is performed, the process proceeds to step 1240. If not re-evaluated, the process proceeds to end.

  In step 1240, the processing data extraction unit 400 is operated to extract data item data extracted by the related item extraction unit 820 from the measurement signal database 310. In step 1250, the classification unit 500 is operated to classify the data extracted in step 1240.

  In step 1260, the cause item extraction unit 810 is operated, and in step 1250, the data item that causes the new category is extracted.

  In step 1270, the number of operations in steps 1220 to 1260 is compared with a threshold value. If the number of operations exceeds the threshold value, the process proceeds to the end. Otherwise, the process returns to step 1220.

  Note that whether or not re-evaluation is possible in step 1230 and the threshold value in step 1270 are set in advance by the operator.

  8A, 8B, and 8C are diagrams for explaining the operation of the cause item extraction unit 810, and explain the operation contents of Step 1200 to Step 1270 in FIG.

  FIG. 8A is a graph showing a relationship between a normal category and a new category in the cause item extraction unit. The item A is taken on the horizontal axis and the item B is taken on the vertical axis.

  FIG. 8B is a graph illustrating a relationship between a measurement position and a measurement value in the example. The horizontal axis represents the measurement position of the data related to the fluid, and the vertical axis represents the measured value.

  FIG. 8C is a graph showing an example of the relationship between the measurement position and the correction coefficient. The horizontal axis represents the measurement position of data relating to the fluid, and the vertical axis represents the correction coefficient f corresponding to each measurement position.

  In ART, as shown in FIG. 8A, data at the reference time is classified into several normal categories. In this operation, when a new category is generated, the similarity between the target data and the normal category is calculated, and the normal category that is most similar (similarity is maximized) is extracted.

  The similarity S is calculated using, for example, the following formula (4), and a normal category having the smallest S (similarity is minimized) is extracted.

  Here, i is a code for identifying a data item, and 1 ≦ i ≦ n (n is the total number of data items). J is a code for identifying a category, and 1 ≦ j ≦ m (m is the total number of normal categories). Further, Sj is the similarity, Di is the value of the data item i of the target data, and Wij is the weighting coefficient of the data item i in category j.

  Next, the contribution Ci of each data item is calculated using, for example, the following formula (5).

  A data item with a higher contribution is far from the normal category, and can be said to be a data item that causes a new category to occur.

  The cause item extraction unit 810 extracts a data item whose contribution is higher than a threshold value as a cause data item. This threshold value is a value set in advance by the operator.

  On the other hand, the contribution may be corrected in consideration of not only the distance to the category but also the measurement position.

  FIG. 8B shows a case where data (measurement value: temperature, for example) is measured from upstream to downstream with respect to the fluid flowing through a certain flow path of the plant 100. α, β, γ, and ω indicate measurement positions from upstream to downstream.

  In this figure, the normal time is indicated by a solid line, and the abnormal time is indicated by a broken line.

  When an abnormality occurs on the upstream side, the measured value on the downstream side is different from the normal value.

  In the case of this figure, an abnormality has occurred between the measurement positions β and γ. Due to this influence, the measured values measured at the positions of γ and ω are changed.

  When data items are extracted by the procedure shown in FIG. 8A, data measured at the positions of γ and ω are extracted, but the contributions are approximately the same.

  However, since γ is closer to the place where the abnormality has occurred, the data measured at the position of γ is more important in estimating the cause of the abnormality.

  Therefore, in this example, a function for correcting the cause item extraction unit 810 so that the contribution becomes larger as the measurement position is upstream is added, and the data measured at the position of γ is used as a more important data item. Judgment can be made.

  FIG. 8C shows an example, in which the correction coefficient f is gradually decreased from upstream to downstream. Thereby, the contribution degree of the upstream data can be increased.

  The degree of contribution Ci ′ when the above correction is performed can be calculated using the following equation (6).

  Here, f is a correction coefficient.

  In this embodiment, the case where an abnormality has occurred in the data relating to the fluid has been described as an example. However, the present invention is not limited to this, and the method of calculating the contribution degree by applying the above correction may be performed from the high temperature part to the low temperature part. Abnormal events that occur, such as heat transfer to the part, diffusion of substances (including gases, solutes in solutions, etc.), propagation of vibrations (including light, sound, etc.) propagate (move) due to movement phenomena. Applicable in all cases. In this case, a measurement position close to the part where the abnormal event has occurred is called an upstream measurement position, and a measurement position far from the part where the abnormal event has occurred is called a downstream measurement position.

  In this embodiment, data items are extracted using ART. However, the present invention is not limited to this, and data items related to calculation of contributions and occurrence of anomalies based on the same concept as in this embodiment. Any other clustering technique other than ART may be used as long as it can be extracted.

  In FIG. 8A, the category is shown as having a substantially circular shape, but the present invention is not limited to this.

  For example, FIG. 8D is a graph showing the relationship between the amount of fuel used as the data item and the amount of power generated by the thermal power generator. The horizontal axis represents fuel consumption, and the vertical axis represents power generation.

  In the case of this figure, since the fuel consumption and the power generation amount are in a proportional relationship, a strong correlation is exhibited during normal operation (correlation coefficient is close to 1). In this case, the category 626 in the graph has an elongated elliptical shape.

  FIG. 8E is a graph showing the relationship between the combustor temperature, which is a data item, and the amount of power generated by the thermal power generator. The horizontal axis represents the temperature of the combustor, and the vertical axis represents the amount of power generation.

  In the case of this figure, there is no clear relationship between the temperature of the combustor and the amount of power generation, and during normal operation, the temperature of the combustor shows a substantially constant value regardless of the amount of power generation. For this reason, the category 627 in this case may be categorized according to the value of the power generation amount and may be circular.

  9A and 9B are diagrams for explaining the operation of the related item extraction unit 820 in FIG.

  FIG. 9A is a block diagram illustrating an example of control logic between items. FIG. 9B is a system diagram illustrating an example of an arrangement of each item.

  In the related item extraction unit 820 in FIG. 1, data items related to the data items extracted by the cause item extraction unit 810 are extracted.

  First, the related item extraction unit 820 selects a predetermined number of data items in descending order of contribution calculated by the cause item extraction unit 810.

  Next, a data item closely related to the selected data item (a data item closely related to the selected data item) is stored in the control logic information 18 stored in the control logic database 350 or the design information database 360. The extracted design information 19 is used for extraction.

  An example of the control logic information 18 stored in the control logic database 350 of FIG. 1 will be described using FIG. 9A. As an example of the plant 100, the thermal power plant of FIG. 6A is used.

  The control logic database 350 in FIG. 1 stores a control logic diagram (data) for calculating the control signal 101 from the operation data 102 in the control device 120 in FIG. 6A.

  FIG. 9A shows an example of a control logic diagram of proportional / integral control widely used for plant control on the display screen of the image display device 950 of FIG.

  This display screen shows a control logic for calculating an item B (control signal) by performing a proportional-integral operation on an error between the item A (operation data) and a set value.

  When the data item previously selected by the related item extraction unit 820 is the item B in FIG. 9A, the item A that is the basis for calculating the data item is a data item that is closely related to the abnormal event (closely related to the abnormal event). Data item having a certain relationship). Therefore, the related item extraction unit 820 extracts item A as a related data item.

  FIG. 9B shows an example of the design information 19 stored in the design information database 360 of FIG. 1 on the display screen of the image display device 950 of FIG.

  The design information database 360 of FIG. 1 stores design information of the gas turbine generator 110 of FIG. 6A. For example, the relationship between the fluid path and the sensor location (T: temperature sensor, P: pressure sensor). A system diagram showing is stored.

  First, when the data item selected by the related item extraction unit 820 is the item E (temperature) in FIG. 9B, it is considered that the item F (pressure) measured at the same position is deeply related to the abnormal event. In addition, items C, D, G, and H having close measurement positions are expected to be associated with abnormal events. Therefore, the related item extraction unit extracts items C, D, F, G, and H as related data items.

  Here, “intimately related to abnormal event” means that when an abnormal event occurs, it is strongly affected by the abnormal event, and the data item “deeply related to abnormal event” indicates that the measured data is an abnormal event Is an item (sensor or the like) that has a high possibility of directly indicating. Such an item is called a related data item.

  In this embodiment, item A is a general data item (operation data). Further, items C, D, F, G, and H are data items based on design information such as a specific system diagram, and include data items including arrangement positions (measurement positions) such as specific sensors and measurement data. It is.

  In addition, the related item extraction unit 820 uses the reference signal 8 stored in the reference signal database 320 to obtain the correlation coefficient between the data item selected based on the contribution and the other data items, and the correlation Data items whose number is higher than the threshold value can also be extracted.

  As shown in FIG. 7, after extracting the related item in step 1220, whether the data item extracted by the related item extracting unit 820 is caused by the operation of steps 1240, 1250, and 1260 is the cause of the occurrence of the new category. Reassess whether.

  If the data item extracted by the related item extraction unit 820 is irrelevant to the occurrence of a new category, it is determined that the data item is not related to the abnormality because the contribution when the step 1260 is operated is reduced. it can. Accordingly, this data item can be excluded from the data items to be confirmed, and the remaining data items need only be confirmed.

  10A, 10B, and 10C show examples in which the degree of contribution is displayed on the display screen of the image display device 950. FIG.

  FIG. 10A displays the result regarding the relationship between the data item and the contribution.

  From this relationship, data with a high contribution can be grasped, and data to be confirmed when an abnormality occurs can be narrowed down.

  FIG. 10B is a display screen of the image display device 950 showing the result of adding the contributions (total contribution points) for the data items of the measuring instruments arranged for each element device constituting the diagnosis target (plant etc.). is there. In this figure, the component devices are described as device 1 and device 2.

  In the figure, it can be seen that the added value of the contribution is higher in the device 1 than in the device 2. This indicates that there is a high possibility that an abnormality has occurred in the device 1.

  In the figure, a device having a high contribution value (in this case, device 1) can be highlighted. Thereby, it is possible to easily identify a device in which an abnormality has occurred.

  When the device 1 in FIG. 10B is clicked (selected) using the mouse, the details of the device 1 can be displayed on the display screen of the image display device 950 as shown in FIG. 10C.

  As illustrated in FIG. 10C, the device 1 includes one pump and four measuring instruments. Of the data items measured by the measuring instrument, the measuring instrument with the largest contribution can be highlighted as shown in FIG. 10C. Thereby, the location where abnormality has occurred can be identified.

  As described above, with the function of displaying related data items on the display screen of the image display device 950, even an unskilled operator can immediately select data items to be confirmed when an abnormality occurs in a plant or the like. Can be (easily) judged.

  In the above embodiments, the apparatus or method for diagnosing the plant has been described. However, the equipment to be diagnosed (diagnostic object) is not limited to the plant, but one or a plurality of measuring instruments. It is possible to apply to a facility that needs to accumulate and manage (control) a plurality of data generated from the network.

  In this case, a program for causing a computer to realize the functions of the diagnostic device, a program for causing the computer to execute a procedure (step) relating to the diagnostic method, or for causing the computer to function as each part of the diagnostic device. These programs are also included in the present invention. In addition, the present invention includes computer-readable recording media, databases, and the like in which the various data or programs described above are recorded.

  According to the present invention, it is possible to automatically extract data items related to an abnormality and reduce the time for estimating the cause of the abnormality.

  The present invention can be widely applied to various plants and the like as a diagnostic apparatus for plants and the like.

  100: plant, 200: diagnostic device, 210: external input interface, 220: external output interface, 310: measurement signal database, 320: reference signal database, 330: classification result database, 340: diagnosis result database, 350: control logic database 360: Design information database, 400: Processing data extraction unit, 500: Classification unit, 600: Diagnosis unit, 700: Alarm generation unit, 800: Diagnosis result visualization unit, 810: Cause item extraction unit, 820: Related item extraction unit 900: external input device, 910: keyboard, 920: mouse, 950: image display device.

Claims (11)

A measurement signal database for storing a measurement signal to be diagnosed, a processing data extraction unit for extracting a diagnosis signal used for diagnosing the state of the diagnosis target from the measurement signal database, and a reference signal database for storing the diagnosis signal A classification unit that classifies data stored in the reference signal database into categories, a classification result database that stores the categories as normal categories, and the latest diagnostic signal extracted by the processing data extraction unit A diagnostic unit that generates and classifies a new category different from the category when it does not belong to the normal category stored in a result database, a diagnostic result database that stores a classification result classified by the diagnostic unit, and the diagnostic result Using the database and information on the normal category, A diagnostic device including an alarm generation unit for generating an alarm when the frequency of generation of a new category exceeds a threshold, and creating a diagnostic result display information for visualizing the cause of the occurrence of the new category a visualization unit, wherein the diagnostic result visualization unit is viewed including the cause item extraction section for extracting a data item which is the cause, the cause item extracting unit, and the classified into normal category as the new category data The normal category that maximizes the similarity is extracted as the maximum similarity category, and the data items classified into the maximum similarity category and the new category are used for each of the data items. The distance is calculated as the contribution to the occurrence of the new category, and the data item whose contribution is higher than the threshold is the cause data item. Diagnostic apparatus and extracting Te.   The diagnostic result visualization unit further includes a control logic database in which control logic information for controlling the diagnosis target is stored, and a design information database in which design information of the diagnosis target is stored. The diagnostic apparatus according to claim 1, further comprising: an item extraction unit; and a related item extraction unit that extracts a related data item closely related to the data item from the control logic database or the design information database. The cause item extraction unit has a function of performing correction to increase the degree of contribution of the data item or the related data item at an upstream measurement position when an abnormal event that has occurred is propagated by a movement phenomenon. The diagnostic device according to claim 1 . Further comprising an image display unit, the image display unit, diagnostic apparatus according to claim 2 or 3, characterized in that it is capable of displaying the relation between the contribution and the data item. The diagnostic apparatus according to claim 4 , wherein the image display unit is capable of displaying a result obtained by adding the contributions of the data items for each element device constituting the diagnosis target. A first step of extracting a diagnostic signal used for diagnosing the state of the diagnostic target from a measurement signal database that stores the measurement signal of the diagnostic target; a second step of classifying the diagnostic signal into a category; A third step of generating and classifying a new category different from the category when the diagnostic signal does not belong to the category, and a fourth step of extracting a data item causing the new category to be generated; , look including a fifth step of extracting the relevant data item having closely related to the data item, it said fifth step is to calculate the similarity between the classified data in the category as the new category The category having the maximum similarity is extracted as the maximum similarity category, and the categories classified into the maximum similarity category and the new category are extracted. Using the data to calculate the respective distances of the data item as a contribution to the generation of the new category, diagnosis the degree of contribution and extracting high the data item from the threshold cause data items Method. The diagnostic method according to claim 6, wherein the steps from the first step to the fifth step are repeated. A first procedure for extracting a diagnostic signal used for diagnosing the state of the diagnostic target from a measurement signal database that stores the diagnostic signal for a diagnostic target in a computer; a second procedure for classifying the diagnostic signal into a category; A third procedure for generating and classifying a new category different from the category when the latest diagnostic signal does not belong to the category, and a fourth procedure for extracting a data item that causes the new category to be generated. A program for executing a procedure and a fifth procedure for extracting a related data item closely related to the data item , wherein the fifth procedure is classified into the category and the new category Calculate the similarity with data, extract the category with the maximum similarity as the maximum similarity category, and maximize the similarity The distance between each data item is calculated as a contribution to the occurrence of the new category using a category and data classified into the new category, and the data item whose contribution is higher than the threshold is the cause data item A program characterized by being extracted as The program according to claim 8 for causing a computer to execute a procedure of repeating the first procedure to the fifth procedure. A first procedure for extracting a diagnostic signal used for diagnosing the state of the diagnostic target from a measurement signal database that stores the diagnostic signal for a diagnostic target in a computer; a second procedure for classifying the diagnostic signal into a category; A third procedure for generating and classifying a new category different from the category when the latest diagnostic signal does not belong to the category, and a fourth procedure for extracting a data item that causes the new category to be generated. A computer-readable recording medium recording a program for executing a procedure and a fifth procedure for extracting a related data item closely related to the data item , the fifth procedure comprising: The similarity between the category and the data classified into the new category is calculated, and the category with the maximum similarity is similar. The maximum category is extracted, and the distance between the data items is calculated as the contribution to the occurrence of the new category using the maximum similarity category data and the data classified into the new category. A recording medium, wherein the data item higher than a threshold is extracted as a cause data item . The computer-readable recording medium according to claim 10, wherein a program for causing a computer to execute a procedure for repeating the first procedure to the fifth procedure is recorded.

Images