Maintenance system and maintenance method

Abstract

To provide a maintenance system capable of easily updating an information group.SOLUTION: A maintenance system 10 includes: storage units 1 and 3 that store check items for a maintenance target device 60 and an occurrence probability of failures detected with the check items; a display unit 22 that displays the check items on the basis of the occurrence probability of the failures; an occurrence probability estimation unit 8 that estimates the occurrence probability of the failures; and an input unit 24 that receives an input of an addition request for a check item. The occurrence probability estimation unit 8 estimates the occurrence probability of failures corresponding to the check item for which the addition request was made and adds the occurrence probability to the storage units 1 and 3.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present invention relates to a maintenance system and a maintenance method. [Background technology]

[0002] With regard to this technical field, claim 3 of Patent Document 1 below states, "...an equipment status monitoring system characterized in that the physical model is configured as a Bayesian network by calculating a prior probability, which is the number of equipment failures per hour, from the equipment failure probability database, calculating a conditional probability, which is the probability that a parameter abnormality will occur when an equipment failure occurs, from the equipment failure record database, and setting the prior probability and conditional probability in the equipment deterioration model." [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2020-9080 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the above-mentioned technology, there may be cases where the Bayesian network needs to be updated. However, Patent Document 1 does not particularly describe how to easily update information groups such as a Bayesian network. The present invention has been made in view of the above-mentioned circumstances, and has as its object to provide a maintenance system and a maintenance method that can easily update various types of information. [Means for solving the problem]

[0005] In order to solve the above problems, the maintenance system of the present invention comprises a memory unit that stores check items for equipment to be maintained and the probability of occurrence of failures detected by the check items, a display unit that displays the check items based on the probability of occurrence of the failure, an occurrence probability estimation unit that estimates the probability of occurrence of the failure, and an input unit that accepts input of a request to add the check item, wherein the occurrence probability estimation unit estimates the probability of occurrence of the failure corresponding to the check item that has been requested to be added and adds it to the memory unit. [Effects of the Invention]

[0006] According to the present invention, various types of information can be easily updated. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 2 is a diagram illustrating an example of a data configuration of Bayesian network information applied to the first embodiment. [Figure 2] FIG. 10 is a diagram illustrating an example of node content information. [Figure 3] FIG. 10 is a diagram illustrating an example of link information. [Figure 4] FIG. 10 is a diagram illustrating an example of maintenance knowledge data. [Figure 5] FIG. 10 is a diagram illustrating an example of case data. [Figure 6] FIG. 10 is a diagram showing a failure mode occurrence probability table. [Figure 7] FIG. 10 is a diagram showing a child node abnormality occurrence probability table when an abnormality occurs in a parent node. [Figure 8] FIG. 10 is a diagram showing a table of child node abnormality occurrence probabilities when the parent node is normal. [Figure 9] 1 is a block diagram of an equipment abnormality cause estimating system according to a first embodiment. [Figure 10] FIG. 1 is a block diagram of a computer. [Figure 11] 10 is a flowchart of a network structure generation processing routine. [Figure 12] 10 is a flowchart of a network probability setting routine. [Figure 13] 10 is a flowchart of a cause estimation processing routine. [Figure 14] FIG. 10 is a diagram showing a failure mode occurrence probability table. [Figure 15] FIG. 10 is a diagram showing a check item input result table. [Figure 16] FIG. 10 is a diagram illustrating an example of an estimation result notification screen. [Figure 17] FIG. 10 is a diagram illustrating an example of maintenance work report data. [Figure 18] FIG. 10 is a diagram showing an example of additional content data. [Figure 19] FIG. 10 is a diagram showing an example of a question screen. [Figure 20] FIG. 10 is a diagram showing an example of an additional check item / failure mode correspondence table. [Figure 21] FIG. 10 is a diagram illustrating an example of a specific process for creating a check item combination table. [Figure 22] 10 is a flowchart of a check item table generation processing routine. [Figure 23] 10 is a flowchart of a question answer processing routine. [Figure 24] FIG. 10 is a diagram showing an example of a failure mode ranking table. [Figure 25] FIG. 10 is a diagram illustrating an example of a probability information table. [Figure 26] FIG. 10 is a block diagram of an equipment abnormality cause estimating system according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] [Prerequisites for the embodiment] The above-mentioned Patent Document 1 does not specifically describe a method for updating Bayesian network information used to estimate the cause of a failure. Therefore, if a change is required to the causal relationships in the Bayesian network information, it may become impossible to perform appropriate estimation. Therefore, in the embodiment described below, it is possible to update the maintenance knowledge as needed by adding new check items to the maintenance knowledge with a simple operation, and to reconstruct the Bayesian network information based on this updated maintenance knowledge. By reconstructing the network information as needed in this way, it is possible to improve the accuracy of estimating the cause of an abnormality.

[0009] [Data structure of the first embodiment] <Bayesian Network Information 80> 1 is a diagram showing an example of the data configuration of Bayesian network information 80 (information group) of maintenance knowledge applied to the first embodiment. The Bayesian network information 80 is data that specifies the causal relationship between an abnormal event that occurs in a maintenance target device 60 (see FIG. 9), such as a gas engine, an equipment failure that is the cause of this abnormal event, a failure mode that is the cause of this equipment failure, and check items that should be confirmed corresponding to the failure mode.

[0010] 1, the Bayesian network information 80 includes a plurality of nodes 811 to 844 and a plurality of links 921 to 947. The nodes 811 to 844 belong to either a hierarchy 81 of abnormal event groups, a hierarchy 82 of functional failure groups, a hierarchy 83 of failure mode groups, or a hierarchy 84 of check items. Each node is assigned a node ID (not shown) which is a unique identification number, and each link is assigned a link ID (not shown) which is a unique identification number.

[0011] Links 921 to 947 represent the causal relationships between these nodes. That is, the origin of the arrow in the diagram is the "cause" and the end of the arrow is the "result." Hereinafter, the node that is the "cause" will be referred to as the "parent node" of the link, and the node that is the "result" will be referred to as the "child node." Links 921 to 947 include the probability that the child node will become abnormal when an abnormality occurs in the parent node, the probability that the child node will become normal when an abnormality occurs in the parent node, the probability that the child node will become abnormal when the parent node is normal, and the probability that the child node will become normal when the parent node is normal.

[0012] The abnormal event group hierarchy 81 includes nodes for multiple "abnormal events" that occur in the maintenance target equipment 60 (see FIG. 9). However, in the example of FIG. 1, for the sake of simplicity, only one node 811 is shown. This node 811 is a node related to "room temperature rise," and indicates, for example, that the room temperature of the room in which the maintenance target equipment 60 is installed has risen abnormally.

[0013] Furthermore, layer 82 of the functional failure group includes nodes for multiple "functional failures" that can be the cause of an abnormal event. In the example shown, layer 82 includes node 821 related to insufficient capacity of a heat exchanger, node 822 related to insufficient capacity of a cooling pump, and node 823 related to insufficient capacity of a cooling tower. Links 921 to 923 directed toward node 811 are connected to these nodes 821 to 823. This indicates that nodes 821 to 823 can be the cause of node 811.

[0014] Furthermore, level 83 of the failure mode group is a level of "failure modes," i.e., detailed failure causes, and includes nodes 831 to 834 (failure cause information). More specifically, level 83 includes node 831 relating to a design flaw in the heat exchanger, node 832 relating to fouling and clogging of the heat exchanger, node 833 relating to fouling and clogging of the cooling pump, and node 834 relating to a deterioration in cooling tower performance.

[0015] Links 931 and 932 are connected to nodes 831 and 832, respectively, which point to node 821. This indicates that nodes 831 and 832 can be the cause of node 821. Furthermore, links 933 and 934 are connected to nodes 822 and 823, respectively, which indicates that node 833 can be the cause of node 822, and node 834 can be the cause of node 823.

[0016] Furthermore, level 84 of the check item group is a level of "check items," i.e., items that users such as maintenance personnel should check, and includes nodes 841 to 844 (check item information). More specifically, level 84 of the check item group includes nodes 841, 842, and 843 related to measurement values ​​D61, D62, and D63 of sensors 61, 62, and 63 attached to maintenance target equipment 60 (see FIG. 9), and node 844 related to the damage status of maintenance target equipment 60. Links 941 and 942 pointing to nodes 841 and 842 are connected to node 831. Link 943 pointing to node 842 is connected to node 832. Links 944 and 945 pointing to nodes 842 and 843 are connected to node 833. Links 946 and 947 pointing to nodes 842 and 844 are connected to node 834.

[0017] In addition to those shown in the figure, the check item group hierarchical level 84 may include items that users such as maintenance personnel should check in the environment of the maintenance target equipment 60, the ancillary equipment, and the components that make up the maintenance target equipment 60. The nodes in hierarchical level 84 correspond to each node in hierarchical level 83 of the failure mode group, and are used to check whether the phenomenon has occurred. As in the example shown in the figure, there may be multiple nodes in hierarchical level 84 that correspond to one node in hierarchical level 83.

[0018] In FIG. 1, "Yes / No" and "Normal / Abnormal" shown inside each node 811 to 844 indicate the state that each node can be in. "Yes" indicates a state in which an abnormality has occurred in the node, and "No" indicates a state in which an abnormality has not occurred in the node. "Normal" indicates that the measurement values ​​D61 to D63 are within the normal range, or that no damage has occurred in the equipment 60 to be maintained. "Abnormal" indicates that the measurement values ​​D61 to D63 are outside the normal range, or that damage has occurred in the equipment 60 to be maintained.

[0019] The network information 80 can be considered to include network structure information 80A and network probability information 80B. Here, the network structure information 80A is information indicating the connections between each node and each link in the Bayesian network information 80. The network probability information 80B is information indicating various probabilities in the network information 80. Note that the structure of the Bayesian network information 80 and the meaning of each layer are not limited to those shown in FIG. 1.

[0020] <Asset Knowledge Database 1> (Node content information 140) FIG. 2 is a diagram showing an example of the node content information 140. As shown in FIG. The node content information 140 is information contained in an asset knowledge database 1 (see FIG. 9) described later. 1 is constructed based on various information contained in the asset knowledge database 1. Node content information 140, which is a part of the information, contains multiple records (rows), and each record corresponds to a combination of one node and the state of that node ("Yes / No" or "normal / abnormal").

[0021] In the example shown in FIG. 1, the Bayesian network information 80 includes 12 nodes 811 to 844, and each node can have two states (Yes / No or normal / abnormal). Therefore, to construct this Bayesian network information 80, the node content information 140 includes 12 x 2 = 24 records. However, only four of these records are displayed in FIG. 2.

[0022] Each record of the node content information 140 includes node information 141, type information 142, status information 143, and component information 144. The node information 141 is the name of any of the nodes 811 to 844 related to the record. The type information 142 is information indicating the type of the node, that is, any of the layers 81 to 84 to which the node belongs (abnormal event, functional failure, failure mode, or check item).

[0023] The state information 143 is information that indicates one of two states ("Yes / No" or "normal / abnormal") that the node can be in. The component information 144 is information that lists identification information for one or more components related to the node when the node belongs to the failure mode hierarchy 83 or the functional failure hierarchy 82. For example, if the number of heat exchangers (not shown) for the node 831 (see FIG. 1) related to a heat exchanger design defect is three, the identification information for these three heat exchangers will be included in the component information 144.

[0024] FIG. 3 is a diagram showing an example of the link information 150. As shown in FIG. This link information 150 is also information included in the asset knowledge database 1 (see FIG. 9). The link information 150 includes multiple records (rows), and each record corresponds to an individual link 921 to 947. In the example shown in FIG. 1, the Bayesian network information 80 includes "14" links 921 to 947. Therefore, the link information 150 for generating this Bayesian network information 80 includes "14" records. However, only two of these records (the records for links 941 and 942) are displayed in FIG. 3.

[0025] Each record of the link information 150 includes parent node information 151 and child node information 152. The parent node information 151 is information indicating the parent node (the base of the arrow in FIG. 1) of the link related to the record, and the child node information 152 is information indicating the child node (the tip of the arrow in FIG. 1) of the link.

[0026] FIG. 4 is a diagram showing an example of the maintenance knowledge data 710. As shown in FIG. This maintenance knowledge data 710 is also information contained in the asset knowledge database 1 (see FIG. 9). The maintenance knowledge data 710 includes a plurality of records (rows), and each record includes abnormal event information 711 , functional failure information 712 , component information 713 , failure mode information 714 , and check item information 715 .

[0027] In Figure 1, the route from layer 81 to layer 84 by following links and nodes is called a "path." For example, the route following links 921, 931, and 941 is one path. Also, the route following links 921, 931, and 942 is another path. Each record in the maintenance knowledge data 710 shown in Figure 4 corresponds to a path, and the maintenance knowledge data 710 includes the same number of records as the number of paths.

[0028] 4 is the name of a node in layer 81 of the abnormal event group in the path, for example, "room temperature rise" at node 811. Functional failure information 712 is the name of a node in layer 82 of the functional failure group in the path, for example, "insufficient heat exchanger capacity" at node 821.

[0029] Component information 713 is information indicating the name of a component corresponding to a node in layer 83 of the failure mode group, for example, "heat exchanger." Failure mode information 714 is the name of a node in layer 83 of the failure mode group in the path, for example, "faulty heat exchanger design" at node 831. Check item information 715 is the name of a node in layer 84 of the check item group in the path, for example, "measurement value D61" at node 841.

[0030] 4 has a cause-and-effect relationship between abnormal event information 711 and functional failure information 712. Similarly, the functional failure information 712 and failure mode information 714 also have a cause-and-effect relationship, and the failure mode information 714 and check item information 715 also have a cause-and-effect relationship.

[0031] FIG. 5 is a diagram showing an example of the case data 720. As shown in FIG. This case data 720 includes case ID information 721 and abnormal event information 722. The case data 720 is data generated by the diagnostic net generation unit 2 (see FIG. 9) when some abnormal event occurs in the maintenance target equipment 60. The case ID information 721 is a unique identification number assigned to the abnormal event. Furthermore, the abnormal event information 722 is information indicating the name of the abnormal event (for example, "room temperature rise").

[0032] FIG. 6 is a diagram showing a failure mode occurrence probability table 110. As shown in FIG. This failure mode occurrence probability table 110 is also information contained in the asset knowledge database 1 (see FIG. 9). The failure mode occurrence probability table 110 includes a plurality of records (rows), and each record includes failure mode information 111, state information 112, and probability information 113.

[0033] The records of the failure mode occurrence probability table 110 are provided corresponding to combinations of nodes belonging to the failure mode group hierarchy 83 (see FIG. 1) and the state of the node. In the example shown in FIG. 1, hierarchy 83 includes four nodes 831 to 834. Each of these nodes takes one of two states: "Yes" (an abnormality has occurred) or "No" (an abnormality has not occurred). Therefore, the failure mode occurrence probability table 110 for generating this Bayesian network information 80 includes "4 x 2 = 8" records. However, FIG. 6 shows only four of these records.

[0034] The failure mode information 111 is the name of the node corresponding to the record. The state information 112 is information indicating one of two states that the node can take ("Yes" or "No" in the above example). In FIG. 6 and other figures, "Y" represents "Yes" and "N" represents "No." The probability information 113 is the probability that the node will be in the state.

[0035] For the same failure mode information 111, the sum of the probability information 133 of records whose status information 112 is "Yes (Y)" and the probability information 133 of records whose status information 112 is "No (N)" is set to be 1.0 (100%).

[0036] In the illustrated example, the probability information 733 is a fixed value of "50%" for all records. However, the initial value of the probability information 133 is not limited to "50%" and may be calculated from past history, etc. For example, the probability information 133 of a record in which the status information 112 is "Yes (Y)" may be "N_b / N_a" where N_a is the total number of failure histories and N_b is the number of occurrences of the failure mode related to the failure mode information 111.

[0037] FIG. 7 is a diagram showing a child node abnormality occurrence probability table 120 (abnormality occurrence probability data) when an abnormality occurs in a parent node. This child node abnormality occurrence probability table 120 is also information contained in the asset knowledge database 1 (see FIG. 9). The child node abnormality occurrence probability table 120 includes multiple records (rows), and each record includes parent node information 121, state information 122, child node information 123, child node state information 124, and probability information 125. Each record corresponds to a combination of any one link and the state of the child node in that link.

[0038] In the example shown in FIG. 1, the Bayesian network information 80 includes 14 links 921 to 947. Furthermore, each child node in these links takes one of two states (for example, "abnormal" or "normal"). Therefore, the child node abnormality occurrence probability table 120 used to generate this Bayesian network information 80 includes 14 x 2 = 28 records. However, only six of these records are displayed in FIG. 7.

[0039] Parent node information 121 is the parent node (the origin of the arrow in FIG. 1) of the link related to the record, and state information 122 is the state of the parent node. In child node abnormality occurrence probability table 120, all state information 122 is "Y (an abnormality has occurred)". Child node information 123 is information indicating the child node of the link (the tip of the arrow in FIG. 1). Child node state information 124 is information indicating one of two states (for example, "abnormal" or "normal") that the child node can take. Furthermore, probability information 125 is information indicating the probability that the child node will take that state.

[0040] The sum of the probability information 125 of two records (for example, the first and second lines in FIG. 7) in which the parent node information 121, status information 122, and child node information 123 are the same but the child node status information 124 is different is "100%." ​​In the illustrated example, if both child node status information 124 is "abnormal," the probability information 125 is "100%," and if both child node status information 124 is "normal," the probability information 125 is "0%." However, the probability information 125 is not limited to these values, and may be calculated from past failure history.

[0041] FIG. 8 is a diagram showing a child node abnormality occurrence probability table 130 (abnormality occurrence probability data) when the parent node is normal. This child node abnormality occurrence probability table 130 is also information contained in the asset knowledge database 1 (see FIG. 9). Child node abnormality occurrence probability table 130 includes multiple records (rows), and each record includes child node information 131, child node state information 132, and probability information 133. Note that the child node abnormality occurrence probability table 130 in Fig. 8 does not include information corresponding to parent node information 121 and parent node state information 122 in Fig. 7. Each record corresponds to a combination of a node in hierarchical layer 84 of the check item group and the state of that node.

[0042] Therefore, the child node anomaly occurrence probability table 130 for generating the Bayesian network information 80 shown in FIG. 1 contains 4×2=8 records. However, FIG. 8 shows only six of these records. Child node information 131 is information indicating the node related to the check item. Child node state information 132 is information indicating one of two states (for example, "abnormal" or "normal") that the child node can take. Furthermore, probability information 133 is information indicating the probability that the child node will take that state.

[0043] The sum of the probability information 133 of two records (for example, the first and second lines in FIG. 8) that have the same child node information 131 but different child node status information 132 is "100%." ​​In the illustrated example, if both child node status information 132 are "abnormal," the probability information 133 is "0%," and if both are "normal," the probability information 133 is "100%." ​​However, the probability information 133 is not limited to these values, and may be calculated from past failure history.

[0044] [First embodiment] <Configuration of the first embodiment> FIG. 9 is a block diagram of an equipment abnormality cause estimating system 10 (maintenance system, computer) according to the first embodiment. In Figure 9, the equipment abnormality cause estimation system 10 includes an asset knowledge database 1 (storage unit), a diagnostic network generation unit 2, a maintenance knowledge Bayesian network database 3 (storage unit) (hereinafter sometimes abbreviated as database 3), an abnormality cause estimation unit 4, a maintenance work report database 5, a GUI question generation unit 6, a probability information input unit 7, a probability information generation unit 8, and an input / output unit 20.

[0045] The input / output unit 20 also includes a display 22 (display unit) and an input device 24 (input unit). The maintenance target device 60 is, for example, a gas engine, and is equipped with sensors 61 to 63. These sensors 61 to 63 output measurement values ​​D61 to D63, respectively.

[0046] The asset knowledge database 1 is a database in which maintenance knowledge is accumulated. Here, maintenance knowledge refers to information extracted from, for example, a maintenance manual, an FT (Fault Tree) diagram, etc. The asset knowledge database 1 also includes the above-mentioned node content information 140 (see FIG. 2), link information 150 (see FIG. 3), maintenance knowledge data 710 (see FIG. 4), failure mode occurrence probability table 110 (see FIG. 6), and child node abnormality occurrence probability tables 120 and 130 (see FIGS. 7 and 8).

[0047] When an abnormality occurs in the equipment 60 to be maintained, the user 50, who is a maintenance worker, inputs the case data 720 shown in Fig. 5 via the input / output unit 20. When this case data 720 is input, the diagnostic net generation unit 2 generates Bayesian network information 80 (see Fig. 1) based on the asset knowledge database 1.

[0048] The database 3 stores the Bayesian network information 80 generated by the diagnostic network generation unit 2. When an abnormality occurs in the maintenance target equipment 60, the user 50 inputs maintenance work report data 520 (see FIG. 17) of various check items to the abnormality cause estimation unit 4. As described above, the check items are the measurement values ​​D61 to D63 of the sensors 61 to 63, the damage state of the maintenance target equipment 60, etc.

[0049] The abnormality cause estimation unit 4 estimates the cause of the abnormality based on the case data 720 (see FIG. 5), the maintenance work report data 520, and the "normal / abnormal" determination results of the measurement values ​​D61 to D63, and calculates the occurrence probability of the failure mode (corresponding to nodes 831 to 834 in FIG. 1). Then, the abnormality cause estimation unit 4 displays the calculated occurrence probability of the failure mode on the display 22 (see FIG. 9).

[0050] The maintenance work report database 5 is a database that stores maintenance work report data 520 (see FIG. 17). That is, the maintenance work report database 5 stores the abnormality cause estimated by the abnormality cause estimation unit 4, the case data 720 (see FIG. 5), and the results of the check items entered by the user 50.

[0051] The GUI question generation unit 6 determines whether or not there is an abnormality cause to be updated in the maintenance work report database 5. If the abnormality cause is present, an update command is sent to the asset knowledge database 1 to update the asset knowledge database 1. The GUI question generation unit 6 also clusters the data of the abnormality causes and check item results stored in the maintenance work report database 5, thereby outputting the abnormality cause detailing targets and the detailing results.

[0052] 10 is a block diagram of the computer 980. The equipment abnormality cause estimating system 10 shown in FIG. 9 includes one or more computers 980 shown in FIG. 10, a computer 980 includes a CPU 981, a storage unit 982, a communication I / F (interface) 983, an input / output I / F 984, and a media I / F 985. Here, the storage unit 982 includes a RAM 982a, a ROM 982b, and an HDD 982c. The communication I / F 983 is connected to a communication circuit 986. The input / output I / F 984 is connected to an input / output device 987. The media I / F 985 reads and writes data from a recording medium 988.

[0053] The ROM 982b stores an IPL (Initial Program Loader) executed by the CPU, etc. The HDD 982c stores application programs, various data, etc. The CPU 981 executes application programs, etc. loaded from the HDD 982c to the RAM 982a, thereby realizing various functions. The interior of the equipment abnormality cause estimation system 10 shown in FIG. 9 is a block diagram showing functions realized by application programs, etc.

[0054] <Operation of the First Embodiment> (Network generation process) Next, the operation of the first embodiment will be described. FIG. 11 is a flowchart of the network structure generation processing routine. This routine is executed in the diagnostic net generation unit 2 when case data 720 (see FIG. 5), i.e., a combination of case ID information 721 and abnormal event information 722, is supplied to the diagnostic net generation unit 2. The task of inputting the case data 720 into the diagnostic net generation unit 2 may be performed, for example, by a user 50. When the process proceeds to step S10 in FIG. 11, the diagnostic net generation unit 2 generates network ID information in the form of a character string or the like. This network ID information is an identification number corresponding to the received case data 720, and is also an identification number assigned to the Bayesian network information 80 generated in the subsequent process.

[0055] Next, when the processing proceeds to step S11, the diagnostic net generation unit 2 searches for abnormal event information 711 relating to the same abnormal event as the abnormal event information 722 from the maintenance knowledge data 710 (see Figure 4) in the asset knowledge database 1 (maintenance knowledge data acquisition). Next, when the process proceeds to step S12, the diagnostic net generator 2 generates network structure information 80A based on the contents of the asset knowledge database 1.

[0056] More specifically, the diagnostic net generation unit 2 generates nodes 811 to 844 and links 921 to 947 in the Bayesian network information 80 (see FIG. 1) based on the acquired maintenance knowledge data 710 (see FIG. 4) of the asset knowledge database 1. This completes the processing of this routine.

[0057] FIG. 12 is a flow chart of the network probability setting routine. When the process proceeds to step S40 in FIG. 12, the diagnostic net generating section 2 acquires the probability information 113 in the failure mode occurrence probability table 110 (see FIG. 6).

[0058] Next, when the processing proceeds to step S41, the diagnostic net generation unit 2 uses the parent node information 151 and child node information 152 in the link information 150 (see Figure 3) to search for and obtain the probability information 125 in the child node abnormality occurrence probability table 120 (see Figure 7) in the asset knowledge database 1.

[0059] Next, when the processing proceeds to step S42, the diagnostic net generation unit 2 uses the parent node information 151 and child node information 152 in the link information 150 (see Figure 3) to search for and obtain the probability information 133 in the child node abnormality occurrence probability table 130 (see Figure 8) in the asset knowledge database 1.

[0060] Next, when the process proceeds to step S43, the diagnostic net generation unit 2 outputs the acquired probability information 113, 125, 133 (see FIGS. 6 to 8) to the maintenance knowledge Bayesian network database 3. As a result, the diagnostic net generation unit 2 stores in the database 3 network probability information 80B that specifies the probability of each part in the Bayesian network information 80.

[0061] (Cause estimation process) FIG. 13 is a flowchart of the cause estimation processing routine. This routine is started in the diagnostic net generating unit 2 after the network structure generating process routine (FIG. 11) and the network probability setting routine (FIG. 12) have been executed.

[0062] 13, when the process proceeds to step S20, the diagnostic net generation unit 2 acquires network ID information, network structure information 80A, network probability information 80B, and case data 720. Note that the network ID information, network structure information 80A, and network probability information 80B were generated in the network structure generation processing routine (FIG. 11) and the network probability setting routine (FIG. 12).

[0063] Next, when the process proceeds to step S21, the diagnostic net generation unit 2 displays the contents of the nodes of the check item group in the acquired Bayesian network information 80 (for example, nodes 841 to 844 in FIG. 1) on the display 22. While looking at the display screen of the display 22, the user 50 checks the maintenance target device 60 in accordance with the displayed check items and inputs the check results to the diagnostic net generation unit 2.

[0064] Next, when the process proceeds to step S22 (information group generation step), the diagnostic net generation unit 2 generates Bayesian network information 80 based on the network structure information 80A and the network probability information 80B.

[0065] Next, when the process proceeds to step S23, the diagnostic net generation unit 2 calculates the probability of occurrence of the failure mode and displays it on the display 22. The procedure is as follows: First, in the generated Bayesian network information 80, the diagnostic net generation unit 2 sets the state of a node (for example, node 811 in FIG. 1) in the hierarchy 81 of the abnormal event group that has the same content as the abnormal event information 722 included in the case data 720 (FIG. 5) to "Yes" (an abnormality has occurred).

[0066] Next, the diagnostic net generation unit 2 calculates the occurrence probability of each node (for example, nodes 831 to 834 in FIG. 1) in the hierarchy 83 of the failure mode group using the check result input by the user 50 in the above-mentioned step S21 and the Bayesian network information 80. Then, the diagnostic net generation unit 2 displays the calculated occurrence probability on the display 22, generates estimation result data ED (see FIGS. 14 and 15), and assigns an estimation result ID to the estimation result data ED.

[0067] The data displayed on the display 22 may include case ID information 721, check results of the check items, the probability of each state of all nodes belonging to the Bayesian network information 80, network ID information, network structure information 80A, and network probability information 80B. For example, a failure mode occurrence probability table 730 (see FIG. 14) and an estimation result notification screen 86 (see FIG. 16), which will be described later, may be displayed.

[0068] FIG. 14 is a diagram showing the failure mode occurrence probability table 730. The failure mode occurrence probability table 730 is part of the estimation result data ED, and has the same data configuration as the failure mode occurrence probability table 110 (see FIG. 6). That is, the failure mode occurrence probability table 730 includes multiple records (rows), and each record includes failure mode information 731, state information 732, and probability information 733. The records of the failure mode occurrence probability table 730 are provided corresponding to combinations of nodes belonging to the failure mode group hierarchy 83 (see FIG. 1) and the state of the nodes.

[0069] 14, state information 732 is "Yes (Y)" or "No (N)", and for the same failure mode information 731, the sum of the probability information 733 of the record where the state information 732 is "Yes (Y)" and the probability information 733 of the record where the state information 732 is "No (N)" is 1.0 (100%). However, the probability information 733 in the failure mode occurrence probability table 730 is a value estimated in the cause estimation processing routine (FIG. 13), and therefore differs from the probability information 113 in the failure mode occurrence probability table 110 (see FIG. 6).

[0070] When displaying the failure mode occurrence probability table 730 on the display 22, the diagnostic net generation unit 2 sorts the failure mode occurrence probability table 730 in descending order of occurrence probability of the failure mode related to the failure mode information 731 (see FIG. 14) in the failure mode occurrence probability table 730, that is, in descending order of the probability information 733 when the status information 732 is "Yes (Y)". This allows the diagnostic net generation unit 2 to display the contents of the failure mode occurrence probability table 730 on the display 22 in descending order of occurrence probability (i.e., the probability information 733) so as to emphasize the items with higher probabilities.

[0071] 1, the node related to the failure mode information 731 belongs to the hierarchy 83 of the failure mode group, and its child nodes exist in both the hierarchy 82 of the functional failure group and the hierarchy 84 of the check item group. Therefore, when displaying the failure mode occurrence probability table 730 on the display 22, the diagnostic net generation unit 2 may also highlight the nodes belonging to the hierarchy 82 of the functional failure group, which are child nodes of the node related to the failure mode information 731. The diagnostic net generation unit 2 may also highlight the child nodes of the hierarchy 84 of the check item group whose status is "abnormal." This allows the user 50 to determine the cause of the abnormality in the maintenance target equipment 60 while looking at the screen displayed on the display 22, and to select the node of the failure mode that has been determined to be the cause of the abnormality.

[0072] FIG. 15 is a diagram showing the check item input result table 740. This check item input result table 740 is also part of the estimation result data ED. The check item input result table 740 includes multiple records (rows), and each record includes check item information 741 and status information 742. Each record in the check item input result table 740 corresponds to each check item previously entered by the user 50 in step S23. That is, the check item information 741 indicates the content of the check item, and the status information 742 indicates the status of the check item, for example, "abnormal" or "normal."

[0073] FIG. 16 is a diagram showing an example of the estimation result notification screen 86. The inference result notification screen 86 is a screen displayed on the display 22, and displays the contents of the Bayesian network information 80. That is, the inference result notification screen 86 displays node images 311 to 344 corresponding to nodes 811 to 844 (see FIG. 1) and link images (without symbols) corresponding to links 921 to 947. Selection buttons 871 to 874 are displayed on each of the node images 331 to 334 of the failure modes. In addition, a manual input button 88 for inferred cause is displayed at the bottom of the inference result notification screen 86.

[0074] The diagnostic net generation unit 2 classifies the probability of a failure mode occurring into three ranks: "high," "medium," and "low" by comparing the probability of the failure mode occurring with a threshold value. Then, the diagnostic net generation unit 2 displays each node image in a different display mode according to the rank on the estimation result notification screen 86 so that these ranks can be distinguished. For example, the diagnostic net generation unit 2 colors each node as follows: "red" for a "high" probability, "pink" for a "medium" probability, and "white" for a "low" probability. In FIG. 16, "dark hatching" represents "red," "light hatching" represents "pink," and "no hatching" represents "white."

[0075] Here, the contents of the diagnostic processing executed by the diagnostic net generating unit 2 will be described using an example. While looking at the screen of Fig. 16, the user 50 checks the equipment 60 to be maintained in accordance with the node images 341 to 344 of the check item group. For example, if the measurement value D61 of the sensor 61 is abnormal, the user 50 inputs that the measurement value D61 is abnormal in association with the node image 341. The diagnostic net generating unit 2 calculates the probability of occurrence of each failure mode based on the input results by the user 50, classifies the probability of occurrence into "high," "medium," and "low" depending on the magnitude of the probability of occurrence, and displays the screen as shown in Fig. 16.

[0076] 16, when the user 50 clicks on any of the selection buttons 871 to 874 in the failure mode node images 331 to 334, the diagnostic net generating unit 2 displays on the display 22 a dialog (not shown) for selecting the failure mode (estimated failure mode) that the user 50 has determined to be the cause of the abnormality. This allows the user 50 to input that he or she has estimated that any of the failure modes corresponding to the node images 331 to 334 is the cause of the abnormality.

[0077] Furthermore, when the user 50 clicks the "Manually Input Estimated Cause" button 88, the anomaly cause estimation unit 4 selects one or more failure modes estimated to be the cause of the anomaly. In this case, the diagnostic net generation unit 2 displays on the display 22 a dialog box for manually inputting one of the one or more failure modes that are the estimation results. In this way, the equipment anomaly cause estimation system 10 accepts input of the estimated failure mode that the user 50 has determined to be the cause of the anomaly through the GUI (Graphical User Interface) of the input / output unit 20. Here, the user 50 performs maintenance work on the maintenance target equipment 60 and investigates the "confirmed failure mode," i.e., the failure mode that is the true cause of the abnormal event.

[0078] 13, when the process proceeds to step S24, the user 50 inputs the results of the maintenance work via the input / output unit 20. That is, the user 50 inputs, via the input / output unit 20, the confirmed failure mode and estimation result evaluation information indicating whether or not the estimated failure mode and the confirmed failure mode match.

[0079] Then, the diagnostic net generation unit 2 displays the input results on the display 22 of the input / output unit 20. The information displayed on the display 22 may include the estimated failure mode with the highest occurrence probability (estimated failure mode information 524 in FIG. 17, which will be described later), the confirmed failure mode input by the user 50 (confirmed failure mode information 525 in FIG. 17, which will be described later), the network ID, the selection result ID, the network ID, etc.

[0080] Next, when the process proceeds to step S25, the diagnostic net generating unit 2 outputs a work report, that is, maintenance work report data 520 (see FIG. 17). With the above, the process of the cause estimation process routine (FIG. 13) ends.

[0081] FIG. 17 is a diagram showing an example of the maintenance work report data 520. As shown in FIG. The maintenance work report data 520 includes multiple records (rows), and each record is work report data generated in the above-mentioned step S25 (see FIG. 13). In FIG. 17, each record includes report ID information 521, network ID information 522, estimation result ID information 523, estimated failure mode information 524, confirmed failure mode information 525, and match information 526.

[0082] The report ID information 521 is identification information that uniquely identifies the work report data (record). The network ID information 522 is identification information of the corresponding Bayesian network information 80. The estimation result ID information 523 is identification information of the estimation result data ED (see FIGS. 14 and 15). The estimated failure mode information 524 is the estimated abnormality cause input in the above-mentioned step S24 (see FIG. 13). The confirmed failure mode information 525 is the confirmed failure mode input in the above-mentioned step S24 (see FIG. 13). The agreement information 526 is information that indicates whether the estimated failure mode information 524 and the confirmed failure mode information 525 agree with each other.

[0083] (Question generation process) FIG. 18 is a diagram showing an example of the additional content data 530. As shown in FIG. When a new symptom that is not included in the existing check items appears as a result of the user 50 performing maintenance work on the maintenance target device 60, a record (row) of the additional content data 530 is added for the additional symptom. In that record, the cause of the appearance of the additional symptom is cause determination result information 535. The additional content data 530 may be manually picked up by the user 50, for example, at regular intervals or after a predetermined number of records have been accumulated.

[0084] Each record of the additional content data 530 includes report ID information 531, network ID information 532, estimation result ID information 533, additional symptom information 534, cause determination result information 535, and match information 536. Of these, the contents of the information 531, 532, and 533 are the same as those of the information 521, 522, and 523 in the maintenance work report data 520 described above.

[0085] The additional symptom information 534 is information that indicates the additional symptom described above. The cause determination result information 535 is the result of an investigation by the user 50 into the cause of the appearance of the additional symptom. The match information 536 is information that indicates whether the inference result data ED related to the inference result ID information 533 matches the cause determination result information 535.

[0086] FIG. 19 is a diagram showing an example of a question screen 610 displayed on the display 22. 19, a question screen 610 includes a question display section 612 (question text) and an answer selection section 614. The question display section 612 displays a question text asking where to start investigating (what is the failure mode) when an abnormality occurs related to a newly added check item. In the illustrated example, a question text is displayed asking where to start investigating when an abnormality such as "the power lamp does not light up and the motor does not turn" occurs.

[0087] The answer selection unit 614 displays answer options for the question displayed in the question display unit 612. In the illustrated example, the options displayed are "1. Power cable," "2. Control panel," and "3. Engine connector," which correspond to failure modes. The user 50 can select one or more options while assigning a probability order to these options, which is the order in which the options are most likely to be the cause of the abnormality displayed in the question display unit 612. The probability information input unit 7 (see FIG. 9) receives the options selected by the user 50 and their probability order.

[0088] FIG. 20 is a diagram showing an example of the additional check item / failure mode correspondence table 540. As shown in FIG. As described above, the check items applied to the Bayesian network information 80 are included in the node content information 140 (see FIG. 2). In contrast, when adding a new check item, the GUI question generation unit 6 (see FIG. 9) generates an added check item / failure mode correspondence table 540 shown in FIG.

[0089] 20, the additional check item / failure mode correspondence table 540 contains multiple records (rows), each corresponding to a new check item to be added. Each record contains a check item number 541, additional check item name information 542, additional check item ID information 543, and link source failure mode information 544 (linked failure mode).

[0090] The check item number 541 is an ascending number assigned to the new check item. The added check item name information 542 is information indicating the name of the new check item. The added check item ID information 543 is the node ID of the new check item. The link source failure mode information 544 is information indicating the node ID of the failure mode linked to the new check item (for example, FM1, FM2, FM3 in the figure).

[0091] Next, the GUI question generation unit 6 creates a check item combination table in the following steps. Step SA1: The number of new check items is n, and the node IDs of these new check items (additional check item ID information 543) are "CK1 to CKn." n-1 Create a table with n rows and n columns. Here, for the cells in the CK1 column, 2 n-2 "N" and 2 n-2 Also, for the cells in the CK2 column, n-3 "N" and 2 n-3 In the same way, substitute "N" and "Y" into each cell up to the column CK(n-1). However, the value of the column CKn is the inverse of the value of the column CK1.

[0092] Step SA2: Move the CK2 column to the left and search for rows where the combination patterns of "N" and "Y" in columns other than CK2 (CK1, CK3 to CKn) overlap. If there are overlapping rows, invert the value of CKn in the row above (i.e., change "N" to "Y" and "Y" to "N").

[0093] Step SA3: Move the CK3 column to the left and search for rows where the combination pattern of "N" and "Y" in columns other than CK3 (CK1, CK2, CK4 to CKn) overlaps. If there are overlapping rows, invert the value of CKn in the row above.

[0094] Step SA4: Thereafter, the columns Ck4 to CKn are sequentially moved to the leftmost position, and the same processing as in steps SA2 and SA3 is repeated.

[0095] FIG. 21 is a diagram showing an example of a specific process for creating a check item combination table. 21 has a number column 205, a CK1 column 201, a CK2 column 202, and a CK3 column 203. The number column 205 indicates pattern numbers in ascending order. For pattern numbers #1 and #2, the CK1 column 201 is "N (no abnormality has occurred)," and for pattern numbers #3 and #4, the CK1 column 201 is "Y (an abnormality has occurred)." That is, the contents of the CK1 column 201 are "N, N, Y, Y" from top to bottom. For these pattern numbers #1 to #4, the CK3 column 203 has values ​​that are the inverse of the CK1 column 201, i.e., "Y, Y, N, N." Furthermore, for pattern numbers #1 to #4, the CK2 column 202 is "N, Y, N, Y" from top to bottom.

[0096] 21 has a number column 215, a CK1 column 211, a CK2 column 212, and a CK3 column 213. This table 210 is obtained by swapping the positions of the rows and columns of table 200. That is, the second and third rows in table 200 have been swapped, so that number column 215 is "1, 3, 2, 4" from top to bottom. Also, CK2 column 212 is located to the left of CK1 column 211. In table 210, the patterns of "CK1, CK3" are, from top to bottom, "N, Y," "Y, N," "N, Y," and "Y, N." That is, the pattern "N, Y" occurs twice, and the pattern "Y, N" also occurs twice.

[0097] Therefore, to prevent the occurrence of identical patterns, the contents of the column located as high as possible, i.e., the CK3 column 213, are changed to generate table 220 in FIG. 21. Table 220 has a number column 225, a CK1 column 221, a CK2 column 222, and a CK3 column 223. In table 220, the pattern of "CK1, CK3" is "N, N" in pattern number #1 and "Y, Y" in pattern number #3 in the second row. The rest of the contents of table 220 are the same as table 210. In table 220, the patterns of "CK1, CK3" are "N, N", "Y, Y", "N, Y", and "Y, N" from top to bottom, with no duplication.

[0098] Table 230 is created by swapping the positions of the columns in table 220. Table 230 has a number column 235, a CK1 column 231, a CK2 column 232, and a CK3 column 233. The contents of these CK1 column 231, CK2 column 232, and CK3 column 233 are the same as the CK1 column 221, CK2 column 222, and CK3 column 223 of table 220.

[0099] In table 230, the patterns of "CK1,CK2" are "N,N", "Y,N", "N,Y", and "Y,Y" from top to bottom, and there are no overlaps. If there are no overlaps, proceed to the next step without making any changes.

[0100] The positions of the columns in this table 230 are swapped to create a check item table 240. The check item table 240 has a number column 245, a CK1 column 241, a CK2 column 242, and a CK3 column 243. The contents of these CK1 column 241, CK2 column 242, and CK3 column 243 are the same as the CK1 column 231, CK2 column 232, and CK3 column 233 in table 230.

[0101] In check item table 240, the patterns for "CK2, CK3" are, from top to bottom, "N,N," "N,Y," "Y,Y," and "Y,N," with no overlaps. If no overlaps occur, proceed to the next step without making any changes. However, because the column swapping has been completed in this check item table 240, check item table 240 is the final check item table.

[0102] FIG. 22 is a flowchart of the check item table generation processing routine. 22, when the process proceeds to step S26, the GUI question generation unit 6 acquires the additional content data 530 (see FIG. 18). That is, a predetermined question screen (not shown) is displayed on the display 22, and the additional content data 530 is acquired in accordance with the input result by the user 50.

[0103] Next, when the process proceeds to step S27, the GUI question generation unit 6 generates a check item table (for example, check item table 240 in FIG. 21) that combines "N (no abnormality has occurred)" and "Y (an abnormality has occurred)" for each check item.

[0104] Next, when the process proceeds to step S28, the GUI question generator 6 generates a GUI question based on the new check items to be added and the failure modes linked to these check items. This GUI question is, for example, the text in the question display section 612 on the question screen 610 (see FIG. 19). In step S28, a question is generated for each combination in accordance with all combinations in the check item table 240.

[0105] If the status of the new check item in the check item table 240 is "Y (an abnormality has occurred)", the question will be, for example, "The power light is not on." If the status of the new check item is "N (no abnormality has occurred)", the question will be, for example, "The power light is on."

[0106] (Question answer processing) FIG. 23 is a flowchart of the question answer processing routine. When the user 50 selects one or more options from the options displayed in the answer selection section 614 (see FIG. 19) while assigning an order to the options, this routine is started in the GUI question generation section 6. When the process proceeds to step S29 (information acquisition step) in FIG. 23, the GUI question generation section 6 creates, for example, a failure mode ranking table 560 shown in FIG. 24 based on the answers from the user 50.

[0107] FIG. 24 is a diagram showing an example of the failure mode ranking table 560. As shown in FIG. The failure mode ranking table 560 includes check item combination information 561 and ranking information (occurrence probability correspondence information) 562. The contents of the check item combination information 561 are the same as those of the check item table 240 (see FIG. 21).

[0108] Furthermore, the ranking information 562 is information indicating ranking values ​​such as "1," "2," and "3" assigned to the node IDs of the failure modes (for example, FM1, FM2, and FM3 in the figure). The larger the ranking value, the higher the possibility that the failure mode has occurred. For example, the ranking value "3" may be assigned to the first place in the ranking assigned by the user 50, "2" to the second place, and "1" to the third place. Then, when a request to add a check item is input, the ranking information 562 may be updated to include the occurrence probability of the failure corresponding to the requested check item.

[0109] Returning to FIG. 23, when the process proceeds to step S30 (probability information generation step), the probability information generation unit 8 (see FIG. 9) generates a probability information table 570 (see FIG. 25). FIG. 25 is a diagram showing an example of the probability information table 570. The probability information table 570 shows the probability that failure modes FM1, FM2, and FM3 will be "Y (an abnormality has occurred)" when check items CK1, CK2, and CK3 are "Y (an abnormality has occurred)" or "N (an abnormality has not occurred)."

[0110] Here, when check item CKp (where p=1, 2, 3, ...) is "Y (an abnormality has occurred)", the probability that failure mode FMq (where q=1, 2, 3, ...) will be "Y" is expressed as "probability information P(CKp=Y|FMq=Y)".

[0111] The contents of the probability information table 570 are the average values ​​of the ranking values ​​in the failure mode ranking table 560. For example, in the table 560, the check item CK1 is "N" in the first and second rows. In these rows, the failure mode FM1 is "1" and "1", and the average value of both is "1". Therefore, the probability information P(CK1=N|FM1=N) shown in the probability information table 570 is "1".

[0112] In addition, in table 560, check item CK1 is "Y" in the third and fourth rows. In these rows, failure mode FM1 is "2" and "3", and the average value of both is "2.5". Therefore, the probability information P(CK1=Y|FM1=Y) shown in probability information table 570 is "2.5".

[0113] When the probability that the state of each check item CKp will be "N" is higher than the probability that it will be "Y," the values ​​of the "probability information P(CKp=Y|FMq=Y)" to be used and the "probability information P(CKp=N|FMq=Y)" to be used may be predetermined values. In this way, according to this embodiment, the probability information P is determined based on the average value of the ranking values ​​corresponding to the check items requested to be added. In other words, this operation is an operation of estimating the probability of a failure occurring based on information regarding the occurrence probability of the check items (for example, ranking values ​​in descending order of occurrence probability).

[0114] After the probability information table 570 shown in Figure 25 is completed, the probability information generation unit 8 modifies the probability information table 570 so that the sum of the probability information P(CKp=Y|FMq=Y) and the probability information P(CKp=N|FMq=Y) becomes "1". 25, the probability information P(CK1=Y|FM1=Y) is "2.5", the probability information P(CK1=N|FM1=Y) is "1.0", and the ratio between the two is "5:2". Therefore, the probability information generator 8 corrects the probability information P(CK1=Y|FM1=Y) to "5 / 7=0.714" and the probability information P(CK1=N|FM1=Y) to "2 / 7=0.286".

[0115] Returning to Figure 23, when the process proceeds to step S31, the probability information generation unit 8 updates the asset knowledge database 1 based on the contents of the probability information table 570 (see Figure 25). That is, the probability information generation unit 8 adds new records corresponding to the new check items to the above-mentioned node content information 140 (see Figure 2), link information 150 (see Figure 3), maintenance knowledge data 710 (see Figure 4), failure mode occurrence probability table 110 (see Figure 6), and child node abnormality occurrence probability tables 120, 130 (see Figures 7 and 8). This completes the processing of this routine.

[0116] In this way, the probability information generation unit 8 updates the asset knowledge database 1 based on the question answer processing routine (FIG. 23). That is, after maintenance work is performed, the probability information generation unit 8 updates the failure probability information corresponding to each check item. Therefore, compared to the previous information, the Bayesian network information 80 generated when the diagnostic net generation unit 2 (see FIG. 9) is executed will include new check items CK1, CK2, and CK3 and new link information connected to these. Therefore, when a request to add a check item is input, the probability information generation unit 8 updates the ranking information to include the occurrence probability of the failure corresponding to the requested check item.

[0117] [Second embodiment] Next, a second embodiment will be described. 26 is a block diagram of an equipment abnormality cause estimating system 10 according to the second embodiment. In the following description, parts corresponding to those in the first embodiment described above are given the same reference numerals, and their description may be omitted. 26, the equipment abnormality cause estimating system 10 of this embodiment differs from that of the first embodiment (see FIG. 9) in that an update effect evaluating unit 9 is provided.

[0118] The update effect evaluation unit 9 reads out the maintenance work report data 520 (see FIG. 17) from the maintenance work report database 5. Then, it compares the accuracy rate of the failure mode based on the Bayesian network information 80 after updating by applying the probability information table 570 (see FIG. 25) with the accuracy rate of the failure mode based on the Bayesian network information 80 before updating, and presents the result to the user 50 via the input / output unit 20.

[0119] If the accuracy rate of the updated Bayesian network information 80 is lower than the accuracy rate of the Bayesian network information 80 before the update, the update effect evaluation unit 9 returns the contents of the asset knowledge database 1 to the state before the probability information table 570 was applied. As a result, the Bayesian network information 80 before the update will be applied again from then on. On the other hand, if the accuracy rate of the updated Bayesian network information 80 is higher than the accuracy rate of the Bayesian network information 80 before the update, the update effect evaluation unit 9 maintains the contents of the asset knowledge database 1 in its current state (the state in which the probability information table 570 has been applied).

[0120] As a result, thereafter, the updated Bayesian network information 80 is applied again. Therefore, according to this embodiment, it is possible to prevent the performance of the Bayesian network information 80 from deteriorating when the probability information table 570 is inappropriate.

[0121] [Effects of the embodiment] As described above, according to each of the above-described embodiments, the maintenance system (10) is characterized by having a storage unit (1, 3, 982) that stores check items for the equipment (60) to be maintained and the occurrence probabilities (113, 125, 133) of failures detected by the check items, an input unit (24) that receives input of a request to add a check item, a question generation unit (6) that generates questions to inquire of a maintenance technician about the most likely causes of the failure, and an occurrence probability estimation unit (8) that estimates the occurrence probability of the failure corresponding to the added check item based on the answers to the questions. This makes it possible to easily update various information such as the occurrence probability (P).

[0122] It is more preferable that the question generator 6 generates questions to inquire about the most likely cause of a failure using the data on the causes of the abnormality, the check items, and the check results stored in the maintenance work report database 5. This allows various information such as the occurrence probability (P) to be updated more appropriately using the data on the causes of the abnormality, the check items, and the check results.

[0123] It is even more preferable to further provide a display unit (22) that displays the check items in descending order of the probability of a malfunction, thereby enabling the user to easily recognize check items with a high probability of a malfunction.

[0124] Furthermore, it is more preferable that the storage unit (1,3,982) stores ranking information (562) relating to the probability of occurrence (P) of a failure corresponding to a check item, and when a request for adding a check item is input, the ranking information (562) is updated to include the probability of occurrence of a failure corresponding to the requested check item. In this way, when a request for adding a check item is input, the ranking information (562) can be updated to include the probability of occurrence of a failure corresponding to the requested check item.

[0125] Furthermore, it is more preferable that the occurrence probability estimation unit (8) updates the occurrence probability (P) of a failure corresponding to each check item after the maintenance work is performed, so that the results of the maintenance work can be reflected in the occurrence probability (P).

[0126] Furthermore, it is more preferable that the occurrence probability estimation unit (8) derives the failure occurrence probability (P) by statistically processing the accumulation of past causes and results, which allows for a more appropriate occurrence probability (P).

[0127] Furthermore, it is more preferable that the occurrence probability estimation unit (8) derives the failure occurrence probability (P) using Bayesian network information (80), thereby making it possible to effectively use various data contained in the Bayesian network information (80).

[0128] Also, from another viewpoint, according to each of the above-described embodiments, the abnormality cause estimation device (10) includes: an information group generation unit (2) that generates an information group (80) including a plurality of failure cause information (831 to 834) corresponding to a plurality of failure causes and check item information (841 to 844) corresponding to a plurality of check items, based on abnormality occurrence probability data (120) describing probability information P that an abnormality will occur in each check item when each of the failure causes occurs, for each combination of a plurality of failure causes in the maintenance target device 60 and a plurality of check items that are items to be checked; an information acquisition unit (6) that acquires new check items that are new check items to be added based on an operation on the input / output unit 20, and acquires occurrence probability correspondence information (562) for the new check items that corresponds to the magnitude of the occurrence probability of each of the plurality of failure causes, based on the operation on the input / output unit 20; and a probability information generation unit 8 that generates probability information P for the new check items based on the occurrence probability correspondence information (562) and adds the check item information related to the new check items to the abnormality occurrence probability data (120). This allows the probability information generating unit 8 to add a new check item to the abnormality occurrence probability data 120 by simply specifying the new check item to be added and the occurrence probability correspondence information (562) of multiple failure modes for the new check item, thereby allowing the information group 80 to be easily updated.

[0129] Furthermore, it is more preferable that the information acquisition unit (6) generates a check item table 240 including at least one pattern among the combinations of the multiple new check items, and generates a question (612) based on the generated check item table 240. This allows the user to input occurrence probability correspondence information (562) based on the question (612), making it possible to update the information group (80) even more easily.

[0130] Furthermore, it is more preferable that the abnormality cause estimation device (10) further comprises a probability information input unit 7 that receives, as a response to the question sentence (612), an order (probability order) of multiple failure causes corresponding to multiple new check items from the input / output unit 20. This allows the user to input the relationship between the new check items and the failure causes by a simple operation of assigning an order to the new check items, thereby making it possible to update the information group (80) even more easily.

[0131] Furthermore, it is more preferable that the probability information generating unit 8 calculates the probability of each failure cause occurring when the new check item is in an abnormal state and the probability of each failure cause occurring when the new check item is in a normal state based on the average value of the occurrence probability correspondence information (562) of each failure cause when the new check item is in an abnormal state and the average value of the occurrence probability correspondence information (562) of each failure cause when the new check item is in a normal state, based on the occurrence probability correspondence information (562) corresponding to the order (probability order) input from the probability information input unit 7. This allows the user to generate various types of probability information by a simple operation of assigning an order (probability order) to the failure causes corresponding to the new check items, making it even easier to update the information group (80).

[0132] Furthermore, as in the second embodiment, it is more preferable that the anomaly cause estimation device (10) further comprises an update effect evaluation unit 9 that calculates the accuracy rate of the updated information group (80) based on the maintenance work history of the maintenance target equipment 60 accumulated in the maintenance work report database 5, presents the calculated accuracy rate via the input / output unit 20, and adopts the updated information group (80) as the information group (80) to be applied in the future if the accuracy rate of the updated information group (80) is higher. This makes it possible to prevent a decrease in the performance of the information group (80) when the probability determined by the probability information generation unit 8 in response to a new check item is inappropriate.

[0133] [Variations] The present invention is not limited to the above-described embodiments and various modifications are possible. The above-described embodiments are provided as examples to facilitate understanding of the present invention and are not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to delete part of the configuration of each embodiment, or to add or replace other configurations. Furthermore, the control lines and information lines shown in the figures are those considered necessary for explanation, and do not necessarily represent all control lines and information lines necessary for the product. In reality, it is acceptable to consider that almost all components are interconnected. Possible modifications of the above-described embodiments include, for example, the following:

[0134] (1) The hardware of the equipment abnormality cause estimation system 10 in the above embodiment can be realized by a general-purpose computer. Therefore, the flowcharts shown in the figures and programs for executing the various processes described above may be stored in a storage medium (a computer-readable storage medium on which a program is recorded) such as a hard disk, a solid-state drive (SSD), a flash memory card, or a digital versatile disk (DVD), or may be distributed via a transmission path.

[0135] (2) In the above embodiment, the processes shown in the flowcharts and the other processes described above are described as software processes using a program, but some or all of them may be replaced with hardware processes using an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array), etc.

[0136] (3) The various processes executed in the above embodiment may be executed by a server computer via a network (not shown), and the various data stored in the above embodiment may also be stored in the server computer.

[0137] (4) In the above embodiment, the "check items" are items that are checked by the user 50. However, the equipment abnormality cause estimation system 10 may automatically collect the check items. [Explanation of symbols]

[0138] 1. Asset knowledge database (storage section) 3 Maintenance knowledge Bayesian network database (memory section) 5 Maintenance work report database 6 GUI question generation section (question generation section) 8. Probability information generation unit (occurrence probability estimation unit) 10 Equipment abnormality cause estimation system (maintenance system) 22 Display (display unit) 24 Input device (input section) 60 Equipment subject to maintenance 80 Bayesian Network Information (Information Group) 113,125,133 Probability information (occurrence probability) 125,133 (Occurrence probability) P Probability information (occurrence probability) 562 Ranking information (occurrence probability information) 612 Question display section (question) P Probability information

Claims

1. a storage unit that stores check items for a maintenance target device and the occurrence probability of a failure detected by the check items; an input unit that receives an input of a request to add the check item; a question generation unit that generates a question to inquire of a maintenance person about a likely cause of the failure; an occurrence probability estimation unit that estimates the occurrence probability of a failure corresponding to the check item to be added based on the answer to the question; A maintenance system characterized by:

2. 2. The maintenance system according to claim 1, The question generation unit generates the question for inquiring about the most likely cause of the failure, using the abnormality causes stored in the maintenance work report database, the check items, and check result data. A maintenance system characterized by:

3. 2. The maintenance system according to claim 1, A display unit is further provided to display the check items in descending order of the probability of occurrence of the fault. A maintenance system characterized by:

4. 2. The maintenance system according to claim 1, The storage unit stores ranking information regarding the probability of occurrence of a failure corresponding to the check item, and when a request to add the check item is received, the storage unit updates the ranking information to include the probability of occurrence of a failure corresponding to the requested check item. A maintenance system characterized by:

5. 2. The maintenance system according to claim 1, The occurrence probability estimation unit updates the occurrence probability of the failure corresponding to each of the check items after the maintenance work is performed. A maintenance system characterized by:

6. 6. The maintenance system according to claim 1, The occurrence probability estimation unit statistically processes the accumulation of causes and effects that have occurred in the past to derive the occurrence probability of the failure. A maintenance system characterized by:

7. 6. The maintenance system according to claim 1, The occurrence probability estimation unit derives the occurrence probability of the failure by utilizing Bayesian network information. A maintenance system characterized by:

8. a step of storing check items for the equipment to be maintained and the occurrence probability of failures detected by the check items in a storage unit; a step of displaying the check items on a display unit based on the occurrence probability of the fault; estimating the probability of occurrence of the fault; and a step of receiving an input of a request to add the check item via an input unit, The step of estimating the occurrence probability is a step of estimating the occurrence probability of the fault corresponding to the check item requested to be added and adding the probability to the storage unit. A maintenance method characterized by:

Images