Maintenance assistance system and maintenance assistance procedure

Abstract

A maintenance assistance system comprises: a maintenance log information storage unit configured to store maintenance log information obtained by diagnosing a diagnostic target device in the past, wherein the maintenance log information comprises data from several data elements and defective component information; a priority learning unit configured to split the maintenance log information stored in the maintenance log information storage unit according to a split condition predetermined based on the data element, to learn a relationship between the data element and a defective component in the maintenance log information for each split, and to generate a priority level for the data element for each split condition;a similar case extraction unit configured to split diagnostic data for diagnosing the diagnostic target device according to the split condition and to extract a similar case, which is similar to the diagnostic data, from the maintenance log information storage unit based on a severity level of the data element associated with the split condition; and a defective component estimation unit configured to estimate a defective component of the diagnostic target device based on the similar case extracted by the similar case extraction unit.

Description

Technical area

[0001] The present disclosure relates to a maintenance assistance system and a maintenance assistance procedure. background

[0002] In recent years, a maintenance assistance system has been known which presents candidates for replacement components using collected data, which includes past malfunctions and components replaced in response to the malfunctions (see, for example, patent document 1). Citation list patent document

[0003] Patent document 1: Japanese unexamined patent application, first publication number 2020 - 009 068 Brief description of the invention: Technical problem

[0004] However, in the maintenance assistance system described above, which represents the relevant state of the art, candidates for replacement components are selected without considering the difference in failure tendency for each data element, such as regionality. Therefore, the problem with this maintenance assistance system, according to the relevant state of the art, is that the failure tendency for each data element, such as regionality, is overlooked, and the accuracy of failure diagnosis is reduced.

[0005] The present disclosure was made to solve the problem described above, and one objective of the present disclosure is to provide a maintenance assistance system and a maintenance assistance procedure which can improve the accuracy of fault diagnosis. Solution to the problem

[0006] To achieve the aforementioned goal, according to one aspect of the present disclosure, a maintenance assistance system is provided which comprises: a maintenance log information storage unit configured to store maintenance log information obtained by diagnosing a diagnostic target device in the past, wherein the maintenance log information comprises data from several data elements and defective component information; a priority learning unit configured to split the maintenance log information stored in the maintenance log information storage unit according to a split condition predetermined based on the data element, to learn a relationship between the data element and a defective component in the maintenance log information for each split, and to generate a priority level of the data element for each split condition;a similar case extraction unit configured to classify diagnostic data for diagnosing the diagnostic target device according to the split condition, and to extract a similar case from the maintenance log information storage unit based on the importance level of the data element associated with the classification; and a defective component estimation unit configured to estimate a defective component of the diagnostic target device based on the similar case extracted by the similar case extraction unit.

[0007] Furthermore, according to another aspect of the present disclosure, a maintenance assistance method for a maintenance assistance system is provided, comprising a maintenance log information storage unit configured to store maintenance log information obtained by diagnosing a diagnostic target device in the past, wherein the maintenance log information comprises data from several data elements and defective component information. The maintenance assistance method includes: causing a priority learning unit to split the maintenance log information stored in the maintenance log information storage unit according to a split condition predetermined based on the data element, learning a relationship between the data element and a defective component in the maintenance log information for each split, and generating a priority level of the data element for each split condition.To cause a similar-case extraction unit to classify diagnostic data for diagnosing the diagnostic target device according to the split condition and to extract a similar case, which is similar to the diagnostic data, from the maintenance log information storage unit based on a degree of importance of the data element assigned to the classification; and to cause a defective-component estimation unit to estimate a defective component of the diagnostic target device based on the similar case extracted by the similar-case extraction unit. Advantageous effects of the invention

[0008] According to the present disclosure, it is possible to improve the accuracy of a fault diagnosis. Brief description of the drawings Fig. 1 A functional block diagram showing an example of a maintenance assistance system according to the present embodiment. Fig. 2 A diagram showing an example of data in a maintenance log information storage unit in the present embodiment. Fig. 3 A diagram showing an example of data in a weight storage unit in the present embodiment. Fig. 4 A diagram showing an example of data in an estimation result storage unit in the present embodiment. Fig. 5 A diagram showing an example of data in an output information storage unit in the present embodiment. Fig. 6 A flowchart showing an example of a model learning process of a model learning device according to the present embodiment. Fig. 7 A flowchart showing an example of a weight learning process of the model learning device according to the present embodiment. Fig. 8 A flowchart showing an example of a diagnostic process of a diagnostic device according to the present embodiment. Fig. 9 A diagram showing a hardware configuration of the diagnostic device and the model learning device of the maintenance assistance system according to the present embodiment. Description of embodiments

[0009] A maintenance assistance system and a maintenance assistance method according to an embodiment of the present disclosure are described below with reference to the drawings.

[0010] Fig. Figure 1 is a functional block diagram showing an example of a maintenance assistance system 1 according to the present embodiment. As shown in Fig. As shown in Figure 1, the maintenance assistance system 1 according to the present embodiment comprises a diagnostic device 10, several diagnostic target devices 20, several maintenance terminals 30 and a model learning device 40.

[0011] In the present embodiment, in the following description, of the several diagnostic target devices 20, a device that has been diagnosed in the past or a device that is in normal operation is referred to as a diagnostic target device 21, and a device that is currently to be diagnosed is referred to as a diagnostic target device 22. Furthermore, in the following description, in the maintenance assistance system 1, if a diagnostic target device is specified or if a diagnostic target device is not specifically distinguished, the diagnostic target device is referred to as the diagnostic target device 20.

[0012] Furthermore, in the present embodiment, in the following description, of the several maintenance terminals 30, a terminal that has transmitted maintenance log information, which is the past diagnostic result, is referred to as a maintenance terminal 31, and a terminal that is currently performing a diagnosis is referred to as a maintenance terminal 32. Additionally, in the maintenance assistance system 1, if a maintenance terminal is specified or if a maintenance terminal is not specifically distinguished, the maintenance terminal is referred to as maintenance terminal 30.

[0013] Furthermore, the diagnostic device 10, the multiple diagnostic target devices 21, the multiple maintenance terminals 30 (31, 32) and the model learning device 40 can be connected to a network NW1 and can communicate with each other via the network NW1.

[0014] In addition, the diagnostic target device 22 and the maintenance terminal 32 can be connected via a network NW2 and communicate with each other via the network NW2.

[0015] Network NW1, for example, is a wide area network (WAN). Furthermore, network NW2, for example, is a local area network (LAN) in a building where diagnostic target device 22 is installed.

[0016] The diagnostic target device 20 (21, 22) is, for example, a household appliance, such as an air conditioner. The diagnostic target device 20 (21, 22) is a device that is to be subjected to a fault diagnosis.

[0017] The maintenance terminal 30 (31, 32) is a terminal device for maintaining the diagnostic target device 20 and is, for example, a smartphone, a tablet terminal, a mobile PC (mobile personal computer), or the like. The maintenance terminal 30 (31, 32) is a device for diagnosing the diagnostic target device 20 when a maintenance service provider diagnoses and maintains the diagnostic target device 20 at the site or before the maintenance service provider goes to the site.

[0018] Furthermore, the maintenance terminal 32 is a terminal which diagnoses the diagnostic target device 22 to be diagnosed (maintained) and includes a network (NW) communication unit 321, an input unit 322, a display unit 323, a terminal storage unit 324 and a terminal control unit 325.

[0019] The NW communication unit 321, for example, is a functional unit implemented by a communication device, such as a network adapter. The NW communication unit 321 is connected to the NW2 network and can communicate with the diagnostic target device 22. Furthermore, the NW communication unit 321 is connected to the NW1 network and can communicate with, for example, the diagnostic device 10.

[0020] The input unit 322 is an input device, such as a keyboard, keypad, or button. The input unit 322 receives various types of input information in response to user operation (service provider). For example, the service provider uses the input unit 322 to input diagnostic data. This diagnostic data includes, for example, the model name, years since installation, installation area, malfunction symptoms, and similar information about the diagnostic target device 22.

[0021] The display unit 323 is, for example, a display device, such as a liquid crystal display. The display unit 323 shows, for example, an input screen for entering diagnostic data and output information received from the diagnostic device 10 described below. In this case, the output information is, for example, a diagnostic result for the diagnostic data and is a candidate for a defective component or the like.

[0022] The terminal storage unit 324 stores various types of information used by the maintenance terminal 32. For example, the terminal storage unit 324 stores input information from the input unit 322, information displayed on the display unit 323, information transmitted to and received from the diagnostic device 10, and the like.

[0023] The terminal control unit 325, for example, is a functional unit implemented by instructing a processor, which includes a central processing unit (CPU), to execute a program. The terminal control unit 325 transmits, for example, the diagnostic data received by the input unit 322 to the diagnostic device 10 via network NW1. Furthermore, the terminal control unit 325 transmits, for example, operating data obtained from the diagnostic target device 22 via network NW2 to the diagnostic device 10 via network NW1. Finally, the terminal control unit 325 displays the output information received from the diagnostic device 10 via network NW1 on the display unit 323.

[0024] Furthermore, the operating data described above includes detection data from various sensors (not shown) contained in the diagnostic target device 22, fault code information, and the like.

[0025] The model learning device 40, for example, is a server device that can be connected to the network NW1. The model learning device 40 performs a weight learning process and a process for learning a defect component detection model. Furthermore, the model learning device 40 includes an NW communication unit 41, a learning storage unit 42, and a learning processing unit 43.

[0026] The NW communication unit 41 is a functional unit implemented by a communication device, such as a network adapter. The NW communication unit 41 is connected to the NW1 network and can communicate with the diagnostic target device 21, the maintenance terminal 31, and the diagnostic device 10.

[0027] The learning memory unit 42, for example, is a storage device, such as RAM, flash memory, or a hard disk drive (HDD), and stores various types of information used by the model learning device 40. The learning memory unit 42 comprises a maintenance log information storage unit 421, an operational data storage unit 422, a weight storage unit 423, and a model storage unit 424.

[0028] The maintenance log information storage unit 421 stores maintenance log information collected from several maintenance terminals 31. This maintenance log information is, for example, a maintenance work report generated by the maintenance service provider. The maintenance log information storage unit 421 also stores, for example, maintenance log information obtained in the past by diagnosing the diagnostic target device 21, which includes data from a variety of data elements and defective component information. With reference to Fig. 2 an example of data in the maintenance log information storage unit 421 is described.

[0029] Fig. Figure 2 is a diagram showing an example of the data in the maintenance log information storage unit 421 in the present embodiment. As shown in Fig. As shown in Figure 2, the maintenance log information storage unit 421 stores maintenance log information in which a number (Nr), a model name, years since installation, a range, a symptom, a replacement component P1 and a replacement component P2 are associated with each other.

[0030] In Fig. 2 is an example of individual identification information for the diagnostic target device 21 (20). Furthermore, the model name specifies the model name of the diagnostic target device 21 (20). The model name is also an example of device identification information for identifying the diagnostic target device 21 (20). Additionally, the years since installation and the area indicate the number of years (duration) and the area where the diagnostic target device 21 (20) is installed. Furthermore, the symptom indicates a symptom of a malfunction or failure when the diagnostic target device 21 (20) was diagnosed in the past. Additionally, the replacement component P1 and the replacement component P2 indicate components that were replaced during past maintenance. Furthermore, the model name, the years since installation, the area, and the symptom correspond to data elements.

[0031] For example, there is in the Fig. In the second example shown, the maintenance log information assigned to number "1" indicates that the model name is "MSZXXX01S" and the number of years since installation is "5" (5 years). Furthermore, the maintenance log information indicates that the region is "Tokyo" and the symptom is "not cold." The maintenance log information also indicates that replacement component P1 is a "compressor" and replacement component P2 is an "expansion valve."

[0032] Returning to the description of Fig. 1. The operating data storage unit 422 stores the operating data collected by each diagnostic target device 21 (20). This operating data includes detection data from various sensors (not shown) contained in each diagnostic target device 21 (20), fault code information, and the like. For example, the operating data storage unit 422 stores the number and model name described above, along with the operating data, in association with each other.

[0033] The weight storage unit 423 (an example of an importance-degree storage unit) divides the maintenance log information stored in the maintenance log information storage unit 421 according to a division condition, which is predetermined based on the data element, and stores a learning result obtained by learning a relationship between the data element and the defective component in the maintenance log information for each division. Furthermore, the learning result specifies a weight (importance degree) of the data element for each division condition. In the present case, with reference to Fig. 3 an example of data in the weight storage unit 423 is described.

[0034] Fig. Figure 3 is a diagram showing an example of the data in the weight storage unit 423 in the present embodiment. As in Fig. As shown in Figure 3, the weight storage unit 423 stores the data elements and the weights in association with each other for each partition. The data elements include, for example, a model, a range, a capacity range, elapsed years, and the like.

[0035] In the Fig. In example 3, the data element's area is divided into subdivision A of a coastal area and subdivision B of an inland area. The subdivision condition is that the data element's area must be either the coastal area or the inland area.

[0036] In allocation A (coastal area), the weight of the model is "0.12", the weight of the area is "0.83", the weight of the capacity area is "0.26", and the weight of elapsed years is "0.38". Furthermore, in allocation B (inland area), the weight of the model is "0.34", the weight of the area is "0.34", the weight of the capacity area is "0.44", and the weight of elapsed years is "0.59".

[0037] Furthermore: The larger the value of the respective weight, the higher the level of importance. The smaller the value of the respective weight, the lower the level of importance.

[0038] Returning to the description of Fig. 1 The model storage unit 424 stores the defect component detection model, which is a result of training using maintenance log information and operational data as training data. The defect component detection model is, for example, an estimation model that estimates a defective component from a similar case of malfunction (failure) of the diagnostic target device 22 (20).

[0039] The learning processing unit 43, for example, is a functional unit implemented by instructing a processor, comprising a CPU, to execute a program. The learning processing unit 43 performs a learning process of learning the weight of the data element for each partition and learning the defect component detection model. The learning processing unit 43 comprises a maintenance log information collection unit 431, an operational data collection unit 432, a weight learning unit 433, and a model learning unit 434.

[0040] The maintenance log information collection unit 431 collects the maintenance log information from the maintenance terminal 31 (30) and stores the collected maintenance log information in the maintenance log information storage unit 421. The maintenance log information collected by the maintenance log information collection unit 431 is used as training data for learning the weight of the data element and the defect component detection model.

[0041] The operational data collection unit 432 collects the operational data from the diagnostic target device 21 (20) and stores the collected operational data in the operational data storage unit 422. The operational data collected by the operational data collection unit 432 can be used as part of the training data for training the weight of the data element and the defect component detection model.

[0042] The weight learning unit 433 (an example of an importance learning unit) divides the maintenance log information stored in the maintenance log information storage unit 421 according to a division condition predetermined based on the data element, learns the relationship between the data element and the defective component in the maintenance log information for each division, and generates the weight (importance) of the data element for each division condition. The weight learning unit 433 generates the weight of the data element for each division condition using a machine learning method, such as LightGBM, which calculates the importance of each data element using the past maintenance log information stored in the maintenance log information storage unit 421 and the operational data stored in the operational data storage unit 422 as the training data.Furthermore, the weight learning unit 433 can generate the weight of the data element for each splitting condition using the maintenance log information as the learning data, without using the operational data.

[0043] The splitting condition is, for example, whether the data element of the Fig. The area shown in Figure 3 is either the coastal area or the inland area. In this case, the weight learning unit 433 generates the weight of each data element in the coastal area (split A) and the weight of each data element in the inland area (split B).

[0044] This means that the weight learning unit 433 divides the learning data depending on whether the area in which the diagnostic target device 21 (20) is installed is the coastal area (Division A) or the inland area (Division B), and performs the learning process on each learning data element to generate the weight of each data element.

[0045] Furthermore, the maintenance service provider determines the allocation condition in advance, taking into account the maintenance log information and the operating data, for example the data element of the area, which has a major influence on the defective component (replacement component).

[0046] The weight learning unit 433 stores the generated weight of the data element for each distribution condition in the weight storage unit 423. Furthermore, the weight learning unit 433 transmits, for example, the weight of the data element for each distribution condition, which is stored in the weight storage unit 423, via the NW communication unit 41 to the diagnostic device 10.

[0047] The model learning unit 434 learns the maintenance log information as the training data and generates the defective component detection model, which estimates a defective component from similar cases. For example, the model learning unit 434 generates the defective component detection model using a machine learning method, such as LightGBM or a support vector machine (SVM), with the past maintenance log information stored in the maintenance log information storage unit 421 and the operational data stored in the operational data storage unit 422 as the training data. Alternatively, the model learning unit 434 can generate the defective component detection model using the maintenance log information as the training data without using the operational data.

[0048] The model learning unit 434 stores the generated defect component detection model in the model storage unit 424. Furthermore, the model learning unit 434 transmits, for example, the defect component detection model stored in the model storage unit 424 to the diagnostic device 10 via the NW communication unit 41.

[0049] Diagnostic device 10, for example, is a server device that can be connected to network NW1. Diagnostic device 10 estimates a defective component using the defect component detection model generated by model learning device 40 and the weight (importance level) for each data element. Diagnostic device 10 estimates the defective component of the diagnostic target device 22 using the diagnostic and operational data obtained from maintenance terminal 32 via network NW1 as input data. Furthermore, diagnostic device 10 generates a display screen as output information based on the defective component estimate and transmits the display screen to maintenance terminal 32 via network NW1.

[0050] Furthermore, the diagnostic device 10 includes a network communication unit 11, a device storage unit 12 and a diagnostic processing unit 13.

[0051] The NW communication unit 11 is a functional unit implemented by a communication device, such as a network adapter. The NW communication unit 11 is connected to the NW1 network and can communicate with the maintenance terminal 32 and the model learning device 40.

[0052] The device storage unit 12, for example, is a storage device, such as RAM, flash memory, or HDD, and stores various types of information used by the diagnostic device 10. The device storage unit 12 comprises a diagnostic data storage unit 121, an operational data storage unit 122, a weight storage unit 123, a model storage unit 124, a similar-case storage unit 125, an estimation result storage unit 126, and an output information storage unit 127.

[0053] The diagnostic data storage unit 121 stores the diagnostic data obtained from the maintenance terminal 32. The diagnostic data has the same data elements as the input data, except for the elements of the replacement components in the maintenance log information, which is used by the model learning device 40 in the learning process. For example, the diagnostic data storage unit 121 stores diagnostic data of the data elements such as the model, the area, the capacity range, and the elapsed years.

[0054] The operating data storage unit 122 stores the operating data of the diagnostic target device 22, which is obtained from the maintenance terminal 32. The operating data has the same data elements as the operating data used by the model learning device 40 in the learning process. For example, the operating data includes data elements such as detection data (indoor temperature, indoor humidity, outdoor temperature, outdoor humidity, and the like) from various sensors and an error code.

[0055] The weight storage unit 123 stores the weight of the data element for each partitioning condition, which is obtained from the model learning device 40. The weight storage unit 123 stores, for example, the same information as the one in Fig. 3 Weight storage unit 423 shown.

[0056] Model storage unit 124 stores the defect component detection model, which is obtained from model learning device 40. Model storage unit 124 stores the same information as model storage unit 424 of model learning device 40.

[0057] The Similar Case Storage Unit 125 stores information on past similar cases, which are extracted by a Similar Case Extraction Unit 133, described below. The Similar Case Storage Unit 125 stores a large number of similar cases (for example, about 100 similar cases), which are extracted using the weight of the data element obtained from the model learning device 40.

[0058] The estimation result storage unit 126 stores candidates for the defective components of the diagnostic target device 22, which are estimated by a defect component estimation unit 134, described below. The estimation result storage unit 126 stores information in which the candidates for the defective components estimated by the defect component estimation unit 134 and their failure probabilities (likelihoods) are associated with each other, and an average value of the failure probabilities of all the similar cases as the estimation result, which is estimated for each similar case using the defect component detection model. Hereinafter, with reference to Fig. 4 An example of data in the estimation result storage unit 126 is described.

[0059] Fig. Figure 4 is a diagram showing an example of the data in the estimation result storage unit 126 in the present embodiment. As in Fig. As shown in Figure 4, the estimation result storage unit 126 stores the candidates for the defective components and the failure probabilities in association with each other for each similar case. Furthermore, the estimation result storage unit 126 stores the average value of the failure probabilities of all similar cases.

[0060] In the Fig. In the four examples shown, for instance, in case study 1 of the similar cases, the candidates for the defective components are a compressor, a four-way valve, a coil, a fan motor, and an electronic substrate, and their failure probabilities are 52.00%, 3.70%, 24.00%, 9.20%, and 4.00%, respectively. Furthermore, the average failure probabilities in case studies 1 to N of the similar cases are 58.20%, 2.70%, 23.10%, 11.60%, and 4.10%, respectively.

[0061] Returning to the description of Fig. 1. The output information storage unit 127 also stores output information from an output information generation unit 135, which is described below. For example, the output information storage unit 127 stores output information which is based on the in Fig. The 5 shown estimate result is based on this.

[0062] Fig. Figure 5 is a diagram showing an example of data in the output information storage unit 127 in the present embodiment. As shown in Fig. As shown in Figure 5, the output information storage unit 127 stores display information in which the estimation result and the failure probability are associated. In this case, the estimation result shows the top three components with the highest failure probabilities among the candidates for the defective components.

[0063] In the Fig. The 5 examples shown are, for instance, the components with the highest average failure probabilities in case studies 1 to N among the similar cases stored in the estimation result storage unit 126: the “compressor”, the “coil” and the “fan motor”, and their failure probabilities are “50.20%”, “23.10%” and “11.60%”.

[0064] Returning to the description of Fig. Furthermore, the diagnostic processing unit 13, for example, is a functional unit implemented by instructing a processor, comprising a CPU, to execute a program. The diagnostic processing unit 13 includes a diagnostic data reference unit 131, an operational data reference unit 132, the similar case extraction unit 133, the defect component estimation unit 134, and the output information generation unit 135.

[0065] The diagnostic data reference unit 131 receives the diagnostic data from the maintenance terminal 32 via the NW communication unit 11. The diagnostic data reference unit 131 stores the received diagnostic data in the diagnostic data storage unit 121.

[0066] The operating data reference unit 132 receives the operating data of the diagnostic target device 22 via the NW communication unit 11 from the maintenance terminal 32. The operating data reference unit 132 stores the received operating data of the diagnostic target device 22 in the operating data storage unit 122.

[0067] The similar case extraction unit 133 classifies diagnostic data for diagnosing the diagnostic target device 22 according to a predetermined split condition and extracts similar cases from the maintenance log information storage unit 421 of the model learning device 40 based on the weight of the data element assigned to the classification (split condition).

[0068] The similarity-case extraction unit 133 classifies, depending on the area in which the diagnostic target device 22 is installed, whether the acquired diagnostic and operational data are assigned to, for example, the coastal area or the inland area. For instance, in the coastal area there is a strong tendency for failures to occur due to metal corrosion, and the weight (importance) of the data element for failure diagnosis differs. Therefore, it is considered effective to classify the data into the aforementioned divisions (the coastal area or the inland area).

[0069] The Similarity Case Extraction Unit 133 retrieves the weight of each data element assigned to the classification (division condition) from the Weight Storage Unit 123 and calculates the degree of similarity to the past case stored in the Maintenance Log Information Storage Unit 421, using the weight of the respective data element assigned to the classification (division condition). The Similarity Case Extraction Unit 133 calculates the degree of similarity (Sim n ) using, for example, the following equation (1). Simn=α⋅fx(x¯−x)+β⋅fy(y¯−y)+⋯

[0070] Here Sim nThe degree of similarity to an nth past case is given by α, β, and ..., and α, β, ..., which specify the weight of a respective data element. Furthermore, x~, y~, ..., provide diagnostic data for the input value of a respective data element, and x, y, ..., provide diagnostic data for the past case of the respective data element. Finally, the function f is a function that returns "1" if the past case and the data of the input element match, and returns "0" if the past case and the data of the input element do not match.

[0071] Furthermore, in the present embodiment, a variable with a horizontal line over the letter “x” is represented by x~ and a variable with a horizontal line over the letter “y” is represented by y~.

[0072] The similarity case extraction unit 133 multiplies the degree of similarity of each data element (“1” is used if the data element matches, and “0” is used if the data element does not match) by the weight (α, β) of the respective data element to calculate the sum of the degrees of similarity of all data elements as the degree of similarity (Sim n ) to calculate the nth past case using equation (1) described above.

[0073] For example, the similar-case extraction unit 133 extracts 100 cases as the similar cases in descending order of the calculated degree of similarity (Sim). n As described above, the Similar Case Extraction Unit 133 extracts a multitude of similar cases. The Similar Case Extraction Unit 133 stores the extracted similar cases in the Similar Case Storage Unit 125.

[0074] The Defect Component Estimator 134 estimates a defective component of the diagnostic target device 22 based on the similar cases extracted by the Similar Case Extractor 133. The Defect Component Estimator 134 estimates the defective component from the similar cases using the Defect Component Detection Model stored in the Model Storage Unit 124. For example, the Defect Component Estimator 134 estimates candidates for the defective components and failure probabilities for each of the multiple similar cases (for example, 100 cases) stored in the Similar Case Storage Unit 125, using the Defect Component Detection Model as shown in Fig. Figure 4 is shown. In addition, the defect component estimation unit 134 stores the estimation results (the candidates for the defective components and the failure probabilities) in the estimation result storage unit 126.

[0075] The output information generation unit 135 generates output information based on the estimation results of the defect component estimation unit 134. The output information generation unit 135 calculates the average value of the failure probabilities of the defective components in the multitude of similar cases, which are stored in the estimation result storage unit 126. As in Fig. As shown in Figure 4, the output information generation unit 135, for example, stores the calculated average value of the failure probabilities in the estimation result storage unit 126.

[0076] Furthermore, the output information generation unit 135 selects a specific number (for example, three) of candidates for the defective components in descending order of their average failure probabilities and generates output information that includes the selected candidates for the defective components. For example, the output information generation unit 135 generates output information (a display screen) that is shown in Fig. 5 is shown, and stores the output information in the output information storage unit 127.

[0077] The output information generation unit 135 transmits the generated output information to the maintenance terminal 32 via the network communication unit 11. As described above, the output information generation unit 135 generates the output information, which includes the selected candidates for the defective components and the average value of the failure probabilities, and transmits the generated output information to the maintenance terminal 32.

[0078] Next, with reference to the drawings, the operation of the maintenance assistance system 1 according to the present embodiment will be described.

[0079] First, in relation to Fig. 6 a model learning process of the model learning device 40 is described.

[0080] Fig. Figure 6 is a flowchart showing an example of the model learning process of the model learning device 40 in the present embodiment.

[0081] As in Fig. As shown in Figure 6, the model learning device 40 collects the maintenance work report and the operating data via the network NW1 (step S101). The maintenance log information collection unit 431 of the model learning device 40 collects the maintenance work report as the maintenance log information from the maintenance terminal 31 via the NW communication unit 41 and stores the collected maintenance log information (maintenance work report) in the maintenance log information storage unit 421.

[0082] In addition, the operating data collection unit 432 of the model learning device 40 collects the operating data from the diagnostic target device 21 via the NW communication unit 41 and stores the collected operating data in the operating data storage unit 422.

[0083] Then, the model learning unit 434 of the model learning device 40 classifies the input and output data from the maintenance work report and the operational data (step S102). The model learning unit 434 classifies the maintenance log information (maintenance work report), which is stored in the maintenance log information storage unit 421, and the operational data, which is stored in the operational data storage unit 422, as the training data into input data and output data. In the Fig. In the case shown, for example, the “model name”, the “years since installation”, the “area” and the “symptom” are classified as the input data, and the “replacement component P1” and the “replacement component P2” are classified as the output data.

[0084] The model learning unit 434 then learns the relationship between the input and output data and generates the defect component detection model (step S103). The model learning unit 434 generates the defect component detection model from the training data described above using a machine learning method, such as LightGBM or SVM. The model learning unit 434 stores the generated defect component detection model in the model storage unit 424.

[0085] Then, the model learning unit 434 transmits the defect component detection model to the diagnostic device 10 (step S104). The model learning unit 434 transmits the defect component detection model stored in the model storage unit 424 to the diagnostic device 10 via the network communication unit 41. The transmitted defect component detection model is also stored in the model storage unit 124 of the diagnostic device 10. After the process in step S104, the model learning unit 434 terminates the model learning process.

[0086] Next, with reference to Fig. 7 a weight learning process of the model learning device 40 is described.

[0087] Fig. Figure 7 is a flowchart showing an example of the weight learning process of the model learning device 40 in the present embodiment.

[0088] As in Fig. As shown in Figure 7, the model learning device 40 first extracts a data element with a significantly different fault tendency depending on the distribution from the maintenance work report and the operational data (step S201). For example, the weight learning unit 433 of the model learning device 40 extracts the "area" as the data element with the significantly different fault tendency.

[0089] Then, the weight learning unit 433 divides the data of the data element according to the division condition of the designated data element (step S202). For example, the weight learning unit 433 divides the training data described above into the coastal area and the inland area according to the "area" of the designated data element.

[0090] The weight learning unit 433 then learns the relationship between the diagnostic data and the replacement component for each partition and calculates the weight for each data element (step S203). For example, the weight learning unit 433 calculates the weight of each data element in the coastal area from the diagnostic data (training data) whose partition condition is the coastal area, using a machine learning method such as LightGBM. Similarly, the weight learning unit 433 calculates, for example, the weight of the respective data element in the inland area from the diagnostic data (training data) whose partition condition is the inland area, using a machine learning method such as LightGBM. The weight learning unit 433 stores the calculated weight of each data element for each partition condition in the weight storage unit 423, for example, as shown in Fig. 3 shown.

[0091] Then, the weight learning unit 433 transmits the weight of each data element to the diagnostic device 10 (step S204). The weight learning unit 433 transmits the weight of each data element for each distribution condition, which is stored in the weight storage unit 423, to the diagnostic device 10 via the network communication unit 41. Additionally, the transmitted weight of each data element for each distribution condition is stored in the weight storage unit 123 of the diagnostic device 10. After the process in step S204, the weight learning unit 433 terminates the weight learning process.

[0092] Next, with reference to Fig. 8 a diagnostic process of the diagnostic device 10 described.

[0093] Fig. Figure 8 is a flowchart showing an example of the diagnostic process of the diagnostic device 10 in the present embodiment.

[0094] As in Fig. As shown in Figure 8, the diagnostic device 10 first extracts the designated data element from the diagnostic data in response to receiving the diagnostic data and the operational data (step S301). The diagnostic data reference unit 131 of the diagnostic device 10 obtains the diagnostic data via the network communication unit 11 from the maintenance terminal 32, and the operational data reference unit 132 obtains the operational data of the diagnostic target device 22 via the network communication unit 11 from the maintenance terminal 32. The similar-case extraction unit 133 of the diagnostic device 10 extracts the designated data element (for example, the "area") in response to receiving the diagnostic data and the operational data.

[0095] Then, the Similar Case Extraction Unit 133 classifies the data of the designated data element according to the partitioning condition during weight learning (step S302). Similar Case Extraction Unit 133 classifies the diagnostic data and the operational data into, for example, the coastal area or the inland area.

[0096] Then, the Similar Case Extraction Unit 133 extracts the weight in the classification from the result of weight learning (step S303). Similar Case Extraction Unit 133 extracts the weight of each data element assigned to the classification (splitting condition), which is classified according to the diagnostic and operational data, from Weight Storage Unit 123. For example, if the classification is the coastal area, Similar Case Extraction Unit 133 retrieves the weight of each data element assigned to the coastal area from Weight Storage Unit 123. Similarly, if the classification is, for example, the inland area, Similar Case Extraction Unit 133 retrieves the weight of each data element assigned to the inland area from Weight Storage Unit 123.

[0097] Then, the Similar Case Extraction Unit 133 calculates the degree of similarity to the previous case using the extracted weight (step S304). The Similar Case Extraction Unit 133 calculates the degree of similarity (Sim n ) between the past case, which is stored in the maintenance log information storage unit 421, and the diagnostic data and the operational data using the equation (1) described above.

[0098] Then, the Similar Case Extraction Unit 133 sorts the similarity levels in descending order and selects the past cases (for example, 100 cases) with the highest similarity levels (step S305). The Similar Case Extraction Unit 133 stores the selected past cases (for example, 100 cases) with the highest similarity levels as the similar cases in the Similar Case Storage Unit 125.

[0099] Then, the Defect Component Estimator 134 of the diagnostic device 10 estimates a defective component from each selected past case using the Defect Component Detection Model (step S306). The Defect Component Estimator 134 estimates candidates for the defective components and the failure probabilities for each of the similar cases stored in the Similar Case Memory Unit 125, using the Defect Component Detection Model stored in the Model Memory Unit 124. The Defect Component Estimator 134 stores the estimation results in the Estimation Result Memory Unit 126, for example, as shown in Case Examples 1 to N. Fig. 4.

[0100] Then, the output information generation unit 135 of the diagnostic device 10 aggregates the defective components and the failure probabilities estimated from the respective past case (step S307). The output information generation unit 135 calculates the average failure probabilities for the defective components across a large number of past cases (similar cases). As in Fig. As shown in Figure 4, the output information generation unit 135, for example, calculates the average value of the failure probabilities for each defective component and stores the average value in the estimation result storage unit 126.

[0101] Then, the output information generation unit 135 generates output information from the aggregation result and transmits the output information to the maintenance terminal 32 (step S308). The output information generation unit 135 sorts the average failure probabilities for each defective component in descending order and identifies candidates for the top three defective components with the highest average failure probabilities. For example, the output information generation unit 135 generates the output information that is displayed in Fig. Figure 5 shows the process using the top three candidates for defective components with the highest average failure probabilities. The output information generation unit 135 stores the generated output information in the output information storage unit 127 and transmits the output information to the maintenance terminal 32 via the network communication unit 11. After the process in step S308, the output information generation unit 135 terminates the diagnostic process of the diagnostic device 10.

[0102] As described above, the maintenance assistance system 1 according to the present embodiment comprises the maintenance log information storage unit 421, the weight learning unit 433 (importance level learning unit), the similar case extraction unit 133, and the defective component estimation unit 134. The maintenance log information storage unit 421 stores the maintenance log information obtained in the past by diagnosing the diagnostic target device 21, which includes data from a variety of data elements and defective component information.The weight learning unit 433 (importance learning unit) divides the maintenance log information stored in the maintenance log information storage unit 421 according to a predetermined division condition (for example, coastal area or inland area) based on the data element, learns the relationship between the data element and the defective component in the maintenance log information for each division, and generates the weight (importance) of the data element for each division condition. The similarity case extraction unit 133 classifies the diagnostic data for diagnosing the diagnostic target device 22 according to the division condition and extracts the similar cases that are similar to the diagnostic data from the maintenance log information storage unit 421 based on the importance of the data element assigned to the classification (division condition).The Defect Component Estimator 134 estimates a defective component of the diagnostic target device 22 based on the similar cases extracted by the Similar Case Extractor 133.

[0103] Therefore, according to the present embodiment, the maintenance assistance system 1 extracts similar cases using the weight (importance) of the data element for each division condition, which has been divided (classified) by the predetermined division condition (for example, the coastal area or the inland area). This makes it possible to extract suitable similar cases with higher accuracy. Therefore, according to the present embodiment, the maintenance assistance system 1 can improve the accuracy of a fault diagnosis and can improve the quality of a diagnosis by the maintenance service provider.

[0104] Furthermore, according to the present embodiment, the maintenance assistance system 1 can extract suitable similar cases, for example taking into account the difference in the failure tendency for each data element, such as regionality, and can select candidates for replacement components taking regionality into account.

[0105] Furthermore, the maintenance assistance system 1 according to the present embodiment comprises the model learning unit 434. The model learning unit 434 learns the maintenance log information as the training data and generates the defect component detection model, which estimates a defective component from similar cases. The defect component estimation unit 134 estimates a defective component from similar cases using the defect component detection model.

[0106] Therefore, according to the present embodiment, the maintenance assistance system 1 estimates a defective component from the same cases using the defective component detection model. This makes it possible to estimate the defective component more accurately and improve the quality of the diagnosis by the maintenance service provider.

[0107] Furthermore, the maintenance assistance system 1 according to the present embodiment includes the output information generation unit 135. The output information generation unit 135 generates the output information based on the estimation results of the defect component estimation unit 134. The similar case extraction unit 133 extracts a multitude of similar cases. The defect component estimation unit 134 estimates the defective component and the failure probability for each of the multitude of similar cases using the defect component detection model.The output information generation unit 135 calculates the average value of the failure probabilities for the defective components in the multitude of similar cases, selects a specific number of (for example, the top three) candidates for the defective components in descending order of the average value of the failure probabilities, and generates the output information which includes the selected candidates for the defective components.

[0108] The maintenance assistance system 1 according to the present embodiment therefore selects candidates for the defective components using the average failure probabilities for the defective components in a multitude of similar cases. This makes it possible to estimate the candidates for the defective components with higher accuracy. Furthermore, the maintenance assistance system 1 according to the present embodiment outputs a specific number of candidates for the defective components (for example, the top three) in descending order of the average failure probabilities. This makes it possible to provide the maintenance service provider with criteria for determining the defective components and to improve the quality of the diagnosis.

[0109] Furthermore, in the present embodiment, the similar-case extraction unit 133 extracts the similar cases that are similar to the diagnostic data received from the maintenance terminal. The output information generation unit 135 generates the output information, which includes the selected candidates for the defective components and the average value of the failure probabilities, and transmits the generated output information to the maintenance terminal 32.

[0110] Therefore, according to the present embodiment, the maintenance assistance system 1 transmits the output information, which includes the selected candidates for the defective components and the average value of the failure probabilities, to the maintenance terminal 32. This makes it possible to provide the maintenance service provider with the criteria for determining the defective components.

[0111] Furthermore, in the present embodiment, the weight learning unit 433 generates the weight (importance) of the data element for each partitioning condition using the learning data, which includes the operational data (e.g., the sensor detection data, the fault code, and the like) of the diagnostic target device 21, which has been collected by the diagnostic target device 21 in the past, and the maintenance log information. The model learning unit 434 generates the defect component detection model using the learning data, which includes both the operational data of the diagnostic target device 21 and the maintenance log information.

[0112] Therefore, according to the present embodiment, the maintenance assistance system 1 generates the defect component detection model and the weight (importance) of the data element for each allocation condition, taking into account the operational data (for example, the sensor detection data, the error code, and the like). This makes it possible to estimate the defective component more accurately.

[0113] Furthermore, in the present embodiment, the weight learning unit 433 divides the maintenance log information according to the area in which the diagnostic target device 21 is installed. The similar-case extraction unit 133 classifies the diagnostic data by area and extracts the similar cases based on the weight (importance) of the data element assigned to the classified area.

[0114] Therefore, according to the present embodiment, the maintenance assistance system 1 can extract suitable similar cases taking into account the difference in regional failure tendency and appropriately select candidates for replacement components taking into account regionality.

[0115] Furthermore, the maintenance assistance method according to the present embodiment is a maintenance assistance method for the maintenance assistance system 1, which includes the maintenance log information storage unit 421, and comprises a weight learning step, a similar-case extraction step, and a defective component estimation step. The maintenance log information storage unit 421 stores the maintenance log information obtained in the past by diagnosing the diagnostic target device 21, which includes data from a multitude of data elements and defective component information.In the weight learning step, the weight learning unit 433 divides the maintenance log information stored in the maintenance log information storage unit 421 according to a division condition, which is predetermined based on the data element. It learns the relationship between the data element and the defective component in the maintenance log information for each division and generates the weight (importance) of the data element for each division condition. In the similar-case extraction step, the similar-case extraction unit 133 classifies the diagnostic data for diagnosing the diagnostic target device 22 according to the division condition and extracts the similar cases that are similar to the diagnostic data from the maintenance log information storage unit 421 based on the weight of the data element assigned to the classification (division condition).In the defect component estimation step, the defect component estimation unit 134 estimates the defective component of the diagnostic target device 22 based on the similar cases extracted by the similar case extraction unit 133. Therefore, the maintenance assistance method according to the present embodiment has the same effect as the maintenance assistance system 1 described above, can extract suitable similar cases with higher accuracy, and can improve the quality of the diagnosis.

[0116] Fig. Figure 9 is a diagram showing a hardware configuration of the diagnostic device 10 and the model learning device 40 of the maintenance assistance system 1 according to the present embodiment. A Fig. Device 9 shown shows a hardware configuration of each device (the diagnostic device 10 and the model learning device 40) of the Maintenance Assistance System 1.

[0117] As in Fig.As shown in Figure 9, each device (the diagnostic device 10 and the model learning device 40) of the maintenance assistance system 1 comprises a communication device H11, a memory H12 and a processor H13.

[0118] The communication device H11, for example, is a communication device that can be connected to the network NW1, such as a LAN card.

[0119] The memory H12, for example, is a storage device, such as RAM, flash memory or HDD, and stores various types of information and programs used by the respective device (the diagnostic device 10 and the model learning device 40).

[0120] The processor H13, for example, is a processing circuit comprising a CPU and similar components. The processor H13 executes the program stored in memory H12 to carry out various processes of the respective device (the diagnostic device 10 and the model learning device 40).

[0121] Furthermore, the present disclosure is not limited to the embodiment described above and can be modified without abandoning the core idea of ​​the present disclosure.

[0122] For example, the embodiment described above illustrates how the weight allocation condition is divided into the coastal area and the inland area according to the "area" of the data element. However, the present disclosure is not limited to this, and other data elements and allocation conditions can be used. For example, if the "years since installation" is used as the data element, the data can be divided (classified) under an allocation condition of five years or more and less than five years.

[0123] Furthermore, the embodiment described above includes an example in which the maintenance assistance system 1 comprises the diagnostic device 10 and the model learning device 40. However, the present disclosure is not limited to this. The diagnostic device 10 can include the functions of the model learning device 40, and the maintenance assistance system 1 can be implemented by a single device.

[0124] Furthermore, in the embodiment described above, the model learning device 40 can include some of the functions of the diagnostic device 10, or the diagnostic device 10 can include some functions of the model learning device 40. In addition, the diagnostic device 10 and the model learning device 40 can be implemented by three or more devices.

[0125] Furthermore, the embodiment described above includes an example in which the similar-case extraction unit 133 extracts a specific number of similar cases (for example, 100 cases). However, the present disclosure is not limited to this, and a single past case with the highest degree of similarity can be extracted as the similar case.

[0126] Furthermore, the embodiment described above includes an example in which the model learning device 40 generates the weight of each data element and the defect component detection model using the maintenance log information and the operational data. However, the present disclosure is not limited to this, and the model learning device 40 can generate the weight of each data element and the defect component detection model without using the operational data. In this case, the diagnostic device 10 also extracts the similar cases from the diagnostic data without using the operational data.

[0127] Furthermore, the embodiment described above includes an example in which the communication of each device is implemented using two networks, NW1 and NW2. However, the present disclosure is not limited to this, and the communication can be implemented using a single network or three or more networks.

[0128] Furthermore, the embodiment described above includes an example in which the diagnostic target device 20 is an air conditioner. However, the present disclosure is not limited to this. The diagnostic target device 20 can, for example, be another household appliance, an IoT device, or the like.

[0129] Furthermore, each component of Maintenance Assistance System 1 includes a computer system. A program for implementing the functions of the respective component provided in Maintenance Assistance System 1 can be recorded on a computer-readable recording medium. The program recorded on the recording medium can then be loaded into a computer system and executed to perform the processes in the respective component contained in Maintenance Assistance System 1. In this context, the phrase "program recorded on the recording medium is loaded into the computer system and executed" refers to installing the program in the computer system. Here, the "computer system" includes an operating system and hardware, such as a peripheral device.

[0130] Furthermore, the "computer system" can comprise a variety of computer devices connected via a network, which includes a communication link, such as the internet, a WAN, a LAN, or a dedicated line. Additionally, the "computer-readable recording medium" means a storage device, such as a portable medium like a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, or a hard drive provided within the computer system. As described above, the recording medium that stores the program can be a non-volatile recording medium, such as a CD-ROM.

[0131] Furthermore, the recording medium also includes an internal or external recording medium, which is made available by a distribution server for distributing the program. The program can also be divided into multiple parts, and these parts can be downloaded at different times and then combined in the respective component provided in Maintenance Assistance System 1. Alternatively, the divided programs can be distributed by different distribution servers. The "computer-readable recording medium" also includes a medium that stores the program for a specific period of time, such as volatile memory (RAM) in a server or client computer system when the program is transmitted over a network. Finally, the program described above can be a program for implementing some of the aforementioned functions.Furthermore, the program can be a so-called difference file (difference program), which can implement the functions described above in combination with a program that is already recorded on the computer system. Reference symbol list 1 Maintenance assistance system 10 Diagnostic device 11, 41, 321 NW communication unit 12 Device storage unit 13 Diagnostic Processing Unit 20, 21, 22 Diagnostic target device 30, 31, 32 Maintenance terminal 40 Model learning device 42 learning memory units 43 Learning Processing Unit 121 Diagnostic Data Storage Unit 122,422 Operating Data Storage Unit 123,423 weight storage units 124,424 model storage units 125 Similar-Case Storage Unit 126 Estimation result storage unit 127 Output information storage unit 131 Diagnostic Data Reference Unit 132 Operational data reference unit 133 Similar-case extraction unit 134 Defect Component Estimation Unit 135 Output Information Generation Unit 322 Input unit 323 Display unit 324 Terminal storage unit 325 Terminal control unit 421 Maintenance log information storage unit 431 Maintenance Log Information Collection Unit 432 Operational Data Collection Unit 433 Weight learning unit 434 Model learning unit NW1, NW2 Network QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] JP 2020 - 009 068

[0003]

Claims

[1] Maintenance assistance system, comprising: a maintenance log information storage unit configured to store maintenance log information obtained by diagnosing a diagnostic target device in the past, wherein the maintenance log information includes data of several data elements and defective component information; an importance level learning unit configured to split the maintenance log information stored in the maintenance log information storage unit according to a split condition predetermined based on the data element, to learn a relationship between the data element and a defective component in the maintenance log information for each split, and to generate an importance level of the data element for each split condition; a similar case extraction unit configured to classify diagnostic data for diagnosing the diagnostic target device according to the split condition and to extract a similar case, which is similar to the diagnostic data, from the maintenance log information storage unit based on a degree of importance of the data element assigned to the classification; and a defect component estimation unit which is configured to estimate a defective component of the diagnostic target device based on the similar case extracted by the similar case extraction unit. [2] Maintenance assistance system according to claim 1, further comprising: a model learning unit configured to learn the maintenance log information as training data and to generate a defect component detection model that estimates the defective component from similar cases, where the defect component estimation unit estimates the defective component from the similar case using the defect component detection model. [3] Maintenance assistance system according to claim 2, further comprising: an output information generation unit configured to generate output information based on an estimate result estimated by the defect component estimator, where the similar-case extraction unit extracts a plurality of similar cases, wherein the defect component estimator estimates the defective component and a failure probability for each of the multitude of similar cases using the defect component detection model, and wherein the output information generation unit calculates an average value of the failure probabilities for the defective components in the multitude of similar cases, selects a specific number of candidates for the defective components in descending order of the average value of the failure probabilities, and generates the output information which includes the selected candidates for the defective components. [4] Maintenance assistance system according to claim 3, wherein the similar case extraction unit extracts the similar case which is similar to the diagnostic data received from a maintenance terminal, and wherein the output information generation unit generates the output information which includes the selected candidates for the defective components and the average value of the failure probabilities, and transmits the generated output information to the maintenance terminal. [5] Maintenance assistance system according to one of claims 2 to 4, wherein the importance level learning unit generates the importance level of the data element for each splitting condition using the learning data, which includes both operational data of the diagnostic target collected from the diagnostic target in the past and the maintenance log information, and wherein the model learning unit generates the defect component detection model using the learning data, which includes both the operating data of the diagnostic target device and the maintenance log information. [6] Maintenance assistance system according to any one of claims 1 to 5, wherein the importance level learning unit divides the maintenance log information according to an area in which the diagnostic target device is installed and generates the importance level of the data element for each area, and wherein the similar case extraction unit classifies the diagnostic data according to the area and extracts the similar case based on the importance level of the data element assigned to the classified area. [7] Maintenance assistance procedure for a maintenance assistance system comprising a maintenance log information storage unit configured to store maintenance log information obtained by diagnosing a diagnostic target device in the past, wherein the maintenance log information comprises data of several data elements and defective component information, wherein the maintenance assistance procedure comprises: To cause an importance-level learning unit to split the maintenance log information stored in the maintenance log information storage unit according to a split condition predetermined based on the data element, to learn a relationship between the data element and a defective component in the maintenance log information for each split, and to generate an importance level of the data element for each split condition; To cause a similar-case extraction unit to classify diagnostic data for diagnosing the diagnostic target device according to the split condition and to extract a similar case, which is similar to the diagnostic data, from the maintenance log information storage unit based on a degree of importance of the data element assigned to the classification; and To cause a defective component estimator to estimate a defective component of the diagnostic target device based on the similar case extracted by the similar case extraction unit.

Images