WO2024195157A1 - Diagnosis system, diagnosis method, and program - Google Patents

Abstract

The present invention can realize the construction of a diagnosis model that enables identifying of a failure factor with higher accuracy and sharing of the failure knowledge of other industries.　This diagnosis system has at least one processor and at least one memory resource. The memory resource stores a failure diagnosis program for diagnosing failure of electronic/electromotive systems. The processor executes the failure diagnosis program to perform diagnosis involving: identifying, on the basis of acquisition of failure-related information indicating the failure state of the electronic/electromotive systems, the industry and the failure type of an electronic/electromotive system to be diagnosed; calculating, as the occurrence probability of a failure factor, a causal coefficient which is outputted by using a prescribed mathematical algorithm having set therein parameter information corresponding to a first industry having been identified and which indicates the strength of the relationship between failure factors corresponding to the failure type; and identifying the failure factor on the basis of the occurrence probability.

Description

Diagnostic system, diagnostic method, and program

The present invention relates to a diagnostic system, a diagnostic method, and a program. The present invention claims priority to Japanese Patent Application No. 2023-045145, filed on March 22, 2023, and the contents of that application are incorporated by reference into this application in designated countries where incorporation by reference of literature is permitted.

In recent years, with the growing need for autonomous driving, there has been an increase in the use of semiconductors and communication standards for information devices in in-vehicle equipment. Furthermore, as this trend spreads to the AI (Artificial Intelligence) industrial sector and mission-critical fields, it is predicted that the adoption rate of common semiconductors and communication standards across various industrial sectors will increase.

In addition, in light of these trends, the IEEE P2851 functional safety standard for electronic and electric systems is promoting the standardization of data and verification processes in the functional safety design and verification process across industries.

On the other hand, diagnostic technologies aimed at ensuring the reliable and safe operation of electronic and electrified service systems are currently being developed on an industry-by-industry basis. Specifically, each vendor operates diagnostic systems that are designed and developed in accordance with industry-standard safety and reliability standards, customizing them based on industry-standard data and processes.

Even if systems have similar system configurations or use the same components, if they are in different industries, there are differences in safety and reliability standards. This makes it difficult to share knowledge from other industries when generating diagnostic models, and makes it necessary to build machine learning-based diagnostic models from scratch.

The challenge is therefore to build an efficient diagnostic model that spans multiple industries and shares knowledge from each industry.

Patent document 1 discloses a system for diagnosing error conditions in a complex environment. Specifically, patent document 1 states, "A system for generating diagnoses of probable causes of a detected fault, advantageously provided with a user interface and using a Bayesian network, where the probabilities are generated automatically and manual processes are used to construct the probability table. The system provides multiple hypotheses and/or diagnoses to an operator simultaneously."

JP 2000-356696 A

The system in Patent Document 1 utilizes a Bayesian network to diagnose the causes of failures that occur in complex systems. However, the system in this document does not take into consideration the generation of an efficient diagnostic model that shares defect knowledge from different industries. Therefore, the technology in Patent Document 1 makes it difficult to solve the problem of building an efficient diagnostic model that spans multiple industries and shares knowledge from each industry.

The present invention was made in consideration of the above problems, and aims to realize the construction of efficient diagnostic models by sharing defect knowledge across industries.

The present application includes multiple means for solving at least part of the above problems, examples of which are as follows: A diagnostic system according to one aspect of the present invention for solving the above problems is a diagnostic system having one or more processors and one or more memory resources, the memory resources storing a fault diagnosis program for diagnosing faults in an electronic/electric system, and the processor executing the fault diagnosis program to: identify the industry and fault type of the electronic/electric system to be diagnosed based on acquisition of fault-related information indicating a fault state of the electronic/electric system, calculate a causal coefficient indicating the strength of the relationship between fault factors corresponding to the fault type output by a predetermined mathematical algorithm in which parameter information corresponding to the identified first industry is set, as the occurrence probability of the fault factor, and perform a diagnosis to identify the fault factor based on the occurrence probability.

The present invention makes it possible to identify the cause of defects with greater accuracy and to build a diagnostic model that can share defect knowledge with other industries.

　Problems, configurations and advantages other than those mentioned above will become clear from the description of the embodiments below.

FIG. 1 is a diagram illustrating an example of a schematic configuration of a diagnostic system. FIG. 11 is a diagram illustrating an example of defect correlation information. FIG. 13 is a diagram showing an example of a mathematical algorithm (neural network). FIG. 13 illustrates an example of a parameter element. FIG. 13 is a diagram illustrating an example of normalized transform coefficients. FIG. 2 is a diagram illustrating an example of a functional configuration of a diagnostic system. FIG. 4 is a flow chart showing an example of each process. FIG. 11 is a flow diagram showing details of a diagnostic model generation process. FIG. 4 is a flow chart showing details of a diagnostic process. FIG. 13 is a diagram showing the relationship between defect causes and occurrence probabilities. FIG. 13 is a diagram showing the relationship between defect causes and occurrence probabilities. FIG. 11 is a diagram showing an example of a diagnosis result. FIG. 11 is a diagram illustrating an example of normalized transformation coefficients according to the second embodiment. FIG. 23 is a diagram showing an example of a determination result of a similarity according to the fifth embodiment. FIG. 13 is a diagram showing an example of a usage environment for each service of similar electronic/electric systems according to the sixth embodiment. FIG. 1 is a schematic diagram of a service form 1. FIG. 13 is a schematic diagram of a service form 2.

Each embodiment of the present invention will be explained below with reference to the drawings.

First Embodiment
1 is a diagram showing an example of a schematic configuration of a diagnostic system 100 according to this embodiment. The diagnostic system 100 (hereinafter, sometimes referred to as "this system") is a device that diagnoses malfunctions that occur in electronic and electric systems.

Specifically, the diagnostic system 100 uses a diagnostic model to diagnose malfunctions that occur in various electronic/electric systems (e.g., recognition systems/wheel drive systems/connected systems in automobiles, recognition systems/arm drive systems/connected systems in robots) installed in moving objects (e.g., automobiles including EVs: Electric Vehicles) and robots (e.g., industrial robots and service robots), and outputs the identified causes of the malfunction as the diagnostic results.

More specifically, the diagnostic system 100 diagnoses malfunctions occurring in the electronic/electric system using a diagnostic model consisting of the following elements:
* "When there is industry parameter information that corresponds to the defect category to be diagnosed"
**Failure correlation information: Information showing the correlation of failures in the specialized fields of each electronic and electric system. For example, FTA: Fault Tree Analysis, FMEA: Failure Mode and Effects Analysis, etc.
** Mathematical algorithm: A mathematical algorithm such as a neural network. An information model that outputs a causal coefficient α that indicates the strength of the relationship between nodes in defect correlation information.
** Parameter information: Parameter values set in the mathematical algorithm.

* "When there is no industry parameter information corresponding to the defect category to be diagnosed"
**Failure correlation information **Mathematical algorithm **Parameter information **Normalization conversion coefficient β: A coefficient that normalizes the safety and reliability standard values for each industry as cross-industry relative values.

The diagnostic system 100 also updates the parameter information for executing the mathematical algorithm used to calculate the causal coefficient α depending on the measures taken by the user in response to the diagnostic results and whether or not the defect has been resolved.

In addition, when there is no parameter information of the industry corresponding to the electronic/electric system to be diagnosed, the diagnostic system 100 normalizes the causal coefficient _α1 output by the mathematical algorithm using parameter information of another industry based on the ratio between the normalization conversion coefficient _β1 of the other industry and the normalization conversion coefficient _β2 of the industry corresponding to the electronic/electric system to be diagnosed, and identifies the cause of the malfunction using the normalized causal coefficient _α2 .

This system makes it possible to identify the cause of defects with greater accuracy, and to build a diagnostic model that can share defect knowledge with other industries.

The electronic/electrical systems to be diagnosed are not limited to mobile objects or robots, but various types of electronic/electrical systems are targeted. In this embodiment, however, the diagnosis of a malfunction that occurs in the electronic/electrical system of an automobile, which is a mobile object, will be used as an example.

<Configuration of diagnostic system 100>
As shown in FIG. 1, a diagnostic system (processor system) 100 is connected to an external device 10 so as to be able to communicate with each other via, for example, a communication cable or a predetermined communication network N (for example, the Internet, a LAN (Local Area Network) or a WAN (Wide Area Network)).

<<External device 10>>
The external device 10 is a device that transmits various information to the diagnostic system 100. In this case, the external device 10 includes a vehicle system that transmits defect-related information to the diagnostic system 100 and a processing unit that performs the processing in the diagnostic system 100. This includes operators' computers that provide a variety of useful information to be used.

The external device 10 is a device that displays the diagnostic results output by the diagnostic system 100. In this case, the external device 10 includes computers of businesses that receive the diagnostic services provided by this system 100, such as product manufacturing and maintenance businesses such as automobile manufacturers, and infrastructure management businesses.

<<Details of diagnostic system 100>>
The diagnostic system 100 executes various processes by the processor 20 reading the program 210 and various information stored in the memory resource 30. Specifically, the diagnostic system 100 executes a process of identifying a malfunction category in the electronic/electric system using the malfunction-related information (hereinafter, may be referred to as a "malfunction category identification process").

The diagnostic system 100 also executes a process for generating a diagnostic model based on the identified defect categories (hereinafter, sometimes referred to as the "diagnostic model generation process").

The diagnostic system 100 also executes a diagnostic process to diagnose the cause of the defect based on the diagnostic model.

The diagnostic system 100 also performs machine learning of a mathematical algorithm (hereinafter sometimes referred to as "learning processing") based on the measures taken by the user in response to the diagnostic results and whether or not the defect has been resolved.

Details of each of these processes will be given later.

The diagnostic system 100 is, for example, a server computer, a cloud server, a personal computer, a tablet terminal, or a smartphone, and is a system that includes at least one of these computers.

Specifically, the diagnostic system 100 has a processor 20, a memory resource 30, an NI (Network Interface Device) 40, and a UI (User Interface Device) 50.

The processor 20 is an arithmetic device that reads the program 210 stored in the memory resource 30 and executes the processing corresponding to the program 210. Examples of the processor 20 include a microprocessor, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), or other semiconductor devices capable of performing calculations.

Memory resource 30 is a storage device that stores various information. Specifically, memory resource 30 is a non-volatile or volatile storage medium such as RAM (Random Access Memory) or ROM (Read Only Memory). Note that memory resource 30 may also be a rewritable storage medium such as a flash memory, a hard disk, or an SSD (Solid State Drive), or a USB (Universal Serial Bus) memory, a memory card, or a hard disk.

The NI 40 is a communication device that communicates information with the external device 10. The NI 40 communicates information with the external device 10 via a predetermined communication network N, such as a LAN or the Internet. Unless otherwise specified below, it is assumed that information communication between the diagnostic system 100 and the external device 10 is performed via the NI 40.

The UI 50 is an input device that inputs instructions from the user (operator) to the diagnostic system 100, and an output device that outputs information generated by the diagnostic system 100. Examples of input devices include pointing devices such as a keyboard, a touch panel, and a mouse, and a voice input device such as a microphone.

In addition, output devices include, for example, displays, printers, and voice synthesizers. Unless otherwise specified below, user operations on the diagnostic system 100 (for example, inputting and outputting information, and issuing instructions to execute processing, etc.) are assumed to be performed via the UI 50.

Furthermore, each configuration, function, processing means, etc. of the present system 100 may be realized in part or in whole in hardware, for example by designing it as an integrated circuit. Furthermore, the present system 100 may also realize each function in part or in whole in software, or through a combination of software and hardware. Furthermore, the present system 100 may use hardware having fixed circuits, or may use hardware having at least some of the circuits that are changeable.

The diagnostic system 100 can also be realized by a user (operator) implementing some or all of the functions and processes realized by each program. Note that the diagnostic system may entrust output processing to the user and some of the input processing from the user to a processor system outside the system (called an external processor system) such as a smartphone or tablet instead of the processor system itself. In such cases, the diagnostic system (or its processor 20, program) may do the following to execute each process or other parts of the program.

Instead of outputting to the user using the UI 50, data required for outputting to the user is sent to the external processor system via the NI 40. Examples of such data include the data to be output itself, data for generating output data in another processor system, but it may also be a program or web data that describes the process of performing user output in the external processor system.

In addition, instead of receiving input or operation from the user using the UI 50, the diagnostic system 100 receives data indicating a user input or operation from an external processor system via the NI 40. From another perspective, the meaning of outputting data to the user may include the diagnostic system 100 itself outputting the data, as well as having another entity other than the system 100 output the data (asserting it). In addition, the meaning of receiving input or operation from the user may include the diagnostic system 100 indirectly receiving the input or operation, as well as directly outputting or receiving the input to the user from the diagnostic system 100.

Furthermore, the database and various information within the memory resource 30 described below may be a data structure other than a database or a file, etc., as long as it is an area capable of storing data. Furthermore, one program may perform the functions of multiple programs, or vice versa. In other words, one or more programs may perform the processing of each program shown in the figure.

The programs executed by the diagnostic system may be stored in a non-volatile storage medium that can be read by the system 100. The programs stored in the non-volatile storage medium may be read directly by the diagnostic system 100, or a processor system for program distribution may read the programs from the medium and then transmit (distribute) the programs to the diagnostic system 100. An example of the non-volatile storage medium is the non-volatile memory described as the memory resource 30, but other optical disk media may also be used.

<<Defect-related information DB 110>>
The malfunction-related information DB110 is a database that stores malfunction-related information. The malfunction-related information is information that indicates the operating state (malfunction state) of an electronic/electric system due to the influence of a malfunction. Specifically, the malfunction-related information includes internal data of the electronic/electric system in a moving body (hereinafter, sometimes referred to as "system internal data"), probe data, user information, environmental information, infrastructure information, and product/service information.

System internal data is, for example, information such as error logs output from the ECU (Electronic Control Unit). Probe data is, for example, location information while driving and video information captured by a drive recorder. User information is, for example, questionnaire information that records the user's complaints (comments) regarding malfunctions. Environmental information is, for example, information indicating the weather and road conditions while driving. Infrastructure information is, for example, information such as server logs when using connected services. Product and service information is, for example, product design documents and service manuals.

The diagnostic system 100 acquires defect-related information from businesses (e.g., automobile manufacturers, infrastructure management businesses, businesses that provide environmental information, etc.) that correspond to the type of information or computers owned by users of mobile vehicles (e.g., external devices 10), and stores the information in the defect-related information DB 110.

<<Defect correlation information DB 120>>
The defect correlation information DB 120 is a database that stores defect correlation information. The defect correlation information is a so-called knowledge graph, FTA information, or FMEA information. Specifically, the defect correlation information is composed of nodes corresponding to knowledge items related to defects and links indicating the relationships between the nodes.

Figure 2 shows an example of malfunction correlation information. The malfunction correlation information shown in the figure shows an example of a knowledge graph related to an emergency brake malfunction (TOP event). Specifically, the example shown includes knowledge items (nodes) related to malfunctions such as camera system failures and control ECU failures, and related nodes are connected by links (paths). The start point of the path corresponds to the cause of the malfunction, and the end point of the path corresponds to the result of the malfunction, and the path represents the relationship between what factors can cause a certain malfunction (result).

Such malfunction correlation information exists for each top malfunction event in electronic and electric systems in each industry. Note that FIG. 2 shows an example of malfunction correlation information corresponding to a top malfunction event, an emergency brake malfunction, in electronic and electric systems in the automotive industry. In other words, the malfunction correlation information DB 120 stores multiple pieces of malfunction-related information corresponding to various top malfunction events in electronic and electric systems in each industry, such as the automotive industry. Furthermore, the top malfunction event corresponds to the type of malfunction (details of the malfunction) indicated by the malfunction category.

Such defect correlation information may be acquired in advance from the external device 10 and stored in the defect correlation information DB 120.

<<Mathematical Algorithm 130>>
The mathematical algorithm 130 is an algorithm executed by a predetermined mathematical model, and in this embodiment, a case where a neural network is used will be described as an example. Note that the mathematical algorithm 130 is not limited to a neural network, and may be another regression model such as XGBoost (eXtreme Gradient Boosting/gradient boosting regression tree).

The mathematical algorithm 130 calculates a causality coefficient α that indicates the strength of the relationship between nodes in the defect correlation information.

Figure 3 shows an example of a neural network. As shown in the figure, the neural network is composed of an input layer that corresponds to defect-related information (internal system data, etc.), a certain number of intermediate layers (in the example shown, there are three layers, from layer 1 to layer 3), and an output layer that indicates possible defect causes. The model configuration of such a neural network is defined by parameter information, which will be described later.

The mathematical algorithm 130 takes the defect-related information and the defect correlation information as input, and outputs a causality coefficient α that indicates the strength of the relationship between the upper node and the lower node in the defect correlation information.

Note that Figure 3 shows the state in which the strengths of relationships αw, αx, αy, and αz between the upper node, i.e., the camera system failure, and each of the lower nodes, i.e., the sensor failure, the recognition AI _HW (Hardware) failure, the communication _LSI failure, and the _{communication cable} failure, are output in the failure correlation information illustrated in Figure 2.

<<Parameter Information DB 140>>
The parameter information DB 140 is a database that stores parameters of the mathematical algorithm 130. Specifically, the parameters have predetermined parameter elements such as an industry, a defect category, a model configuration, a weighting coefficient, a bias, and an activation function.

Figure 4 shows an example of a parameter element. Here, the industry is the industry to which the electronic/electric system to be diagnosed belongs. The malfunction type is information equivalent to the TOP event in the malfunction correlation information (for example, "emergency brake malfunction" as shown in Figure 2).

The model configuration is information that indicates the configuration of the mathematical algorithm 130. In the case of a neural network, the model configuration defines the number of input layers, intermediate layers, and output layers.

The weighting coefficient is a coefficient that indicates the weight of the connections between nodes (neurons) in the input layer, intermediate layer, and output layer. The bias is information that indicates the bias (ease of flow) of the connections between each node (neuron). The activation function is information that indicates the type of nonlinear transformation function used in the mathematical algorithm 130. Examples of activation functions include ReLU (Rectified Linear Unit) and Step function.

Such parameter information exists for each industry of the electronic/electric system being diagnosed, and is stored in the parameter information DB 140.

<<Normalization Transformation Coefficient DB 150>>
The normalization conversion coefficient DB 150 is a database that stores a normalization conversion coefficient β. The normalization conversion coefficient β is a coefficient obtained by normalizing a standard value determined based on safety and reliability standards defined for each industry as a cross-industry relative value.

5 is a diagram showing an example of the normalization conversion coefficient β. As shown in the figure, the normalization conversion coefficient β is a value (such as 0.001 for β 1-1 and 0.01 for β 3-2) normalized as a cross-industry relative value, for example, ASIL-D, which is a standard value for communication _IF in the automobile industry, or SIL- ₂ , which is a standard value for cameras in the industrial robot industry.

In this way, the normalization conversion coefficient β exists for each industry of the electronic/electric system being diagnosed, and is stored in the normalization conversion coefficient DB 150.

In addition, each of the defect-related information, defect correlation information, mathematical algorithm, parameter information, and normalization conversion coefficients may be obtained directly from the external device 10 when each process described below is executed, and used in each process.

<<Defect History Related Information DB160>>
The defect history related information DB 160 is a database that stores defect history related information. The defect history related information includes, for example, defect related information acquired from the external device 10, a defect diagnosis result, measures taken by the user in response to the diagnosis result, and information on whether the defect has been resolved.

<<Failure diagnosis program 211>>
The fault diagnosis program 211 is a program for diagnosing faults in electronic/electric systems. Specifically, the fault diagnosis program 211 identifies the industry and fault category of the electronic/electric system in which the fault has occurred based on fault-related information. The fault diagnosis program 211 also executes diagnosis to identify the cause of the fault using fault correlation information, the mathematical algorithm 130, parameter information, and normalization conversion coefficients corresponding to the identified industry and fault category. Details of these processes (fault category identification process, diagnostic model generation process, and diagnosis process) will be described later.

<<Learning Program 212>>
The learning program 212 is a program that performs machine learning of the mathematical algorithm 130 depending on the user's response to the diagnosis result and whether or not the defect has been resolved. The details of the learning process will be described later.

The above describes the details of the diagnostic system 100.

<Functional configuration of diagnostic system 100>
6 is a diagram showing an example of the functional configuration of the diagnostic system 100. The illustrated functional units are classified according to the main processing contents in order to facilitate understanding of the functions realized by the processor 20 reading each program stored in the memory resource 30. Therefore, the present invention is not limited by the way in which the functions are classified or the names of the functional units.

The fault diagnosis unit 300 is a functional unit classified to facilitate understanding of the functions realized by the processor 20 reading the fault diagnosis program 211. The fault diagnosis unit 300 is further classified into a fault category identification unit 302 and a diagnosis unit 301 according to the processing content.

The learning unit 310 is a functional unit that is classified to facilitate understanding of the functions that are realized by the processor 20 reading the learning program 212. The learning unit 310 is further classified into an information registration unit 311 and a relearning unit 312 according to the processing content.

Note that some of the functional units may be constructed using hardware (such as an integrated circuit such as an ASIC) implemented in a computer. Furthermore, the processing of each functional unit may be executed by a single piece of hardware, or may be executed by multiple pieces of hardware.

<Processing Description>
7 is a flow diagram showing an example of the defect category identification process, the diagnostic model generation process, the diagnostic process, and the learning process. The details of these processes will be described in order below.

Note that the subject of each process is the processor 20 that has loaded the program stored in the memory resource 30, but to make it easier to understand the characteristics of each process, the following description will focus on the functional units realized by each program as the subject of the process.

The defect category identification process is started, for example, when an instruction to execute the process is received from a user (operator).

When the process starts, the defect category identification unit 302 acquires defect-related information from the defect-related information DB 110 (step S010). Note that the defect-related information to be acquired may be specified by the user, or may be acquired in the order in which it is stored in the defect-related information DB 110 (e.g., oldest first).

Next, the defect category identification unit 302 analyzes the defect-related information to identify a defect category that indicates the target field (industry) of the electronic/electric system and the type of defect (defect content) (step S011). The defect category identification unit 302 identifies the defect category based on, for example, an analysis of information recorded in the system internal data and probe data.

Once the defect category is identified, the defect category identification unit 302 proceeds to step S020.

In step S020, the diagnosis unit 301 executes a diagnostic model generation process. Specifically, the diagnosis unit 301 generates a diagnostic model that is a set of defect correlation information corresponding to a defect category, a mathematical algorithm 130, parameter information, and the like.

FIG. 8 is a flow diagram showing the details of the diagnostic model generation process. As shown in the figure, the diagnostic unit 301 determines whether or not there is parameter information for the industry (in this case, the first industry) corresponding to the defect category (step S021). Specifically, the diagnostic unit 301 searches the parameter information DB 140 and determines whether or not there is parameter information corresponding to the first industry.

If it is determined that there is corresponding parameter information (Yes in step S021), the diagnosis unit 301 proceeds to step S022. On the other hand, if it is determined that there is no corresponding parameter information (No in step S021), the diagnosis unit 301 proceeds to step S024.

In step S022, the diagnosis unit 301 acquires defect correlation information corresponding to the defect category, parameter information for the first industry, and the mathematical algorithm 130 from the corresponding databases.

Then, the diagnosis unit 301 generates a diagnosis model that sets the acquired information (step S023). That is, if there is parameter information for an industry (first industry) that corresponds to the defect category to be diagnosed, the diagnosis model is composed of a set of information including the parameter information for that industry, defect correlation information that corresponds to the defect category (particularly, the type (content) of the defect), and the mathematical algorithm 130.

In step S024, the diagnosis unit 301 acquires defect correlation information corresponding to the defect category, existing parameter information of another industry (a second industry different from the first industry), the mathematical algorithm 130, and normalization conversion coefficients of the first industry and the second industry. Then, the diagnosis unit 301 generates a diagnosis model that sets the acquired information (step S023).

In other words, if there is no parameter information for an industry (first industry) that corresponds to the defect category to be diagnosed, the diagnostic model is composed of a set of information including existing parameter information for an industry (second industry) different from the industry, defect correlation information corresponding to the defect category, a mathematical algorithm 130, and normalization conversion coefficients corresponding to the first industry and the second industry.

Next, the diagnosis unit 301 sets the mathematical algorithm 130 based on the acquired parameter information (step S024). After setting the mathematical algorithm 130, the diagnosis unit 301 transitions to step S030.

In step S030, the diagnosis unit 301 executes a diagnosis process using the diagnosis model. Specifically, the diagnosis unit 301 uses the diagnosis model to calculate the occurrence probability of each defect cause.

FIG. 9 is a flow diagram showing the details of the diagnosis process. As shown in the figure, the diagnosis unit 301 determines whether the mathematical algorithm 130 is set based on parameter information of the industry (first industry) corresponding to the defect category (step S0301).

If it is determined that the setting is based on the parameter information of the first industry (Yes in step S0301), the diagnosis unit 301 proceeds to step S0302. On the other hand, if it is determined that the setting is not based on the parameter information of the first industry (No in step S0301), the diagnosis unit 301 proceeds to step S0304.

In step S0302, the diagnosis unit 301 inputs the defect-related information and the defect correlation information to the mathematical algorithm 130 set based on the parameter information of the first industry, and obtains an output of the causal coefficient α between the defect factors. Specifically, the diagnosis unit 301 inputs the defect-related information and the defect correlation information corresponding to the defect category to the mathematical algorithm 130 (neural network) shown in FIG. 3, and obtains an output of the causal coefficients _αa to _αz between the defect factors.

The diagnosis unit 301 also calculates the occurrence probability of each defect factor based on the causal coefficients α _a to α _z (step S0303).

10 is a diagram showing the relationship between defect factors and occurrence probability. The diagnosis unit 301 calculates the occurrence probability γ _a to γ _z of each defect factor based on the causal coefficients α _a to α _z . In this example, the mathematical algorithm 130 is set based on parameter information corresponding to the defect category, so the relationship of occurrence probability γ=α holds. Therefore, the diagnosis unit 301 regards α _a to α _z output by the mathematical algorithm 130 as the occurrence probability of each defect factor branched off from the upper node.

In step S0304, the diagnosis unit 301 inputs the defect-related information and defect correlation information to the mathematical algorithm 130 set based on the parameter information of the second industry, and obtains an output of the causal coefficient α between defect factors.

In this case, the diagnosis unit 301 normalizes (multiplies) the causal coefficient α based on the ratio between the normalized conversion coefficient for the first industry and the normalized conversion coefficient for the second industry to calculate the occurrence probability γ of each defect cause.

11 is a diagram showing the relationship between defect factors and occurrence probability. The diagnosis unit 301 calculates occurrence probability γ _a to γ _z of each defect factor by multiplying the causal coefficient α by the ratio between the normalization conversion coefficient in the first industry and the normalization conversion coefficient in the second industry. Specifically, the diagnosis unit 301 calculates a value obtained by multiplying the corresponding α _n by β _n = normalization conversion coefficient in the first industry/normalization conversion coefficient in the second industry, as the occurrence probability γ _n of the defect factor.

When parameter information of an industry (first industry) corresponding to the defect category to be diagnosed exists (step S0303), the causal coefficients α _a to α _z output from the mathematical algorithm 130 set based on the parameter information can be set as the defect occurrence probability γ. On the other hand, in the case of step S0304, the mathematical algorithm 130 is set based on parameter information of a second industry different from the first industry. Therefore, the diagnosis unit 301 multiplies the causal coefficient α by the ratio between the normalization conversion coefficient corresponding to the first industry and the normalization conversion coefficient corresponding to the second industry, thereby absorbing the difference in the standards for the functional safety levels in each industry and calculating the occurrence probability of the defect cause.

Note that after performing the processing of step S0303 or step S0305, the diagnosis unit 301 transitions the processing to step S031.

In step S031, the diagnostic unit 301 outputs the malfunction factors with a high occurrence probability as the diagnosis result. Specifically, the diagnostic unit 301 identifies a predetermined number (e.g., three) of malfunction factors in descending order of occurrence probability. The diagnostic unit 301 also generates display information of the diagnosis result that associates the identified malfunction factors with their occurrence probabilities. The diagnostic unit 301 also transmits the generated display information to the external device 10.

In addition, the display information of the diagnosis results may be output, for example, to a display device possessed by the diagnosis system 100.

FIG. 12 shows an example of the diagnosis result. As shown in the figure, the diagnosis result corresponds to a specific number of identified defect causes and their respective occurrence probabilities.

The user inputs the measures taken in response to the diagnosis results and whether or not the problem has been resolved via the input device of the external device 10, thereby transmitting (feeding back) input information regarding the response status to the diagnosis system 100.

Next, the information registration unit 311 of the diagnostic system 100 associates the diagnostic results with the feedback from the user and registers them in the defect history related information DB 160 (step S040).

Next, the re-learning unit 312 executes machine learning of the mathematical algorithm 130 based on the input information such as the response status (step S040). The re-learning unit 312 also updates the parameter information of the mathematical algorithm 130 based on the machine learning (step S041). Specifically, the re-learning unit 312 updates parameter values of the parameter information, such as weighting coefficients and biases, based on the machine learning.

If parameter information for an industry corresponding to a defect category exists and a diagnosis is made using that parameter information, the re-learning unit 312 stores updated information for that parameter information set in the mathematical algorithm 130 in the parameter information DB 140 (overwriting and saving the existing parameter information). On the other hand, if parameter information for an industry corresponding to a defect category does not exist, i.e., if a defect is diagnosed using parameter information from another industry, the re-learning unit 312 stores the updated parameter information in the parameter information DB 140 as new parameter information for the industry corresponding to the defect category to be diagnosed.

The above describes in detail each process performed by the diagnostic system 100.

In this way, the diagnostic system can identify the cause of defects with greater accuracy, and can also create a diagnostic model that can share defect knowledge with other industries.

In particular, this system can calculate the causal coefficient between defect factors using a mathematical algorithm set based on parameter information that exists for each industry, making it possible to identify defect factors with greater accuracy.

In addition, even if there is no industry parameter information corresponding to the defect category to be diagnosed, this system can calculate the occurrence probability of defect factors that absorbs the differences in standards for functional safety levels in each industry by using α normalized based on the ratio of the normalization conversion coefficient β. Therefore, with this system, even when diagnosing defects in a new industry, it is possible to efficiently build a diagnostic model by utilizing defect knowledge from other industries.

The malfunction correlation information included in the configuration of the diagnostic model is not limited to malfunction correlation information of the industry corresponding to the electronic/electric system to be diagnosed. For example, if there is no malfunction correlation information of a TOP event corresponding to the type (content) of the malfunction in the industry corresponding to the electronic/electric system to be diagnosed (e.g., the first industry), the diagnosis unit 301 may include malfunction correlation information of the same or a similar other industry (e.g., the second industry) in the diagnostic model.

Whether defect correlation information from different industries is the same or similar can be determined based on, for example, the TOP events and defect causes (nodes) included in the defect correlation information, or the branching between nodes. Whether defect correlation information is the same or similar can be determined by linking and managing the information in advance.

As a result, even if no fault correlation information is registered for the industry corresponding to the electronic/electric system being diagnosed, the diagnostic system 100 can calculate the occurrence probability of the fault cause using a diagnostic model that includes the same or similar fault correlation information for a different industry. As a result, the diagnostic system 100 can perform fault diagnosis that is compatible with a wide range of industries.

Second Embodiment
Next, a second embodiment of the system 100 will be described. In the first embodiment described above, β is used, which is a standard value of functional safety levels in each industry of electronic/electric systems normalized as a cross-industry relative value. In contrast, in the present embodiment, β is used, which is a design method normalized as a cross-industry relative value by utilizing safety/reliability design information of manufacturers of electronic/electric systems.

Note that the configuration and each process of the diagnostic system 100 in this embodiment and the following embodiments are similar to those in the first embodiment, so detailed explanations will be omitted.

13 is a diagram showing an example of the normalization conversion coefficient β. As shown in the figure, the normalization conversion coefficient β in this embodiment is a value (such as 0.01 for β 1-1 and 0.0001 for β _{2-2 )} normalized as an industry-wide relative value, for example, a communication IF design method (shielded twisted pair cable) at electronic/electric system manufacturer A and a camera design method (Z company camera) at manufacturer B.

By using this normalization conversion coefficient β, the diagnostic system can reflect the defect factor ratio in the diagnostic model based on the manufacturer's reliability design information. As a result, it is possible to reduce error factors due to differences in manufacturers and achieve high-precision defect diagnosis.

Third Embodiment
The diagnostic system 100 according to the third embodiment uses a value calculated based on a calculation formula for a failure rate prediction model defined in, for example, FIDES or IEC/TR62380 and the operational environment information (user information, environmental information, infrastructure information, and product/service information) of the defect-related information as the normalization conversion coefficient β.

Normally, for components used in electronic and electric systems, a formula is established to calculate the failure rate, which changes depending on the usage conditions. The diagnosis unit 301 calculates a normalization conversion coefficient β each time a diagnosis is performed based on the failure rate prediction model (formula) and defect-related information, which includes the usage environment and operating time, and uses this coefficient to calculate the occurrence probability of the defect cause.

By using this normalization conversion coefficient β, the diagnostic system can reflect the failure probability according to the usage state of each component that makes up the electronic/electric system in the diagnostic model. As a result, it is possible to achieve high accuracy in fault diagnosis.

<Fourth embodiment>
The diagnostic system 100 according to the fourth embodiment changes the normalization conversion coefficient β corresponding to the updated part when the system components of the electronic/electric system are updated (changed) through an OTA (Over The Air)/HW (Hardware) update.

For example, if an OTA/HW update changes the components used in the electronic/electric system, communication LSI 1 and communication cable 1, to communication LSI 2 and communication cable 2, the diagnosis unit 301 calculates the occurrence probability of the malfunction factor using a normalized conversion coefficient β that normalizes the failure rate corresponding to the changed components.

The diagnostic system uses the normalization conversion coefficient β that corresponds to the changed parts as appropriate in response to updates to the components of such electronic and electric systems. This allows the diagnostic system to achieve high-precision fault diagnosis even if there are changes to the component configuration of the electronic and electric systems.

Fifth Embodiment
When diagnosing faults in an electronic/electric system for which no diagnostic model has been generated or an electronic/electric system with a small number of diagnostic records (machine learning records), the diagnostic system 100 of the fifth embodiment uses a diagnostic model of an electronic/electric system that is highly similar in a different industry and that has already accumulated a large number of diagnostic records.

FIG. 14 shows an example of the similarity judgment result for electronic/electric systems. As shown in the figure, the similarity between electronic/electric systems is judged based on certain viewpoints such as system configuration components, connection topology, and system functions. In the example shown, it is judged that recognition system 1 has a system configuration that is most similar to the system to be diagnosed. In this case, the diagnosis unit 301 performs fault diagnosis using the diagnosis model of recognition system 1. However, for the normalization transformation β included in the diagnosis model, the diagnosis unit 301 performs diagnosis using the normalization transformation β that corresponds to the industry of the electronic/electric system to be diagnosed.

The process of determining the degree of similarity is performed, for example, by the diagnosis unit 301, which performs fault diagnosis using a diagnostic model of the electronic/electric system with the most similar system configuration.

In this way, by using diagnostic models of similar electronic and electric systems, the diagnostic system can achieve high accuracy even when diagnosing electronic and electric systems for which no diagnostic models have been generated or for which there is a small number of diagnostic records.

Sixth Embodiment
The diagnostic system 100 of the sixth embodiment calculates the probability of occurrence of a malfunction factor in similar electronic/electric systems using a normalization conversion coefficient β corresponding to the usage environment associated with the servicing of each electronic/electric system.

Figure 15 shows an example of the usage environment for each service of similar electronic and electric systems. As shown in the figure, a logistics EV (Electric Vehicle) fleet, an autonomous taxi, and an autonomous bus are services that utilize similar electronic and electric systems. However, the environments, continuous operating times, and maintenance frequencies of the electronic and electric systems in these services are mutually different. Therefore, even if electronic and electric systems have similar system configurations, the parts that are prone to failure differ depending on the usage environment associated with the service.

Taking these characteristics into consideration, the diagnostic system 100 stores in the normalization conversion coefficient DB a normalization conversion coefficient β normalized according to the failure rate of each component for each service. The diagnostic unit 301 then identifies the service environment of the electronic/electric system to be diagnosed based on the malfunction-related information, and performs malfunction diagnosis using the normalization conversion coefficient β according to the environment.

This allows the diagnostic system to achieve high-precision fault diagnosis.

<Service type 1>
16 is a schematic diagram of service form 1. Service form 1 is an example of launching a fault diagnosis service for a new industry. Specifically, it is an example of building a new fault diagnosis service for service robots.

The diagnostic system 100 is based on the premise that a fault diagnosis service for automobiles and industrial robots has already been established, and that the mathematical algorithm 130 based on actual fault history has progressed in learning. It is also based on the premise that industry standard values for service robots exist as safety and reliability criteria.

First, the diagnostic system 100 registers in the normalization conversion coefficient DB 150 the normalization conversion coefficient β corresponding to the service robot industry, which is calculated by normalizing the standard values determined based on the safety and reliability standards defined in the service robot industry as cross-industry relative values.

Note that during the launch of a fault diagnosis service for a new industry, parameter information for that industry (service robots) does not exist. Therefore, the diagnosis system 100 utilizes the causality coefficient α of an electronic/electric system in another industry that is similar to the electronic/electric system of the service robot (in the illustrated example, the arm drive system and connected system).

Specifically, the diagnosis system 100 uses the causality coefficient _α1 existing in the automobile industry for the recognition system of the service robot, and uses the causality coefficient _α3 existing in the industrial robot industry for the arm drive system of the service robot.

In this way, when the diagnostic system 100 launches a fault diagnosis service for a new industry, it starts the fault diagnosis service by using the causal coefficient α of a similar electronic/electric system that has already been learned.

In addition, by advancing machine learning as a service robot through operation, the diagnostic system 100 can generate and update parameter information that matches the actual situation of the service robot. As a result, the diagnostic system 100 can build and operate a highly accurate defect diagnosis service even for new industries.

Furthermore, such a service form 1 can be realized, for example, by applying and utilizing the diagnostic system 100 according to the fifth embodiment described above.

<Service type 2>
17 is a schematic diagram of service form 2. Service form 2 is similar to service form 1 in that it is an example of a case where a new fault diagnosis service for service robots is constructed. However, this service form 2 assumes a case where there are no standard values for safety and reliability standards that are the industry standard in the service robot industry.

In this case, the diagnostic system 100 uses both the causality coefficient α of similar electronic/electric systems in the automotive and industrial robot industries, where the learning of the mathematical algorithm 130 is advanced, and the normalization conversion coefficient β of similar electronic/electric systems in these industries.

Specifically, the diagnosis system 100 uses the causality coefficient _α1 and normalization conversion coefficient _β1 existing in the automobile industry for the recognition system of the service robot, and uses the causality coefficient _α3 and normalization conversion coefficient _β3 existing in the industrial robot industry for the arm drive system of the service robot.

In this way, when diagnostic system 100 launches a fault diagnosis service for a new industry, it starts the fault diagnosis service by using the causality coefficient α of a similar electronic/electric system that has already undergone learning, and the normalization conversion coefficient β corresponding to that industry. In addition, diagnostic system 100 generates and updates parameter information tailored to the actual situation of the service robot by advancing machine learning as a service robot through operation.

As a result, the diagnostic system 100 can build and operate a highly accurate fault diagnosis service even for new industries.

In addition, when a standard value for safety and reliability standards that will become the industry standard for the service robot industry is established, the diagnostic system 100 can use an industry-wide normalization conversion coefficient that normalizes the relative value between the standard value and the standard values of other industries as the normalization conversion coefficient corresponding to the service robot industry.

In addition, the computer related to such a diagnostic system 100 may function as a program distribution server that distributes at least the programs in the memory resource 30 to other computers so that the programs can be executed on the other computers.

Furthermore, the present invention is not limited to the above-mentioned embodiments and modifications, but includes various modifications within the scope of the same technical idea. For example, the above-mentioned embodiments have been described in detail to clearly explain the present invention, and are not necessarily limited to those having all of the configurations described. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

In addition, in the above explanation, the control lines and information lines are those that are considered necessary for the explanation, and do not necessarily show all the control lines and information lines in the product. In reality, it can be assumed that almost all components are interconnected.

100: diagnostic system, 20: processor, 30: memory resource, 40: NI (network interface device), 50: UI (user interface device), 110: defect-related information DB, 120: defect correlation information DB, 130: mathematical algorithm, 140: parameter information DB, 150: normalization conversion coefficient DB, 160: defect history-related information DB, 210: program, 211: defect diagnosis program, 212: learning program, 300: defect diagnosis unit, 301: diagnosis unit, 302: defect category identification unit, 310: learning unit, 311: information registration unit, 312: relearning unit, 10: external device, N: network

Claims (14)

1. 1. A diagnostic system having one or more processors and one or more memory resources, comprising:
   The memory resource stores a fault diagnosis program for diagnosing a fault in an electronic/electric system;
   The processor executes the fault diagnosis program to:
   Identifying the industry and type of defect of the electronic/electric system to be diagnosed based on the acquired defect-related information indicating the defect state of the electronic/electric system;
   Calculating a causal coefficient indicating a strength of a relationship between defect factors corresponding to the defect types, which is output by a predetermined mathematical algorithm in which parameter information corresponding to the identified first industry is set, as an occurrence probability of the defect factors;
   A diagnostic system comprising: a diagnostic unit for performing a diagnosis to identify a cause of the defect based on the occurrence probability.
2. 2. The diagnostic system of claim 1,
   The processor executes the fault diagnosis program to:
   When there is no parameter information corresponding to the identified first industry, the causal coefficient output by the mathematical algorithm set with parameter information of a second industry different from the first industry is normalized by a normalization conversion coefficient obtained by normalizing an industry-specific standard value from a predetermined perspective as a cross-industry relative value, the normalization conversion coefficient being a ratio between the first industry and the second industry, thereby calculating the occurrence probability of the defect cause.
3. 3. The diagnostic system according to claim 1 or 2,
   The memory resource further stores a learning program for performing machine learning of the mathematical algorithm;
   The processor executes the learning program to:
   updating the parameter information by executing machine learning of the mathematical algorithm based on information regarding measures taken against the identified cause of the malfunction and whether the malfunction has been resolved;
   and storing the diagnostic information in the memory resource as parameter information corresponding to the industry in which the diagnostic was performed.
4. 3. The diagnostic system according to claim 1 or 2,
   The processor executes the fault diagnosis program to:
   A diagnostic system characterized by identifying defect correlation information that indicates a correlation that is identical or similar to the correlation between defect factors corresponding to the identified defect type, the defect correlation information corresponding to an industry different from the first industry, and performing a diagnosis to identify the defect factors using the defect correlation information.
5. 3. The diagnostic system of claim 2,
   A diagnostic system characterized in that the normalization conversion coefficient is a value obtained by normalizing an industry standard specification value regarding the functional safety level of the electronic/electric system as an industry-wide relative value.
6. 3. The diagnostic system of claim 2,
   A diagnostic system characterized in that the normalization conversion coefficient is a value obtained by normalizing a design method related to safety and reliability design of a manufacturer of the electronic/electric system as an industry-wide relative value.
7. 3. The diagnostic system of claim 2,
   A diagnostic system characterized in that the normalization conversion coefficient is a value calculated based on a calculation formula of a failure rate prediction model of components used in the electronic/electric system and information indicating an operating environment.
8. 3. The diagnostic system of claim 2,
   The processor executes the fault diagnosis program to:
   A diagnostic system characterized in that, when a change is made to a system component of the electronic/electric system, the occurrence probability of the malfunction factor is calculated using a normalization conversion coefficient β corresponding to the changed part.
9. 3. The diagnostic system of claim 2,
   The processor executes the fault diagnosis program to:
   determining a similarity between a system configuration of the electronic/electric system in the first industry that is a diagnosis target and a system configuration of the electronic/electric system in the second industry;
   A diagnostic system characterized by calculating a probability of occurrence of the defect cause using the causal coefficient output by the mathematical algorithm in which the parameter information corresponding to the electronic/electric system in the second industry with the high similarity is set, and the normalization conversion coefficient corresponding to the first industry.
10. 3. The diagnostic system of claim 2,
    The processor executes the fault diagnosis program to:
    A diagnostic system comprising: a diagnostic unit for calculating a probability of occurrence of the malfunction factor using a normalization conversion coefficient normalized based on a failure rate of a component that differs depending on a service provided by the electronic/electric system;
11. A diagnostic method performed by a diagnostic system having one or more processors and one or more memory resources, comprising:
    The memory resource stores a fault diagnosis program for diagnosing a fault in an electronic/electric system;
    The processor executes the fault diagnosis program to:
    Identifying an industry and a type of defect of the electronic/electric system to be diagnosed based on acquisition of defect-related information indicating a defect state of the electronic/electric system;
    calculating a causal coefficient indicating a strength of a relationship between defect factors corresponding to the defect types, the causal coefficient being output by a predetermined mathematical algorithm in which parameter information corresponding to the identified first industry is set, as an occurrence probability of the defect factors;
    a diagnosis step of identifying a cause of the defect based on the occurrence probability.
12. The diagnostic method according to claim 11,
    The processor executes the fault diagnosis program to:
    When there is no parameter information corresponding to the identified first industry, the diagnostic method includes a step of calculating a probability of occurrence of the defect cause by normalizing the causal coefficient output by the mathematical algorithm set with parameter information of a second industry different from the first industry by a normalization conversion coefficient obtained by normalizing an industry-specific standard value from a predetermined perspective as a cross-industry relative value, the normalization conversion coefficient corresponding to the first industry and the second industry.
13. A program for a diagnostic system having one or more processors and one or more memory resources, the program being read from the memory resource and executed by the processor,
    The memory resource stores a fault diagnosis program for diagnosing a fault in an electronic/electric system;
    The fault diagnosis program executed by the processor includes:
    Identifying the industry and type of defect of the electronic/electric system to be diagnosed based on the acquired defect-related information indicating the defect state of the electronic/electric system;
    Calculating a causal coefficient indicating a strength of a relationship between defect factors corresponding to the defect types, which is output by a predetermined mathematical algorithm in which parameter information corresponding to the identified first industry is set, as an occurrence probability of the defect factors;
    A program for performing a diagnosis to identify a cause of the defect based on the occurrence probability.
14. The program according to claim 13,
    The fault diagnosis program executed by the processor includes:
    When there is no parameter information corresponding to the identified first industry, the program calculates the occurrence probability of the defect cause by normalizing the causal coefficient output by the mathematical algorithm in which parameter information of a second industry different from the first industry is set, using a normalization conversion coefficient obtained by normalizing an industry-specific standard value from a predetermined perspective as a cross-industry relative value, the normalization conversion coefficient being a ratio between the first industry and the second industry.

Images