System and method for diagnosing failure cause of work machine at fault

Abstract

To rapidly and accurately diagnose a failure cause of a work machine at fault.SOLUTION: A system for diagnosing a failure cause of a work machine at fault includes: a data acquisition section; a learned model of generative AI; and an output section. The data acquisition section acquires machine data indicating a state of a work machine at fault. The learned model has learned according to a failure handling manual showing a state of a work machine at fault and candidates of a failure cause, and according to failure history data showing a state of other work machines that have failed in the past. The learned model generates answers containing at least text data showing a failure cause of a work machine at fault, with respect to inputted machine data. The output section outputs the answer generated by the learned model.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to a system and method for diagnosing the cause of a malfunction in a failed work machine. [Background technology]

[0002] Conventionally, systems for diagnosing the cause of a breakdown in a work machine are known. For example, a fault diagnosis device disclosed in Patent Document 1 stores data such as diagnostic procedures, fault items, and reference values ​​that are previously written in a repair manual. The fault diagnosis device receives a fault code signal from a controller of the work machine. The fault diagnosis device performs a fault diagnosis based on the fault code signal and the stored data, and displays the diagnostic procedure and other information on a display unit. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 06-011419 Summary of the Invention [Problem to be solved by the invention]

[0004] In the above-described diagnostic procedure, the cause of the failure is not immediately identified; instead, multiple candidate parameters estimated to be related to the failure are presented, and analysis of the data for each parameter is required. Therefore, it is not easy to quickly identify the cause of the failure. Furthermore, it is not easy to accurately identify the cause of the failure from the analysis results. An object of the present disclosure is to quickly and accurately diagnose the cause of a failure in a broken work machine. [Means for solving the problem]

[0005] A system according to one aspect of the present disclosure is a system for diagnosing the cause of a failure of a broken down work machine. The system includes a data acquisition unit, a trained model of generative AI (artificial intelligence), and an output unit. The data acquisition unit acquires machine data indicating the state of the broken down work machine. The trained model has been trained using a failure response manual indicating the state of the broken down work machine and potential causes of the failure, and failure history data indicating the states of other work machines that have broken down in the past. The trained model is configured to generate an answer including at least text data indicating the cause of the failure of the broken down work machine in response to input machine data. The output unit outputs the answer generated by the trained model.

[0006] A method according to another aspect of the present disclosure is a computer-executed method for diagnosing the cause of a failure of a broken down work machine, the method comprising: acquiring machine data indicating the state of the broken down work machine; inputting the machine data into a trained model of a generation AI that has been trained using a failure response manual indicating the state of the broken down work machine and candidate causes of the failure and failure history data indicating the states of other work machines that have broken down in the past, and that is configured to generate, in response to the input machine data, an answer including at least text data indicating the cause of the failure of the broken down work machine; and outputting the answer generated by the trained model. [Effects of the Invention]

[0007] According to the present invention, the trained model of the generation AI generates a response to input machine data that includes at least text data indicating the cause of a fault in a broken-down work machine. The trained model has been trained using a fault response manual that indicates the state of the broken-down work machine and potential causes of the fault, and fault history data that indicates the state of other work machines that have broken down in the past. As a result, the cause of a fault in a broken-down work machine can be diagnosed quickly and accurately. [Brief explanation of the drawings]

[0008] [Figure 1]1 is a block diagram showing the configuration of a fault diagnosis system for a work machine. [Figure 2] FIG. 10 is a diagram showing an example of a failure response manual. [Figure 3] 10 is a flowchart showing a process for performing fault diagnosis by a generation AI. [Figure 4] FIG. 10 is a diagram illustrating an example of an application screen for fault diagnosis. [Figure 5] FIG. 10 is a diagram showing an example of an answer to a first diagnostic result displayed in an answer field. [Figure 6] FIG. 10 is a diagram showing an example of a second question. [Figure 7] FIG. 10 is a diagram showing an example of a program code answer from the generation AI displayed in the answer field. [Figure 8] FIG. 10 is a diagram showing an example of a second diagnostic result displayed on an application screen. [Figure 9] 10 is a flowchart showing a process for performing fault diagnosis by a generation AI according to a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0009] A fault diagnosis system according to an embodiment will now be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a fault diagnosis system 1 according to an embodiment. The fault diagnosis system 1 is a system for diagnosing the cause of a fault in a broken down work machine 100. The work machine 100 is, for example, a construction machine such as a bulldozer, shovel, wheel loader, or grader. However, the work machine 100 may be a machine other than a construction machine. As shown in FIG. 1, the fault diagnosis system 1 includes a local computer 2 and a server 3.

[0010] The local computer 2 is a mobile communication terminal such as a smartphone or a tablet, or may be another computer such as a desktop computer or a laptop computer. The local computer 2 communicates with the server 3 via a communication network 4 such as the Internet.

[0011] The local computer 2 includes a data acquisition unit 11 and an output unit 12. The data acquisition unit 11 acquires machine data. The machine data is data that indicates the status of the broken down work machine 100. The data acquisition unit 11 is realized by, for example, a processor and memory. The machine data includes a fault code, status data of the work machine 100, and still images or videos of the work machine 100.

[0012] A fault code is an identifier assigned to each fault phenomenon of the work machine 100. The fault code is output by the controller 101 of the work machine 100 in accordance with the fault phenomenon detected in the work machine 100. For example, the data acquisition unit 11 acquires the fault code from the controller 101 of the work machine 100 via data communication. Alternatively, the data acquisition unit 11 may acquire the fault code by having the user input the fault code displayed on the display of the work machine 100 into the local computer 2.

[0013] The status data is time-series data of multiple status parameters of the work machine 100. The multiple status parameters are variables that indicate the status of the work machine 100. The multiple status parameters include, for example, the voltage, temperature, or pressure of various devices equipped on the work machine 100. Alternatively, the multiple status parameters may include the engine rotation speed, motor rotation speed, or vehicle speed of the work machine 100.

[0014] The controller 101 of the work machine 100 detects and records status data at a predetermined sampling period. The data acquisition unit 11 acquires status data for a predetermined time before and after the occurrence of a failure. The data acquisition unit 11 acquires the status data from the controller 101 of the work machine 100 via data communication. Alternatively, the data acquisition unit 11 may acquire the status data from the controller 101 of the work machine 100 via another computer.

[0015] Still images or video of the work machine 100 are acquired, for example, by a camera provided in the local computer 2. Alternatively, the data acquisition unit 11 may acquire still images or video of the work machine 100 via another computer. The output unit 12 outputs the diagnosis results of the cause of a fault in the work machine 100. The output unit 12 includes, for example, a display, and displays the diagnosis results as text data and images.

[0016] The server 3 receives machine data from the local computer 2. The server 3 is equipped with a trained model 13 of the generation AI (hereinafter simply referred to as "generation AI 20"). The generation AI 20 includes, for example, a large-scale language model. The server 3 uses the generation AI 20 to diagnose the cause of a failure of the work machine 100 based on the machine data. The generation AI 20 is configured to generate an answer indicating the cause of a failure of the work machine 100 for the input machine data. The answer indicating the cause of the failure includes a sentence explaining the cause of the failure. The generation AI 20 has been trained using a failure response manual and failure history data. The failure response manual indicates the state of the work machine 100 and potential causes of failure for each state.

[0017] FIG. 2 is a diagram showing an example of the failure response manual 14. The failure response manual 14 is in the form of text data. The failure response manual 14 includes an action level, a failure code, a failure phenomenon, a failure content, a cause, and a procedure. The action level indicates a level requiring urgency. The failure code corresponds to the failure code in the machine data described above. The failure phenomenon indicates a phenomenon occurring in the broken down work machine 100. The failure content indicates the failure content of the equipment that is causing the failure of the work machine 100. The cause indicates more detailed possible causes of the failure. The procedure indicates the procedure for confirming each cause or a reference value. In the example shown in FIG. 2, the DEF pump is a pump for supplying urea water in a device that purifies exhaust gas.

[0018] The failure history data indicates the status of other work machines that have broken down in the past, and includes failure codes, status data, and still images or video of a plurality of other work machines that have broken down in the past.

[0019] The local computer 2 communicates with the server 3 via, for example, a web application, thereby performing fault diagnosis of the work machine 100. Alternatively, the local computer 2 may perform fault diagnosis of the work machine 100 by communicating with the server 3 via an application installed on the local computer 2.

[0020] Fig. 3 is a flowchart showing the processing of the local computer 2 for performing fault diagnosis using the generation AI 20. Fig. 4 is a diagram showing an example of an application screen 15 for fault diagnosis of the work machine 100. The application screen 15 is displayed on the output unit 12. As shown in Fig. 3, in step S101, the local computer 2 acquires machine data.

[0021] As shown in FIG. 4, the application screen 15 includes a fault code input field 21. A user of the local computer 2 selects a fault code to input from a list of fault codes displayed in the input field 21. Alternatively, the user may be able to input the fault code directly into the input field 21. Alternatively, if the fault code is included in the machine data, a fault code extracted from the machine data may be used. The application screen 15 also includes a file upload button 22. By operating the upload button 22, the user can specify image data and machine data indicating a still image or video to be sent to the server 3. The application screen 15 may include an input field for the model of the work machine 100. Alternatively, the image data and machine data may be uploaded automatically.

[0022] In step S102, the local computer 2 determines whether image data is present. If image data to be uploaded is specified, in step S103, the local computer 2 transmits the fault code, image data, and first question to the server 3. As shown in FIG. 4 , the application screen 15 includes an execute button 23 for AI fault diagnosis. When the user operates the execute button 23, the local computer 2 transmits the fault code, image data, and first question to the server 3. If image data to be uploaded is not specified, the local computer 2 transmits the fault code and the first question to the server 3.

[0023] The first question is expressed, for example, as text data such as the following: "For fault code ****, please list the action level, fault phenomenon, and fault content in itemized form. Also, please list the monitoring code and monitoring parameter names that are presumed to be related to identifying the cause. Please list the cause in itemized form." Fault code **** is the fault code entered on application screen 15.

[0024] In the server 3, the fault code, image data, and first question are input to the generation AI 20. The generation AI 20 generates an answer to the first question as a first diagnostic result based on the fault code and image data. The server 3 transmits the answer generated by the generation AI 20 to the local computer 2.

[0025] In step S105, the local computer 2 acquires the answer of the first diagnostic result generated by the generation AI 20. The local computer 2 displays the answer of the first diagnostic result generated by the generation AI 20 on the application screen 15. As shown in FIG. 4, the application screen 15 includes an answer field 24. The local computer 2 displays the answer of the first diagnostic result generated by the generation AI 20 in the answer field 24.

[0026] Fig. 5 is a diagram showing an example of a response to the first diagnostic result displayed in the response field 24. As shown in Fig. 5, the response to the first diagnostic result is displayed as text data. The response to the first diagnostic result includes a fault code, an action level, a fault phenomenon, a fault content, a monitoring code, a monitoring parameter name, and a probable cause.

[0027] The monitoring parameter name is the name of a status parameter, out of multiple status parameters, that is presumed to be related to the failure of each failure code. If there are multiple status parameters that are presumed to be related to the failure, multiple monitoring parameter names are displayed in the answer field 24. The monitoring code indicates an identification code assigned to each monitoring parameter. The presumed cause indicates the failure content of equipment that is a candidate cause of the failure of the work machine 100. If there are multiple candidate causes of the failure, multiple presumed causes are displayed in the answer field 24.

[0028] In step S106, the local computer 2 transmits the second question and the status data of the monitoring parameters to the server 3. FIG. 6 is a diagram showing an example of the second question. As shown in FIG. 6, the second question is displayed as text data. The second question includes an answer to the first diagnostic result and an instruction to cause the generation AI 20 to generate program code for an analysis program for analyzing the monitoring parameters indicated in the first diagnostic result. The local computer 2 also transmits the status data of the monitoring parameters to the server 3.

[0029] In the server 3, the second question and the status data of the monitoring parameters are input to the generation AI 20. The generation AI 20 generates program code for an analysis program as an answer based on the monitoring parameters and probable causes in the second question and the status data of the monitoring parameters. The server 3 transmits the analysis program code generated by the generation AI 20 to the local computer 2.

[0030] For example, as shown in Figure 5, the first diagnostic result includes multiple monitoring parameters (DEF pump state, DEF pump pressure, DEF pump temperature, DEF pump pressure sensor voltage, and DEF injection amount) as monitoring parameters. The generation AI 20 generates program code for a frequency analysis program that detects the maximum amount of change in the waveform of each monitoring parameter.

[0031] In step S107, the local computer 2 acquires the program code of the analysis program generated by the generation AI 20. For example, as shown in Fig. 7, the local computer 2 displays the program code of the analysis program generated by the generation AI 20 in the answer field 24 as text data.

[0032] In step S108, the local computer 2 performs an analysis on the status data of each monitoring parameter. For example, the local computer 2 performs an analysis on each of the status data of the multiple monitoring parameters using the multiple analysis programs described above.

[0033] In step S109, the local computer 2 transmits the third question and the analysis results of each monitoring parameter to the server 3. The local computer 2 converts the analysis results of each monitoring parameter into text data and transmits it to the server 3. The third question is displayed as text data. The third question includes an instruction to identify the corresponding cause of the failure from the estimated cause of the first diagnosis result based on the analysis results of each monitoring parameter.

[0034] For example, the generation AI 20 determines which monitoring parameters deviate from the normal state based on the analysis results of each monitoring parameter, and identifies, from among the estimated causes of the first diagnosis result, the one that corresponds to the monitoring parameter that deviates from the normal state as the cause of the failure.

[0035] The generation AI 20 also determines the degree of abnormality of the monitoring parameters based on the analysis results of the monitoring parameters by the analysis program. For example, the generation AI 20 compares the analysis results of each monitoring parameter with a reference value, and determines that a value that deviates greatly from the reference value has a high degree of abnormality. The generation AI 20 generates a second diagnostic result response that includes the cause of the failure. The server 3 transmits the second diagnostic result response to the local computer 2.

[0036] In step S110, the local computer 2 acquires the answer to the second diagnostic result generated by the generation AI 20. In step S111, the local computer 2 displays the analysis results of each monitoring parameter and the answer to the second diagnostic result on the application screen 15. As shown in FIG. 4, the application screen 15 includes an analysis result display field 25. The local computer 2 displays a waveform indicating the analysis result of each monitoring parameter in the analysis result display field 25.

[0037] The local computer 2 displays the waveforms of the monitoring parameters in descending order of the degree of abnormality. For example, among the multiple monitoring parameters, a waveform 26 of the analysis result of a first monitoring parameter (pump state), a waveform 27 of the analysis result of a second monitoring parameter (pump pressure), and a waveform 28 of the analysis result of a third monitoring parameter (voltage of the pump pressure sensor) are displayed in a display field 25.

[0038] Furthermore, in step S110, the local computer 2 displays the answer of the second diagnostic result on the application screen 15. Fig. 8 is a diagram showing an example of the second diagnostic result displayed on the application screen 15. As shown in Fig. 8, the local computer 2 displays the cause of the failure as text data in the answer field 24.

[0039] In the fault diagnosis system 1 according to the present embodiment described above, the generation AI 20 generates a response to input machine data that includes at least text data indicating the cause of a fault in the work machine 100. The generation AI 20 has already learned using the fault response manual 14, which indicates the state of the work machine 100 and potential causes of the fault, and fault history data, which indicates the state of other work machines that have broken down in the past. Therefore, the cause of a fault in a broken work machine 100 can be diagnosed quickly and accurately.

[0040] Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications are possible without departing from the gist of the invention.

[0041] The configuration of the fault diagnosis system 1 is not limited to that of the above embodiment and may be modified. For example, the generation AI 20 may be installed in the local computer 2.

[0042] The process for performing fault diagnosis using the generation AI 20 is not limited to that of the above embodiment and may be modified. For example, Fig. 9 is a flowchart showing the process for performing fault diagnosis using the generation AI 20 according to a modified example. In the process for performing fault diagnosis using the generation AI 20 according to the modified example, in step S206, the second question includes an instruction to select an analysis program for analyzing the state data of the monitoring parameters from a plurality of analysis programs.

[0043] A plurality of analysis programs are input to the generation AI 20 in advance. For example, the plurality of analysis programs include a first analysis program, a second analysis program, and a third analysis program. The first analysis program is a program that calculates the standard deviation of the status data. The second analysis program is a program that calculates the difference between the maximum and minimum values ​​of the status data. The third analysis program is a program that calculates the maximum amount of change in the status data. The generation AI 20 selects an analysis program suitable for fault diagnosis from the plurality of analysis programs based on the results of the first analysis and the status data of the monitoring parameters.

[0044] 9, in step S207, the local computer 2 acquires the analysis program selected by the generation AI 20. In step S208, the local computer 2 analyzes the status data of the monitoring parameters using the analysis program selected by the generation AI 20. Note that other processes shown in FIG. 9 are similar to those shown in FIG. 3. [Industrial Applicability]

[0045] According to the present disclosure, the cause of a malfunction in a working machine can be diagnosed quickly and accurately. [Explanation of symbols]

[0046] 11: Data acquisition section 12: Output section 13: Generation AI 100:Work machinery

Claims

1. A system for diagnosing a cause of a malfunction of a broken down work machine, comprising: a data acquisition unit that acquires machine data indicating the state of the broken down work machine; a trained model of a generation AI that has been trained using a failure response manual that indicates the state of the broken-down work machine and candidate causes of the failure, and failure history data that indicates the states of other work machines that have broken down in the past, and that is configured to generate, in response to the input machine data, an answer that includes at least text data that indicates the cause of the failure of the broken-down work machine; an output unit that outputs the answer generated by the trained model; A system comprising:

2. the machine data includes time-series data of a plurality of state parameters that indicate the state of the failed work machine; The trained model determines, as monitoring parameters, one or more state parameters estimated to be related to the fault among the plurality of state parameters. The system of claim 1 .

3. The trained model is generating a program code for an analysis program for analyzing the time-series data of the monitoring parameters; The system of claim 2 .

4. The trained model is selecting an analysis program for analyzing the time-series data of the monitoring parameters from a plurality of analysis programs; The system of claim 2 .

5. the trained model generates the answer indicating the cause of the failure of the broken down work machine from the analysis results of the time series data of the monitoring parameters by the analysis program.

5. The system according to claim 3 or 4.

6. the trained model determines the degree of abnormality of the monitoring parameter from the analysis result of the time-series data of the monitoring parameter by the analysis program; the output unit displays waveforms of the time-series data of the monitoring parameters in descending order of abnormality in the response.

5. The system according to claim 3 or 4.

7. the machine data includes a fault code output by the failed work machine and time-series data of a plurality of status parameters that indicate the status of the failed work machine; the failure response manual includes a plurality of failure codes and monitoring parameters to be analyzed corresponding to each failure code among the plurality of status parameters; The trained model diagnoses the cause of the failure of the broken down work machine by analyzing the time series data of the monitoring parameters. The system of claim 1 .

8. the machine data includes a fault code output by the broken-down work machine, time-series data of a plurality of status parameters indicating the status of the broken-down work machine, and an image or video of the broken-down work machine, the failure history data includes a failure code of the other work machine that has previously failed, time-series data of a plurality of status parameters that indicate the status of the other work machine that has previously failed, and an image or video of the failed work machine. The system of claim 1 .

9. 1. A computer-implemented method for diagnosing a cause of a fault in a failed work machine, comprising: obtaining machine data indicative of a condition of the failed work machine; inputting the machine data into a trained model of a generation AI that has been trained using a failure response manual that indicates the state of the broken-down work machine and candidate causes of the failure, and failure history data that indicates the states of other work machines that have broken down in the past, and that is configured to generate, in response to the input machine data, a response that includes at least text data that indicates the cause of the failure of the broken-down work machine; outputting the answer generated by the trained model; A method for providing the above.

10. the machine data includes time-series data of a plurality of state parameters that indicate the state of the failed work machine; determining, by the learned model, one or more state parameters estimated to be related to the fault among the plurality of state parameters as monitoring parameters; 10. The method of claim 9.

11. generating a program code of an analysis program for analyzing the time series data of the monitoring parameters using the trained model; The method of claim 10.

12. selecting an analysis program for analyzing the time series data of the monitoring parameters from a plurality of analysis programs using the trained model. The method of claim 10.

13. generating the answer indicating a cause of failure of the broken-down work machine from an analysis result of the time-series data of the monitoring parameters by the analysis program using the trained model; 13. The method of claim 11 or 12.

14. determining an abnormality degree of the monitoring parameter from an analysis result of the time-series data of the monitoring parameter by the analysis program using the trained model; In the response, waveforms of the time-series data of the monitoring parameters are displayed in descending order of abnormality; 13. The method of claim 11 or 12, comprising:

15. the machine data includes a fault code output by the failed work machine and time-series data of a plurality of status parameters indicating the status of the failed work machine, the failure response manual includes a plurality of failure codes and monitoring parameters to be analyzed corresponding to each failure code among the plurality of status parameters; diagnosing a cause of failure of the broken down work machine by analyzing time series data of the monitoring parameters using the trained model.

10. The method of claim 9.

16. the machine data includes a fault code output by the faulty work machine, time-series data of a plurality of status parameters indicating the status of the faulty work machine, and an image or video of the faulty work machine, the failure history data includes a failure code of the other work machine that has previously failed, time-series data of a plurality of status parameters that indicate the status of the other work machine that has previously failed, and an image or video of the failed work machine.

10. The method of claim 9.

Images