WO2022208894A1 - 機械診断装置、機械診断方法、および記録媒体 - Google Patents
機械診断装置、機械診断方法、および記録媒体 Download PDFInfo
- Publication number
- WO2022208894A1 WO2022208894A1 PCT/JP2021/014402 JP2021014402W WO2022208894A1 WO 2022208894 A1 WO2022208894 A1 WO 2022208894A1 JP 2021014402 W JP2021014402 W JP 2021014402W WO 2022208894 A1 WO2022208894 A1 WO 2022208894A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- machine
- time
- series data
- knowledge
- data
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims description 5
- 238000003745 diagnosis Methods 0.000 title abstract description 47
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000002405 diagnostic procedure Methods 0.000 claims description 4
- 230000008447 perception Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 15
- 230000005856 abnormality Effects 0.000 description 13
- 238000007689 inspection Methods 0.000 description 6
- 208000024891 symptom Diseases 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000010365 information processing Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000446 fuel Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 239000002243 precursor Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M17/00—Testing of vehicles
- G01M17/007—Wheeled or endless-tracked vehicles
Definitions
- the present invention relates to a machine diagnostic device, a machine diagnostic method, and a recording medium, and for example, to a machine diagnostic device, a machine diagnostic method, and a recording medium for remotely diagnosing a machine based on time-series data received from equipment of the machine. .
- Patent Document 1 provides a diagnostic system for remotely diagnosing automobiles.
- the diagnostic system transmits diagnostic program data from the information center to the onboard computer.
- the diagnostic program displays guidance on the in-vehicle display. By following the guidance and answering the questions from the diagnostic program, the user can identify the cause of the vehicle abnormality (or no failure).
- Patent Document 1 With the related technology described in Patent Document 1, the user has to answer many questions until the diagnostic program can identify the cause of the abnormality, which imposes a heavy burden on the user.
- the present invention has been made in view of the above problems, and its purpose is to reduce the user's burden in diagnosing a machine remotely.
- a machine diagnostic apparatus comprises acquisition means for acquiring time-series data obtained from equipment of a machine, prediction means for predicting the state of the machine based on the time-series data, and the time-series data. search means for searching knowledge information related to the state of the machine from a knowledge database storing records of past diagnoses using data as a query; and providing means for providing the result of the prediction and the knowledge information.
- a machine diagnosis method acquires time-series data obtained from equipment of a machine, predicts the state of the machine based on the time-series data, uses the time-series data as a query, retrieving knowledge information associated with the state of the machine from a knowledge database storing records in the diagnosis of the machine, and providing results of the prediction and the knowledge information.
- a recording medium acquires time-series data obtained from equipment of a machine, predicts the state of the machine based on the time-series data, uses the time-series data as a query, Knowledge information associated with the state of the machine is retrieved from a knowledge database containing records in the diagnosis to provide results of the prediction and the knowledge information.
- FIG. 1 schematically shows an example of a diagnostic system to which the machine diagnostic device according to Embodiment 1 or 2 is applied;
- 2 is a diagram illustrating functions of an analysis engine included in the diagnostic system shown in FIG. 1;
- FIG. 2 is a diagram showing an example of the data structure of a knowledge database provided in the diagnostic system shown in FIG. 1;
- FIG. 2 is a diagram showing an example of the data structure of a data dictionary provided in the diagnostic system shown in FIG. 1;
- FIG. 1 is a block diagram showing the configuration of a machine diagnostic device according to Embodiment 1;
- FIG. 4 is a flow chart showing the operation of the machine diagnostic device according to the first embodiment;
- FIG. 11 is a block diagram showing the configuration of a machine diagnostic device according to Embodiment 2; 9 is a flow chart showing the operation of the machine diagnostic device according to the second embodiment;
- FIG. 5 is a diagram schematically showing a modified example of a diagnostic system to which the machine diagnostic device according to Embodiment 1 or 2 is applied; It is a figure which shows an example of the time-series data to which the comment was added.
- FIG. 1 schematically shows an example of a diagnostic system 1 to which a machine diagnostic device 10 (FIG. 4) or a machine diagnostic device 20 (FIG. 6) according to Embodiments 1 or 2, which will be described later, is applied.
- machine diagnostic device 10 or machine diagnostic device 20 is referred to as “machine diagnostic device 10 (20)”.
- the diagnostic system 1 includes a machine diagnostic device 10 (20), an analysis engine 100, a knowledge database 200, and a data dictionary 300.
- the knowledge database 200 hereinafter sometimes referred to as knowledge DB (Data Base) 200.
- the machine diagnosis device 10 (20) remotely diagnoses the machine based on the time-series data obtained from the equipment of the machine.
- a machine is an industrially applicable device that operates by converting a power source such as electric power or fuel into power.
- the machine is an automobile, a watercraft, an agricultural machine, an industrial machine, an unmanned aerial vehicle, or a personal airplane.
- an automobile is taken as an example.
- a machine's equipment is equipment, devices, equipment, or a combination thereof.
- Mechanical equipment outputs time series data continuously or intermittently.
- the time-series data is, for example, sensor data output from various sensors.
- various accessories such as power windows, seat belts, door locks, fuel gauges, wipers, fog lamps, air conditioners, and interior lights output sensor data.
- the machine diagnosis device 10 (20) uses the analysis engine 100 to analyze the time-series data obtained from the equipment of the machine.
- the machine diagnosis device 10 (20) diagnoses the state of the machine based on the analysis results of the time-series data. After that, the machine diagnosis device 10 (20) transmits the diagnosis result to the machine, user device, or the like, such as whether the machine has a symptom of failure or needs to be inspected.
- the user decides whether to continue driving the car or bring the car to a dealer based on the diagnostic results sent from the machine diagnostic device 10 (20).
- the analysis engine 100 includes a computer program such as LSTM (Long short-term memory) that performs machine learning on feature amounts of normal time-series data and feature amounts of abnormal time-series data. Analysis engine 100 further comprises computer hardware such as a processor and memory for executing computer programs and interfaces.
- the analysis engine 100 analyzes the time series data input by the machine diagnosis device 10 (20), searches for comments having feature amounts similar to the feature amounts of the time series data, and outputs search results. For example, the analysis engine 100 searches for comments having feature quantities similar to those of the time-series data based on a known similarity concept such as Euclidean distance in the feature vector space.
- the comment is text data representing the state of the machine implied by the time-series data.
- the analysis engine 100 returns search results having a data format different from that of the query. This is done so that the position of a feature extracted from a segment of time-series data is close to the position of a feature extracted from text data related to this segment of time-series data in the feature space. , by training a computer program (LSTM, etc.) to correlate time-series data and text data. For example, a comment (text data) are associated.
- LSTM computer program
- FIG. 2 is a diagram explaining the functions of the analysis engine 100 included in the diagnostic system 1 (FIG. 1).
- time-series data is input to the analysis engine 100 as a search query.
- the analysis engine 100 extracts feature quantities from the input time-series data.
- the analysis engine 100 searches the data dictionary 300 (FIG. 4) for comments having feature quantities similar to those of the time-series data.
- the analysis engine 100 then outputs comments as search results.
- FIG. 3 shows an example of the data structure of the knowledge DB 200 included in the diagnostic system 1 (FIG. 1).
- the knowledge DB 200 stores knowledge information associated with feature amounts or their hash values.
- the knowledge information includes audio data and video data.
- Knowledge information is not limited to these.
- the knowledge information is related to the state of the machine implied by the feature quantity of the time-series data, and is provided from the diagnostic system 1 to the user.
- Knowledge information includes arbitrary content data and arbitrary text data.
- the knowledge information is a video showing actions to be taken to check the state of the machine (such as "perform a specific operation on the machine").
- the knowledge information is text data of comments (such as "sensor is broken") that describe the state of the machine.
- Auxiliary information may accompany the time-series data sent from the automobile to the machine diagnostic device 10 (20).
- Auxiliary information attached to the time-series data is stored in the knowledge DB 200 as knowledge information.
- the auxiliary information stored as knowledge information in the knowledge DB 200 is provided from the diagnostic system 1 to the user as knowledge information related to the state of the machine during another diagnosis later.
- Auxiliary information is perceived using the user's senses or intuition, or obtained using a user device.
- the auxiliary information may be structured data or unstructured data. Also, the content of the auxiliary information is not limited.
- Auxiliary information includes content data generated by the user device.
- the auxiliary information also includes text data describing the user's perception.
- the auxiliary information may be structured data, such as acceleration data output by a smartphone's gyro sensor.
- the auxiliary information may be unstructured data, such as voice data entered by the user into a microphone or text data entered into the smartphone by the user.
- FIG. 4 is a diagram showing an example of the data structure of the data dictionary 300.
- the data dictionary 300 stores time-series data and comments that are related to each other in association with respective feature amounts.
- “mutually related” means that the position of the feature quantity of the time-series data and the position of the feature quantity of the comment are close to each other in the feature quantity space, that is, that the feature quantities are similar.
- Time-series data and comments are classified into either a normal (no abnormality) group or an abnormality or its precursor group for each abnormality type. More specifically, time-series data and comments from vehicle installations for which a particular sensor is in good condition are grouped in the data dictionary 300 in the normal group for that sensor. On the other hand, the time-series data and comments obtained from the vehicle equipment in which a particular sensor is in an abnormal state are classified into the abnormal group for that sensor in the data dictionary 300 .
- the diagnosis target is an automobile.
- the target of diagnosis is not limited to automobiles.
- the object of diagnosis may be a machine such as an automobile, a ship, an agricultural machine, an industrial machine, an unmanned aerial vehicle, or a private airplane.
- Embodiment 1 Embodiment 1 will be described with reference to FIGS. 5 and 6.
- FIG. 5 An illustration of FIG. 5
- FIG. 5 is a block diagram showing the configuration of the machine diagnostic device 10 according to the first embodiment. As shown in FIG. 5, the machine diagnostic device 10 includes an acquisition unit 11, a prediction unit 12, a search unit 13, and a provision unit .
- the acquisition unit 11 acquires time-series data obtained from equipment of a machine (here, an automobile). Acquisition unit 11 is an example of acquisition means.
- the acquisition unit 11 acquires ECU (Engine Control Unit) output data, sensor data, and/or other time-series data from the user's vehicle equipment via any wireless network such as a mobile network. do.
- the time-series data is obtained to measure the state of the drive train of the vehicle or the state of the accessories.
- the acquisition unit 11 outputs the acquired time-series data to the prediction unit 12 and the search unit 13 .
- auxiliary information attached to the time-series data may be transmitted from the user's vehicle equipment (eg, car navigation device or touch panel display device) or user device (eg, smartphone).
- the acquiring unit 11 acquires auxiliary information attached to the time-series data together with the time-series data. Then, the acquisition unit 11 outputs the time-series data and the auxiliary information to a recording unit (not shown) (second embodiment).
- the prediction unit 12 predicts the state of the machine based on the time-series data.
- the prediction unit 12 is an example of prediction means.
- the prediction unit 12 inputs the time-series data as a query to the analysis engine 100 (FIG. 2) that performs machine learning on the feature amount of normal time-series data and the feature amount of abnormal time-series data, and performs a search. Receive the result of the prediction as a result.
- the analysis engine 100 FIG. 2
- the prediction unit 12 receives time-series data from the acquisition unit 11 .
- the prediction unit 12 inputs the time-series data to the analysis engine 100 .
- the analysis engine 100 extracts the feature amount of the input time-series data, and searches the data dictionary 300 (FIG. 4) for comments (text data) having similar feature amounts.
- the data dictionary 300 the correlated time-series data and comments are classified by type of abnormality (for example, by sensor) into a normal group or an abnormality or its precursor group.
- the analysis engine 100 outputs comments belonging to either the normal group or the abnormal group in the data dictionary 300 .
- the prediction unit 12 predicts the state of the automobile based on whether the comments output from the analysis engine 100 belong to the normal group or the abnormal group in the data dictionary 300 . Specifically, if the comment output from the analysis engine 100 belongs to the normal group in the data dictionary 300, the prediction unit 12 predicts that the car is normal.
- the prediction unit 12 predicts that the vehicle is abnormal.
- the prediction unit 12 may predict the state of the automobile by analyzing the comments output from the analysis engine 100. For example, if the output comment includes the word "stable” or a synonym thereof, the prediction unit 12 predicts that the car is normal. On the other hand, if the output comment includes the word "abnormality" or a synonym thereof, it is predicted that there is an abnormality in the vehicle. The prediction unit 12 outputs the comment output from the analysis engine 100 to the provision unit 14 together with the prediction result.
- the search unit 13 uses the time-series data as a query to search the knowledge DB 200 that stores records of past diagnoses for knowledge information related to machine abnormalities.
- the search unit 13 is an example of search means.
- the search unit 13 receives time-series data from the acquisition unit 11 .
- the search unit 13 inputs the time-series data to the analysis engine 100 (FIG. 2).
- the search unit 13 acquires feature data extracted from the time-series data by the analysis engine 100 .
- the search unit 13 searches the knowledge DB 200 (FIG. 3) for knowledge information associated with a feature quantity similar to the feature quantity of the time-series data.
- the search unit 13 outputs knowledge information obtained as a result of the search to the provision unit 14 .
- the providing unit 14 provides prediction results and knowledge information.
- the providing unit 14 is an example of providing means.
- the providing unit 14 receives prediction results (and comments) from the prediction unit 12 .
- the providing unit 14 also receives knowledge information obtained as a result of the search from the searching unit 13 .
- the providing unit 14 outputs the prediction result and the knowledge information to a subsequent processing unit (not shown).
- the providing unit 14 transmits prediction results and knowledge information to vehicle equipment (e.g., car navigation device or display device) or user device (e.g., smartphone) via any wireless network such as a mobile network. do.
- the provision unit 14 may display the knowledge information on the screen of the car navigation device or smartphone.
- the providing unit 14 may output the knowledge information as audio from a speaker provided in the vehicle or the user device.
- FIG. 6 is a flow chart showing the flow of processing executed by each unit of the machine diagnostic apparatus 10. As shown in FIG.
- the user turns on the diagnostic button installed in advance on the car to start the on-demand diagnosis.
- the machine diagnosis device 10 is notified from the diagnosis button that the diagnosis button has been pressed.
- the diagnostic button may be an in-vehicle device.
- the user may input voice into the microphone or use a smartphone application.
- a voice AI Artificial Intelligence
- chatbot may perform a simple medical question to the user.
- the machine diagnostic system 10 then initiates remote diagnostics as described below.
- the acquisition unit 11 acquires time-series data obtained from equipment of the machine (S101).
- the prediction unit 12 predicts the state of the machine based on the time series data (S102).
- the search unit 13 searches the knowledge DB 200 storing records of past diagnoses for knowledge information related to the state of the machine (S103).
- the providing unit 14 provides the prediction result and knowledge information to the user by transmitting the prediction result and knowledge information to the equipment of the vehicle or the user device (S104).
- the acquisition unit 11 acquires time-series data obtained from equipment of the machine.
- a prediction unit 12 predicts the state of the machine based on the time-series data.
- the search unit 13 uses the time-series data as a query to search knowledge information related to the state of the machine from a knowledge database that stores records of past diagnoses.
- the providing unit 14 provides prediction results and knowledge information.
- the user can obtain knowledge information along with prediction results without having to answer numerous questions. For example, the user obtains, as knowledge information, a video showing actions to be taken to confirm the state of the machine. The user refers to the contents of the video (knowledge information) to confirm whether the machine is normal. This reduces the burden on the user in diagnosing the machine remotely.
- Embodiment 2 will be described with reference to FIGS. 7 and 8.
- FIG. 7 In the second embodiment, how to expand the knowledge DB 200 (FIG. 3) of the diagnostic system 1 (FIG. 1) will be explained.
- FIG. 7 is a block diagram showing the configuration of the machine diagnostic device 20 according to the second embodiment.
- the machine diagnostic device 20 includes an acquisition unit 11, a prediction unit 12, a search unit 13, and a provision unit .
- the machine diagnostic device 20 further includes a recording section 25 .
- the recording unit 25 associates the feature amount extracted from the time-series data by the analysis engine or the data of the hash value thereof with the auxiliary information and saves it in the knowledge database.
- the recording unit 25 is an example of recording means.
- the recording unit 25 receives time-series data and auxiliary information from the acquisition unit 11 .
- the recording unit 25 inputs the received time-series data to the analysis engine 100 (FIG. 2).
- the recording unit 25 acquires feature data extracted from the time-series data by the analysis engine 100 .
- the recording unit 25 associates the auxiliary information attached to the time-series data with the feature amount data of the time-series data. Then, the recording unit 25 saves the feature amount data and the auxiliary information in the knowledge DB 200 .
- FIG. 8 is a flow chart showing the flow of processing executed by each part of the machine diagnosis device 20. As shown in FIG.
- the user turns on a diagnostic button installed in advance on the vehicle to start on-demand diagnostics.
- the machine diagnostic device 20 then initiates remote diagnostics as described below.
- the acquisition unit 11 acquires the time-series data obtained from the equipment of the machine and auxiliary information attached to the time-series data (S201).
- auxiliary information includes content data generated by a user device such as a smart phone.
- the auxiliary information includes text data describing the user's perception.
- the prediction unit 12 predicts the state of the machine based on the time series data (S202).
- the recording unit 25 stores the auxiliary information linked with the feature amount of the time-series data in the knowledge DB (Fig. 3) (S203). It should be noted that the recording unit 25 skips the process of step S203 when auxiliary information is not attached to the time-series data.
- the search unit 13 searches the knowledge DB 200 storing records of past diagnoses for knowledge information related to the state of the machine (S204).
- the providing unit 14 provides the prediction result and knowledge information to the user by transmitting the prediction result and knowledge information to the vehicle equipment or user device (S205).
- the acquisition unit 11 acquires time-series data obtained from equipment of the machine.
- a prediction unit 12 predicts the state of the machine based on the time-series data.
- the search unit 13 uses the time-series data as a query to search knowledge information related to the state of the machine from a knowledge database that stores records of past diagnoses.
- the providing unit 14 provides prediction results and knowledge information.
- the user can obtain knowledge information along with prediction results without having to answer numerous questions. For example, the user obtains, as knowledge information, a video showing actions to be taken to confirm the state of the machine. The user refers to the contents of the video (knowledge information) to confirm whether the machine is normal. This reduces the burden on the user in diagnosing the machine remotely.
- the recording unit 25 associates the feature amount extracted from the time-series data by the analysis engine or its hash value data with the auxiliary information and saves it in the knowledge database.
- Auxiliary information stored in the knowledge database is referred to as knowledge information in future diagnoses. In this way, the knowledge database can be expanded.
- the machine diagnosis device 10 introduces a candidate dealer to the user (driver) to inspect the automobile.
- FIG. 9 schematically shows a modified example of the diagnostic system 1 shown in FIG. 1 (referred to as "diagnostic system 1'").
- the machine diagnostic device 10 (20) of the diagnostic system 1' receives a reaction indicating whether the diagnostic result is appropriate or not from the user.
- the machine diagnostic device 10 (20) receives a reaction including information indicating whether or not the diagnostic result is appropriate from the car navigation device or smart phone.
- the machine diagnosis device 10 (20) responds differently depending on the content of the reaction.
- the machine diagnostic device 10 (20) receives (1) a diagnostic result indicating that there is an abnormality in the vehicle and a reaction corresponding to the diagnostic result, Provide the user with candidate information.
- the machine diagnosis device 10 (20) receives (2) a diagnosis result indicating that there is an abnormality in the vehicle and a reaction indicating that the diagnosis result does not apply, the machine diagnosis device 10 (20) is a high-level dealer capable of performing a highly difficult inspection. Provide the user with candidate information.
- the machine diagnosis device 10 (20) does not provide information on dealer candidates to the user when (3) there is a diagnosis result that there is no abnormality in the automobile and a reaction that corresponds to the diagnosis result is received. Instead, the machine diagnostic device 10 (20) transitions from the on-demand diagnostic mode to the constant diagnostic mode. In the constant diagnosis mode, the machine diagnosis device 10 (20) executes automobile diagnosis periodically or at a predetermined timing without receiving a request for on-demand diagnosis from the user (that is, turning on the diagnosis button).
- the dealer will inspect the car and, if necessary, repair the car.
- the dealer sends an inspection record containing information such as findings, symptoms, and countermeasures related to the inspection of the automobile to the management system or machine diagnostic device 10 (20).
- the inspection record sent from the dealer is linked to the feature amount data of the time-series data or its hash value (index) and stored in the knowledge DB 200 (FIG. 3).
- the machine diagnostic device 10 (20) may generate a comment from the inspection record.
- the comment is text data representing the state of the machine implied by the time-series data.
- the machine diagnosis device 10 (20) extracts descriptions of symptoms and measures from inspection records, and generates comments on symptoms and measures.
- the machine diagnostic device 10 (20) stores the generated comment (text data) in the data dictionary 300 (FIG. 4) in association with the time-series data.
- the comments saved in the data dictionary 300 can be used for learning the analysis engine 100 (FIG. 11).
- FIG. Training of the analysis engine 100 requires time-series data and comments representing the state of the machine implied by the time-series data.
- FIG. 10 is a diagram showing an example of time-series data with comments added.
- the time-series data are four sensor data obtained from four sensors AD.
- different comments are added to four segment data of a predetermined time width extracted from time-series data at different times.
- the segment data at the left end has a comment of "normal + working". This indicates that the time series data imply that all four sensors AD are operating normally.
- FIG. 11 is an explanatory diagram showing the learning flow of the analysis engine 100.
- a plurality of mutually related time-series data and comments are prepared. Note that the time-series data and comments that are related to each other are not limited to one-to-one pairs. One or more pieces of time-series data may be linked (associated) with one or more comments.
- time-series data is input to the analysis engine 100 .
- the analysis engine 100 outputs the feature amount of the input time-series data.
- comments associated with the previously input time-series data are input to the analysis engine 100 .
- the analysis engine 100 outputs the feature amount of the input comment. In this way, the time-series data feature amount and the comment feature amount that are related to each other are obtained.
- the operator causes the analysis engine 100 to learn so that the feature amount of the time-series data and the feature amount of the comment are brought closer to each other. For example, the operator makes the analysis engine 100 learn to maximize the similarity between the feature amount of the time-series data and the feature amount of the comment.
- Learning of the analysis engine 100 is completed by repeating the above procedure for mutually related time-series data and comments. As described with reference to FIG. 2, the analysis engine 100 that has completed learning can return comments as search results when time-series data is input as a query.
- FIG. 12 is a block diagram showing an example of the hardware configuration of the information processing device 900. As shown in FIG. 12
- the information processing device 900 includes the following configuration as an example.
- a program 904 that implements the function of each component is stored in advance in, for example, the storage device 905 or the ROM 902, and is loaded into the RAM 903 and executed by the CPU 901 as necessary.
- the program 904 may be supplied to the CPU 901 via the communication network 909 or may be stored in the recording medium 906 in advance, and the drive device 907 may read the program and supply it to the CPU 901 .
- the machine diagnostic devices 10 and 20 described in the first and second embodiments are implemented as hardware. Therefore, the same effects as those described in the first and second embodiments can be obtained.
- (Appendix 1) Acquisition means for acquiring time-series data obtained from equipment of the machine; prediction means for predicting the state of the machine based on the time-series data; a search means for searching knowledge information related to the state of the machine from a knowledge database storing records of past diagnoses using the time-series data as a query; a providing means for providing said prediction result and said knowledge information.
- the prediction means inputs the time-series data as a query to an analysis engine that machine-learns the feature amount of normal time-series data and the feature amount of abnormal time-series data, and outputs the result of the prediction as a search result.
- Appendix 3 The machine diagnostic apparatus according to appendix 1 or 2, wherein the acquiring means acquires auxiliary information attached to the time-series data together with the time-series data.
- Appendix 4 The machine diagnostic apparatus according to appendix 3, wherein the auxiliary information includes content data generated by a user device.
- Appendix 5 The machine diagnostic apparatus according to appendix 3 or 4, wherein the auxiliary information includes text data describing the user's perception.
- appendix 6 Any one of appendices 3 to 5, further comprising recording means for storing in the knowledge database the data of the feature amount or its hash value extracted from the time-series data in association with the auxiliary information. 2. The machine diagnostic device according to claim 1.
- Appendix 7 The machine diagnosis device according to any one of appendices 1 to 6, wherein the result of the prediction includes a comment regarding the state of the machine.
- Appendix 8 The machine diagnostic apparatus according to any one of appendices 1 to 7, wherein the providing means provides the knowledge information via a third party's platform.
- a non-transitory recording medium storing a program that causes a computer to perform: providing the prediction result and the knowledge information.
- the present invention can be used, for example, as a machine diagnostic device for remotely diagnosing machines such as automobiles, ships, agricultural machines, industrial machines, unmanned aerial vehicles, and private airplanes.
Abstract
Description
図1は、後述する実施形態1または2に係わる機械診断装置10(図4)または機械診断装置20(図6)が適用される診断システム1の一例を概略的に示す。なお、以下では、「機械診断装置10または機械診断装置20」を、「機械診断装置10(20)」と記載する。
分析エンジン100は、正常な時系列データの特徴量と異常な時系列データの特徴量とを機械学習したLSTM(Long short-term memory)などのコンピュータプログラムを備えている。分析エンジン100は、コンピュータプログラムを実行するプロセッサとメモリ、および、インタフェースなどのコンピュータハードウェアをさらに備えている。分析エンジン100は、機械診断装置10(20)が入力した時系列データを分析して、時系列データの特徴量と類似する特徴量を有するコメントを検索し、検索結果を出力する。例えば、分析エンジン100は、特徴ベクトル空間におけるユークリッド距離など、既知の類似度の概念に基づいて、時系列データの特徴量と類似する特徴量を有するコメントを検索する。コメントは、時系列データが暗示する機械の状態を表すテキストデータである。
図3は、診断システム1(図1)が備えた知識DB200のデータ構造の一例を示す。図3に示すように、知識DB200は、特徴量またはそのハッシュ値と紐づけられた知識情報を格納している。図3に示す一例では、知識情報は、音声データおよびビデオデータを含む。しかしながら、知識情報はこれらに限定されない。知識情報は、時系列データの特徴量が暗示する機械の状態と関連するものであり、診断システム1からユーザへ提供される。知識情報は、任意のコンテンツデータ、および任意のテキストデータを含む。例えば、知識情報は、機械の状態を確認するための対処(「機械に対して特定の操作を行う」など)を表したビデオである。あるいは、知識情報は、機械の状態を表したコメント(「センサが故障」など)のテキストデータである。
図4は、データ辞書300のデータ構造の一例を示す図である。図4に示すように、データ辞書300には、相互に関連する時系列データおよびコメントが、それぞれの特徴量と紐づけられて格納されている。ここでの「相互に関連する」とは、特徴量空間上における時系列データの特徴量の位置とコメントの特徴量の位置とが近い、つまり特徴量が類似することを意味する。時系列データおよびコメントは、異常の種別ごとに、正常(異常なし)のグループと、異常またはその前兆ありのグループとのどちらか一方に分類されている。より詳細には、ある特定のセンサが正常な状態にある自動車の設備より得られた時系列データおよびコメントは、データ辞書300における当該センサに関する正常のグループに分類されている。一方、ある特定のセンサが異常な状態にある自動車の設備より得られた時系列データおよびコメントは、データ辞書300における当該センサに関する異常のグループに分類されている。
図5~図6を参照して、実施形態1について説明する。
図5は、本実施形態1に係わる機械診断装置10の構成を示すブロック図である。図5に示すように、機械診断装置10は、取得部11、予測部12、検索部13、および提供部14を備えている。
図6を参照して、本実施形態1に係わる機械診断装置10の動作を説明する。図6は、機械診断装置10の各部が実行する処理の流れを示すフローチャートである。
本実施形態の構成によれば、取得部11は、機械の設備から得られる時系列データを取得する。予測部12は、時系列データに基づいて、機械の状態を予測する。検索部13は、時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、機械の状態と関連する知識情報を検索する。提供部14は、予測の結果および知識情報を提供する。ユーザは、多量の質問に返答することなしに、予測の結果とともに、知識情報を得ることができる。例えば、ユーザは、知識情報として、機械の状態を確認するための対処を表したビデオを得る。ユーザは、ビデオ(知識情報)の内容を参考にして、機械が正常であるのかどうかを確認する。これにより、遠隔での機械の診断において、ユーザの負担を低減することができる。
図7~図8を参照して、実施形態2について説明する。本実施形態2では、診断システム1(図1)の知識DB200(図3)を、どのようにして拡充してゆくのかについて説明する。
図7は、本実施形態2に係わる機械診断装置20の構成を示すブロック図である。図7に示すように、機械診断装置20は、取得部11、予測部12、検索部13、および提供部14を備えている。また、機械診断装置20は、記録部25をさらに備えている。
図8を参照して、本実施形態2に係わる機械診断装置20の動作を説明する。図8は、機械診断装置20の各部が実行する処理の流れを示すフローチャートである。
本実施形態の構成によれば、取得部11は、機械の設備から得られる時系列データを取得する。予測部12は、時系列データに基づいて、機械の状態を予測する。検索部13は、時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、機械の状態と関連する知識情報を検索する。提供部14は、予測の結果および知識情報を提供する。ユーザは、多量の質問に返答することなしに、予測の結果とともに、知識情報を得ることができる。例えば、ユーザは、知識情報として、機械の状態を確認するための対処を表したビデオを得る。ユーザは、ビデオ(知識情報)の内容を参考にして、機械が正常であるのかどうかを確認する。これにより、遠隔での機械の診断において、ユーザの負担を低減することができる。
前記実施形態1~2の一変形例では、機械診断装置10(20)は、自動車を検査するために、ユーザ(ドライバー)に、ディーラー候補を紹介する。
(補足:分析エンジン100の学習)
図10および図11を参照して、診断システム1(図1)あるいは診断システム1´(図9)が備えた分析エンジン100の学習について説明する。分析エンジン100の学習には、時系列データと、時系列データが暗示する機械の状態を表すコメントが必要である。
前記実施形態1~2で説明した機械診断装置10,20の各構成要素は、機能単位のブロックを示している。これらの構成要素の一部又は全部は、例えば図12に示すような情報処理装置900により実現される。図12は、情報処理装置900のハードウェア構成の一例を示すブロック図である。
・ROM(Read Only Memory)902
・RAM(Random Access Memory)903
・RAM903にロードされるプログラム904
・プログラム904を格納する記憶装置905
・記録媒体906の読み書きを行うドライブ装置907
・通信ネットワーク909と接続する通信インタフェース908
・データの入出力を行う入出力インタフェース910
・各構成要素を接続するバス911
前記実施形態1~2で説明した機械診断装置10,20の各構成要素は、これらの機能を実現するプログラム904をCPU901が読み込んで実行することで実現される。各構成要素の機能を実現するプログラム904は、例えば、予め記憶装置905やROM902に格納されており、必要に応じてCPU901がRAM903にロードして実行される。なお、プログラム904は、通信ネットワーク909を介してCPU901に供給されてもよいし、予め記録媒体906に格納されており、ドライブ装置907が当該プログラムを読み出してCPU901に供給してもよい。
本発明の一態様は、以下の付記のようにも記載されるが、以下に限定されない。
機械の設備から得られる時系列データを取得する取得手段と、
前記時系列データに基づいて、前記機械の状態を予測する予測手段と、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索する検索手段と、
前記予測の結果および前記知識情報を提供する提供手段と
を備えた機械診断装置。
前記予測手段は、正常な時系列データの特徴量と異常な時系列データの特徴量とを機械学習した分析エンジンへ、クエリとして前記時系列データを入力して、検索結果として前記予測の結果を受信する
ことを特徴とする付記1に記載の機械診断装置。
前記取得手段は、前記時系列データとともに、前記時系列データに付帯する補助情報を取得する
ことを特徴とする付記1または2に記載の機械診断装置。
前記補助情報は、ユーザデバイスが生成したコンテンツデータを含む
ことを特徴とする付記3に記載の機械診断装置。
前記補助情報は、ユーザの知覚を記述したテキストデータを含む
ことを特徴とする付記3または4に記載の機械診断装置。
前記時系列データから抽出された特徴量またはそのハッシュ値のデータを、前記補助情報に紐づけて、前記知識データベースに保存する記録手段をさらに備えた
ことを特徴とする付記3から5のいずれか1項に記載の機械診断装置。
前記予測の結果は、前記機械の状態に関するコメントを含む
ことを特徴とする付記1から6のいずれか1項に記載の機械診断装置。
前記提供手段は、第三者のプラットフォームを介して、前記知識情報を提供する
ことを特徴とする付記1から7のいずれか1項に記載の機械診断装置。
機械の設備から得られる時系列データを取得し、
前記時系列データに基づいて、前記機械の状態を予測し、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索し、
前記予測の結果および前記知識情報を提供する
機械診断方法。
機械の設備から得られる時系列データを取得することと、
前記時系列データに基づいて、前記機械の状態を予測することと、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索することと、
前記予測の結果および前記知識情報を提供することと
をコンピュータに実行させるプログラムを格納した、一時的でない記録媒体。
1´ 診断システム
10 機械診断装置
11 取得部
12 予測部
13 検索部
14 提供部
20 機械診断装置
25 記録部
100 分析エンジン
200 知識DB(データベース)
Claims (10)
- 機械の設備から得られる時系列データを取得する取得手段と、
前記時系列データに基づいて、前記機械の状態を予測する予測手段と、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索する検索手段と、
前記予測の結果および前記知識情報を提供する提供手段と
を備えた機械診断装置。 - 前記予測手段は、正常な時系列データの特徴量と異常な時系列データの特徴量とを機械学習した分析エンジンへ、クエリとして前記時系列データを入力して、検索結果として前記予測の結果を受信する
ことを特徴とする請求項1に記載の機械診断装置。 - 前記取得手段は、前記時系列データとともに、前記時系列データに付帯する補助情報を取得する
ことを特徴とする請求項1または2に記載の機械診断装置。 - 前記補助情報は、ユーザデバイスが生成したコンテンツデータを含む
ことを特徴とする請求項3に記載の機械診断装置。 - 前記補助情報は、ユーザの知覚を記述したテキストデータを含む
ことを特徴とする請求項3または4に記載の機械診断装置。 - 前記時系列データから抽出された特徴量またはそのハッシュ値のデータを、前記補助情報に紐づけて、前記知識データベースに保存する記録手段をさらに備えた
ことを特徴とする請求項3から5のいずれか1項に記載の機械診断装置。 - 前記予測の結果は、前記機械の状態に関するコメントを含む
ことを特徴とする請求項1から6のいずれか1項に記載の機械診断装置。 - 前記提供手段は、第三者のプラットフォームを介して、前記知識情報を提供する
ことを特徴とする請求項1から7のいずれか1項に記載の機械診断装置。 - 機械の設備から得られる時系列データを取得し、
前記時系列データに基づいて、前記機械の状態を予測し、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索し、
前記予測の結果および前記知識情報を提供する
機械診断方法。 - 機械の設備から得られる時系列データを取得することと、
前記時系列データに基づいて、前記機械の状態を予測することと、
前記時系列データをクエリとして、過去の診断における記録を格納した知識データベースから、前記機械の状態と関連する知識情報を検索することと、
前記予測の結果および前記知識情報を提供することと
をコンピュータに実行させるプログラムを格納した、一時的でない記録媒体。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2021/014402 WO2022208894A1 (ja) | 2021-04-02 | 2021-04-02 | 機械診断装置、機械診断方法、および記録媒体 |
JP2023510159A JPWO2022208894A5 (ja) | 2021-04-02 | 機械診断装置、機械診断方法、およびプログラム |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2021/014402 WO2022208894A1 (ja) | 2021-04-02 | 2021-04-02 | 機械診断装置、機械診断方法、および記録媒体 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022208894A1 true WO2022208894A1 (ja) | 2022-10-06 |
Family
ID=83457570
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2021/014402 WO2022208894A1 (ja) | 2021-04-02 | 2021-04-02 | 機械診断装置、機械診断方法、および記録媒体 |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2022208894A1 (ja) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007248070A (ja) * | 2006-03-13 | 2007-09-27 | Nissan Motor Co Ltd | 車両走行試験装置 |
JP2008216113A (ja) * | 2007-03-06 | 2008-09-18 | Toyota Infotechnology Center Co Ltd | 不具合情報集約システムおよび車両 |
JP2015176285A (ja) * | 2014-03-14 | 2015-10-05 | 株式会社デンソー | 故障情報提示システム |
WO2020110446A1 (ja) * | 2018-11-27 | 2020-06-04 | 住友電気工業株式会社 | 車両故障予測システム、監視装置、車両故障予測方法および車両故障予測プログラム |
-
2021
- 2021-04-02 WO PCT/JP2021/014402 patent/WO2022208894A1/ja active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007248070A (ja) * | 2006-03-13 | 2007-09-27 | Nissan Motor Co Ltd | 車両走行試験装置 |
JP2008216113A (ja) * | 2007-03-06 | 2008-09-18 | Toyota Infotechnology Center Co Ltd | 不具合情報集約システムおよび車両 |
JP2015176285A (ja) * | 2014-03-14 | 2015-10-05 | 株式会社デンソー | 故障情報提示システム |
WO2020110446A1 (ja) * | 2018-11-27 | 2020-06-04 | 住友電気工業株式会社 | 車両故障予測システム、監視装置、車両故障予測方法および車両故障予測プログラム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2022208894A1 (ja) | 2022-10-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108563214B (zh) | 车辆诊断方法、装置及设备 | |
US20230186697A1 (en) | Systems and methods for in-vehicle predictive failure detection | |
US7987028B2 (en) | Method and apparatus for reading and erasing diagnostic trouble codes from a vehicle | |
US10332319B2 (en) | Methods and systems for updating diagnostic and repair information | |
EP3267400B1 (en) | Vehicle-diagnostic client-server system | |
USRE45815E1 (en) | Method for simplified real-time diagnoses using adaptive modeling | |
JP2009265104A (ja) | 診断データマイニング | |
WO2014001799A1 (en) | Vehicle diagnostic system and method | |
WO2014174791A1 (ja) | 車両修理支援システム、サーバ及びコンピュータプログラム | |
WO2022020453A1 (en) | Mobile vehicle inspection system | |
Gültekin et al. | A novel deep learning approach for intelligent fault diagnosis applications based on time-frequency images | |
CN112254983A (zh) | 一种车辆检测方法、装置、设备及存储介质 | |
CN105468667A (zh) | 用于车辆的诊断方法和调查方法 | |
US20220108569A1 (en) | Automated detection of vehicle data manipulation and mechanical failure | |
WO2022208894A1 (ja) | 機械診断装置、機械診断方法、および記録媒体 | |
US20110125364A1 (en) | Method and system for diagnosing a malfunction of an automobile | |
CN113885476B (zh) | 一种车辆诊断方法及相关装置 | |
JPH0715427B2 (ja) | 自動車用故障診断装置 | |
US20230239639A1 (en) | Abnormal sound diagnosis system | |
US20240104977A1 (en) | Abnormal sound diagnostic system | |
JPH05172702A (ja) | 自動車用故障診断装置 | |
US20240029479A1 (en) | Abnormal sound diagnosis system | |
Rizzoni et al. | A survey of automotive diagnostic equipment and procedures | |
CN114896168B (en) | Quick debugging system, method and memory for automatic driving algorithm development | |
CN117972508A (zh) | 故障诊断方法、装置、终端设备以及存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21935054 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18283942 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2023510159 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 21935054 Country of ref document: EP Kind code of ref document: A1 |