WO2023095236A1 - Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien - Google Patents

Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien Download PDF

Info

Publication number
WO2023095236A1
WO2023095236A1 PCT/JP2021/043147 JP2021043147W WO2023095236A1 WO 2023095236 A1 WO2023095236 A1 WO 2023095236A1 JP 2021043147 W JP2021043147 W JP 2021043147W WO 2023095236 A1 WO2023095236 A1 WO 2023095236A1
Authority
WO
WIPO (PCT)
Prior art keywords
equipment
inspection
information
inference model
data
Prior art date
Application number
PCT/JP2021/043147
Other languages
English (en)
Japanese (ja)
Inventor
光貴 岩村
守真 横田
剛久 三輪
智子 富田
恵大 太田
恭平 西出
仁 川▲崎▼
洋介 金子
Original Assignee
三菱電機ビルソリューションズ株式会社
三菱電機株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 三菱電機ビルソリューションズ株式会社, 三菱電機株式会社 filed Critical 三菱電機ビルソリューションズ株式会社
Priority to JP2023546399A priority Critical patent/JP7466788B2/ja
Priority to PCT/JP2021/043147 priority patent/WO2023095236A1/fr
Publication of WO2023095236A1 publication Critical patent/WO2023095236A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management

Definitions

  • the present disclosure relates to a configuration of an equipment maintenance support device that presents a failure cause and a replacement part when equipment installed in a building fails, and a method for supporting equipment maintenance work.
  • the present disclosure aims to improve the efficiency of maintenance work by accurately identifying the cause of failure and replacement parts under various conditions.
  • the equipment maintenance support device of the present disclosure is an equipment maintenance support device that presents the cause of the failure and replacement parts based on the data of the inspection report created by the local worker when the equipment installed in the building breaks down, an inspection report data acquisition unit that reads data of equipment attribute information and equipment main operation information from the inspection report; A specific rule database that stores a plurality of specific rules that associate maintenance information with a priority rank, and a combination of the device attribute information read by the inspection report data acquisition unit and the device main operation information matches. a computing unit that selects a specific rule and uses the selected specific rule to identify and output the cause of the failure and the replacement part, and the priority rank indicates whether the specific accuracy of the specific rule is high or low. , wherein the calculation unit identifies the cause of failure and the replacement part using the identification rule selected when the priority rank is equal to or higher than a predetermined rank.
  • the cause of failure and the replacement part are identified using the identification rule selected when the priority rank is equal to or higher than the predetermined rank, so the cause of failure and the replacement part can be identified with high accuracy.
  • the inspection report data acquisition unit reads the device attribute information, the device operation information including the main device operation information, and the inspection report data including the inspection information from the inspection report. and store it in an inspection report database, read out the equipment attribute information, the equipment operation information including the equipment main operation information, and the inspection information from the inspection report database, and based on these information, a skilled person an inspection report analysis unit that generates inspection report analysis data by combining the input estimated failure cause, estimated replacement part, and priority rank, and stores the inspection report analysis data in an inspection report analysis database; The equipment attribute information, the equipment main operation information, the estimated maintenance information including the estimated failure cause and the estimated replacement parts, and the priority rank are extracted from the report analysis data and associated with each other. a specific rule generation unit that generates the specific rule and stores it in the specific rule database.
  • the expert inputs the estimated failure cause, estimated replacement parts, and priority rank to generate the specific rule.
  • the specific rule can reflect the experience of the expert, and the specific rule can be used to accurately identify the cause of failure and replacement parts.
  • a maintenance report data acquisition unit that reads maintenance report data including actual failure causes and actual replacement parts from a maintenance report created by an on-site worker and stores the data in a maintenance report database; a priority update unit that updates the priority rank of the specific rule database based on a comparison between the actual replacement part read from the maintenance report database and the estimated replacement part defined by the specific rule. good.
  • the identification accuracy of the identification rule is defined by an accuracy rate
  • the identification rule includes the equipment main operation information, the estimated failure cause, and the estimated replacement part.
  • the maintenance information is associated with specific rule information including the priority rank, the number of correct answers, the number of incorrect answers, and the correct answer rate, and the priority updating unit extracts the device attribute information from the maintenance report database.
  • the specific rule matching the combination thereof is set as an update target specific rule
  • the actual replacement part is read from the maintenance report database
  • the read actual replacement part is the update target specific rule If it matches with the estimated replacement part specified by , the number of correct answers is incremented by 1, and if not, the number of incorrect answers is incremented by 1, and the accuracy rate of the update target identification rule may be calculated, and the priority rank of the update target specific rule may be updated according to the accuracy rate and stored in the specific rule database.
  • the priority update unit reads the actual replacement part from the maintenance report database, and the read actual replacement part does not match the estimated replacement part defined by the update target identification rule.
  • the device attribute information read from the maintenance report database, the device main operation information, the actual failure cause read from the maintenance report database, the actual replacement parts, the priority rank, and the The new specific rule that associates the number of correct answers, the number of incorrect answers, and the correct answer rate may be stored in the specific rule database.
  • the inspection report data acquisition unit reads the device attribute information, the device operation information including the main device operation information, and the inspection information from the inspection report, and reads the device attribute information. information, the equipment operation information, and the inspection information as inputs, and an inference model database that stores a learned inference model that outputs the estimated failure cause and the estimated replacement part;
  • the priority rank of the specific rule is lower than a predetermined rank, or when the combination of the device attribute information read by the inspection report data acquisition unit and the device main operation information matches, the specific rule is If not, the device attribute information read by the inspection report data acquisition unit, the device operation information, and the inspection information are input to the learned inference model, and the cause of failure and the replacement part are input. can be output.
  • the learned inference model is Since the cause of the failure and the replacement parts are output using the method, the cause of the failure and the replacement parts can be estimated with high accuracy even if there is no suitable specific rule.
  • the computing unit computes a plurality of the failure causes, a plurality of the replacement parts, and a prediction probability of each of the failure causes and each of the regular replacement parts, and The failure cause and the replacement parts may be extracted and output.
  • the inference model database stores a plurality of learned inference models classified by a set of device type, logical format, and required memory capacity, and read out from the inspection report.
  • a model selection unit that selects one of the learned inference models from among the plurality of learned inference models based on device attribute information or hardware configuration, wherein the calculation unit is selected by the model selection unit; The cause of failure and the replacement part may be calculated using the learned inference model.
  • the cause of the failure and the replacement part are calculated using the learned inference model selected according to the target equipment of the maintenance work or the computing power of the equipment maintenance support device. Parts can be output.
  • the inspection report data acquisition unit obtains inspection report data including the equipment attribute information, the equipment operation information including the equipment main operation information, and the inspection information from the inspection report. is read and stored in an inspection report database, the equipment attribute information, the equipment operation information, and the inspection information are read from the inspection report database, and the estimation input by the expert based on these information an inspection report analysis unit that generates inspection report analysis data by combining the failure cause, the estimated replacement part, and the priority rank, and stores the data in an inspection report analysis database; and the equipment attribute information read from the inspection report analysis database. and generating training data having as input the equipment operation information and the inspection information and outputting the estimated maintenance information including the estimated failure cause and the estimated replacement parts read from the inspection report analysis database, and generating training data.
  • a training data generation unit for storing in a database; and an inference model learning that causes an inference model to perform learning based on the training data stored in the training data database, generates the trained inference model, and stores the trained inference model in the inference model database.
  • a maintenance and inspection report data generation unit that also generates maintenance and inspection report data and stores it in a maintenance and inspection report database; and reads out the actual replacement parts from the maintenance and inspection report database, and reads out the actual replacement parts and the learned inference model.
  • an inference model accuracy determination unit that calculates an accuracy index of the learned inference model based on the replacement parts output by is equal to or less than a predetermined threshold, the teacher data is updated based on the inspection report analysis data and stored in the teacher data database, and the inference model learning unit updates the teacher data based on the updated teacher data.
  • the inference model may be re-learned to update the learned inference model and store it in the inference model database.
  • the accuracy rate of the learned inference model is calculated in advance, and when the accuracy rate is equal to or less than a predetermined threshold, the teacher data is updated and learning is performed again using the updated teacher data. Since the trained inference model is updated, the accuracy rate of the trained inference model can be kept high, and the cause of the failure and replacement parts can be specified with high accuracy.
  • the accuracy index of the learned inference model is an accuracy rate of the learned inference model
  • the inference model database includes device types of the learned inference models, logical formats, and , the required memory capacity, the number of correct answers when estimating the replacement part using the learned inference model, the number of incorrect answers, and the accuracy rate are stored in association with each other, and the inference model accuracy determination unit
  • the number of correct answers of the learned inference model is incremented by 1; may calculate the accuracy rate of the learned inference model by incrementing the number of incorrect answers of the learned inference model by one.
  • the correct answer rate of the learned inference model is calculated based on this, so the correct answer rate of the learned inference model can be accurately measured.
  • the inference model database stores the plurality of learned inference models classified by a set of device type, logical format, and required memory capacity, and the a model selection unit that selects one of the trained inference models from among the plurality of trained inference models based on the logical format of the trained inference model or hardware configuration, wherein the inference model learning unit is , the learned inference model selected by the model selection unit may be re-learned based on the updated teacher data, and the learned inference model may be updated and stored in the inference model database.
  • the inspection report analysis data includes the installation location of the device, the device attribute information, and the inspection date
  • the teacher data generation unit generates the inspection report analysis from the inspection report analysis database.
  • the inspection report analysis data to be extracted when generating the teacher data may be selected based on the installation location, the device attribute information, and the inspection date.
  • the accuracy index of the learned inference model is an accuracy rate of the learned inference model
  • the inference model accuracy determination unit calculates the accuracy rate of the learned inference model. is lower than the re-learning prohibition threshold, an alarm is issued, the teacher data generation unit does not update the teacher data, and the inference model learning unit re-learns the learned inference model. You don't have to.
  • the inference model learning unit when the calculation unit is executing calculations using the learned inference model, No need to update.
  • the inspection report data acquisition unit obtains inspection report data including the equipment attribute information, the equipment operation information including the equipment main operation information, and the inspection information from the inspection report. is read and stored in an inspection report database, the equipment attribute information, the equipment operation information, and the inspection information are read from the inspection report database, and the estimation input by the expert based on these information an inspection report analysis unit that generates inspection report analysis data by combining the failure cause, the estimated replacement part, and the priority rank, and stores the data in an inspection report analysis database; and the equipment attribute information read from the inspection report analysis database. and generating training data having as input the equipment operation information and the inspection information and outputting the estimated maintenance information including the estimated failure cause and the estimated replacement parts read from the inspection report analysis database, and generating training data.
  • a training data generation unit for storing in a database; and an inference model learning that causes an inference model to perform learning based on the training data stored in the training data database, generates the trained inference model, and stores the trained inference model in the inference model database.
  • past failure data that associates the equipment attribute information of failures that occurred in the past, the equipment operation information, the inspection information, and maintenance information including actual failure causes and actual replacement parts.
  • a past failure case database that stores a plurality of past failure case examples
  • a past failure case extraction unit that extracts the past failure data in which the device attribute information and the device main operating information match the inspection report data and outputs the past failure case data to the teacher data generation unit
  • the teacher data generation unit receives the device attribute information, the device operation information, and the inspection information of the past failure data input from the past failure case extraction unit, and the Additional teaching data outputting the maintenance information including the actual failure cause and the actual replacement parts may be added to the teaching data database.
  • the inspection report data acquisition unit reads the device attribute information, the device operation information including the main device operation information, and the inspection information from the inspection report, and reads the device attribute information.
  • An incidental replacement part inference model database storing a replacement part inference model is provided, and the computing unit stores the equipment attribute information read by the inspection report data acquisition unit and the equipment operation data in the learned incidental replacement part inference model.
  • Information, the inspection information, and the failure cause and the replacement part specified by the calculation unit using the specification rule may be input, and the incidental replacement part may be output.
  • the incidental replacement parts that may be replaced in some cases can be effectively estimated. It can be displayed as a display, and the work and preparation efficiency of local workers can be improved.
  • the inspection report data acquisition unit obtains inspection report data including the equipment attribute information, the equipment operation information including the equipment main operation information, and the inspection information from the inspection report. is read and stored in an inspection report database, the equipment attribute information, the equipment operation information, and the inspection information are read from the inspection report database, and the estimation input by the expert based on these information an inspection report analysis unit that generates inspection report analysis data by combining the failure cause, the estimated replacement part, and the priority rank, and stores the data in an inspection report analysis database; and the equipment attribute information read from the inspection report analysis database. , the equipment operation information, the inspection information, the estimated maintenance information including the estimated failure cause and the estimated replacement parts, and input by a skilled person based on the information set read from the inspection report analysis database.
  • an incidental replacement part training data generation unit for generating incidental replacement part training data outputting the incidental replacement parts obtained and storing the data in the incidental replacement part training data database;
  • Incidental replacement part inference for causing an incidental replacement part inference model to learn based on incidental replacement part teaching data to generate the learned incidental replacement part inference model and storing the incidental replacement part inference model database in the incidental replacement part inference model database.
  • a model learning unit for generating incidental replacement part training data outputting the incidental replacement parts obtained and storing the data in the incidental replacement part training data database.
  • the equipment maintenance support device of the present disclosure includes an equipment structure information database including inspection manual information and drawing information of the equipment, and the incidental replacement part inference model learning unit learns the incidental replacement part inference model.
  • the inspection manual information and the drawing information acquired from the equipment structure information database may be referred to.
  • the maintenance support method of the present disclosure is a maintenance support method that presents the cause of the failure and replacement parts based on the data of the inspection report created by the local worker when the equipment installed in the building breaks down.
  • Data of equipment attribute information and equipment main operation information is read from the report, and the equipment attribute information, the equipment main operation information, presumed maintenance information including the presumed cause of failure and presumed replacement parts, and high or low specific accuracy.
  • the device attribute information, the device operation information including the device main operation information, and the inspection report data including the inspection information are read from the inspection report, and the read device attribute information, generating inspection report analysis data by combining the estimated failure cause, the estimated replacement part, and the priority rank input by the expert based on the equipment operation information and the inspection information, and generating the inspection report analysis;
  • the equipment attribute information, the equipment main operation information, the estimated maintenance information including the estimated failure cause and the estimated replacement part, and the priority rank are extracted from the data, and the identification is made by associating them. You can generate rules.
  • maintenance report data including actual failure causes and actual replacement parts is read from a maintenance report prepared by a field worker, and the actual replacement parts and the specific rule in the maintenance report data are read.
  • the priority rank of the specific rule may be updated based on the comparison with the probable replacement part that defines.
  • the identification accuracy of the identification rule is defined by an accuracy rate
  • the identification rule includes the equipment main operation information, the estimated failure cause, and the estimated replacement parts. information, the priority rank, the number of correct answers, the number of incorrect answers, and the correct answer rate.
  • the identification rule that matches the combination of the attribute information and the equipment main operation information is defined as an update target identification rule, and the actual replacement part in the maintenance report data matches the estimated replacement part defined by the update target identification rule. If yes, the number of correct answers is incremented by 1, and if there is no match, the number of incorrect answers is incremented by 1 to calculate the accuracy rate of the update target specific rule, and according to the accuracy rate may be used to update the priority rank of the update target specifying rule.
  • the device attribute information read from the inspection report, the device operation information including the device main operation information, and the inspection information are input, and the estimated failure cause and the estimated replacement part are obtained.
  • inspection report data including the device attribute information, the device operation information including the device main operation information, and the inspection information are read from the inspection report, and the device attribute information and , generating inspection report analysis data by combining the estimated failure cause, the estimated replacement part, and the priority rank input by the expert based on the equipment operation information and the inspection information, and generating the inspection report;
  • the device attribute information, the device operation information, and the inspection information in the analysis data are input, and the estimated maintenance information including the estimated failure cause and the estimated replacement parts in the inspection report analysis data is output.
  • the inference model is trained based on the generated training data to generate the learned inference model, and the actual failure cause and actual replacement parts are obtained from the maintenance report created by the local worker.
  • reading out the maintenance report data including the above, reading out the inspection results of the equipment subject to maintenance work from the inspection report data, generating maintenance inspection report data together with the maintenance report data, and generating maintenance inspection report data in the generated maintenance inspection report data
  • An accuracy index of the learned inference model is calculated based on the actual replacement part and the replacement part output by the learned inference model, and if the accuracy index is equal to or less than a predetermined threshold, the inspection report analysis data and re-learning the inference model based on the updated teacher data to update the learned inference model.
  • the accuracy index of the learned inference model is the accuracy rate of the learned inference model, the device type of the learned inference model, the logical format, the required memory capacity, and the Stores the number of correct answers, the number of incorrect answers, and an accuracy rate when estimating the replacement part using the learned inference model in association with each other, and stores the actual replacement part in the maintenance inspection report data and the inference.
  • the replacement part output by the model matches, the number of correct answers of the learned inference model is incremented by 1, and when the model does not match, the number of incorrect answers of the learned inference model is incremented by 1.
  • Clement may be used to calculate the accuracy rate of the learned inference model.
  • the device attribute information read from the inspection report device operation information including the device main operation information, inspection information, the failure cause identified using the identification rule, and the Prepare a learned incidental replacement part inference model having replacement parts as input and incidental replacement parts as outputs, and store the device attribute information, the device operation information, and The inspection information, the cause of failure identified by the identification rule, and the replacement part may be input, and the incidental replacement part may be output.
  • the device attribute information, the device operation information including the device main operation information, and the inspection report data including the inspection information are read from the inspection report, and the read device attributes are read.
  • generating inspection report analysis data by combining the estimated failure cause, the estimated replacement part, and the priority rank input by the expert based on the information, the equipment operation information, and the inspection information;
  • the equipment attribute information, the equipment operation information, the inspection information, and the estimated maintenance information including the estimated failure cause and the estimated replacement parts in the inspection report analysis data are input, and the inspection report analysis data is obtained.
  • incidental replacement part training data whose output is the incidental replacement part input by the expert based on the information set in the above, and training the incidental replacement part inference model based on the incidental replacement part teaching data may be performed to generate the learned incidental replacement part inference model.
  • inspection manual information and drawing information of the equipment may be referred to when making the incidental replacement part inference model learn.
  • FIG. 2 is a diagram showing inspection report data stored in an inspection report database shown in FIG. 1;
  • FIG. FIG. 3 is an explanatory diagram showing generation of inspection report analysis data stored in the inspection report analysis database shown in FIG. 1;
  • FIG. 2 is a diagram showing data extracted for specific rules from the inspection report analysis database shown in FIG. 1;
  • 2 is a diagram showing the configuration of a specific rule database shown in FIG. 1;
  • FIG. 2 is a diagram showing the configuration of a teacher data database shown in FIG. 1;
  • FIG. 2 is a diagram showing the configuration of an inference model database shown in FIG. 1;
  • FIG. 1 is a system diagram showing the configuration of a general-purpose computer
  • FIG. 4 is a flow chart showing the operation of generating a specific rule of the equipment maintenance support device of the embodiment shown in FIG. 1
  • 4 is a flow chart showing the operation of generating a learned inference model of the equipment maintenance support device of the embodiment shown in FIG. 1
  • 4 is a flow chart showing a failure cause and a replacement part identifying operation of the equipment maintenance support device of the embodiment shown in FIG. 1
  • It is a functional block diagram of the equipment maintenance assistance device of other embodiments.
  • It is a functional block diagram of the equipment maintenance assistance device of other embodiments.
  • FIG. 9 is a functional lock diagram showing the configuration of an equipment maintenance support device according to another embodiment
  • 16 is a diagram showing the configuration of a maintenance report database of the equipment maintenance support device of the embodiment shown in FIG. 15
  • FIG. 16 is a diagram showing the configuration of a specific rule database of the equipment maintenance support device of the embodiment shown in FIG. 15
  • FIG. 16 is a flow chart showing the update operation of the priority rank of the specific rule of the equipment maintenance support device of the embodiment shown in FIG. 15
  • FIG. It is a functional block diagram of the equipment maintenance assistance device of other embodiments.
  • 20 is a diagram showing the configuration of a maintenance and inspection report database of the device maintenance support device of the embodiment shown in FIG. 19;
  • FIG. 20 is a diagram showing the configuration of a teacher data database of the device maintenance support device of the embodiment shown in FIG. 19;
  • FIG. 20 is a diagram showing the configuration of an inference model database of the device maintenance support device of the embodiment shown in FIG. 19;
  • FIG. 20 is a flow chart showing an update operation of a learned inference model of the equipment maintenance support device of the embodiment shown in FIG. 19;
  • FIG. FIG. 24 is a flow chart following the flow chart shown in FIG. 23;
  • FIG. It is a functional block diagram which shows the structure of the equipment maintenance assistance apparatus of other embodiment.
  • It is a functional block diagram which shows the structure of the equipment maintenance assistance apparatus of other embodiment.
  • 27 is a diagram showing the configuration of a past failure example database of the device maintenance support device of the embodiment shown in FIG. 26;
  • FIG. 30 is a diagram showing a configuration of an incidental replacement part teaching data database shown in FIG. 29;
  • FIG. 30 is a flow chart showing an operation of generating a learned incidental replacement part inference model of the equipment maintenance support device shown in FIG. 29;
  • FIG. 30 is a flow chart showing a failure cause and replacement part identifying operation using the identification rule of the equipment maintenance support device shown in FIG. 29 and an incidental replacement part estimation operation by a learned incidental replacement part inference model;
  • FIG. This is a continuation of the flow chart shown in FIG.
  • the equipment maintenance support device 100 of the embodiment includes an inspection report data acquisition unit 10, an inspection report database 11, an inspection report analysis unit 12, a display unit 13, an input unit 14, an inspection report Analysis database 15, specific rule generator 16, specific rule database 17a, teacher data generator 20, teacher data database 21, inference model learning unit 22, inference model database 23a, calculator 31, failure
  • an inspection report data acquisition unit 10 an inspection report database 11
  • an inspection report analysis unit 12 a display unit 13, an input unit 14, an inspection report Analysis database 15, specific rule generator 16, specific rule database 17a, teacher data generator 20, teacher data database 21, inference model learning unit 22, inference model database 23a, calculator 31, failure
  • Each functional block of the cause and replacement parts display section 32 is provided.
  • the inspection report data acquisition unit 10 receives an inspection report 71 as shown in FIG. Information and inspection report data including inspection information are read and stored in the inspection report database 11 .
  • the inspection report 71 includes building name, address, telephone number, date of inspection, time of inspection, inspector, message to customer, date of call, content of call, error code, occurrence It is a document in which each code of situation, probable cause, treatment contents, model name of repair equipment, serial number, serial number, special notes, equipment, situation, cause, failure, treatment, and quantity is entered.
  • the inspection report 71 may be created by the local worker 81 using, for example, a tablet terminal or the like, and input to the inspection report data acquisition unit 10 of the equipment maintenance support device 100 via a communication line such as the Internet line.
  • a paper document prepared by the local worker 81 may be read by a scanner or the like at the service center and input.
  • the inspection report data acquisition unit 10 reads each entry in the inspection report 71 as text data and generates inspection report data as shown in FIG.
  • the inspection report data acquisition unit 10 inputs the device type corresponding to the device code of the inspection report 71 .
  • a device code "F” indicating the device type of the air conditioner is described. Therefore, enter “air conditioner” as the device type.
  • the contents of the call are entered in the symptom column of the equipment operation information, and the contents of the treatment are entered in the check result column. Items other than these are entered in the columns corresponding to the respective items of the inspection report 71 .
  • Inspection report data includes call date, installation location including building name and address, equipment attribute information including equipment type, equipment model number and serial number, equipment operation information including symptom, error code and occurrence status, inspection It is a data set that associates inspection information including dates, inspection times, workers, and inspection results. Also, the inspection report database 11 is a database storing a plurality of inspection report data.
  • the inspection report analysis unit 12 reads the equipment attribute information, the equipment operation information including the main equipment operation information, and the inspection information from the inspection report database 11, and based on these information, analyzes the presumed failure entered by the expert 82. By combining the cause, the estimated replacement part, and the priority rank, inspection report analysis data shown in FIG. 3 is generated and stored in the inspection report analysis database 15 .
  • the display unit 13 displays the inspection report data to the expert 82, and the input unit 14 allows the expert 82 to input an estimated cause of failure, an estimated replacement part, and a priority rank based on the displayed inspection report data.
  • the display unit 13 and the input unit 14 may be configured by, for example, a touch pad carried by the expert 82 .
  • the inspection report analysis data consists of the call date and installation location of the inspection report data, equipment attribute information, equipment operation information, and inspection information.
  • This is a data set in which the input presumed cause of failure and presumed replacement parts are used as presumed maintenance information, and the priority rank input by the expert 82 is associated as the priority rank of specific rule information.
  • the priority rank input by the expert 82 indicates the rank of high or low identification accuracy of the estimated cause of failure and the estimated replacement part input by the expert 82.
  • ranks B and C are input as the identification accuracy decreases.
  • the specific rule generating unit 16 obtains from the inspection report analysis database 15, the equipment attribute information including the equipment type and equipment model number in the inspection report analysis data, and the equipment main operation information including symptoms and error codes. , presumed maintenance information including the presumed cause of failure and presumed replacement parts, and the priority rank are extracted, and a specific rule relating them as shown in FIG. 6 is generated and stored in the specific rule database 17a.
  • the specific rule generation unit 16 generates specific rules as specific rule information indicating the level of specific accuracy when estimating a replacement part according to the specific rules for the priority rank in the inspection report analysis data.
  • the training data generation unit 20 generates device attribute information including the device type and device model number read from the inspection report analysis database 15, device operation information including symptoms, error codes, and occurrence conditions,
  • the inspection information is input, and the teacher data is generated and stored in the teacher data database 21 with the estimated maintenance information including the estimated failure cause and the estimated replacement parts read out from the inspection report analysis database 15 as output.
  • the inference model learning unit 22 causes the inference model to learn based on the teacher data stored in the teacher data database 21, generates a learned inference model, and stores it in the inference model database 23a.
  • the inference model database 23a stores a plurality of learned inference models classified by a combination of device type, logical format, and required memory capacity.
  • the calculation unit 31 selects a specific rule in which the combination of the device attribute information including the device type and device model number read by the inspection report data acquisition unit 10 and the device main operation information including the symptom and error code matches, Using the selected identification rule, the cause of the failure and replacement parts are identified and output.
  • the operation unit 31 stores device attribute information including the device type and device model number read by the inspection report data acquisition unit 10 in the learned inference model, device operation information including symptoms, error codes, and occurrence conditions. It inputs inspection information, and outputs the cause of failure and replacement parts.
  • the failure cause and replacement part display unit 32 displays the failure cause and replacement parts input from the calculation unit 31 to the field worker 81 .
  • the failure cause and replacement part display unit 32 is, for example, a tablet terminal carried by the local worker 81, and the failure cause and replacement part display unit 32 is displayed on the tablet carried by the local worker 81 via a communication line such as the Internet. You may display a failure cause and a replacement part on a terminal.
  • a general-purpose computer 150 includes a CPU 151 that is a processor that performs information processing, a ROM 152 and a RAM 153 that temporarily store data during information processing, and a hard disk drive that stores programs and user data. (HDD) 154, a mouse 155 provided as input means, a keyboard 156, a USB port 159, and a display 157 provided as a display device.
  • a data bus 162 connects the CPU 151 , the ROM 152 , the RAM 153 and the HDD 154 .
  • the mouse 155 , keyboard 156 and display 157 are also connected to the data bus 162 via the input/output controller 158 .
  • a network controller 160 provided as communication means is connected to the data bus 162 .
  • the inspection report data acquisition unit 10, the inspection report analysis unit 12, the specific rule generation unit 16, the teacher data generation unit 20, the inference model learning unit 22, and the calculation unit 31 shown in FIG. and a program running on the CPU 151 cooperate with each other.
  • the inspection report database 11, inspection report analysis database 15, specific rule database 17a, teacher data database 21, and inference model database 23a are implemented by storing data in the HDD 154 of the general-purpose computer 150 shown in FIG. Incidentally, instead of the HDD 154, it may be realized by using an external storage means via a network.
  • the display unit 13, the input unit 14, the portion for inputting the inspection report 71 in the inspection report data acquisition unit 10, and the failure cause and replacement parts display unit 32 are carried by the expert 82 and the local worker 81.
  • the display unit 13 and the input unit 14 may be configured by the display 157, the mouse 155 and the keyboard 156 shown in FIG.
  • the part for inputting the inspection report 71 in the inspection report data acquisition unit 10 is configured by an input device connected to the USB port 159 shown in FIG. good too.
  • the cause of failure and replacement parts display section 32 may be configured by the display 157 shown in FIG.
  • the inspection report data acquisition unit 10 acquires the data of the inspection report 71 created by the local worker 81.
  • the inspection report data acquisition unit 10 may acquire the data of the inspection report 71 created by the local worker 81 using a tablet terminal or the like via a communication line such as the Internet line.
  • the inspection report data acquisition unit 10 reads out the entries in each column of the inspection report 71 as text data as shown in Fig. 2 . Then, as shown in FIG. 2, the inspection report data acquisition unit 10 acquires the call date, the installation location including the building name and address, the device attribute information including the device type, device model number, and serial number, symptom and error Inspection report data, which is a data set in which equipment operation information including codes and occurrence conditions, and inspection information including inspection dates, inspection times, workers, and inspection results are associated with each other, is generated. The inspection report data acquisition unit 10 then stores the generated inspection report data in the inspection report database 11 .
  • the inspection report analysis unit 12 reads inspection report data from the inspection report database 11 and displays it on the display unit 13.
  • the display unit 13 may be a touch pad carried by the expert 82 .
  • the expert 82 confirms the inspection report data displayed on the display unit 13, predicts the cause of failure of the equipment and replacement parts, and inputs the input unit 14 to the estimated cause of failure. Enter replacement parts. Further, when the expert 82 inputs the estimated cause of failure and the estimated replacement part, the expert 82 inputs a priority rank indicating the rank of the estimated accuracy of the estimated cause of failure and the estimated replacement part.
  • the inspection report analysis unit 12 generates an inspection data set that associates the inspection report data, the estimated failure cause, the estimated replacement parts, and the priority rank input by the expert 82. Report analysis data is generated and stored in the inspection report analysis database 15 as shown in FIG.
  • the specific rule generation unit 16 acquires the inspection report analysis data from the inspection report analysis database 15, and determines whether the equipment including the equipment type and the equipment model number in the inspection report analysis data Attribute information, equipment main operation information including symptoms and error codes, presumed maintenance information including presumed failure causes and presumed replacement parts, and priority ranks are extracted and identified by associating them as shown in FIG. A rule is generated and stored in the specific rule database 17a.
  • the inspection report analysis unit 12 displays the inspection report data on the display unit 13 in steps S203 to S205 of FIG. and the priority rank to generate inspection report analysis data and store it in the inspection report analysis database 15 .
  • the training data generation unit 20 generates device attribute information including the device type and model number read from the inspection report analysis database 15, symptoms, error codes, and occurrence conditions. and inspection information are input, and estimated maintenance information including estimated failure causes and estimated replacement parts read from the inspection report analysis database 15 is generated and stored in the teacher data database 21. .
  • the inference model learning unit 22 reads the teacher data from the teacher data database 21 and makes the inference model learn with the read teacher data to obtain a learned inference model.
  • the inference model learning unit 22 stores a plurality of inference models for each device type and logical format of the device to be inferred.
  • the target device is an air conditioner
  • the logic format stores an inference model of logistic regression, neural network, or decision tree.
  • Each inference model has a predetermined memory capacity required for execution. Therefore, the inference model learning unit 22 selects an inference model based on the device type included in the teacher data read from the teacher data database 21 . For example, as shown in FIG. 7, when the device type included in the teacher data is an air conditioner, the target device selects an air conditioner inference model. Then, the optimum inference model for learning is selected based on the learning results so far or the hardware configuration of the equipment maintenance support apparatus 100 .
  • the inference model learning unit 22 inputs the device type, device model number, symptom, error code, occurrence situation, and inspection information of the training data to the selected inference model, and replaces the failure cause with the inference model. Output parts. Then, the failure cause and the replacement part output by the inference model are compared with the estimated failure cause and the estimated replacement part in the training data, and a plurality of parameters in the inference model are changed so that they match. Then, when both of them substantially match each other, the learning of the estimation model is terminated, and the model is regarded as a learned estimation model.
  • step S208 of FIG. 11 the inference model learning unit 22 prepares the learned inference model as shown in FIG. store in
  • the calculation unit 31 reads the inspection report data from the inspection report database 11. Then, proceeding to step S302 in FIG. 12, it is determined whether or not the combination of the device type, device model number, symptom, and error code in the read inspection report data matches one of the specific rules.
  • the calculation unit 31 reads out the inspection report data shown in FIG. is.
  • the combination of device type, device model number, symptom, and error code matches air conditioner, ABC01, abnormal stop, and 1234.
  • the calculation unit 31 determines YES in step S302 of FIG. 12 and proceeds to step S303 of FIG.
  • step S303 of FIG. 12 the calculation unit 31 selects the No. that matches the combination of the device type, device model number, symptom, and error code. 1 specific rule is selected and the process proceeds to step S304 in FIG.
  • the calculation unit 31 determines whether the priority rank of the selected specific rule is equal to or higher than a predetermined rank.
  • the predetermined rank can be freely determined. For example, if the predetermined rank is B and the priority rank of the selected specific rule is A or B, the calculation unit 31 performs step S304 in FIG. is determined to be YES. As in the previous example, No. When one specific rule is selected, the priority rank is A, so the calculation unit 31 determines YES in step S304 of FIG. On the other hand, No. shown in FIG. 5 specific rule is selected, the calculation unit 31 determines NO in step S304 of FIG.
  • step S304 in FIG. 12 the operation unit 31 proceeds to step S305 in FIG. to display.
  • the calculation unit 31 may select a learned inference model that matches the device type read from the inspection report data for the target device, and read the selected learned inference model. Further, the calculation unit 31 may select a learned inference model based on the device type read from the inspection report data and the hardware configuration of the device maintenance support device 100, and read the selected learned inference model.
  • the calculation unit 31 inputs the device type, device model number, symptom, error code, occurrence situation, and inspection result of the inspection report data read to the learned inference model in step S307 of FIG. Output the route cause and replacement parts. Then, the output failure cause and replacement parts are displayed on the failure cause and replacement part display section 32 .
  • the expert 82 determines the estimated failure cause, the estimated replacement parts, and the priority rank based on the device attribute information, the device operation information, and the inspection information of the inspection report data. Since the specific rule is generated by inputting, the specific rule can reflect the experience of the expert 82 . When the priority rank is equal to or higher than a predetermined rank, the cause of the failure and the replacement part are specified using the specified rule, so the cause of the failure and the replacement part can be specified with high accuracy.
  • the learned inference model is used to Since the cause and the replacement parts are output, the cause of the failure and the replacement parts can be accurately estimated even if there is no appropriate specific rule.
  • the equipment maintenance support device 100 of the embodiment can accurately identify the cause of failure and replacement parts under various conditions, and improve the efficiency of maintenance work.
  • the learned inference model inputs the device type, device model number, symptom, error code, occurrence situation, and inspection result of the inspection report data, and inputs multiple failure route causes, multiple replacement parts, each failure cause, and each replacement part.
  • the prediction probability may be the probability of outputting the cause of failure and the replacement part defined by the teacher data when inference is executed by inputting a plurality of teacher data during learning.
  • the calculation unit 31 may extract the cause of failure and replacement parts with high prediction probability, and display them on the cause of failure and replacement part display unit 32 .
  • a plurality of failure causes and replacement parts with high prediction probabilities may be extracted and displayed on the failure cause and replacement part display section 32 . As a result, the local worker 81 can easily determine replacement parts.
  • the calculation unit 31 selects the learned inference model to be read based on the device type read from the inspection report data and the hardware configuration of the device maintenance support device 100, but the present invention is not limited to this.
  • a model selection unit 24a that selects a learned inference model may be provided, and the calculation unit 31 may be configured to calculate the cause of failure and replacement parts using the learned inference model selected by the model selection unit 24a.
  • FIG. 15 to 18 Parts similar to those of the equipment maintenance support device 100 described above with reference to FIGS.
  • the device maintenance support device 400 shown in FIG. 15 is configured by adding a maintenance report data acquisition unit 41, a maintenance report database 42, and a priority update unit 43 to the device maintenance support device 100 described with reference to FIGS. is.
  • Other functional blocks are the same as the functional blocks constituting the equipment maintenance support apparatus 100 described above, so descriptions thereof will be omitted.
  • the maintenance report data acquisition unit 41 reads maintenance report data including actual failure causes and actual replacement parts from the maintenance report 72 created by the field worker 81 and stores it in the maintenance report database 42 .
  • the maintenance report 72 is a document prepared by the local worker 81 after performing maintenance work such as parts replacement of the equipment on-site, and is a document in which the actual cause of the failure and the actually replaced parts are entered. As shown in FIG. 16, the maintenance report 72 includes building name, address, telephone number, date of work, work hours, worker, message to customer, date of call, content of call, error code, occurrence It is a document in which each code of situation, cause, action content, model name of repair equipment, serial number, serial number, special notes, machine, situation, cause, failure, action, and quantity is entered.
  • the maintenance report 72 may be created by the local worker 81 using, for example, a tablet terminal or the like, and input to the inspection report data acquisition unit 10 of the equipment maintenance support device 100 via a communication line such as the Internet line.
  • a paper document prepared by the local worker 81 may be read by a scanner or the like at the service center and input.
  • the maintenance report data acquisition unit 41 reads each entry in the maintenance report 72 as text data and generates maintenance report data as shown in FIG.
  • the maintenance report data acquisition unit 41 inputs the air conditioner from the device code F of the maintenance report 72 into the device type, inputs the call content into the symptom column of the device operation information, and identifies the cause. Enter the description in the actual failure cause column, and enter the action content in the actual replacement part column. Items other than this are entered in the corresponding fields of the maintenance report 72 .
  • the maintenance report data includes call date, installation location including building name and address, device attribute information including device type, device model number and serial number, device operation information including symptom, error code and occurrence status, work It is a data set that associates dates, work hours, workers, actual failure causes, actual replacement parts, and maintenance information. Also, the maintenance report database 42 is a database storing a plurality of maintenance report data.
  • the priority update unit 43 updates the priority rank of the specific rule database 17b based on the comparison between the actual replacement parts read from the maintenance report database 42 and the estimated replacement parts defined by the specific rules.
  • the plurality of specific rules stored in the specific rule database 17b of the equipment maintenance support apparatus 400 are the specific rules stored in the specific rule database 17a of the equipment maintenance support apparatus 100 described with reference to FIG.
  • the number of correct answers in which the replacement part specified using the specific rule in the specific rule information matches the actual replacement part actually replaced on site by the local worker 81, the number of incorrect answers, and the number of correct answers and the number of incorrect answers. is a data set that associates the accuracy rate, which is the ratio of the number of correct answers to the total number of times.
  • the accuracy rate is an index that defines the accuracy of specifying a specific rule.
  • the priority update unit 43 increments the number of correct answers by 1 when the replacement part specified using the specification rule matches the actual replacement part actually replaced at the site by the field worker 81, and , the number of incorrect answers is incremented by 1 to calculate the correct answer rate, and the priority rank is updated according to the correct answer rate.
  • the maintenance report data acquisition unit 41 acquires the data of the maintenance report 72, and, as shown in step S402 of FIG.
  • the item is read out as text data, maintenance report data is created, and stored in the maintenance report database 42 .
  • the priority update unit 43 reads maintenance report data from the maintenance report database 42. Then, in step S404 of FIG. 18, the priority update unit 43 updates the combination of the device type, device model number, symptom, and error code that matches the combination of the device type, device model number, symptom, and error code in the read maintenance report data. Determine if there are specific rules to include.
  • step S404 of FIG. 18 If it is determined YES in step S404 of FIG. 18, the process proceeds to step S405 of FIG. 18, and the priority update unit 43 matches the combination of the read maintenance report data device type, device model number, symptom, and error code.
  • a specific rule including a combination of device type, device model number, symptom, and error code is defined as an update target specific rule.
  • step S406 in FIG. 18 it is determined whether or not the estimated replacement part specified in the update target identification rule matches the actual replacement part specified in the maintenance report data. Then, if the estimated replacement part and the actual replacement part match, the process proceeds to step S407 in FIG. 18, increments the number of correct answers of the update target specifying rule by 1, and proceeds to step S408 in FIG.
  • the priority updating unit 43 proceeds to step S408 in FIG. 18 to calculate the accuracy rate of the update target specific rule, and proceeds to step S409 in FIG. Update.
  • step S406 of FIG. 18 the process proceeds to step S410 of FIG.
  • the priority update unit 43 advances to step S411 in FIG. 18, and selects one specification rule based on the device type, device model number, symptom, error code, actual failure cause, and actually replaced parts of the update target specification rule. It is generated and registered in the specific rule database 17b.
  • the identification rule 6 is the update target identification rule, and if the presumed cause of failure is an abnormality in the temperature sensor and the presumed replacement part is the temperature sensor, and if the actual cause of the failure is an abnormality in the remote controller and the actual replacement part is the remote controller, the priority update unit 43 is No. No. 6 in which the presumed cause of failure and the presumed replacement part of the specific rule No. 6 are replaced with the actual cause of failure and the actual replacement part.
  • 100 specific rules are generated and added to the specific rule database 17b.
  • the priority update unit 43 proceeds to step S408 in FIG. 18 to calculate the accuracy rate of the update target specific rule, and in step S409 in FIG. 18, updates the priority rank based on the accuracy rate.
  • the priority update unit 43 determines NO in step S104 of FIG. 18, it determines that there is no specific rule to be updated, and terminates the update operation of the specific rule.
  • priority rank A has an accuracy rate of 80% or more
  • B has an accuracy rate of 70% or more and less than 80%
  • C has an accuracy rate of 60% or more and less than 70%
  • D has an accuracy rate of 50% or more and 60%.
  • E is defined as having a correct answer rate of less than 50%, and when the correct answer rate rises from 79% to 81%, the priority rank is updated from B to A, and the correct answer rate rises from 71% to If the percentage drops to 68%, the priority rank may be lowered from B to C°, and if the accuracy rate fluctuates within the accuracy rate range of each rank, the priority rank may not be updated.
  • the equipment maintenance support device 400 of the embodiment described above reduces the number of correct answers to the update target specifying rule to 1 when the actual replacement part specified in the maintenance report data matches the estimated replacement part specified by the update target specifying rule. If there is no match, the priority rank of the specific rule to be updated is updated by a learning operation in which the number of incorrect answers is incremented by 1, so the accuracy of estimating the specific rule can be improved.
  • the priority update unit 43 updates the estimated cause of failure and the estimated replacement part specified by the update target specification rule. is replaced with the actual failure cause and the actual replacement part, and added to the specific rule database 17b.
  • the scope of application of the identification rule can be expanded, and the cause of failure and replacement parts can be identified with high accuracy under various conditions, thereby improving the efficiency of maintenance work.
  • the accuracy of specifying the specific rule is defined by the accuracy rate.
  • the accuracy is not limited to this. It may be defined by the index of
  • FIG. 19 The same functional blocks as those of the equipment maintenance support apparatuses 100 and 400 described earlier with reference to FIGS. 1 to 9 and FIGS.
  • the equipment maintenance assistance device 500 includes the equipment maintenance assistance device 100 described with reference to FIGS. This is a configuration in which an accuracy determination unit 46 is added.
  • the maintenance and inspection report data generation unit 44 reads out maintenance report data including actual failure causes and actual replacement parts from the maintenance report 72 created by the local worker 81, and obtains inspection results of the target equipment for maintenance work from the inspection report database 11. is read, and together with the maintenance report data, maintenance inspection report data is generated and stored in the maintenance inspection report database 45 .
  • the maintenance inspection report data generation unit 44 reads the installation location, equipment attribute information, etc. from the inspection report database 11, identifies the target equipment for maintenance work described in the maintenance report 72, and determines the inspection result of the target equipment. read out. Then, the inspection results are combined with the maintenance report data to generate maintenance inspection report data as shown in FIG.
  • the inference model accuracy determination unit 46 reads the actual replacement parts from the maintenance inspection report database 45, and calculates the accuracy index of the learned inference model based on the read actual replacement parts and the replacement parts output by the learned inference model.
  • the accuracy index is explained as the accuracy rate when estimating a replacement part using a learned inference model.
  • the inference model database 23b of the device maintenance support device 500 uses the device type, logical format, required memory capacity, and learned inference model of the learned inference model to determine replacement parts. It stores the number of correct answers, the number of incorrect answers, and the accuracy rate when estimation is performed, in association with each other. If the actual replacement part read from the maintenance inspection report database 45 matches the replacement part output by the inference model, the inference model accuracy determination unit 46 increments the number of correct answers of the learned inference model by one. If they do not match, the number of incorrect answers of the learned inference model is incremented by 1 to calculate the accuracy rate of the learned inference model.
  • the teacher data generation unit 20 updates the teacher data based on the inspection report analysis data, stores it in the teacher data database 21, and performs inference.
  • the model learning unit 22 re-learns the inference model based on the updated teacher data, updates the learned inference model, and stores it in the inference model database 23b.
  • the predetermined threshold can be freely set, and may be 90% or 70%, for example.
  • the maintenance inspection report data generator 44 acquires the data of the maintenance report 72. Then, in step S502 of FIG. 23, the maintenance inspection report data generator 44 reads out maintenance report data as shown in FIG. 16 from each entry in the maintenance report 72. The maintenance inspection report data generation unit 44 also identifies the target equipment for maintenance work described in the maintenance report 72 from the inspection report database 11, and reads out the inspection result of the target equipment. Then, the inspection results are combined with the maintenance report data to generate maintenance inspection report data as shown in FIG.
  • the inference model accuracy determination unit 46 determines the number of the learned inference model used from the calculation unit 31, the input data input to the learned inference model, the failure cause calculated by the learned inference model, and the and replacement parts. Then, in step S504 of FIG. 23, the inference model accuracy determination unit 46 inputs the device type, device model number, symptom, error code, occurrence situation, and inspection result from the maintenance inspection report database 45 to the learned inference model. Extract maintenance report data that matches the input data.
  • the inference model accuracy determination unit 46 extracts the actual failure cause and actual replacement parts as shown in FIG. 20 from the maintenance inspection report data extracted in step S505 of FIG. 23, and proceeds to step S506 of FIG.
  • step S506 of FIG. 23 the inference model accuracy determination unit 46 determines whether the cause of failure and the replacement part calculated by the learned inference model match the actual cause of failure and the actual replacement part extracted from the maintenance inspection report data. do.
  • step S506 of FIG. 23 determines YES in step S506 of FIG. 23, it proceeds to step S507 of FIG. 23 and increments the number of correct answers of the learning inference model shown in FIG.
  • the inference model accuracy determination unit 46 proceeds to step S508 in FIG. 24 to calculate the accuracy rate of the learned inference model, proceeds to step S509 in FIG. 24, and determines whether the accuracy rate is equal to or less than a predetermined value.
  • step S510 of FIG. Error code, Occurrence status, Inspection result, Presumed failure cause, Presumed replacement parts columns are extracted, and model type, model number, Symptom, Error code, Occurrence status, Inspection result are input, Presumed failure cause, Presumed replacement
  • the teacher data outputting the part is generated and added to the teacher data database 21 to update the teacher data.
  • step S511 of FIG. 24 the inference model learning unit 22 reads the updated teacher data from the teacher data database 21, makes the inference model learn with the read updated teacher data, and updates the learned inference model. Then, in step S512 of FIG. 24, the inference model learning unit 22 stores the updated learned inference model in the inference model database 23b.
  • step S513 the inference model accuracy determination unit 46 proceeds to step S513 in FIG. 23, increments the number of incorrect answers of the trained inference model by 1, and Proceed to S514.
  • the inference model accuracy determination unit 46 corrects and adds teacher data based on the actually replaced parts.
  • the cause of failure and the replacement parts output when the teacher data No. 6 is input to the learned inference model are the teacher data No. 6 shown in FIG. 6 presumed cause of failure, abnormality of the temperature sensor of the presumed replacement part, and temperature sensor.
  • step S508 of FIG. 23 the correct answer rate of the learned inference model is calculated in step S508 of FIG.
  • the teacher data is updated, the inference model is re-learned in step S511 of FIG. 24 to update the learned inference model, and the updated learned inference model is stored in the inference model database 23b in step S512 of FIG.
  • the equipment maintenance support device 500 counts the number of correct answers and the number of incorrect answers of the learned inference model, and calculates the correct answer rate of the learned inference model based on this count. can be accurately measured.
  • the equipment maintenance support apparatus 500 updates the teacher data and re-learns using the updated teacher data. Since the inference model is updated, the accuracy rate of the learned inference model can be kept high, and the cause of failure and replacement parts can be specified with high accuracy.
  • the correct answer rate is used as the accuracy index of the learned inference model.
  • Indicators may be used.
  • the learned inference model is updated when the accuracy rate of the learned inference model becomes equal to or less than a predetermined threshold value, but this is not the only option.
  • the inference model accuracy determination unit 46 issues an alarm, and the teacher data generation unit 20 updates the teacher data.
  • the inference model learning unit 22 may be configured not to relearn the learned inference model.
  • the relearning prohibition threshold can be set freely, and may be set to 60% or 50%, for example.
  • the inference model learning unit 22 is configured not to update the learned inference model in the inference model database 23b when the calculation unit 31 is executing calculations using the learned inference model. good too.
  • inspection report analysis data includes the installation location of the equipment, equipment attribute information, and the date of inspection. Inspection report analysis data to be extracted when generating teacher data may be selected based on the attribute information and inspection date.
  • the model selection unit 24c selects one trained inference model from among the trained inference models, and the inference model learning unit 22 re-selects the trained inference model selected by the model selection unit 24c based on the updated teacher data. Learning may be performed to update the learned inference model and store it in the inference model database 23b.
  • FIG. 26 The same functional blocks as those of the equipment maintenance support device 100 described above with reference to FIGS.
  • an equipment maintenance support device 700 includes a past failure case extraction unit 51 and a past failure case database 52 in the equipment maintenance support device 100 described above with reference to FIGS. It is an addition.
  • the past failure case database 52 stores equipment attribute information including equipment types and equipment model numbers of past failures, equipment operating information including symptoms, error codes and occurrence conditions, and inspection results.
  • a plurality of past failure data are stored in which inspection information including actual failure causes and maintenance information including actual replacement parts are associated with each other.
  • the past failure case extraction unit 51 extracts past failure data in which the equipment attribute information and the equipment main operation information including symptoms and error codes match the inspection report data, and outputs the data to the teacher data generation unit 20 . Output.
  • the training data generation unit 20 receives the device attribute information, the device operation information, and the inspection information of the past failure data input from the past failure case extraction unit 51, and extracts the actual failure cause and the actual replacement part from the past failure data. Additional teacher data whose output is the maintenance information included is added to the teacher data database 21 .
  • the inference model learning unit 22 makes the inference model learn using the teacher data to which the additional teacher data has been added, and stores the learned inference model in the inference model database 23a.
  • the equipment maintenance support device 700 can increase the number of teacher data, Part estimation accuracy can be improved.
  • the equipment maintenance support device 800 of the embodiment will be described with reference to FIG.
  • the equipment maintenance support device 800 has the teacher data generator 20, the teacher data database 21, and the inference model database 23a of the device maintenance support device 400 described with reference to FIGS.
  • the cause of the failure and replacement parts are identified by the identification rule.
  • the operation of identifying the cause of failure and replacement parts using the specific rule and the operation of updating the priority rank of the specific rule are the same as those of the equipment maintenance support device 400 .
  • the equipment maintenance support device 800 can identify the cause of the failure and replacement parts with a simple configuration.
  • FIG. 29 to 33 The same functional blocks as those of the equipment maintenance support device 100 described above with reference to FIGS. 29 to 33.
  • the equipment maintenance support device 900 includes an incidental replacement parts teaching data generation unit 91 and an incidental replacement parts teaching data database 92 in addition to the equipment maintenance assistance device 100 described with reference to FIGS. , an equipment structure information database 93, an incidental replacement parts inference model learning unit 94, an incidental replacement parts inference model database 95, and a display/input unit 14a.
  • the equipment maintenance support device 900 uses the identification rule to identify the cause of the failure and the replacement part, and when the identified replacement part is replaced using the learned incidental replacement part inference model, the equipment maintenance support device 900 An incidental replacement part to be replaced is estimated, and the incidental replacement part is displayed on the failure cause and replacement part display section 32 together with the cause of failure and the replacement part.
  • the incidental replacement parts teaching data generation unit 91 generates estimated maintenance information including equipment attribute information read from the inspection report analysis database 15 shown in FIG. 5, equipment operation information, inspection information, estimated failure causes, and estimated replacement parts. , and based on the information set read out from the inspection report analysis database 15, the incidental replacement part input by the expert 82 from the display/input unit 14a shown in FIG. Parts teaching data is generated and stored in the incidental replacement parts teaching data database 92 .
  • each cell of "device type, device model number, symptom, error code, probable cause of failure, presumed replacement part" shaded is specified as shown in FIG. It is the same data as "device type, device model number, symptom, error code, probable cause of failure, probable replacement part" in the rule database.
  • the incidental replacement part inference model learning unit 94 causes the incidental replacement part inference model to learn based on the incidental replacement part teaching data stored in the incidental replacement part teaching data database 92 to perform learned incidental replacement part inference.
  • a model is generated and stored in the incidental replacement part inference model database 95 .
  • the equipment structure information database is a database that associates and stores equipment attribute information such as equipment type and equipment model number with drawing information and inspection manual information for each equipment.
  • the incidental replacement part teaching data generation unit 91, the incidental replacement part inference model learning unit 94, and the display/input unit 14a of the equipment maintenance support device 900 operate on the hardware and CPU 151 of the general-purpose computer 150 shown in FIG. It is realized by the cooperative operation of the programs that The incidental replacement part teaching data database 92, the equipment structure information database 93, and the incidental replacement part inference model database 95 are realized by storing data in the HDD 154 of the general-purpose computer 150 shown in FIG.
  • the incidental replacement parts teaching data generation unit 91 acquires inspection report analysis data from the inspection report analysis database 15 shown in FIG. Columns of equipment operation information including symptom, error code and occurrence status, inspection information including inspection result, and estimated maintenance information including estimated failure cause and estimated replacement parts are extracted. Then, the extracted information is displayed on the display/input unit 14a. Based on these displayed information sets, the expert 82 inputs incidental replacement parts to be incidentally replaced when executing the replacement of the estimated replacement parts from the display/input unit 14a.
  • incidental replacement parts are parts or miscellaneous parts that are incidentally replaced when replacing an estimated replacement part.
  • Incidental replacement parts are parts that may or may not need to be replaced.
  • the incidental replacement part training data generation unit 91 associates each extracted information with the incidental replacement part, and provides "device type, device model number, symptom, error code, occurrence status, inspection result, presumed failure cause, presumed replacement part
  • An incidental replacement part teaching database is generated in which "part” is input and “incidental replacement part” is output, and is stored in the incidental replacement part teaching data database 92.
  • the incidental replacement part inference model learning unit 94 reads incidental replacement part teaching data from the incidental replacement part teaching data database 92 shown in FIG. Data such as equipment inspection manual information, drawing information, etc. are acquired, and an incidental replacement part inference model is learned based on these data to obtain a learned incidental replacement part inference model.
  • the incidental replacement part inference model learning unit 94 After completing the learning of the incidental replacement part inference model, the incidental replacement part inference model learning unit 94 advances to step S603 in FIG. It is stored in the parts inference model database 95 .
  • FIG. 32 Ancillary replacement part inference operations will now be described with reference to FIGS.
  • FIG. 32 the same steps as those for identifying the cause of failure and replacement parts of the equipment maintenance support device 100 described above with reference to FIG.
  • the calculation unit 31 reads the inspection report data from the inspection report database 11, selects a specific rule, and if the priority rank of the selected specific rule is equal to or higher than a predetermined rank.
  • step S701 of FIG. 32 the cause of the failure and replacement parts are identified based on the selected identification rule, and the process proceeds to step S702 of FIG.
  • the calculation unit 31 reads the learned incidental replacement part inference model from the incidental replacement part inference model database 95 in step S702 of FIG. 32, and proceeds to step S703 of FIG.
  • step S703 of FIG. 32 the calculation unit 31 reads "device type, device model number, symptom, error code, occurrence situation, inspection result" of the inspection report data read out in step S301 of FIG.
  • the "cause of failure, replacement part” identified by the selected specific rule is input to the learned incidental replacement part estimation model.
  • each cell of "equipment type, equipment model number, symptom, error code, probable cause of failure, probable replacement part” shaded in the incidental replacement parts teaching data shown in FIG. 6 is the same data as "equipment type, equipment model number, symptom, error code, presumed cause of failure, presumed replacement part" of the specific rule database shown in FIG.
  • the cause of failure and the replacement part specified using the specification rule are the same as the estimated cause of failure and the estimated replacement part of the incidental replacement part teaching data. Therefore, the input data of the incidental replacement part teaching data shown in FIG. The input data has the same structure as Therefore, the "device type, device model number, symptom, error code, occurrence situation, inspection result" of the inspection report data read in step S301 of FIG. By inputting "replacement part” into the learned incidental replacement part estimation model, "incidental replacement part" can be output.
  • step S703 of FIG. 32 After outputting the incidental replacement parts in step S703 of FIG. 32, the computing unit 31 proceeds to step S704 of FIG. 32 to output the failure cause, the replacement part and the learned incidental replacement part inference model identified by the identification rule. The resulting additional replacement parts are displayed on the failure cause and replacement part display section 32 .
  • the calculation unit 31 proceeds to steps S306 and S307 of FIG. Then, the inference model is used to output the cause of failure and replacement parts, which are displayed on the display unit 32 for the cause of failure and replacement parts.
  • the equipment maintenance support device 900 described above combines the identification of the failure cause by the identification rule and the estimation of the incidental replacement part by the learned incidental replacement part inference model. It is possible to effectively estimate and display replacement parts, and improve the work and preparation efficiency of local workers.
  • the equipment maintenance support devices 100, 200, 300, 400, 500, 600, 700, 800, and 900 are provided with functional blocks realized by the hardware of the general-purpose computer 150, and the functional blocks operate.
  • the apparatus may be configured with other functional blocks as long as the following maintenance support method can be executed.
  • the maintenance support method of the embodiment is a maintenance support method of presenting the cause of the failure and replacement parts based on the data of the inspection report 71 created by the local worker 81 when the equipment installed in the building breaks down, Data of device attribute information and device main operation information are read from the inspection report 71, and the device attribute information, device main operation information, presumed maintenance information including the presumed cause of failure and presumed replacement parts, and the level of specific accuracy are obtained.
  • a plurality of specific rules associated with a priority rank indicating is selected, a specific rule in which the combination of the device attribute information read from the inspection report 71 and the device main operation information matches is selected, and the priority of the specific rule It is characterized by specifying and outputting the cause of the failure and the replacement part using the specified rule selected when the rank is equal to or higher than a predetermined rank.
  • the device attribute information, the device operation information including the device main operation information, and the inspection report data including the inspection information are read from the inspection report 71, and the read device attribute information and the device operation information are read. and inspection information, the probable failure cause, the presumed replacement part, and the priority rank input by the expert 82 are combined to generate inspection report analysis data, and the equipment attribute information in the generated inspection report analysis data and , equipment main operation information, presumed maintenance information including presumed failure cause and presumed replacement parts, and priority rank may be extracted, and a specific rule associated with them may be generated.
  • the maintenance report data including the actual failure cause and the actual replacement parts is read from the maintenance report 72 created by the field worker 81, and the actual replacement parts in the maintenance report data and the specification rule are specified.
  • the priority rank of the specific rule may be updated based on the comparison with the estimated replacement part.
  • the identification accuracy of the identification rule is defined by the accuracy rate
  • the identification rule is composed of equipment main operation information, estimated maintenance information including estimated failure causes and estimated replacement parts, priority rank and , associated with specific rule information including the number of correct answers, the number of incorrect answers, and the correct answer rate
  • a specific rule that matches the combination of equipment attribute information and equipment main operating information in the maintenance report data is the update target identification rule
  • the actual replacement parts in the maintenance report data is the update target identification rule.
  • the number of correct answers is incremented by 1, and if they do not match, the number of incorrect answers is incremented by 1 to calculate the accuracy rate of the update target identification rule, and the correct answer rate is calculated.
  • the priority rank of the update target specific rule may be updated according to the rate.
  • the device attribute information read from the inspection report 71, the device operation information including the device main operation information, and the inspection information are input, and the estimated failure cause and the estimated replacement parts are output. If the priority rank of the specific rule selected by preparing the inference model is lower than the predetermined rank, or if the combination of the device attribute information read from the inspection report 71 and the device main operation information matches If there is no rule, the device attribute information read from the inspection report 71, the device operation information, and the inspection information may be input to the learned inference model, and the failure cause and replacement parts may be output.
  • the device attribute information, the device operation information including the device main operation information, and the inspection report data including the inspection information are read from the inspection report 71, and the device attribute information and the device operation information are read.
  • inspection information, the probable failure cause, the presumed replacement part, and the priority rank input by the expert 82 are combined to generate inspection report analysis data, and the equipment attribute information in the generated inspection report analysis data, Input equipment operation information and inspection information, generate training data outputting estimated maintenance information including the estimated failure cause and estimated replacement parts in the inspection report analysis data, and generate an inference model based on the generated training data.
  • the accuracy index of the learned inference model is the accuracy rate of the learned inference model
  • the device type, logical format, required memory capacity, and learned inference model of the learned inference model are The number of correct answers, the number of incorrect answers, and the accuracy rate when estimating replacement parts are stored in association with each other. increment the number of correct answers of the learned inference model by 1, and if they do not match, increment the number of incorrect answers of the learned inference model by 1 to calculate the accuracy rate of the learned inference model. good too.
  • the maintenance support method of the embodiment input the device attribute information read from the inspection report 71, the device operation information including the device main operation information, the inspection information, the failure cause and the replacement parts specified using the specification rule. and prepare a learned incidental replacement part inference model that outputs incidental replacement parts, and use equipment attribute information, equipment operation information, inspection information, and specific rules in the learned incidental replacement part inference model
  • the identified cause of failure and replacement parts may be input, and incidental replacement parts may be output.
  • the device attribute information, the device operation information including the device main operation information, and the inspection report data including the inspection information are read from the inspection report 71, and the read device attribute information and the device operation Based on the information and the inspection information, the estimated failure cause, the estimated replacement part, and the priority rank input by the expert 82 are combined to generate inspection report analysis data, and the equipment attribute information in the generated inspection report analysis data.
  • incidental replacement part training data with parts as an output, and train the incidental replacement part inference model based on the incidental replacement part teaching data to generate a trained incidental replacement part inference model.
  • the inspection manual information and the drawing information of the equipment may be referred to when making the incidental replacement part inference model learn.
  • inspection report data acquisition unit 11 inspection report database 12 inspection report analysis unit 13 display unit 14 input unit 14a display/input unit 15 inspection report analysis database 16 specific rule generation unit 17a, 17b specific rule database , 20 teacher data generation unit 21 teacher data database 22 inference model learning unit 23a, 23b inference model database 24a, 24b, 24c model selection unit 31 operation unit 32 failure cause and replacement parts display unit 41 maintenance report data acquisition unit, 42 maintenance report database, 43 priority update unit, 44 maintenance inspection report data generation unit, 45 maintenance inspection report database, 46 inference model accuracy determination unit, 51 past failure case extraction unit, 52 past failure case database, 71 inspection report, 72 maintenance report, 81 field worker, 82 skilled worker, 91 incidental replacement parts training data generator, 92 incidental replacement parts teaching data database, 93 device structure information database, 94 incidental replacement parts inference model Learning unit 95 Incidental replacement part inference model database 100, 200, 300, 400, 500, 600, 700, 800, 900 Equipment maintenance support device 150 General-purpose computer 151 CPU 151 CPU 152 ROM 153 RAM 154

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

L'invention concerne un dispositif d'assistance à l'entretien des appareils (100) permettant de présenter une cause de défaillance et une pièce de rechange sur la base des données d'un rapport d'inspection créé par un travailleur sur site (81) au moment de la défaillance d'un appareil comprenant une unité d'acquisition de données de rapport d'inspection qui lit les données d'un rapport d'inspection, une base de données de règles d'identification (17a) qui stocke une pluralité de règles d'identification associant chacune les données lues, une cause de défaillance estimée, une pièce de rechange estimée et un rang de priorité, et une unité arithmétique (31) qui sélectionne une règle d'identification et utilise la règle d'identification sélectionnée pour identifier et produire une cause de défaillance et une pièce de rechange. Le rang de priorité indique le niveau de précision de l'identification par une règle d'identification, et l'unité arithmétique (31) identifie une cause de défaillance et une pièce de rechange en utilisant la règle d'identification sélectionnée lorsque le rang de priorité est égal ou supérieur à un rang prédéterminé.
PCT/JP2021/043147 2021-11-25 2021-11-25 Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien WO2023095236A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2023546399A JP7466788B2 (ja) 2021-11-25 2021-11-25 機器保守支援装置及びその動作方法
PCT/JP2021/043147 WO2023095236A1 (fr) 2021-11-25 2021-11-25 Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2021/043147 WO2023095236A1 (fr) 2021-11-25 2021-11-25 Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien

Publications (1)

Publication Number Publication Date
WO2023095236A1 true WO2023095236A1 (fr) 2023-06-01

Family

ID=86539116

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2021/043147 WO2023095236A1 (fr) 2021-11-25 2021-11-25 Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien

Country Status (2)

Country Link
JP (1) JP7466788B2 (fr)
WO (1) WO2023095236A1 (fr)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH05242061A (ja) * 1992-01-09 1993-09-21 Nec Corp 診断システム
JPH07271589A (ja) * 1994-03-29 1995-10-20 Fuji Heavy Ind Ltd 故障診断装置
JP2001175327A (ja) * 1999-12-22 2001-06-29 Yaskawa Electric Corp 故障診断装置
JP2006350923A (ja) * 2005-06-20 2006-12-28 Ricoh Co Ltd 交換部品推定システム、交換部品推定方法および交換部品推定プログラム
JP2007293489A (ja) * 2006-04-24 2007-11-08 Mitsubishi Electric Corp 設備機器用故障診断装置および設備機器用故障診断方法
JP2018085589A (ja) * 2016-11-22 2018-05-31 パナソニックＩｐマネジメント株式会社 診断方法、診断装置及び表示装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4916017B2 (ja) * 2007-07-03 2012-04-11 鹿島建設株式会社 構造物の耐震性能評価プログラム
JP2012174125A (ja) * 2011-02-23 2012-09-10 Yamaguchi Univ 防災計画支援システムとそのプログラム
JP2019012316A (ja) * 2017-06-29 2019-01-24 株式会社日立プラントサービス 設備システムのリスク分析方法、及び装置
JP2019207521A (ja) * 2018-05-29 2019-12-05 株式会社日立製作所 臨床試験支援システム、臨床試験支援プログラム及び臨床試験支援方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH05242061A (ja) * 1992-01-09 1993-09-21 Nec Corp 診断システム
JPH07271589A (ja) * 1994-03-29 1995-10-20 Fuji Heavy Ind Ltd 故障診断装置
JP2001175327A (ja) * 1999-12-22 2001-06-29 Yaskawa Electric Corp 故障診断装置
JP2006350923A (ja) * 2005-06-20 2006-12-28 Ricoh Co Ltd 交換部品推定システム、交換部品推定方法および交換部品推定プログラム
JP2007293489A (ja) * 2006-04-24 2007-11-08 Mitsubishi Electric Corp 設備機器用故障診断装置および設備機器用故障診断方法
JP2018085589A (ja) * 2016-11-22 2018-05-31 パナソニックＩｐマネジメント株式会社 診断方法、診断装置及び表示装置

Also Published As

Publication number Publication date
JPWO2023095236A1 (fr) 2023-06-01
JP7466788B2 (ja) 2024-04-12

Similar Documents

Publication Publication Date Title
US8676553B2 (en) Apparatus abnormality diagnosis method and system
WO2020079860A1 (fr) Système et procédé de support de diagnostic de défaillance d'équipement
US8667332B2 (en) Methods for optimizing diagnostics reasoner models
US8676727B2 (en) Abnormality diagnosis filter generator
CN109858823B (zh) 主配网停电计划选择方法及装置
US20110106747A1 (en) Turbine life assessment and inspection system and methods
CN111624986A (zh) 基于案例库的故障诊断方法和系统
JP7292076B2 (ja) 劣化推定装置、学習装置、劣化推定方法、学習方法、劣化推定プログラム、および、学習プログラム
JP4314200B2 (ja) 保守支援システム、保守管理装置およびプログラム
JP2014085730A (ja) 機器の損傷解析支援システム及び損傷解析支援方法
JP2012058885A (ja) 故障原因診断方法及び故障原因診断装置
JP2014139774A (ja) プラント設備保全計画の提案支援システム
JP6851247B2 (ja) 運用計画立案装置、運用制御システム、および、運用計画立案方法
JP2013105221A (ja) 保守部品配送支援システム，保守部品配送支援装置および保守部品配送支援プログラム
JPS62276470A (ja) 機器・部品の寿命管理装置
US20210398087A1 (en) Planning device, planning method, and program
JP3054039B2 (ja) プラントの保全支援装置
WO2023095236A1 (fr) Dispositif d'assistance à l'entretien des appareils et procédé d'assistance à l'entretien
JP2020086495A (ja) 保全装置、及び、電気機器の保全方法
EP4287089A1 (fr) Appareil, procédé et programme
JP2022083474A (ja) 保全リコメンドシステム
JP2016177676A (ja) 診断装置、診断方法、診断システムおよび診断プログラム
JP2018032229A (ja) 設備点検順位設定装置及び設備点検順位設定方法
KR20050107589A (ko) 보수 정보 제공 시스템
CN117151677A (zh) 用于维护设备的装置、方法和非临时性计算机可读介质

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: 21965608

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2023546399

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE