JP2020009068A - Component presentation system - Google Patents

Abstract

To present an appropriate selection result of a component that has a probability of being exchanged when a failure in a machine device is addressed.SOLUTION: A component presentation system 100 presents a component that has a probability of being exchanged to address a failure in a machine device 10 including a plurality of components. The component presentation system includes a storage unit 20, a learning unit 30, and a candidate selection unit 40. The storage unit stores storage data including a failure having occurred in the past, and a component having been exchanged to address the failure. The learning unit leans a component that needs to be exchanged to address a failure in accordance with a part of the storage data stored in the storage unit. The candidate selection unit selects, for an inputted failure, a candidate of a component that needs to be exchanged to address the failure, by using a learning result obtained by the learning unit. The candidate selection unit selects a candidate by using frequency data which is obtained from the storage data with no use of the learning result and which is about the frequency of exchanging components.SELECTED DRAWING: Figure 7

Description

  A component presentation system that presents components that may be replaced to deal with a failure that has occurred in a mechanical device

  Conventionally, as disclosed in Patent Document 1 (Japanese Patent Application Laid-Open No. 2017-33526), a system that uses a machine learning method to cope with a malfunction of a mechanical device has been used. As an example of such a system, conventionally, when a failure occurs in a mechanical device having a plurality of components, a component that is likely to be replaced for handling the failure is selected using a machine learning method. Systems are known.

  However, when a part that can be replaced when a problem is dealt with is selected using only the machine learning method, the work of actually replacing the part is performed due to the influence of input data for learning and the like. May output a selection result that is uncomfortable for the user.

  When a malfunction occurs in a mechanical device having a plurality of components, the component presentation system according to the first aspect presents one or a plurality of components that may be replaced to deal with the malfunction. The component presentation system includes a storage unit, a learning unit, and a candidate selection unit. The accumulating unit accumulates accumulated data including a defect that has occurred in the past and a component that has been replaced in response to the defect. The learning unit learns a part that needs to be replaced in order to deal with a failure according to at least a part of the stored data stored by the storage unit. The candidate selection unit selects a candidate for a component that needs to be replaced in order to deal with the defect using the learning result of the learning unit for the input defect. The candidate selection unit selects a candidate by further using frequency data, which is data obtained from accumulated data without using a learning result and is related to a frequency at which parts are replaced, is used.

  The component presentation system according to the first aspect can present an appropriate selection result of a component that is likely to be replaced when dealing with a malfunction of a mechanical device.

  The component presentation system according to the second aspect presents, when a malfunction occurs in a mechanical device having a plurality of components, one or a plurality of components that may be replaced to deal with the malfunction. The component presentation system includes a storage unit, a learning unit, and a candidate selection unit. The accumulating unit accumulates accumulated data including a defect that has occurred in the past and a component that has been replaced in response to the defect. The learning unit learns a part that needs to be replaced in order to deal with a failure according to at least a part of the stored data stored by the storage unit. The candidate selection unit selects a candidate for a component that needs to be replaced in order to deal with the defect using the learning result of the learning unit for the input defect. The learning unit learns, from the accumulated data, parts that need to be replaced in order to deal with a defect, in accordance with data obtained by removing removal data related to parts that are determined to need to be replaced.

  The component presentation system according to the second aspect can present an appropriate selection result of a component that may be replaced when dealing with a malfunction of a mechanical device.

  A component presentation system according to a third aspect is the system according to the first aspect, wherein the frequency data is data relating to a combination of a plurality of components that are simultaneously exchanged when dealing with a defect.

  The component presentation system according to a fourth aspect is the system according to the first aspect or the third aspect, wherein the candidate selection unit uses a first candidate that is a component candidate obtained using the learning result and frequency data. The obtained part is compared with a second candidate. If the part included in the second candidate is not included in the first candidate, the candidate selection unit selects the second candidate as a candidate for a part that needs to be replaced for handling a defect. . When the part included in the second candidate is included in the first candidate, the candidate selecting unit may use the first candidate or a combination of the first candidate and the second candidate to deal with the defect. As a candidate for parts that need to be replaced.

  The component presentation system according to a fifth aspect is the system according to the second aspect, wherein the removal data is based on data on a component specified from an abnormal code output by the mechanical device and information on a type and a defect of the mechanical device. The data includes at least one of data regarding the part to be determined.

  The component presentation system according to a sixth aspect is the system according to the second aspect or the fifth aspect, wherein the candidate selection unit is a component obtained using the learning result and needs to be replaced for dealing with a defect. A certain part and a part included in the removal data are selected as candidates.

  A parts presentation system according to a seventh aspect is the system according to any one of the first to sixth aspects, wherein the learning result is shared by a plurality of mechanical devices.

  The component presentation system according to the seventh aspect can be used as a system common to a plurality of mechanical devices having different models.

  The component presentation system according to an eighth aspect is the system according to any one of the first aspect to the seventh aspect, wherein the stored data includes a machine identification number, a machine type, a component name, a malfunction identification number, and a malfunction identification number. At least one of the following: the time of occurrence of the machine, the installation location of the machine, the delivery of the machine, and the person in charge of the machine.

  The component presentation system according to a ninth aspect is the system according to any one of the first to eighth aspects, wherein the mechanical device is an air conditioner, an air purifier, a heat pump hot water supply device, or a refrigeration device, and the component is: , Printed circuit board, display board, thermistor, electric motor for fan, remote control assembly, electric valve coil, heat exchanger, drain pump, mixing valve and humidity sensor.

FIG. 2 is a block diagram of the air conditioner 10. 4 is a flowchart of a process of repairing the air-conditioning apparatus 10. FIG. 1 is a block diagram of a component presentation system 100. It is an example of accumulated data. 6 is a flowchart of a learning process of a learning unit 30. It is an example of frequency data. FIG. 3 is a diagram for explaining a flow of data processed by the component presentation system 100. FIG. 9 is a diagram for explaining a method of creating a list of replacement target parts P1, P2,... Which are second candidates. It is a table | surface for explaining a matching process. It is an example of the display of the final candidate of the parts to be replaced P1, P2,. FIG. 14 is a diagram for describing a flow of data processed by a component presentation system 200 according to Modification A.

(1) Overall Configuration When a malfunction occurs in a mechanical device having a plurality of components, the component presentation system 100 presents one or a plurality of components that may be replaced in order to deal with the malfunction. For example, the component presentation system 100 presents to a service engineer who is in charge of repairing a mechanical device in which a failure has occurred, a candidate for a component that may be replaced when the mechanical device is repaired. In the following, a component that constitutes a mechanical device and that may be replaced at the time of repair for dealing with a malfunction that has occurred in the mechanical device is referred to as a replacement target component. The service engineer of the mechanical device replaces the component of the mechanical device based on the candidate for the replacement target component presented by the component presentation system 100, and repairs the mechanical device. Note that the failure of the mechanical device means a state in which a component of the mechanical device cannot function normally due to damage or wear, and a state in which a specific function of the mechanical device cannot be performed as intended. The repair of a mechanical device means an operation performed for dealing with a defect of the mechanical device, and includes an operation for completely resolving the defect as well as an operation for partially resolving the defect.

  The mechanical device is not limited to a specific device. Hereinafter, the mechanical device is assumed to be an air conditioner 10 including a refrigerant circuit. FIG. 1 is a block diagram of the air conditioner 10. The replacement parts P1, P2,... Of the air conditioner 10 are, for example, a printed circuit board, a display board, an inverter board, a remote control assembly, a capability setting adapter, a thermistor, a DC motor for a fan, a stepping motor, a float switch, and a cloth. Connection piping kit, electric valve coil, current fuse, scroll compressor, heat exchanger, crankcase heater, radiation fin, drain pump, air filter, flare nut, outdoor unit bottom frame, outdoor unit side plate, heat insulating tape, mixing valve , Anti-vibration rubber, compressor soundproofing material, pressure sensor and humidity sensor. However, the replacement target parts P1, P2,... Are not limited to those listed here. The parts P1, P2,... That are likely to be replaced when the air conditioner 10 is repaired include printed circuit boards, display boards, thermistors, and fans as shown in FIG. At least one of a motor, a remote control assembly, a motorized valve coil, a heat exchanger, a drain pump, a mixing valve, and a humidity sensor.

  Next, a process in which the service engineer of the air-conditioning apparatus 10 replaces the replacement target parts P1, P2,... With new parts for repairing the air-conditioning apparatus 10 in which a problem has occurred will be described. FIG. 2 is a flowchart of a process of repairing the air conditioner 10. In step S1, the user of the air-conditioning apparatus 10 in which a problem has occurred requests the service station of the manufacturer of the air-conditioning apparatus 10 to repair the air-conditioning apparatus 10. In step S2, the person in charge of the service station obtains information necessary for repair of the air conditioner 10 from the user who has requested the repair in step S1, and inputs the information to the component presentation system 100. The information necessary for repair of the air conditioner 10 is, for example, the model of the air conditioner 10 and an abnormality code output by the air conditioner 10 when a failure occurs. In step S3, the component presentation system 100 performs the replacement target component that needs to be replaced for repair of the air conditioner 10 based on the information input in step S2 and the information regarding the repair performed in the past. The candidates P1, P2,... Are selected and presented. In step S4, a new replacement part is arranged based on the candidates for the replacement target parts P1, P2,... Presented in step S3. In step S5, the service engineer in charge of repair of the air conditioner 10 goes to the installation location of the air conditioner 10 to be repaired, and replaces the parts P1, P2,. The air conditioner 10 is repaired by replacing it with a new replacement part arranged in step S4. In step S6, the service engineer inputs information relating to the repair of the air-conditioning apparatus 10 performed in step S5 to the component presentation system 100.

  The component presentation system 100 accumulates information related to the repair of the air-conditioning apparatus 10 performed in the past, analyzes and processes the accumulated data, and then learns the accumulated data using a machine learning technique. The component presentation system 100 includes candidates for the replacement target components P1, P2,... Obtained from the learned result, and the replacement target components P1, P2 obtained based on the information input in step S2 in FIG. .. Can be presented to the service engineer based on the candidates of P2,.

(2) Detailed Configuration FIG. 3 is a block diagram of the component presentation system 100. The component presentation system 100 is a program stored in, for example, a server installed in a service station or a cloud server. The component presentation system 100 mainly includes a storage unit 20, a learning unit 30, a candidate selection unit 40, a storage unit 50, an input unit 60, and a presentation unit 70.

(2-1) Accumulating Unit The accumulating unit 20 accumulates data on the faults that have occurred in the air conditioner 10 in the past and the replacement target parts P1, P2,... I do. Hereinafter, data stored by the storage unit 20 is referred to as stored data. The stored data is data relating to repairs performed in the past of the air conditioner 10.

  The stored data includes, for example, the identification number of the air conditioner 10, the type of the air conditioner 10, the names of the replacement parts P1, P2,..., The identification number of the defect, the time when the defect occurred, and the installation of the air conditioner 10. It consists of a place, a delivery time of the air conditioner 10, and a person in charge of the air conditioner 10. The type of the air conditioner 10 is information indicating the model of the air conditioner 10. The fault identification number is, for example, a fault code output by the air-conditioning apparatus 10 when a fault occurs. The air conditioner 10 displays and outputs an abnormal code on a display of a remote controller, for example. The person in charge of the air conditioner 10 is a service engineer in charge of repairing the air conditioner 10 in the past. In addition, among the data listed above, the time of occurrence of a defect, the installation location of the air conditioner 10, the delivery time of the air conditioner 10, and the person in charge of the air conditioner 10 are not included in the accumulated data. You may. FIG. 4 is an example of stored data. The stored data is stored and stored in the storage unit 50 described later by the storage unit 20 for each past repair case of the air-conditioning apparatus 10. For example, the accumulated data is accumulated every time the service engineer inputs information regarding the repair of the air conditioner 10 in step S6 of FIG. The accumulated data shown in FIG. 4 also includes the identification number of the repair case.

(2-2) Learning Unit The learning unit 30 is a replacement target component P1, P2, which is a component that needs to be replaced in order to deal with a malfunction of the air-conditioning apparatus 10 according to the accumulated data accumulated by the accumulation unit 20. Learn ... Specifically, the learning unit 30 analyzes and processes the accumulated data to acquire a training data set, and uses the machine learning method to replace the replacement target parts P1, P2,... In accordance with the training data set. learn. The training data set is data in which state variables and determination data are associated with each other. State variables are input values of machine learning. The state variables include at least the identification number of the air conditioner 10, the type of the air conditioner 10, and the identification number of the malfunction. The judgment data is an output value of machine learning. The determination data is the names of the parts P1, P2,... To be replaced.

  Next, an example of the learning process of the learning unit 30 will be described. FIG. 5 is a flowchart of the learning process of the learning unit 30. In step S11, the learning unit 30 acquires a state variable based on the accumulation data accumulated by the accumulation unit 20. In step S12, the learning unit 30 acquires the determination data based on the accumulated data accumulated by the accumulation unit 20. In step S13, the learning unit 30 creates a training data set based on a combination of the state variable acquired in step S11 and the determination data acquired in step S12. In step S14, the replacement target parts P1, P2,... Are learned according to the training data set created in step S13. The processes of steps S11 to S14 are repeated until the learning unit 30 has sufficiently learned the replacement target parts P1, P2,.

  The learning unit 30 learns replacement target parts P1, P2,... According to, for example, a neural network model. A neural network is a mathematical model generally used in so-called supervised learning, which learns the relationship between state variables and determination data according to a training data set. Data relating to the state variables of the training data set is input to the input layer of the neural network. The learning unit 30 adjusts parameters such as weights and biases so that data output from the output layer of the neural network and determination data associated with the state variables input to the input layer match as much as possible.

  When the learning unit 30 performs supervised learning, in order to improve the accuracy of the learning result of the learning unit 30, it is preferable that the number of training data sets is large. Therefore, it is preferable that the amount of data stored by the storage unit 20 is larger. Here, the learning result is a neural network model obtained as a result of learning by the learning unit 30. In general, machine learning using a neural network tends to require more computational resources for learning as the number of state variables and the number of hidden layers increase. For this reason, the learning unit 30 uses the computational resources of the GPU for the calculation of the machine learning using, for example, the technology of the GPGPU, in addition to the general-purpose CPU such as the server, thereby increasing the speed of the learning process. May be planned. In this case, the GPU may be attached to a server or the like, or may be attached to a computer on a cloud.

(2-3) Candidate Selection Unit The candidate selection unit 40 selects candidates for the parts to be replaced P1, P2,... Using both the data obtained using the learning result and the frequency data. The data obtained by using the learning result means that the learning unit 30 that has sufficiently learned the replacement target parts P1, P2,... Has data about the defect input in step S2 of FIG. , Data output by the learning unit 30. The frequency data is data obtained from the accumulated data without using the learning result of the learning unit 30. Specifically, the frequency data is data on the frequency with which parts of the air conditioner 10 are replaced. More specifically, the frequency data is data relating to a combination of a plurality of parts that are simultaneously replaced when dealing with a malfunction of the air conditioner 10. The frequency data is stored in advance in a storage unit 50 described later for each component. FIG. 6 is an example of frequency data. In FIG. 6, the numerical value of “frequency” is set for each component that may be replaced at the same time, and represents the probability of being used at the same time at the time of repair. FIG. 6 shows an example of frequency data of the motor-operated valve coil. For example, when the motor-operated valve coils, which are the parts to be replaced P1, P2,. is there. Such frequency data is obtained from information on past repairs of the air conditioner 10. Specifically, frequency data is acquired by totalizing the number of appearances of a combination of parts used simultaneously in the past based on accumulated data for each repair case performed in the past. The greater the number of past repair cases to be counted, the more frequency data based on actual results can be obtained. Here, the actual result is, specifically, information on the handling (repair) of a defect of the air-conditioning apparatus 10 performed in the past. Further, the candidate selection unit 40 performs a matching process described later in order to select appropriate candidates for the replacement target parts P1, P2,.

(2-4) Storage Unit The storage unit 50 stores programs and data used for processing executed by the component presentation system 100 in a storage device such as a hard disk drive of a server. Specifically, the storage unit 50 causes the storage device to store the accumulated data accumulated by the accumulation unit 20 and the frequency data used by the candidate selection unit 40.

(2-5) Input Unit The input unit 60 receives input of data from the outside. Specifically, the input unit 60 accepts the input when the service station staff inputs information necessary for repair of the air conditioner 10 in step S2 in FIG. Further, the input unit 60 accepts the input when the service engineer inputs information relating to the repair of the air conditioner 10 performed by the service engineer in step S6 of FIG. Input of data is performed via input devices such as a keyboard, a mouse, and a touch panel.

(2-6) Presentation Unit The presentation unit 70 outputs predetermined data. Specifically, the presentation unit 70 outputs the candidates for the replacement target components P1, P2,... Selected by the candidate selection unit 40. Thereby, the presentation unit 70 presents candidates for the replacement target parts P1, P2,... To the service engineer. The output of data is performed via an output device such as a display.

(3) Operation Next, the operation of the component presentation system 100 will be described. FIG. 7 is a diagram for explaining the flow of data processed by the component presentation system 100. In FIG. 7, points at which data is processed are indicated by rectangular blocks B1 to B9. Specifically, FIG. 7 shows the service guide input block B1, data processing block B2, output filter block B3, business screen block B4, input filter block B5, data storage block B6, processing filter block B7, AI (artificial intelligence) 3) A learning block B8 and an AI (artificial intelligence) operation block B9 are shown. In FIG. 7, arrows indicate the flow of data.

  In the data storage block B6, stored data and frequency data are stored. The accumulated data and the frequency data are inputted as needed, and are accumulated or updated. The accumulated data is accumulated every time the service engineer inputs information relating to the repair of the air conditioner 10 in step S6 of FIG. The frequency data is updated periodically based on the accumulated data.

  The stored data stored in the data storage block B6 is analyzed and processed in the processing filter block B7. Specifically, in the processing filter block B7, the accumulated data is converted into a training data set that can be used as input data for machine learning. In the AI learning block B8, learning of the replacement target parts P1, P2,... Is performed using the training data set output from the processing filter block B7. Specifically, in the AI learning block B8, learning using a neural network model is performed. The neural network model obtained as a result of the learning is copied from the AI learning block B8 to the AI operation block B9. In the AI operation block B9, candidates for the parts to be replaced P1, P2,... Are obtained using the data passed from the input filter block B5 and the copied neural network model obtained as a result of the learning. . These candidates are candidates for replacement target parts P1, P2,... Obtained by using a machine learning method. Hereinafter, this candidate is referred to as a “first candidate”. The first candidate is data obtained from prediction by AI (artificial intelligence). The first candidate is a list of components arranged from a component having a higher priority to a component having a lower priority when a plurality of components are included. The higher the priority of a part, the higher the likelihood that the part will actually be used for repair.

  On the other hand, the frequency data stored in the data storage block B6 is processed in the data processing block B2 together with the input data from the service guide input block B1. Thus, in the data processing block B2, candidates for the replacement target parts P1, P2,... Are acquired. The candidates are replacement target parts P1, P2,... Obtained without using the machine learning method. Hereinafter, this candidate is referred to as a “second candidate”. The second candidate is data obtained from actual results. The second candidate is a list of components arranged from a component having a higher priority to a component having a lower priority when a plurality of components are included. The higher the priority of a part, the higher the likelihood that the part will actually be used for repair.

  A specific method for acquiring the second candidate will be described. In step S2 of FIG. 2, the person in charge of the service station inputs information necessary for repair of the air conditioner 10. The input information corresponds to the input data from the service guide input block B1, and includes, for example, information on the model of the air conditioner 10 to be repaired, an abnormality code output by the air conditioner 10, and the air conditioner. It is a classification item of 10 symptoms. The symptom classification item is data that specifically describes the symptom of the air conditioner 10, and is, for example, a character string “No drainage from the outdoor unit”. If an abnormal code can be input, there is no need to input the classification item of the symptom.

  FIG. 8 is a view for explaining a method in which the data processing block B2 creates a list of replacement target parts P1, P2,... Included in the second candidate based on input data from the service guide input block B1. FIG. In FIG. 8, in the column of “input conditions”, identification numbers of conditions that can be input from the service guide input block B1 are listed. In the column of “case”, identification numbers of all past repair cases of the air conditioner 10 are listed. In the column of "used parts", the identification numbers of all parts actually used for dealing with a defect in a past repair case are listed. In FIG. 8, one or a plurality of input conditions and a case satisfying the input condition are connected by a line, and each case and one or a plurality of parts actually used in the case are connected by a line. It is connected by.

  Next, the processing performed by the data processing block B2 will be specifically described with reference to FIG. As shown by the thick line frame in FIG. 8, the condition “AC01” is input as the model of the air conditioner 10 from the service guide input block B1 as the “input condition”, and the abnormality output by the air conditioner 10 is performed. It is assumed that the condition “ER013” is input as a code. As a result, the case “CS0011” and the case “CS0012” are selected as cases that satisfy both the input conditions “AC01” and “ER013”. Then, the parts “PT00101” and “PT00105” are selected as the parts used in the cases “CS0011” and “CS0012”. Accordingly, the data processing block B2 creates a list including the components “PT00101” and “PT00105” based on the input from the service guide input block B1. As described above, the parts included in the list created based on the input from the service guide input block B1 are “important replacement candidate parts” determined to be highly likely to be replaced based on the results.

  Then, the data processing block B2 sorts the important replacement candidate parts in descending order of priority based on the frequency data stored in the data storage block B6. Further, the data processing block B2 is a part that may be replaced due to other factors such as aging after the sorted list of important replacement candidate parts, if necessary. "List. The priority of the non-critical replacement candidate parts is lower than the priority of the important replacement candidate parts. The list of non-critical replacement candidate parts is also rearranged in descending order of priority based on the results. With the above processing, the data processing block B2 acquires the second candidate including the important replacement candidate part and the non-critical replacement candidate part.

  Next, in the output filter block B3, matching processing between the first candidate and the second candidate is performed. The candidate selection unit 40 performs the matching process. In the matching process, first, the first candidate and the second candidate are compared. As a result, when the part included in the second candidate is not included in the first candidate, the candidate selecting unit 40 determines that the final candidate of the replacement target parts P1, P2,. Select two candidates. If the part included in the second candidate is included in the first candidate, the candidate selecting unit 40 determines that the first candidate as the final candidate for the replacement target parts P1, P2,. A candidate or a combination of the first candidate and the second candidate is selected. The maximum value of the number of replacement target parts P1, P2,... Included in the final candidate is, for example, five. That is, the parts brought by the service engineer for the on-site repair of the air conditioner 10 are a maximum of five types.

  Next, a specific example of the matching process between the first candidate and the second candidate will be described. FIG. 9 is a table for explaining the matching process. It is assumed that the first candidate and the second candidate each include at least five parts and are arranged in descending order of priority. The table of FIG. 9 shows “condition”, “policy”, “processing”, and “arrangement” of pattern 1, pattern 2, pattern 3-1 and pattern 3-2 as specific methods of the matching processing. Have been. The “condition” relates to a condition that must be satisfied by the first candidate and the second candidate in order to perform the matching process. The “policy” relates to the outline and concept of the matching process. “Process” relates to a specific process of the matching process. The “arrangement method” relates to the arrangement order of components in the final candidates of the replacement target components P1, P2,. Next, each of the four patterns of the matching process will be specifically described.

  Pattern 1 is a pattern in which there is no common component between the component included in the first candidate and the component included in the second candidate. In this case, the candidate selection unit 40 gives priority to the second candidate. Specifically, the candidate selection unit 40 sets the top five parts included in the second candidate as the final candidates of the replacement target parts P1, P2,. Here, the “top five components” mean five components from the component having the highest priority to the component having the fifth highest priority when the components are rearranged in descending order of priority. In the final candidate obtained by the pattern 1, five parts are arranged in descending order of priority of the second candidate.

  Pattern 2 is a pattern in which at least one common part exists between the top two parts of the first candidate and the top two parts of the second candidate. Here, the “top two parts” mean two parts, a part having the highest priority and a part having the second highest priority when the parts are rearranged in descending order of priority. In this case, the candidate selection unit 40 gives priority to the first candidate. Specifically, the candidate selecting unit 40 first arranges the common parts (common parts) between the first candidate and the second candidate, and then arranges the common parts that are included in the first candidate and are common. Parts that are not parts (non-common parts) are arranged as final candidates for the parts P1, P2,... To be replaced. In the final candidate obtained by pattern 2, when there are a plurality of (two) common parts, they are arranged in descending order of the priority of the first candidate. In the final candidate obtained by the pattern 2, the non-common parts are arranged in descending order of the priority of the first candidate.

  The pattern 3-1 and the pattern 3-2 are patterns other than the pattern 1 and the pattern 2. In these patterns, the candidate selection unit 40 sets the combination of the first candidate and the second candidate as the final candidate of the replacement target parts P1, P2,.

  The pattern 3-1 is a pattern in which at least one of the important replacement candidate parts included in the second candidate is included in the first candidate. In this case, the candidate selecting unit 40 gives priority to the first candidate and corrects the data of the first candidate with the data of the second candidate. Specifically, the candidate selection unit 40 first arranges the important replacement candidate parts included in the second candidate and the parts (first-type parts) included in the first candidate, and then arranges the first candidate And the parts (type 2 parts) that are not the type 1 parts and are not included in the second candidate and are not the important replacement candidate parts (the type 3 .. Are arranged as final candidates for replacement target parts P1, P2,. In the final candidates obtained by the pattern 3-1, the first type components and the second type components are arranged in descending order of the priority of the first candidate, and the third type components are arranged in descending order of the values based on the results. Have been. Here, the value based on the performance is, for example, a value based on the number of times used in a past repair case.

  The pattern 3-2 is a pattern in which the important replacement candidate part included in the second candidate is not included in the first candidate. In this case, the candidate selection unit 40 adds the first candidate data to the second candidate data, giving priority to the second candidate. Specifically, the candidate selecting unit 40 first arranges the components (the fourth type components) included in the second candidate, and then sorts the components (the fifth type components) included in the first candidate. .. Are arranged as final candidates for the parts to be replaced P1, P2,. In the final candidates, the type 4 parts are arranged in descending order of the values based on the above-described results, and the type 5 parts are arranged in descending order of the priority of the first candidate.

  Therefore, in the matching process, the pattern 1 and the pattern 3-2 give priority to data based on the actual results, and the pattern 2 and the pattern 3-1 give priority to AI prediction.

  By the above-described matching processing by the output filter block B3, final candidates for the replacement target parts P1, P2,... Are obtained. The final candidates of the parts to be replaced P1, P2,... Are presented to the service engineer by the business screen block B4. Specifically, the final candidates of the parts to be replaced P1, P2,... Are displayed on the display of the computer operated by the service engineer. FIG. 10 shows an example of the display of the final candidates of the replacement parts P1, P2,. The service engineer arranges necessary parts and repairs the air conditioner 10 according to the information on the parts to be replaced P1, P2,... Presented by the business screen block B4. Information regarding the repair of the air conditioner 10 performed by the service engineer is stored as accumulated data in the data accumulation block B6.

  In the “selection condition” area at the top of the screen shown in FIG. 10, an interface for inputting information regarding the air conditioner 10 to be repaired is displayed. The “model” input box is an interface for inputting information about the model of the air conditioner 10. The “error code” input box is an interface for inputting an abnormal code output by the air conditioner 10. The “classification during symptom” input box is an interface for inputting a classification item of a symptom of the air conditioner 10. The service station person or service engineer can automatically obtain the final candidates of the parts to be replaced P1, P2,... By inputting necessary information to these interfaces.

  In the "candidate parts" area at the bottom of the screen shown in FIG. 10, information on the final candidates of the parts to be replaced P1, P2,... Is displayed. FIG. 10 shows an example of a table listing final candidates. The “part name” is the name of the parts P1, P2,... To be replaced. The “product number” is an identification number of the replacement part P1, P2,. "Usage rate" is the probability of being replaced in actual repairs. The “simultaneous replacement rate” is the probability of replacement at the same time as the part displayed at the top of the table. In the case of FIG. 10, for example, the probability that the scroll compressor is replaced with a new one when the air conditioner 10 is repaired is 80%, and when the scroll compressor is replaced, the scroll compressor is replaced with a new one. At the same time, the probability of exchanging the anti-vibration rubber is 80%. “Arrangement selection” is a check box for selecting a part that needs to be arranged for repair. When the “arrangement of parts” button is pressed, the parts checked in “arrangement selection” are arranged.

  Thus, in the business screen block B4, information necessary for repair of the air conditioner 10 is input, and a list of final candidates of the replacement target parts P1, P2,... Is output. The information input in the business screen block B4 is converted in the input filter block B5 into a format that can be used as input data of the AI operation block B9.

(4) Features (4-1)
Conventionally, when a failure occurs in a mechanical device such as the air conditioner 10 having a plurality of components, a component that is likely to be replaced to deal with the failure is selected using a machine learning method. A method is used. However, when a component that is likely to be replaced at the time of handling a defect is selected using only the machine learning method, the component selection result is easily affected by input data for learning. For this reason, a part that does not match the past results of handling the defect and a part that is uncomfortable for a person who actually replaces the part may be selected.

  The component presentation system 100 repairs the candidates for the replacement target components P1, P2,... Which may be replaced at the time of repair for dealing with a malfunction that has occurred in the air conditioner 10, by repairing the air conditioner 10. To the service engineer in charge. The component presentation system 100 uses the first candidate, which is a candidate for the replacement target component P1, P2,... Obtained by using the machine learning method, and data based on the results of past failure handling. The final candidate for the replacement parts P1, P2,... Is presented to the service engineer based on both the obtained second candidates for the replacement parts P1, P2,. The first candidate is a candidate obtained using a result of learning accumulated data that is accumulated every time a service engineer inputs information regarding repair of the air-conditioning apparatus 10. The second candidate is a candidate obtained without using the result of learning the accumulated data using the data obtained from the accumulated data.

  As described above, the component presentation system 100 selects the candidates for the replacement target components P1, P2,... Based on the second candidates based on the past results, in addition to the first candidates based on the AI prediction. In addition, it is possible to prevent the selection result from being excessively affected by the input data for learning. For this reason, the component presentation system 100 can obtain a selection result that is more suitable based on the results of handling failures that have been performed in the past, and a selection result that has little or no discomfort for a person who actually performs the work of replacing parts. it can.

  Therefore, the component presentation system 100 can present an appropriate selection result of components that may be replaced when dealing with a malfunction of the air conditioner 10. Accordingly, when the service engineer performs the on-site repair of the air conditioner 10, the air conditioner 10 has new parts that need to be replaced with the replacement target parts P1, P2,. Since it is not at hand, it is possible to suppress occurrence of a situation in which the participant returns to get the new part. Therefore, the service engineer of the air-conditioning apparatus 10 can complete the repair of the air-conditioning apparatus 10 in one visit by arranging the parts necessary for the repair based on the selection result of the parts presentation system 100. The maintenance work of the air-conditioning apparatus 10 can be performed efficiently.

(4-2)
The component presentation system 100 uses the second candidates obtained based on the data obtained from the accumulated data to select the final candidates of the replacement target components P1, P2,. Here, the data obtained from the accumulated data is, for example, frequency data relating to a combination of a plurality of parts that are simultaneously replaced when dealing with a malfunction of the air conditioner 10, as shown in FIG. The frequency data is data based on the performance, which is updated periodically based on the accumulated data. The component presentation system 100 corrects the first candidate based on the prediction of the AI using the second candidate obtained based on the frequency data, for example, so that the replacement target components P1, P2,.・ The final candidate can be selected.

(4-3)
The component presentation system 100 performs matching processing between the first candidate based on the prediction of the AI and the second candidate based on the actual result, and selects the final candidate of the replacement target components P1, P2,. The component presentation system 100 performs an appropriate matching process according to the number of components commonly included in the first candidate and the second candidate, the presence or absence of an important replacement candidate component included in the first candidate, and the like. The component presenting system 100 gives priority to the first candidate, gives priority to the second candidate, or obtains the first candidate when obtaining the final candidate of the replacement target component P1, P2,. Select whether to combine the candidate and the second candidate. Further, when combining the first candidate and the second candidate, the component presentation system 100 corrects the data of the second candidate with the data of the first candidate or converts the data of the first candidate with the data of the second candidate. Select whether to correct. Accordingly, the component presentation system 100 finally replaces the replacement target parts P1, P2,... That match the performance based on the first candidate based on the AI prediction and the second candidate based on the performance. Can be selected as a candidate.

(4-4)
The component presentation system 100 uses the first candidate based on the prediction of the AI and the second candidate based on the actual results to select the final candidate for the replacement target component P1, P2,. The second candidate includes a part based on information that is already known regarding a defect, such as an abnormality code of the air-conditioning apparatus 10. On the other hand, the first candidate may include a part that is not included in the second candidate. Parts that are not included in the second candidate and are included in the first candidate are parts that are not clear from the abnormality code of the air conditioner 10 or the like. It is a part that is predicted from know-how related to the case. The component presentation system 100 combines the first candidate and the second candidate to select the final candidate of the replacement target component P1, P2,. It is possible to select a component predicted from know-how relating to past repair cases. As a result, the component presentation system 100 can replace the replacement target components P1, P2,... Which are more suitable for past failures than in the case of selecting only component candidates based on the AI prediction. Selection results can be presented.

(5) Modification (5-1) Modification A
The component presentation system 100 selects the final candidate for the replacement target component P1, P2,... Based on the first candidate based on the AI prediction and the second candidate based on the actual result. Specifically, the component presentation system 100 performs a matching process between the first candidate and the second candidate, and selects a final candidate. In the matching process, by giving priority to one of the first candidate and the second candidate, or by correcting the data of one of the first candidate and the second candidate with the other data, only the candidate of the component based on the AI prediction is obtained. As compared with the case of selecting, a selection result that is more suitable based on the results of dealing with the failure performed in the past is generated.

  However, the component presentation system 100 may select the final candidates of the replacement target components P1, P2,... That match the past results of handling the failure without performing the matching process. Next, a component presentation system 200 which is a modification of the component presentation system 100 will be described. The component presentation system 200 selects the final candidates of the replacement target components P1, P2,... Without performing the matching process.

  Similar to the component presentation system 100, the component presentation system 200 mainly includes a storage unit 20, a learning unit 30, a candidate selection unit 40, a storage unit 50, an input unit 60, and a presentation unit 70. The basic configuration and operation of the component presentation system 200 are substantially the same as those of the component presentation system 100. The main difference between the component presentation system 200 and the component presentation system 100 is the function of the learning unit 30 and the candidate selection unit 40. Next, the operation of the component presentation system 200 will be described focusing on the functions of the learning unit 30 and the candidate selection unit 40.

  The learning unit 30 replaces the parts P1, P2,... That need to be replaced in order to deal with a malfunction of the air conditioner 10 in accordance with the data obtained by removing the removal data from the stored data stored in the storage unit 20.・ ・ Learning. The removal data includes at least one of data relating to a component specified from the abnormality code output by the air conditioner 10 and data relating to a component determined based on information relating to the type and trouble of the air conditioner 10. The information on the type and trouble of the air conditioner 10 is, for example, information on the model of the air conditioner 10 to be repaired, and classification items of symptoms of the air conditioner 10. The removal data is data based on the results of dealing with a defect performed in the past, and is data regarding a component that is determined in advance to need to be replaced when the air-conditioning apparatus 10 is repaired. Further, the data obtained by removing the removal data from the accumulated data is data relating to a part to be replaced based on a factor other than the result of dealing with a defect performed in the past. Factors other than the performance include, for example, aging. Data obtained by removing the removal data from the accumulated data is used as input data for learning by the learning unit 30.

  The candidate selection unit 40 replaces the replacement target parts P1, P2,... Obtained using the learning result of the learning unit 30 and the replacement target parts P1, P2,. .. Are selected as final candidates for the components P1, P2,. More specifically, the candidate selection unit 40 replaces the combination of the component obtained from the learning result of the learning unit 30 with the component specified from the abnormality code or the like output by the air-conditioning apparatus 10 with the replacement target components P1 and P2. ,... Are selected as final candidates.

  FIG. 11 is a diagram for explaining the flow of data processed by the component presentation system 200. In FIG. 11, similarly to FIG. 7, the service guide input block B1, data processing block B2, output filter block B3, business screen block B4, input filter block B5, data storage block B6, processing filter block B7, AI (artificial An intelligent) learning block B8 and an AI (artificial intelligence) operation block B9 are shown. The difference between FIG. 7 and FIG. 11 is mainly the function of the service guide input block B1, data processing block B2, output filter block B3, data storage block B6, and processing filter block B7. Hereinafter, the operation of the component presentation system 200 will be described focusing on this difference.

  In the data storage block B6, stored data is stored. The stored data is input as needed, and stored or updated. The accumulated data is accumulated every time the service engineer inputs information relating to the repair of the air conditioner 10 in step S6 of FIG. The accumulated data includes the above-described removal data.

  The stored data stored in the data storage block B6 is analyzed and processed in the processing filter block B7. Specifically, after the processing for removing the removal data from the accumulated data is performed in the processing filter block B7, the accumulated data that does not include the removed data is converted into a training data set that can be used as input data for machine learning. Is done. In the AI learning block B8, learning of the replacement target parts P1, P2,... Is performed using the training data set output from the processing filter block B7. The neural network model obtained as a result of the learning is copied from the AI learning block B8 to the AI operation block B9. In the AI operation block B9, candidates for the parts to be replaced P1, P2,... Are obtained using the data passed from the input filter block B5 and the copied neural network model obtained as a result of the learning. . This candidate is a list of replacement target parts P1, P2,... Acquired based on factors other than the results of handling the failures performed in the past.

  On the other hand, in the data processing block B2, based on the removal data removed from the accumulated data in the processing filter block B7 and the input data from the service guide input block B1, candidates for replacement target parts P1, P2,. Is obtained. The candidates are input data such as an abnormal code output by the air-conditioning apparatus 10 and a list of replacement target parts P1, P2,... .

  The output filter block B3 outputs final candidates for the replacement target parts P1, P2,... Based on the candidates output from the AI operation block B9 and the candidates output from the data processing block B2. The output filter block B3 outputs the final candidate by simply combining the two candidates without performing the matching process. The final candidates of the parts to be replaced P1, P2,... Are presented to the service engineer by the business screen block B4.

  In the present modified example, a process is performed in which the accumulated data accumulated in the accumulation unit 20 is divided in advance into data used as input data for learning of the learning unit 30 and removal data based on actual results. With this processing, the component presentation system 200 obtains candidates based on the prediction of the AI using a machine learning method for the factors other than the results of the handling of the failure performed in the past, and With respect to the factors related to the results of the above-mentioned measures, candidates based on the results can be acquired using the abnormality code or the like output by the air-conditioning apparatus 10. Therefore, the component presentation system 200 can present a selection result that is more suitable for past failure handling than the case where only component candidates are selected based on AI prediction.

(5-2) Modification B
In the component presentation system 100, the learning result of the learning unit 30 is used to select candidates for the replacement target components P1, P2,... Of the air conditioner 10. That is, the learning result of the learning unit 30 is used for one specific mechanical device. However, the learning result of the learning unit 30 may be shared by a plurality of mechanical devices. For example, a common training data set may be used among a plurality of mechanical devices used in a chemical plant or the like. When the learning unit 30 performs learning using a neural network model, a common neural network model may be used among a plurality of mechanical devices.

(5-3) Modification C
The component presentation system 100 includes a learning unit 30 that learns replacement target components P1, P2,... According to the accumulated data accumulated by the accumulation unit 20. In order to improve the accuracy of the learning result of the learning unit 30, the larger the number of training data sets, the better. However, as the number of training data sets increases, the calculation resources required for learning by the learning unit 30 tend to increase. Therefore, the learning unit 30 may perform learning using a virtual machine service instead of learning using the CPU and GPU attached to the server in which the component presentation system 100 is stored. In the learning using the virtual machine service, for example, a fee for using a computing resource capable of processing at least the training data set in an amount necessary for the learning by the learning unit 30 is paid to the virtual machine service provider, and the virtual machine service is provided. CPU and GPU are used for learning. Thereby, even if the amount of the training data set increases in the future, scalable response according to the amount of the training data set is possible without changing the configuration of the server in which the component presentation system 100 is stored. Becomes Therefore, the cost of the computational resources required for learning by the learning unit 30 can be suppressed.

(5-4) Modification D
As shown in FIG. 8, the component presentation system 100 is likely to be replaced based on the performance using information necessary for repair of the air conditioner 10 input by a service station staff or the like. The determined parts (important replacement candidate parts) are acquired. As shown in FIG. 10, the information necessary for repair of the air conditioner 10 includes information about the model of the air conditioner 10 to be repaired, an abnormality code output by the air conditioner 10, and the air conditioner. It is a classification item of 10 symptoms. When an abnormal code can be acquired, the component presentation system 100 can acquire an important replacement candidate component corresponding to the abnormal code by inputting the abnormal code. However, when the abnormal code cannot be acquired, the component presentation system 100 estimates the abnormal code from, for example, the classification item of the symptom of the air conditioner 10, and acquires the important replacement candidate part suitable for the estimated value of the abnormal code. You may.

(5-5) Modification E
The component presentation system 100 selects replacement target components P1, P2,... Which may be replaced in order to deal with a malfunction that has occurred in the air conditioner 10. However, the component presentation system 100 may select a component that is likely to be replaced for dealing with a malfunction that has occurred in a mechanical device other than the air conditioner 10. Examples of such a mechanical device include an air purifying device, a heat pump hot water supply device, and a refrigerating device such as a refrigerator.

(6) Conclusion While the embodiments of the present disclosure have been described above, it is understood that various changes in form and details can be made without departing from the spirit and scope of the present disclosure described in the claims. Will be.

  The component presentation system can be used to acquire information on components that may be replaced in response to a failure when a failure occurs in a mechanical device having a plurality of components.

10 Air conditioner (mechanical device)
Reference Signs List 20 accumulation unit 30 learning unit 40 candidate selection unit 100 parts presentation system

JP 2017-33526 A

Claims (9)

A component presentation system (100) for presenting one or a plurality of components that may be replaced in response to a failure when a failure occurs in a mechanical device (10) having a plurality of components. ,
A storage unit (20) for storing storage data including the failure that has occurred in the past and the component that has been replaced in response to the failure;
A learning unit (30) that learns the part that needs to be replaced in order to deal with the problem according to at least a part of the stored data stored by the storage unit;
A candidate selection unit (40) for selecting, based on the learning result of the learning unit, a candidate for the component that needs to be replaced in order to deal with the failure,
With
The candidate selection unit is a data obtained from the accumulated data without using the learning result, and further selects the candidate by further using frequency data regarding a frequency at which the part is replaced,
Parts presentation system. A component presentation system (100) for presenting one or a plurality of components that may be replaced in response to a failure when a failure occurs in a mechanical device (10) having a plurality of components. ,
A storage unit (20) for storing storage data including the failure that has occurred in the past and the component that has been replaced in response to the failure;
A learning unit (30) that learns the part that needs to be replaced in order to deal with the problem according to at least a part of the stored data stored by the storage unit;
A candidate selection unit (40) for selecting, based on the learning result of the learning unit, a candidate for the component that needs to be replaced in order to deal with the failure,
With
The learning unit learns, from the stored data, the part that needs to be replaced in order to cope with the defect, according to data obtained by removing removal data related to the part that has been determined to need to be replaced.
Parts presentation system. The frequency data is data relating to a combination of a plurality of the parts that are simultaneously replaced at the time of handling the problem.
The component presentation system according to claim 1. The candidate selection unit compares a first candidate that is a candidate for the component obtained using the learning result with a second candidate that is a candidate for the component obtained using the frequency data, If the part included in the second candidate is not included in the first candidate, the second candidate is selected as the candidate. If the part is included in the first candidate, the part is selected. A first candidate or a combination of the first candidate and the second candidate is selected as the candidate;
The component presentation system according to claim 1. The removal data includes data relating to the component specified from the abnormality code output by the mechanical device, and includes at least one of data relating to the component determined based on information on the type of the mechanical device and the defect.
The component presentation system according to claim 2. The candidate selection unit is configured to select the part obtained by using the learning result, the part that needs to be replaced in order to cope with the defect, and the part included in the removal data, Choose as
The component presentation system according to claim 2. The learning result is shared by a plurality of the mechanical devices,
The component presentation system according to claim 1. The accumulated data, the identification number of the mechanical device, the type of the mechanical device, the name of the component, the identification number of the defect, the time of occurrence of the defect, the installation location of the mechanical device, the delivery time of the mechanical device, and , Including at least one of the personnel of the machinery;
The component presentation system according to any one of claims 1 to 7. The mechanical device is an air conditioner, an air purifier, a heat pump hot water supply device or a refrigeration device,
The parts include at least one of a printed circuit board, a display board, a thermistor, a motor for a fan, a remote control assembly, a motorized valve coil, a heat exchanger, a drain pump, a mixing valve, and a humidity sensor.
The component presentation system according to claim 1.

Images