Learning device, inference device, disconnection position identification system, learning method, inference method, and program

Abstract

A learning device (30) that generates a trained model for inferring a disconnection position and a disconnection state for a sensor network (10) being diagnosed, the learning device comprising: a training data acquisition unit that acquires training data, including actually measured waveforms of signals flowing through a transmission path of the sensor network (10), scenario information for each of a plurality of disconnection scenarios where the resistance value of a resistor and the insertion position of the resistor are changed for a sensor network model which models the sensor network (10) and in which the resistor is inserted at a specific position of the transmission path, and simulated waveforms obtained by executing a simulation of each of the disconnection scenarios; and a model generation unit (32) that generates, through machine learning using the training data acquired by the training data acquisition unit, a trained model for inferring a disconnection position and a disconnection state of the sensor network (10) from an actually measured waveform.

Description

Learning device, inference device, disconnection location identification system, learning method, inference method, and program

[0001] The present disclosure relates to a learning device, an inference device, a disconnection location identification system, a learning method, an inference method, and a program.

[0002] In order to identify the location and state of a break in a transmission line including a cable, an electric wire, etc., physical inspections such as visual inspections, electrical measurements, etc. are carried out.

[0003] Physical inspection can quickly identify the location of a break if it is in a location that is easily visible or if the damage is obvious from the outside, but if this is not the case, it is necessary to check a wide range of transmission lines one by one, which is a significant burden of time and cost. Furthermore, electrical measurements only provide limited data, making it difficult to identify the exact location. Therefore, these methods take time to identify the location and state of the break, and there is a risk that the system will stop during that time.

[0004] Patent Document 1 discloses an abnormality inspection device that inspects a wire harness incorporated in an electromechanical device for abnormalities. This abnormality inspection device includes a reflected wave detector that detects reflected waves generated due to input waves transmitted to the wire harness, and determines whether the wire harness has an abnormality based on an abnormality determination model generated from feature quantities obtained from frequency characteristics of reflected waves when the wire harness has an abnormality and when it does not have an abnormality.

[0005] Patent Document 2 discloses a cable status management device that manages the progression of a cable breakage in a cable used as wiring for a device to be managed. This cable status management device estimates the progression of a cable breakage based on resistance value data, which is a time-series change in the resistance value of the cable when the cable is operated repeatedly and periodically.

[0006] JP 2022-120857 A JP 2023-044616 A

[0007] The devices described in Patent Documents 1 and 2 use only measured waveforms, such as reflected waves and time-series changes in resistance, as learning data. Because waveform data changes depending on the measurement environment, the devices described in Patent Documents 1 and 2 take a long time to learn. Therefore, there is room for improvement in terms of more efficiently identifying the location and state of a disconnection.

[0008] The present disclosure has been made in consideration of the above-mentioned problems, and aims to propose a learning device, an inference device, a disconnection location identification system, a learning method, an inference method, and a program that can more efficiently identify the disconnection location and disconnection state.

[0009] In order to achieve the above object, the learning device according to the present disclosure is a learning device that generates a trained model for inferring the location and state of a disconnection in a sensor network to be diagnosed, and includes a training data acquisition unit that acquires training data including the measured waveform of a signal flowing through a transmission path of the sensor network, scenario information for each of a plurality of disconnection scenarios in which the resistance value of the resistor and the insertion position of the resistor are changed in a sensor network model in which the sensor network is modeled and a resistor is inserted at a specific position in the transmission path, and simulation waveforms obtained by running a simulation for each disconnection scenario, and a model generation unit that generates a trained model for inferring the location and state of the disconnection in the sensor network from the measured waveform by machine learning using the training data acquired by the training data acquisition unit.

[0010] According to the present disclosure, a trained model for inferring the location and state of a disconnection in a sensor network from the measured waveforms is generated using training data including measured waveforms, scenario information for each of multiple disconnection scenarios, and simulation waveforms for each disconnection scenario. Therefore, data acquisition, training, and the like can be performed more efficiently than when using only measured waveforms. Therefore, by performing inference processing using the trained model generated by this training device, the location and state of a disconnection can be efficiently identified.

[0011] FIG. 1 is a diagram illustrating a configuration of a disconnection position identification system according to a first embodiment of the present disclosure. FIG. 2 is a block diagram illustrating an example of the physical configuration of an information processing device. FIG. 3 is a block diagram illustrating the functional configuration of the learning device shown in FIG. 1. FIG. 4 is a flowchart illustrating learning processing by the learning device. FIG. 5 is a block diagram illustrating the functional configuration of the inference device shown in FIG. 1.

[0012] Hereinafter, a learning device, an inference device, a disconnection location identification system, a learning method, an inference method, and a program according to embodiments of the present disclosure will be described with reference to the drawings. Note that the same reference numerals are used to denote identical or corresponding parts in the drawings.

[0013] FIG. 1 is a schematic diagram showing the entirety of a wire break location identification system 100. The wire break location identification system 100 is a system that identifies the location of a wire break in a sensor network 10 to be diagnosed and the extent of the break, i.e., the state of the break, such as the extent to which the break is about to occur. The wire break state is not limited to a state in which the cable is completely broken, but also includes, for example, a state in which the cable is degraded, such as a broken or damaged core wire or element wire included in the cable, or a deterioration state of the cable, such as a deterioration in the electrical characteristics of the cable due to aging. As shown in the figure, the wire break location identification system 100 includes the sensor network 10 to be diagnosed, a simulation device 20 that generates a wire break scenario based on a sensor network model that simulates the configuration of the sensor network 10 and executes a transmission path simulation of the generated wire break scenario, a learning device 30 that generates a trained model for identifying the location and state of the break in the sensor network 10, a trained model storage unit 40 that stores the trained model generated by the learning device 30, and an inference device 50 that infers the location and state of the break in the sensor network 10.

[0014] The sensor network 10 includes a unit 11, which is a master unit that manages the sensor network 10, and sensors 12a, 12b, 12c, and so on that acquire various measurement data, and the unit 11 and the sensors 12a, 12b, 12c, and so on are connected via a transmission path 13. In the following description, the sensors 12a, 12b, 12c, and so on are also collectively referred to as sensors 12. The unit 11 and the sensors 12 have a function of logging waveform data of signals flowing through the transmission path.

[0015] The simulation device 20 generates a disconnection scenario 21 and includes a simulation unit 22 .

[0016] The simulation device 20 inserts resistors at specific positions in the transmission line in a sensor network model that simulates the configuration of the sensor network 10 to be diagnosed, and generates multiple disconnection scenarios 21 by changing the resistance value and the insertion position of the inserted resistor. In the disconnection scenarios 21, the position of the resistor indicates the disconnection position, and the resistance value indicates the disconnection state.

[0017] The simulation unit 22 executes a transmission path simulation for each disconnection scenario 21 and outputs a simulated waveform of a signal flowing through the transmission path for each disconnection scenario 21. The transmission path simulation refers to, for example, simulating a signal at an observation point of the sensor network 10 when a signal with a specific waveform is input from the input terminal of the sensor network 10, or expressing the sensor network 10 as a distributed parameter system and calculating the signal waveform at the observation point from reflections at points where impedance changes and the group velocity of the transmission path.

[0018] The learning device 30 generates a learned model for inferring the disconnection location and disconnection state from the measured waveform using measured waveforms, which are waveform data logged by the units 11 and sensors 12 of the sensor network 10, scenario information including the resistance values ​​and resistance positions of the disconnection scenarios 21 generated by the simulation device 20, and learning data including the simulated waveforms of each disconnection scenario 21.

[0019] The measured waveforms used as learning data may be data acquired during demonstration operation of the sensor network 10 before commercial operation, data acquired in the sensor network 10 in the past and stored in advance, or data acquired from the sensor network 10 currently in operation.

[0020] The trained model storage unit 40 stores the trained model generated by the learning device 30.

[0021] The inference device 50 acquires measured waveforms from the sensor network 10 and uses the trained model generated by the learning device 30 to infer the location and state of a break from the acquired measured waveforms.

[0022] In the disconnection location identification system 100, the information processing device 60 in which the simulation device 20, the learning device 30, the learned model storage unit 40, and the inference device 50 are realized physically includes, as illustrated in FIG. 2, a CPU (Central Processing Unit) 61 that executes processing according to a program, a RAM (Random Access Memory) 62 which is a volatile memory, a ROM (Read Only Memory) 63 which is a non-volatile memory, a storage unit 64 that stores data, an input unit 65 that accepts input of information, a display unit 66 that visualizes and displays information, and a communication unit 67 that transmits and receives information, all of which are connected via an internal bus 99.

[0023] The CPU 61 executes various processes by reading out the programs stored in the storage unit 64 into the RAM 62 and executing them. As main functions provided by the programs, the CPU 61 executes processes by the simulation unit 22 of the simulation device 20, the model generation unit 32, the reward calculation unit 33, and the function update unit 34 of the learning device 30 illustrated in Fig. 3, and the inference unit 52 of the inference device 50 illustrated in Fig. 5.

[0024] The RAM 62 is used as a work area for the CPU 61. The ROM 63 stores the control program executed by the CPU 61, a BIOS (Basic Input Output System), and the like.

[0025] The storage unit 64 includes a hard disk drive, stores the programs executed by the CPU 61, and stores various data used when the programs are executed. The storage unit 64 functions as the trained model storage unit 40.

[0026] The input unit 65 is a user interface that includes a keyboard, a mouse, a communication device, and the like.

[0027] The display unit 66 is a display device such as a liquid crystal display or an organic EL (Electro Luminescence) display that visualizes and displays information.

[0028] The communication unit 67 is a network termination device or a wireless communication device that connects to the network, and a serial interface or a LAN (Local Area Network) interface that connects to them. The communication unit 67 functions as the data acquisition unit 31 of the learning device 30 shown in Fig. 3 and the data acquisition unit 51 of the inference device 50 shown in Fig. 5.

[0029] 3 is a configuration diagram of the learning device 30. The learning device 30 includes a data acquisition unit 31 that acquires learning data, and a model generation unit 32 that generates a trained model based on the learning data acquired by the data acquisition unit 31.

[0030] The data acquiring unit 31 acquires, from the simulation device 20, scenario information including the resistance value and resistor position for each of the plurality of disconnection scenarios 21 generated by the simulation device 20, and simulation waveforms obtained by performing simulations for each of the disconnection scenarios 21. The data acquiring unit 31 also acquires, from the sensor network 10, measured waveforms logged by the units 11 and sensors 12 of the sensor network 10. The data acquiring unit 31 is an example of a learning data acquiring unit.

[0031] The model generation unit 32 generates a trained model that infers the location and state of a disconnection in the sensor network 10 based on the learning data including the scenario information, simulation waveforms, and measured waveforms acquired by the data acquisition unit 31. That is, a trained model that infers the location and state of a disconnection from the measured waveforms of the sensor network 10 is generated.

[0032] The learning algorithm used by the model generation unit 32 can be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where reinforcement learning is applied will be described. In reinforcement learning, an agent, which is the subject of action in a certain environment, observes environmental parameters that indicate the current state and determines the action to be taken. The environment changes dynamically depending on the actions of the agent. The agent is given a reward in accordance with the change in the environment. The agent repeats this process and learns the course of action that will obtain the most reward through a series of actions.

[0033] Known representative methods of reinforcement learning include Q-learning and TD-learning. For example, in the case of Q-learning, a general update formula for the action value function Q(s, a) is expressed by the following formula 1:

[0034]

[0035] In Equation 1, s ｔ represents the state of the environment at time t, and a ｔ represents the action at time t. ｔ Therefore, the state is s ｔ＋１ It changes to r ｔ＋１ represents the reward that can be obtained depending on the change in state, γ represents the discount rate, and α represents the learning coefficient. Note that γ is in the range of 0<γ≦1, and α is in the range of 0<α≦1. The scenario information including the resistance value and the resistance position of each disconnection scenario 21 generated by the simulation device 20, and the simulation waveform output by each disconnection scenario 21 are used to represent the behavior a ｔ The measured waveform of the sensor network 10 is in the state s ｔ and the state s at time t ｔ Best action in a ｔ Learn.

[0036] The update formula expressed by Equation 1 increases the action value Q if the action value Q of the action a with the highest Q value at time t+1 is greater than the action value Q of the action a executed at time t, and decreases the action value Q in the opposite case. In other words, the action value function Q(s, a) is updated so that the action value Q of the action a at time t approaches the best action value at time t+1. As a result, the best action value in a certain environment is propagated sequentially to the action value in the previous environment.

[0037] As described above, when generating a trained model by reinforcement learning, the model generation unit 32 includes a reward calculation unit 33 and a function update unit 34.

[0038] The reward calculation unit 33 calculates a reward based on scenario information including the resistance value and resistor location of each disconnection scenario 21, simulation waveforms output by simulating each disconnection scenario 21, and measured waveforms logged by the units 11 and sensors 12 of the sensor network 10. The reward calculation unit 33 calculates a reward r based on the potential difference between the simulation waveform and the measured waveform. If the potential difference between the simulation waveform and the measured waveform decreases, the reward calculation unit 33 increases the reward r (for example, provides a reward of "1"), and if the potential difference between the simulation waveform and the measured waveform increases, the reward calculation unit 33 reduces the reward r (for example, provides a reward of "-1"). Note that the reward calculation unit 33 may calculate the reward r based on the similarity between the measured waveform and the simulation waveform in terms of long-term or short-term trends, waveform length, frequency, etc.

[0039] The function update unit 34 updates the function for determining the disconnection position and disconnection state in accordance with the reward r calculated by the reward calculation unit 33, and outputs the updated function to the learned model storage unit 40. For example, in the case of Q-learning, the action value function Q(s) expressed by Equation 1 is ｔ , a ｔ ) is used as a function for calculating the disconnection position and disconnection state.

[0040] The function update unit 34 repeatedly executes the above learning to generate a trained model. The trained model storage unit 40 stores the action value function Q(sｔ , a ｔ ), i.e., stores the trained model.

[0041] Next, the operation of the learning device 30 to generate and output a trained model will be described using the flowchart illustrated in Figure 4. The learning device 30 starts executing a learning process when it is powered on. The learning process is executed before performing a diagnostic process on the sensor network 10 for the first time, when a change in the configuration of the sensor network 10 requires updating the trained model, when it is desired to retrain the trained model using more training data and improve the accuracy of inference, etc.

[0042] The data acquisition unit 31 of the learning device 30 acquires learning data from the sensor network 10 and the simulation device 20 (step S11). Specifically, the data acquisition unit 31 acquires, as learning data, scenario information including the resistance value and resistor position of each disconnection scenario 21 generated by the simulation device 20, simulation waveforms output by simulating each disconnection scenario 21, and actual measurement waveforms logged by the units 11 and sensors 12 of the sensor network 10.

[0043] Next, the reward calculation unit 33 calculates the reward based on the scenario information of the disconnection scenario 21 acquired in step S11, the simulation waveform, and the measured waveform (step S12). Specifically, the reward calculation unit 33 acquires the scenario information of the multiple disconnection scenarios 21 and the simulation waveform and the measured waveform for each disconnection scenario 21, and determines whether to increase or decrease the reward based on the potential difference between the signals of the predetermined simulation waveform and the measured waveform.

[0044] The reward calculation unit 33 increases the reward when the potential difference between the simulation waveform and the measured waveform signal decreases (step S13), while the reward calculation unit 33 decreases the reward when the potential difference between the simulation waveform and the measured waveform signal increases (step S14).

[0045] Next, the function update unit 34 updates the action value function Q(s) represented by Equation 1 stored in the trained model storage unit 40 based on the reward calculated by the reward calculation unit 33. ｔ , a ｔ ) is updated (step S15).

[0046] The learning device 30 repeatedly executes the above steps S11 to S15 to generate the action value function Q(s ｔ , a ｔ ) is stored as a trained model.

[0047] The learning device 30 is configured to store the learned model in a learned model storage unit 40 provided outside the learning device 30, but the learned model storage unit 40 may also be provided inside the learning device 30.

[0048] 5 is a configuration diagram of the inference device 50. The inference device 50 includes a data acquisition unit 51 that acquires measured waveforms from the sensor network 10, and an inference unit 52 that infers the location and state of a disconnection in the sensor network 10.

[0049] The data acquisition unit 51 acquires the measured waveforms logged by the units 11 and sensors 12 of the sensor network 10 .

[0050] The inference unit 52 infers the location and state of the disconnection using the trained model generated by the learning device 30. That is, by inputting the measured waveform acquired by the data acquisition unit 51 into this trained model, it is possible to infer the location and state of the disconnection based on the measured waveform.

[0051] In this embodiment, the disconnection location and disconnection state are output using a trained model trained by the model generation unit 32 of the disconnection location identification system 100, but it is also possible to obtain a trained model from another disconnection location identification system and output the disconnection location and disconnection state based on this trained model.

[0052] Next, the inference process for obtaining an inference result using the trained model will be described with reference to FIG. 6 .

[0053] The data acquisition unit 51 of the inference device 50 acquires the measured waveforms from the sensor network 10 (step S21). Specifically, the data acquisition unit 51 acquires the measured waveforms logged by the units 11 and sensors 12 of the sensor network 10.

[0054] Next, the inference unit 52 inputs the measured waveform acquired in step S21 into the trained model stored in the trained model storage unit 40 (step S22), and obtains an inference result including the disconnection position and the disconnection state. The inference unit 52 outputs the obtained inference result including the disconnection position and the disconnection state to the disconnection position identification system 100 (step S23).

[0055] Next, the wire break location identification system 100 uses the wire break location output by the inference unit 52 to display the wire break location on a management terminal (not shown) or notify the facility manager of the wire break location, thereby enabling the wire break location to be identified early.

[0056] As described above, the wire break location identification system 100 performs learning using the measured waveform of the sensor network 10 to be diagnosed, scenario information including the resistance value and resistance location of each of the multiple wire break scenarios 21, and learning data including the simulated waveform of each wire break scenario 21, and infers the wire break location and state using the learned model. This makes it possible to more efficiently identify the wire break location and infer the wire break state, including the degree of cable deterioration due to aging.

[0057] In this embodiment, a case where reinforcement learning is applied to the learning algorithm used by the inference unit 52 has been described, but the present invention is not limited to this. As for the learning algorithm, other than reinforcement learning, supervised learning, unsupervised learning, semi-supervised learning, or the like can also be applied.

[0058] Furthermore, the learning algorithm used in the model generation unit 32 may be deep learning, which learns to extract the features themselves, or machine learning may be performed according to other known methods, such as neural networks, genetic programming, functional logic programming, and support vector machines.

[0059] The learning device 30 and the inference device 50 may be connected to the disconnection location identification system 100 via a network, for example, and may be separate devices from the disconnection location identification system 100. The learning device 30 and the inference device 50 may also be built into the disconnection location identification system 100. Furthermore, the simulation device 20, the learning device 30, and the inference device 50 may reside on a cloud server.

[0060] The model generation unit 32 may also learn the wire break location using learning data acquired from multiple wire break location identification systems 100. The model generation unit 32 may acquire learning data from multiple wire break location identification systems 100 used in the same area, or may learn the wire break location using learning data collected from multiple wire break location identification systems 100 operating independently in different areas. It is also possible to add or remove wire break location identification systems 100 that collect learning data from the target during the process. Furthermore, a learning device 30 that has learned the wire break location for one wire break location identification system 100 may be applied to another wire break location identification system 100, and the wire break location for the other wire break location identification system 100 may be re-learned and updated.

[0061] In the above embodiment, the functions of the wire break location identification system 100 have been described as being executed by individual devices, namely, the simulation device 20, the learning device 30, the trained model storage unit 40, and the inference device 50. However, this is not limiting, and the functions of each device may be executed by an information processing device that is a single computer. Furthermore, each process of the wire break location identification system 100 may be executed by multiple computers that execute the functions of several devices, such as by executing the functions of some devices by one computer and the functions of other devices by another computer.

[0062] Furthermore, data used in the learning process and the inference process, such as the disconnection scenario 21, the simulation waveform, the measured waveform, and the trained model, may be collectively managed by a cloud server on the network, and the learning device 30 and the inference device 50 may access the cloud server as needed to read and write information. In this case, the disconnection location identification system 100 does not need to include the trained model storage unit 40.

[0063] Furthermore, the simulation device 20, the learning device 30, the trained model storage unit 40, and the inference device 50 can be realized using a normal computer system, rather than using dedicated devices. For example, a program for realizing each function may be stored and distributed on a computer-readable recording medium such as a CD-ROM (Compact Disc Read Only Memory) or a DVD-ROM (Digital Versatile Disc Read Only Memory), and a computer capable of realizing each of the above-described functions may be configured by installing the program on the computer.

[0064] Furthermore, when each function is realized by sharing the functions between an OS (Operating System) and an application, or by cooperation between the OS and the application, only the application may be stored on the recording medium.

[0065] The present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to explain the present disclosure and do not limit the scope of the present disclosure. In other words, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and the meaning of equivalent disclosures are considered to be within the scope of the present disclosure.

[0066] 100 Disconnection location identification system, 10 Sensor network, 11 Unit, 12, 12a, 12b, 12c Sensor, 13 Transmission path, 20 Simulation device, 21 Disconnection scenario, 22 Simulation unit, 30 Learning device, 31 Data acquisition unit, 32 Learned model generation unit, 33 Reward calculation unit, 34 Function update unit, 40 Learned model storage unit, 50 Inference device, 51 Data acquisition unit, 52 Inference unit, 60 Information processing device, 61 CPU, 62 RAM, 63 ROM, 64 Storage unit, 65 Input unit, 66 Display unit, 67 Communication unit, 99 Internal bus.

Claims

1. A learning device that generates a trained model for inferring the location and state of a disconnection in a sensor network to be diagnosed, comprising: a training data acquisition unit that acquires training data including: an actual measured waveform of a signal flowing through a transmission path of the sensor network; scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which the sensor network is modeled and a resistor is inserted at a specific position in the transmission path; and a model generation unit that generates a trained model for inferring the location and state of a disconnection in the sensor network from the actual measured waveform by machine learning using the training data acquired by the training data acquisition unit.

2. The learning device described in claim 1, wherein the model generation unit comprises: a reward calculation unit that calculates a reward based on the potential difference between the measured waveform and the simulation waveform for each of the disconnection scenarios; and a function update unit that updates a function for inferring the location and state of a disconnection in the sensor network based on the reward calculated by the reward calculation unit.

3. The learning device according to claim 1 or 2, wherein the model generation unit acquires the measured waveform from the sensor network in operation and performs re-learning using the acquired measured waveform as the learning data.

4. A learning device according to any one of claims 1 to 3, wherein the model generation unit acquires the measured waveforms that are measured during operation of the sensor network and stored in advance, and uses the measured waveforms as the learning data.

5. An inference device that infers the location and state of a disconnection in a sensor network to be diagnosed, comprising: a data acquisition unit that acquires a measured waveform of a signal flowing through a transmission path of the sensor network; and an inference unit that outputs an inference result that includes the location and state of a disconnection in the sensor network by inputting the measured waveform acquired by the data acquisition unit into a trained model generated by machine learning using training data that includes: the measured waveform; scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which a resistor is inserted at a specific position in the transmission path, and simulation waveforms obtained by running a simulation for each of the disconnection scenarios.

6. A disconnection location identification system for identifying the location and state of a disconnection in a sensor network to be diagnosed, comprising: a learning device according to any one of claims 1 to 4; an inference device according to claim 5; and a simulation device that models the sensor network, inserts a resistor at a specific position in a transmission line, generates a plurality of disconnection scenarios by changing the resistance value of the resistor and the insertion position of the resistor in the sensor network model, and outputs a simulation waveform by running a simulation for each of the disconnection scenarios.

7. A learning method in which a computer that generates a trained model for inferring the location and state of a disconnection in a sensor network to be diagnosed executes the following steps: acquiring training data including measured waveforms of signals flowing through a transmission path of the sensor network, scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which the sensor network is modeled and a resistor is inserted at a specific position in the transmission path, and simulation waveforms obtained by running a simulation for each of the disconnection scenarios; and generating a trained model for inferring the location and state of a disconnection in the sensor network from the measured waveforms by machine learning using the acquired training data.

8. An inference method in which a computer that infers the location and state of a disconnection in a sensor network to be diagnosed executes the following steps: acquire a measured waveform of a signal flowing through a transmission path of the sensor network; and output an inference result including the location and state of the disconnection in the sensor network by inputting the acquired measured waveform into a trained model generated by machine learning using training data including the measured waveform, scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which a resistor is inserted at a specific position in the transmission path, and simulation waveforms obtained by running a simulation for each of the disconnection scenarios.

9. A program that causes a computer that generates a trained model for inferring the location and state of a disconnection in a sensor network to be diagnosed to execute the following steps: acquire training data including measured waveforms of signals flowing through the transmission path of the sensor network, scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which the sensor network is modeled and a resistor is inserted at a specific position in the transmission path, and simulated waveforms obtained by running a simulation for each of the disconnection scenarios; and generate a trained model for inferring the location and state of a disconnection in the sensor network from the measured waveforms through machine learning using the acquired training data.

10. A program that causes a computer that infers the location and state of a disconnection in a sensor network to be diagnosed to execute the following steps: acquire a measured waveform of a signal flowing through a transmission path of the sensor network; and output an inference result that includes the location and state of a disconnection in the sensor network by inputting the acquired measured waveform into a trained model generated by machine learning using training data that includes the measured waveform, scenario information for each of a plurality of disconnection scenarios in which the resistance value of a resistor and the insertion position of the resistor are changed in a sensor network model in which a resistor is inserted at a specific position in the transmission path, and simulation waveforms obtained by running a simulation for each of the disconnection scenarios.

Images