JP2023167883A - Failure prediction system, failure prediction device, learning device, failure prediction method, and program - Google Patents

Abstract

[Problem] Predict failures of factory equipment with high accuracy. [Solution] A learning information acquisition unit 340 receives sensor detection information that is information that a sensor 100 detects the state of the equipment, and a worker explains the state of the equipment generated by a mobile terminal 200 of the worker using the equipment. The learned model generation unit 350 acquires learning information based on the worker explanation information, and the learned model generation unit 350 generates a learned model that shows the relationship between the equipment status and failure by machine learning using the learning information acquired in advance. A model is generated, and the learned model output unit 370 outputs the learned model. The learned model acquisition unit 510 acquires the learned model, and the inference result information generation unit 560 inputs inference information based on the newly acquired sensor detection information and worker explanation information into the learned model, and The inference result information output unit 570 generates inference result information that is information indicating the result of inferring the failure of the inference result. [Selection diagram] Figure 2

Description

The present disclosure relates to a failure prediction system, a failure prediction device, a learning device, a failure prediction method, and a program.

BACKGROUND ART Conventionally, in production factories, devices and systems are used to detect malfunctions and breakdowns in equipment and equipment within the factory in order to prevent accidents from occurring. Such devices and systems detect failures by detecting deviations from normal conditions from information obtained from devices and equipment.

Patent Document 1 discloses that failure information indicating the presence or absence of a failure of an industrial machine or the degree of failure is output in response to input of a current state variable obtained from output data of a sensor that detects the state of the industrial machine or the surrounding environment. A failure prediction device is disclosed. The failure prediction device of Patent Document 1 detects failures of industrial machines according to a training data set created based on a combination of state variables observed from sensor output data and determination data representing the presence or absence of failures of industrial machines or the degree of failure. The conditions associated with are learned in advance.

JP 2021-2398 Publication

The device described in Patent Document 1 learns and makes inferences based on state variables obtained from sensor output data. However, since industrial machines fail frequently, it is difficult to prepare a sufficient number of state variables that indicate the failure state of the industrial machine. For this reason, the device described in Patent Document 1 cannot perform sufficient learning in advance, and there is a possibility that the accuracy of detecting a failure may be reduced. As a result, the device described in Patent Document 1 may have lower accuracy in predicting failures than, for example, a skilled worker who handles industrial machinery on site.

The present disclosure has been made in view of the above circumstances, and aims to accurately predict failures of equipment in factories.

In order to achieve the above object, a failure prediction system according to the present disclosure predicts failures of equipment in a factory. A failure prediction system according to the present disclosure includes a sensor that detects the state of equipment, a mobile terminal carried by a worker who works with the equipment, and a learning device that learns the relationship between the equipment state and equipment failure. and an inference device that infers a failure of the device based on the state of the device. The learning device acquires learning information based on sensor detection information, which is information that a sensor detects the state of the equipment, and worker explanation information, which is information generated by the mobile terminal and is information that the worker explains about the state of the equipment. A learning information acquisition unit, a trained model generation unit that generates a trained model that shows the relationship between equipment status and equipment failure through machine learning using previously acquired learning information, and outputs the trained model. and a trained model output unit. The inference device includes a learned model acquisition unit that acquires a trained model, a sensor detection information acquisition unit that acquires sensor detection information from a sensor, and a worker explanation information acquisition unit that acquires worker explanation information from a mobile terminal. Inference result information generation that inputs inference information based on newly acquired sensor detection information and worker explanation information into the trained model and generates inference result information that is information indicating the result of inferring equipment failure. and an inference result information output unit that outputs inference result information.

According to the present disclosure, the learning device generates a learned model by machine learning using learning information based not only on sensor detection information but also on worker explanation information. In other words, the learning device learns the relationship between the equipment status and equipment failure by taking into account not only the equipment status detected by the sensor, but also the equipment status explained by the worker using the equipment. . As a result, the failure prediction system according to the present disclosure is more effective than a failure prediction system in which the learning device does not generate a learned model by machine learning using learning information based on sensor detection information and worker explanation information. Failures can be predicted with high accuracy.

Overall explanatory diagram of the failure prediction system according to Embodiment 1 A diagram showing a functional configuration of a failure prediction system according to Embodiment 1 Block diagram showing the hardware configuration of each device according to Embodiment 1 (A) Display example of learning information according to Embodiment 1, (B) Display example of inference information according to Embodiment 1, (C) Display example of inference result information according to Embodiment 1 Flowchart of learned model generation processing according to Embodiment 1 Flowchart of inference result information generation processing according to Embodiment 1 A diagram showing the functional configuration of a failure prediction system according to Embodiment 2 Explanatory diagram of a neural network according to Embodiment 2 (A) Display example of learning information according to Embodiment 2, (B) Display example of new learning information output by the failure prediction device according to Embodiment 2

Hereinafter, a failure prediction system, a failure prediction device, a learning device, a failure prediction method, and a program according to an embodiment of the present disclosure will be described in detail with reference to the drawings. In addition, the same reference numerals are given to the same or corresponding parts in the figures.

[Embodiment 1]
(About failure prediction system 1 according to Embodiment 1)
A failure prediction system 1 according to Embodiment 1 of the present disclosure is a system that predicts failures of equipment in a factory. As shown in FIG. 1, the failure prediction system 1 includes a sensor 100 that detects the state of equipment, a mobile terminal 200 carried by a worker who performs work using the equipment, and a sensor 100 that detects the relationship between the equipment state and equipment failure. It includes a learning device 300 for learning and a storage device 400 for storing information. Furthermore, the failure prediction system 1 includes a failure prediction device 500 as an example of an inference device that infers a failure of a device based on the state of the device. The sensor 100, the mobile terminal 200, the learning device 300, the storage device 400, and the failure prediction device 500 are capable of transmitting and receiving information via a wireless LAN (Local Area Network), not shown.

In the failure prediction system 1, the sensor 100 is installed at a predetermined position in the factory and collects information about the device itself and the surrounding environment of the device, and from this information, the sensor 100 is used as information that detects the condition of the device. Outputting sensor detection information. Furthermore, the mobile terminal 200 is always carried by the worker while working, and stores location information that is information indicating the worker's location and worker explanation information that is information that the worker explains about the status of the equipment. Collected and output.

In the failure prediction system 1, the learning device 300 acquires the sensor detection information that has been output from the sensor 100, the position information and the worker explanation information that has been output from the mobile terminal 200, and stores it in the storage device 400. It is accumulated in Further, the learning device 300 generates a learned model by machine learning using learning information based on these pieces of information stored in advance, and outputs the generated learned model to the storage device 400 for storage. .

In the failure prediction system 1, the failure prediction device 500 acquires the learned model from the storage device 400, and newly acquires sensor detection information, position information, and worker explanation information from the sensor 100 and the mobile terminal 200. Therefore, the failure prediction device 500 inputs inference information based on this information into a learned model and generates and outputs inference result information, which is information indicating the result of inferring a device failure. Can be predicted.

(About sensor 100 according to Embodiment 1)
The sensor 100 is, for example, a camera, thermometer, hygrometer, ammeter, voltmeter, speed sensor, acceleration sensor, or displacement sensor that can transmit sensor detection information via wireless LAN. That is, the sensor 100 detects information regarding the equipment itself in the factory or the surrounding environment of the equipment, such as an image of the equipment, temperature, current, voltage, speed, acceleration, and displacement inside the equipment, and temperature and humidity of the surrounding environment. This sensor can output the information as sensor detection information. As shown in FIG. 2, the sensor 100 includes a sensor detection information generation section 110 that generates sensor detection information, and a sensor detection information output section 120 that outputs the sensor detection information.

(About mobile terminal 200 according to Embodiment 1)
The mobile terminal 200 is a so-called wearable terminal worn by a worker, such as a smart watch or smart glasses. Note that the mobile terminal 200 is not limited to a wearable terminal as long as it is portable by a worker, and may be a computer device such as a smartphone or a notebook personal computer. The mobile terminal 200 includes a position information generation unit 210 that generates worker position information, a worker explanation information generation unit 220 that generates worker explanation information, a position information output unit 230 that outputs position information, and a position information output unit 230 that outputs worker explanation information. It includes a worker explanation information output unit 240 to output.

(About learning device 300 according to Embodiment 1)
The learning device 300 is, for example, a computer device such as a personal computer, a server computer, or a supercomputer. The learning device 300 includes a sensor detection information acquisition section 310 that acquires sensor detection information, a position information acquisition section 320 that acquires position information, and a worker explanation information acquisition section 330 that acquires worker explanation information. The learning device 300 also includes a worker explanation information determination unit 335 that determines worker explanation information, a learning information acquisition unit 340 that acquires learning information, a learned model generation unit 350 that generates a learned model, and a learning model generation unit 350 that generates a learned model. It includes a learning information output section 360 that outputs information and a learned model output section 370 that outputs a learned model.

(About storage device 400 according to Embodiment 1)
The storage device 400 is, for example, a HDD (Hard Disk Drive) on a communication network connected via a wireless LAN, or a so-called NAS (Network Attached Storage). The storage device 400 includes a trained model storage unit 410 that stores trained models, and a learning information storage unit 420 that stores learning information.

(About failure prediction device 500 according to Embodiment 1)
Failure prediction device 500 is a computer device similar to learning device 300. The failure prediction device 500 includes a trained model acquisition unit 510 that acquires a trained model, a sensor detection information acquisition unit 520 that acquires sensor detection information, a position information acquisition unit 530 that acquires position information, and a determination of worker explanation information. It includes a worker explanation information determination section 535 and a worker explanation information acquisition section 540 that acquires worker explanation information. The failure prediction device 500 also includes an inference information acquisition unit 550 that acquires inference information, an inference result information generation unit 560 that generates inference result information, and an inference result information output unit 570 that outputs inference result information.

(About the hardware configuration of learning device 300 according to Embodiment 1)
As shown in FIG. 3, the learning device 300 includes a control unit 51 that executes processing according to a control program 59. The control unit 51 includes a CPU (Central Processing Unit). The control unit 51 functions as the worker explanation information determination unit 335, the learning information acquisition unit 340, and the learned model generation unit 350 shown in FIG. 2 according to the control program 59.

Returning to FIG. 3, the learning device 300 includes a main storage section 52 that loads a control program 59 and is used as a work area for the control section 51. The main storage unit 52 includes a RAM (Random Access Memory).

The learning device 300 also includes an external storage unit 53 that stores a control program 59 in advance. The external storage section 53 supplies the data stored by this program to the control section 51 according to instructions from the control section 51, and stores the data supplied from the control section 51. The external storage unit 53 includes nonvolatile memory such as a flash memory, an HDD (Hard Disk Drive), and an SSD (Solid State Drive).

The learning device 300 also includes an operation unit 54 operated by the user. The input information is supplied to the control section 51 via the operation section 54 . The operation unit 54 includes information input components such as a keyboard, a mouse, and a touch panel.

The learning device 300 also includes a display section 55 that displays information input via the operation section 54 and information output from the control section 51. The display unit 55 includes a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display.

The learning device 300 includes a transmitting/receiving section 56 that transmits and receives information. The transmitting/receiving unit 56 includes information communication components such as a communication network termination device and a wireless communication device connected to the network. The transmitting/receiving section 56 functions as a sensor detection information acquisition section 310, a position information acquisition section 320, a worker explanation information acquisition section 330, a learning information output section 360, and a learned model output section 370 shown in FIG.

Returning to FIG. 3, in the learning device 300, the main storage section 52, external storage section 53, operation section 54, display section 55, and transmission/reception section 56 are all connected to the control section 51 via the internal bus 50.

In the learning device 300, the control section 51 uses the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transmission/reception section 56 as resources, so that the above-mentioned sections 310, 320, 330, shown in FIG. 335, 340, 350, 360, and 370 functions. For example, the learning device 300 executes a sensor detection information acquisition step performed by the sensor detection information acquisition unit 310, a position information acquisition step performed by the position information acquisition unit 320, and a worker explanation information acquisition step performed by the worker explanation information acquisition unit 330. do. Further, for example, the learning device 300 performs a worker explanation information determination step performed by the worker explanation information determination unit 335, a learning information acquisition step performed by the learning information acquisition unit 340, and a learned model performed by the learned model generation unit 350. Execute the generation step. Further, for example, the learning device 300 executes a learning information output step performed by the learning information output unit 360 and a learned model output step performed by the learned model output unit 370.

(About the hardware configuration of failure prediction device 500 according to Embodiment 1)
As shown in FIG. 3, like the learning device 300, the failure prediction device 500 also includes a control section 51, a main storage section 52, an external storage section 53, an operation section 54, and a display section connected via an internal bus 50. 55, a transmitting/receiving section 56. The control unit 51 functions as the worker explanation information determination unit 535, the inference information acquisition unit 550, and the inference result information generation unit 560 shown in FIG. 2 according to the control program 59. Furthermore, the display section 55 functions as the inference result information output section 570 shown in FIG. Further, the transmitting/receiving unit 56 functions as a trained model acquisition unit 510, a sensor detection information acquisition unit 520, a position information acquisition unit 530, and a worker explanation information acquisition unit 540 shown in FIG.

Returning to FIG. 3, the failure prediction device 500 is configured such that the control section 51 uses the main storage section 52, external storage section 53, operation section 54, display section 55, and transmission/reception section 56 as resources. The functions of 510, 520, 530, 535, 540, 550, 560, and 570 are realized. For example, the failure prediction device 500 executes a learned model acquisition step performed by the learned model acquisition unit 510 and a sensor detection information acquisition step performed by the sensor detection information acquisition unit 520. For example, the failure prediction device 500 also includes a position information acquisition step performed by the position information acquisition unit 530, a worker explanation information determination step performed by the worker explanation information determination unit 535, and a worker explanation performed by the worker explanation information acquisition unit 540. Execute the information retrieval step. For example, the failure prediction device 500 includes an inference information acquisition step performed by the inference information acquisition unit 550, an inference result information generation step performed by the inference result information generation unit 560, and an inference result information output performed by the inference result information output unit 570. Execute the steps.

(About the hardware configuration of sensor 100 according to Embodiment 1)
Although not shown, the sensor 100 includes a control section 51, a main storage section 52, an external storage section 53, an operation section 54, and a transmitting/receiving section 56, which are connected via an internal bus 50. The control unit 51 functions as the sensor detection information generation unit 110 shown in FIG. 2 according to the control program 59. Further, the transmitting/receiving section 56 functions as the sensor detection information output section 120 shown in FIG. 2. In the sensor 100, the control section 51 uses the main storage section 52, the external storage section 53, the operation section 54, and the transmission/reception section 56 as resources, thereby realizing the functions of the above-mentioned sections 110 and 120 shown in FIG. For example, the sensor 100 executes a sensor detection information generation step performed by the sensor detection information generation section 110 and a sensor detection information output step performed by the sensor detection information output section 120.

(About the hardware configuration of mobile terminal 200 according to Embodiment 1)
Although not shown, the mobile terminal 200 includes a control section 51, a main storage section 52, an external storage section 53, an operation section 54, and a transmitting/receiving section 56, which are connected via an internal bus 50. The control unit 51 functions as the position information generation unit 210 and the worker explanation information generation unit 220 shown in FIG. 2 according to the control program 59. Further, the transmitting/receiving section 56 functions as a position information output section 230 and a worker explanation information output section 240 shown in FIG. In the mobile terminal 200, the control section 51 uses the main storage section 52, the external storage section 53, the operation section 54, and the transmitting/receiving section 56 as resources, thereby realizing the functions of the above-mentioned sections 210 to 240 shown in FIG. For example, the mobile terminal 200 executes a position information generation step performed by the position information generation unit 210 and a worker explanation information generation step performed by the worker explanation information generation unit 220. Further, for example, the mobile terminal 200 executes a position information output step performed by the position information output unit 230 and a worker explanation information output step performed by the worker explanation information output unit 240.

(Details of functional configuration of mobile terminal 200 according to Embodiment 1)
The position information generation unit 210 generates the position information of the worker using, for example, a GPS (Global Positioning System) function included in the mobile terminal 200. Note that the position information includes, for example, information indicating the position of the worker detected by the mobile terminal 200, and date and time information indicating the date and time of the detection by the mobile terminal 200.

The worker explanation information generation unit 220 generates worker explanation information as audio information, for example, by collecting and recording the voice uttered by the worker using a microphone included in the mobile terminal 200. The worker explanation information includes, for example, audio information that includes words related to equipment failure, such as "It's not working well," "It's stopped again," and "I think it's nearing the end of its lifespan," as well as a date and time indicating the date and time when the audio information was recorded. Contains information.

(Details of functional configuration of learning device 300 according to Embodiment 1)
The sensor detection information acquisition unit 310 acquires sensor detection information output from the sensor 100. Note that the sensor detection information includes numerical information indicating a numerical value indicating the state of the device detected by the sensor 100, and date and time information indicating the date and time of detection by the sensor 100. Further, the position information acquisition unit 320 acquires the position information of the worker output from the mobile terminal 200, and the worker explanation information acquisition unit 330 acquires the worker explanation information output from the mobile terminal 200.

The worker explanation information determination unit 335 determines whether the worker explanation information acquired from the mobile terminal 200 is information in which the worker actually explained the state of the equipment, based on the position information. The worker explanation information determination unit 335 determines whether the worker is working with the equipment if the distance from the worker's location to the location where the equipment is installed is smaller than a predetermined threshold at the date and time specified from the location information. Therefore, it is determined that the operator explanation information acquired from the mobile terminal 200 at the relevant date and time is the information in which the operator actually explained the state of the equipment.

Note that if a predetermined time has not elapsed since the date and time when the worker explanation information determination unit 335 determines that the worker explanation information is information in which the worker actually explained the state of the equipment, the worker There is a high possibility that the user continues to make comments about the device even after leaving the device. Therefore, even if the worker explanation information determination unit 335 determines that the worker explanation information acquired from the mobile terminal 200 before the above-mentioned time has elapsed is information in which the worker actually explained the state of the equipment, good. Further, when a first device and a second device different from the first device are provided as equipment in the factory, the worker explanation information determination unit 335 determines whether the worker It is determined whether the explanation information is first worker explanation information explaining the state of the first device or second worker explanation information explaining the state of the second device.

The learning information acquisition unit 340 acquires learning information based on sensor detection information and worker explanation information. For example, regarding sensor detection information and worker explanation information with the same date and time information, the learning information acquisition unit 340 acquires numerical analysis information obtained by analyzing numerical information included in the sensor detection information and audio information included in the worker explanation information. Obtain learning information including audio analysis information that was analyzed. Note that the learning information is information that can be displayed in the form of a table shown in FIG. 4(A). The learning information is, for example, information including numerical analysis information LND1 and audio analysis information LMD1 having the same date and time information. Note that the audio information is analyzed using, for example, a known analysis engine and analysis algorithm of so-called natural language processing, such as morphological analysis and syntactic analysis. Further, when a first device and a second device are provided as devices in a factory, the learning information acquisition unit 340 collects the first learning information that is the learning information of the first device and the learning information of the second device. The second learning information is acquired.

The learned model generation unit 350 generates a learned model that indicates the relationship between the state of the device and the failure of the device by machine learning using learning information acquired in advance. The learned model generation unit 350 generates a learned model by machine learning using a plurality of types of learning information acquired in advance and stored in the learning information storage unit 420. Here, the machine learning method performed by the trained model generation unit 350 is the K-means method as an example of unsupervised learning. Note that unsupervised learning is a method of learning characteristics of data by using learning information that does not include so-called labels. Further, the K-means method is a non-hierarchical clustering algorithm, and is a method of classifying data, that is, learning information, into k clusters, and using the average of each cluster.

Specifically, in the K-means method, clusters are first randomly assigned to n pieces of data xi (i=1, 2, . . . , n). Next, the center Vj (j=1, 2, . . . , k) of each cluster is calculated based on the allocated data. Next, the distance between each data xi and the center Vj of each cluster is calculated, and the process of reassigning each data xi to the cluster with the nearest center Vj is repeated. Then, if the cluster allocation of all data xi does not change, or if the amount of change does not exceed a predetermined threshold, it is determined that the allocation has converged, and the process ends. Further, when a first device and a second device are provided as devices in a factory, the learned model generation unit 350 generates a first learned model that is a learned model of the first device and a learned model of the second device. A second learned model, which is a model, is generated.

The learning information output unit 360 outputs the learning information to the storage device 400 and stores it in the learning information storage unit 420. Further, when a first device and a second device are provided as equipment in the factory, the learning information output unit 360 outputs the first learning information and the second learning information to the storage device 400 for learning. information storage unit 420. Note that the trained model generation unit 350 causes the learning information acquisition unit 340 to read the learning information stored in the learning information storage unit 420 when generating the trained model.

The trained model output unit 370 outputs the trained model to the storage device 400 and stores it in the trained model storage unit 410. Further, when a first device and a second device are provided as devices in the factory, the learned model output unit 370 outputs the first learned model and the second learned model to the storage device 400 for learning. The completed model is stored in the completed model storage unit 410.

(Details of functional configuration of failure prediction device 500 according to Embodiment 1)
The trained model acquisition unit 510 acquires the trained model by reading out the trained model stored in the trained model storage unit 410. Further, when a first device and a second device are provided as devices in a factory, the learned model acquisition unit 510 acquires a first learned model, which is a learned model of the first device, and a learned model of the second device. A second learned model that is a model is obtained.

The sensor detection information acquisition section 520 acquires sensor detection information output from the sensor 100, similarly to the sensor detection information acquisition section 310. Further, the position information acquisition unit 530 acquires the position information of the worker output from the mobile terminal 200, similarly to the position information acquisition unit 320.

Similarly to the worker explanation information determination section 335, the worker explanation information determination section 535 determines whether the information acquired together with the location information from the mobile terminal 200 is worker explanation information based on the location information. Further, when a first device and a second device are provided as devices in the factory, the worker explanation information determination unit 535 determines whether the information acquired from the mobile terminal 200 is the worker explanation information of the first device. It is determined whether the information is the worker explanation information or the second worker explanation information which is the worker explanation information of the second device.

The worker explanation information acquisition unit 540 acquires the worker explanation information output from the mobile terminal 200. When it is determined that the information acquired together with the position information from the mobile terminal 200 is worker explanation information, the worker explanation information acquisition unit 540 acquires the information as worker explanation information. Further, when a first device and a second device are provided as equipment in the factory, the worker explanation information acquisition unit 540 determines that the information acquired from the mobile terminal 200 together with the position information is the first worker explanation information. If determined, the information is acquired as the first worker explanation information, and if it is determined that the information acquired together with the position information from the mobile terminal 200 is the second worker explanation information, the information is acquired as the second worker explanation information. Obtained as descriptive information.

The inference information acquisition unit 550 acquires inference information based on the sensor detection information of the date and time when the new information was acquired and the worker explanation information. Similar to the learning information acquisition unit 340, the inference information acquisition unit 550 includes, for example, numerical analysis information and voice analysis information analyzed from the sensor detection information and worker explanation information on the date and time when the information was newly acquired. Obtain information for inference. Note that the inference information is information that can be displayed in the form of a table shown in FIG. 4(B). The inference information is, for example, information in which the date and time information includes numerical analysis information IND1 and audio analysis information IMD1 of the date and time when the information was newly acquired. Further, when a first device and a second device are provided as devices in a factory, the inference information acquisition unit 550 acquires first inference information, which is inference information of the first device, and inference information of the second device. The second inference information is obtained.

The inference result information generation unit 560 inputs inference information to the learned model and generates inference result information. Specifically, the inference result information generation unit 560 infers which cluster the inference information is classified into using the trained model, and determines whether the equipment will malfunction based on multiple types of data included in the inferred clusters. Generates inference result information indicating whether or not the Here, the inference result information is information including, for example, any one of information such as the probability that the device is malfunctioning, the type of failure when the device is malfunctioning, and the degree of failure. Note that the inference result information is information that can be displayed in the form of a table shown in FIG. 4(C). The inference result information is, for example, information in which the date and time information includes failure degree information RMD1 indicating the degree of failure at the date and time when the information was newly acquired. Further, when a first device and a second device are provided as devices in the factory, the inference result information generation unit 560 inputs the first inference information to the first learned model and generates the inference result of the first device. First inference result information, which is information, is generated. Furthermore, in this case, the inference result information generation unit 560 inputs the second inference information to the second trained model and generates second inference result information that is the inference result information of the second device.

The inference result information output unit 570 outputs inference result information by displaying the inference result information on the display unit 55. Furthermore, when a first device and a second device are provided as devices in the factory, the inference result information output unit 570 displays the first inference result information and the second inference result information on the display unit 55.

(About trained model generation processing according to Embodiment 1)
Next, the operation of the learning device 300 to generate and output a trained model will be described using a flowchart. When the learning device 300 is powered on, it starts executing the learned model generation process shown in FIG. 5 . First, the trained model generation unit 350 causes the learning information acquisition unit 340 to read the learning information stored in the learning information storage unit 420 to acquire learning information (step S101). Here, the learning information is information including numerical analysis information obtained by analyzing the sensor detection information acquired by the learning information acquisition section 340 and audio analysis information obtained by analyzing the worker explanation information, and the learning information output section 360 is the information output to the storage device 400 and stored in the learning information storage section 420.

Next, the learned model generation unit 350 generates a learned model by machine learning using all the acquired learning information (step S102). Specifically, the trained model generation unit 350 generates a trained model that classifies all the learning information (x _i (i=1, 2, . . . , n)) into one of k clusters. The trained model output unit 370 then outputs the generated trained model to the storage device 400 (step S103), and ends the process. Note that the trained model output by the trained model output unit 370 is stored in the trained model storage unit 410 of the storage device 400.

(Regarding inference result information generation processing according to Embodiment 1)
Next, the operation of the failure prediction device 500 to generate and output inference result information will be described using a flowchart. When the failure prediction device 500 is powered on, it starts executing the inference result information generation process shown in FIG. 6 . First, the trained model acquisition unit 510 acquires a trained model by reading out the trained model stored in the trained model storage unit 410 (step S201). Next, the sensor detection information acquisition unit 520 acquires the sensor detection information output from the sensor 100 (step S202), and the position information acquisition unit 530 acquires the worker's position information output from the mobile terminal 200. (Step S203).

Next, the worker explanation information determination unit 535 determines whether the information acquired together with the location information from the mobile terminal 200 is worker explanation information based on the location information (step S204). Specifically, the worker explanation information determination unit 535 determines whether the information acquired from the mobile terminal 200 is worker explanation information by determining whether the distance from the worker to the device is smaller than a threshold value. Determine whether If it is determined that the information is not worker explanation information (step S204; N), the failure prediction device 500 returns to step S202 and repeats the processing of steps S202 to S204 until it is determined that it is worker explanation information.

On the other hand, if it is determined that the information is worker explanation information (step S204; Y), the worker explanation information acquisition unit 540 acquires the information acquired together with the position information from the mobile terminal 200 as worker explanation information (step S205). ), the inference information acquisition unit 550 acquires inference information based on the newly acquired sensor detection information and worker explanation information (step S206). Here, the inference information is information including numerical analysis information obtained by analyzing sensor detection information and audio analysis information obtained by analyzing worker explanation information.

Next, the inference result information generation unit 560 inputs inference information to the learned model and generates inference result information (step S207). Specifically, the inference result information generation unit 560 uses the trained model to infer to which cluster of k clusters the inference information belongs, and infers to which of the k clusters the inference information belongs, and infers the multiple types of data included in the cluster to which it belongs. Based on this, inference result information indicating whether or not the device is out of order is generated. Then, the inference result information output unit 570 displays the generated inference result information on the display unit 55 (step S208), and ends the process.

As described above, according to the failure prediction system 1 according to the present embodiment, in the learning device 300, the learning information acquisition unit 340 acquires learning information based on the sensor detection information and the worker explanation information. . Further, the learned model generation section 350 generates a learned model by machine learning using learning information acquired in advance, and the learned model output section 370 outputs the learned model. Further, in the failure prediction device 500, the learned model acquisition unit 510 acquires the learned model, the sensor detection information acquisition unit 520 acquires sensor detection information from the sensor 100, and the worker explanation information acquisition unit 540, Obtain worker explanation information from the mobile terminal 200. Then, the inference result information generation unit 560 inputs inference information based on the newly acquired sensor detection information and worker explanation information into the trained model to generate inference result information, and the inference result information output unit 570 generates inference result information. , outputs inference result information.

By doing so, the learning device 300 can generate a learned model by machine learning using learning information based on sensor detection information and worker explanation information. That is, the learning device 300 considers not only the state of the equipment detected by the sensor 100 but also the state of the equipment explained by the worker who uses the equipment to determine the relationship between the equipment state and equipment failure. You can learn. Therefore, the failure prediction device 500 can predict equipment failures using a trained model that incorporates new elements such as the intuition and knowledge of skilled workers who actually handle equipment at the site. As a result, the failure prediction system 1 according to the present embodiment is superior to a failure prediction system in which the learning device does not generate a learned model by machine learning using learning information based on sensor detection information and worker explanation information. equipment failures can be predicted with high accuracy.

Further, according to the failure prediction system 1 according to the present embodiment, the worker explanation information is information including audio information in which the worker explains the state of the equipment.
By doing so, the mobile terminal 200 can generate worker explanation information by recording the worker's utterances while working with the device. As a result, in the failure prediction system 1 according to the present embodiment, the burden on the worker for generating the worker explanation information can be reduced more than in a failure prediction system in which the worker explanation information does not include audio information.

Further, according to the failure prediction system 1 according to the present embodiment, in the failure prediction device 500, the worker explanation information determination unit 535 determines that the information acquired from the mobile terminal 200 based on the position information is the worker explanation information. Determine whether or not. Then, when it is determined that the information is worker explanation information, worker explanation information acquisition section 540 acquires the information obtained from mobile terminal 200 as worker explanation information.
By doing so, in the failure prediction system 1 according to the present embodiment, it is possible to generate and acquire inference information based on the information acquired by the failure prediction device 500 from the mobile terminal 200. can be identified.

[Embodiment 2]
In the first embodiment, the learning device 300 generates the learned model by unsupervised learning, but the present invention is not limited to this. In the failure prediction system 1 according to the second embodiment, the learning device 300 generates a learned model by supervised learning. Hereinafter, the failure prediction system 1 according to the second embodiment will be described in detail with reference to FIGS. 3, 7, and 8. Note that in the second embodiment, configurations that are different from those in the first embodiment will be described, and descriptions of the same configurations as in the first embodiment will be omitted because they are redundant.

(About the learning device 300 according to the second embodiment)
As shown in FIG. 7, the learning device 300 according to the second embodiment further includes a failure record information acquisition unit 380 that acquires failure record information that is information indicating the failure record of the device.

(About storage device 400 according to Embodiment 2)
The storage device 400 according to the second embodiment further includes a failure history information storage unit 430 that stores failure history information.

(About failure prediction device 500 according to Embodiment 2)
Failure prediction device 500 according to the second embodiment further includes a learning information output unit 580 that outputs learning information.

(About the hardware configuration of the learning device 300 according to the second embodiment)
As shown in FIG. 3, the transmitting/receiving section 56 according to the second embodiment functions as the failure record information acquisition section 380 shown in FIG. 7. In the learning device 300, the control section 51 uses the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transmitting/receiving section 56 as resources, so that the function of the failure history information acquisition section 380 shown in FIG. Realize. For example, the learning device 300 executes a failure record information acquisition step performed by the failure record information acquisition unit 380.

(About the hardware configuration of failure prediction device 500 according to Embodiment 2)
The transmitting/receiving section 56 according to the second embodiment functions as the learning information output section 580 shown in FIG. 7. In the failure prediction device 500, the function of the learning information output unit 580 shown in FIG. Realize. For example, the learning device 300 executes a learning information output step performed by the learning information output unit 580.

(Details of functional configuration of learning device 300 according to Embodiment 2)
The failure record information acquisition unit 380 acquires failure record information by reading the failure record information stored in the failure record information storage unit 430. Here, the failure record information is, for example, information generated by input from a user such as an administrator, worker, or equipment manager of the failure prediction system 1 on a terminal owned by the user, and output from the terminal. be.

The learning information acquisition unit 340 acquires learning information based on sensor detection information, operator explanation information, and failure record information. Note that the learning information is information that can be displayed in the form of a table shown in FIG. 9(A). The learning information is, for example, information including numerical analysis information LND1, voice analysis information LMD1, and failure degree information LMA1 included in failure record information, which have the same date and time information. The learning information acquisition unit 340 acquires learning information including, for example, the above-mentioned numerical analysis information, audio analysis information, and failure record information. Further, the learning information acquisition unit 340 acquires new learning information output from the failure prediction device 500.

The learned model generation unit 350 generates a learned model for outputting optimal inference result information based on failure record information from the inference information by machine learning using learning information acquired in advance. The learned model generation unit 350 generates a learned model by machine learning using a plurality of types of learning information acquired in advance and stored in the learning information storage unit 420. Here, the machine learning method performed by the trained model generation unit 350 is a neural network as an example of supervised learning.

Here, supervised learning refers to a method in which a data set of input and a label as a result is given to the learning device 300, and the characteristics in the learning information are learned, and the result is inferred from the input. . Further, a neural network is composed of an input layer including a plurality of neurons, an intermediate layer also called a hidden layer including a plurality of neurons, and an output layer including a plurality of neurons. In addition, although the intermediate layer is one layer, it may be two or more layers.

For example, in a three-layer neural network as shown in FIG. 8, when multiple inputs are input to the input layer (X1, X2, X3), the first weight W1 (w11, w12,... , w16) and input to the intermediate layer (Y1, Y2), and the result is further multiplied by the second weight W2 (w21, w22, ..., w26) and output from the output layer (Z1, Z2, Z3). Ru. Therefore, the output result changes depending on the values of each weight W1 and W2.

In this embodiment, the neural network learns inference result information based on failure record information by so-called supervised learning according to the learning information acquired by the learning information acquisition unit 340. That is, by adjusting each weight W1 and W2, the neural network inputs inference information based on sensor detection information and worker explanation information into the input layer, and converts the results output from the output layer into failure record information. Learn to get closer. The learned model generation unit 350 generates a learned model by performing the above learning.

(Details of functional configuration of failure prediction device 500 according to Embodiment 2)
The inference result information generation unit 560 inputs a plurality of types of inference information into the learned model and generates inference result information based on failure record information. Specifically, the inference result information generation unit 560 generates input values indicating multiple types of inference information in the input layer (X1, X2, X3), middle layer (Y1, Y2), output layer (Z1, Inference result information is generated based on the output values that have changed depending on the values of the weights W1 and W2 through the weights W1 and W2.

The learning information output unit 580 outputs new learning information based on the inference information and inference result information to the learning device 300. In addition, the new learning information specifically includes numerical analysis information and audio analysis information based on sensor detection information and worker explanation information included in the inference information, and inference result information similar to the failure record information. This is the information that it contains. Note that the new learning information is information that can be displayed in the form of a table shown in FIG. 9(B). The new learning information is, for example, information in which the date and time information includes numerical analysis information IND1, voice analysis information IMD1, and failure degree information RMD1 included in the inference result information.

As explained above, according to the failure prediction system 1 according to the present embodiment, in the learning device 300, the learning information acquisition unit 340 acquires the learning information based on the sensor detection information, the worker explanation information, and the failure record information. Get information. Then, the learned model generation unit 350 generates a learned model for outputting optimal inference result information based on the failure record information from the inference information by machine learning using the learning information acquired in advance.

By doing so, the failure prediction device 500 can predict equipment failures using a learned model that has learned the history of equipment failures through supervised learning. As a result, the failure prediction system 1 according to the present embodiment is capable of failure prediction in which the learning device does not generate a learned model by machine learning using learning information based on sensor detection information, worker explanation information, and failure record information. It can predict failures of factory equipment with better accuracy than the system.

Further, according to the failure prediction system 1 according to the present embodiment, in the failure prediction device 500, the learning information output unit 580 outputs new learning information to the learning device 300 based on the inference information and the inference result information. Output. Then, the learning information acquisition unit 340 acquires new learning information output from the failure prediction device 500 in the learning device 300, and the learned model generation unit 350 generates multiple types of learning information including the new learning information. Generate a trained model by machine learning using learning information.

By doing so, the learning device 300 can perform machine learning by feeding back the results inferred by the failure prediction device 500. As a result, the failure prediction system 1 according to the present embodiment is more effective at preventing factory equipment failures than a failure prediction system that does not generate a trained model through machine learning using multiple types of learning information including new learning information. can be predicted with high accuracy.

[Example of change]
In the first and second embodiments described above, the devices 100, 300 to 500 are separate devices, but they may be integrated. For example, all the devices 100, 300 to 500 may be integrated into one device. Good too. Furthermore, for example, the learning device 300 and the storage device 400 may be integrated, and the remaining devices 100 and 500 may be separate devices. Further, for example, the learning device 300 and the failure prediction device 500 may be integrated into one device, and the remaining devices 100 and 400 may be separate devices.

Note that in the first and second embodiments described above, the sensor 100, the mobile terminal 200, the learning device 300, the storage device 400, and the failure prediction device 500 are capable of transmitting and receiving data via wireless LAN. The configuration is not limited to this. For example, while the sensor 100, the learning device 300, the storage device 400, and the failure prediction device 500 can transmit and receive information via a communication cable that connects them to each other, the mobile terminal 200 and each device 100, 300 to 500 can communicate with each other via the Internet. It may also be possible to send and receive information via the network. Further, for example, while the learning device 300, the storage device 400, and the failure prediction device 500 can transmit and receive information via a LAN, the sensor 100 or the mobile terminal 200 and each device 300 to 500 can communicate via the Internet. It may also be possible to send and receive information. In this case, for example, the learning device 300, the storage device 400, and the failure prediction device 500 may function as a so-called cloud server. In this case, the cloud server may generate and store the learned model by machine learning using learning information based on information acquired from the sensor 100 and the mobile terminal 200. Further, in this case, the cloud server may input inference information based on information acquired from the sensor 100 and the mobile terminal 200 into the learned model, and generate and output inference result information.

Note that in the first embodiment described above, the K-means method is employed as the machine learning method performed by the trained model generation unit 350, but the method is not limited to this, and unsupervised learning different from the K-means method may be employed. Good too. For example, known algorithms for unsupervised learning include the shortest distance method as an example of hierarchical clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) as an example of density-based clustering, principal component analysis, and self-organization. It is also possible to employ a conversion mapping or the like. Further, in the second embodiment described above, a neural network is employed as a machine learning method performed by the trained model generation unit 350, but the method is not limited to this, and supervised learning different from a neural network may be employed. . For example, deep learning, which is a well-known algorithm for supervised learning, may be employed. Note that a method different from the above-described unsupervised learning and supervised learning may be employed, and for example, known algorithms such as semi-supervised learning and reinforcement learning may be employed.

Note that in the first and second embodiments described above, learning information is generated and acquired using sensor detection information and worker explanation information with the same date and time information, but sensor detection information and worker explanation information with similar date and time information are used to generate and acquire learning information. Learning information may be generated and acquired using the information. For example, the learning information may be generated and acquired using sensor detection information and worker explanation information in which the date and time difference specified from the date and time information is smaller than a predetermined threshold.

Note that in the first and second embodiments described above, the learning information is information that includes numerical analysis information and voice analysis information, but the learning information is acquired based on sensor detection information and worker explanation information. It is not limited to this as long as it is. For example, the learning information may be information regarding a failure of a device identified based on sensor detection information and worker explanation information. Further, for example, the learning information may be information including sensor detection information and worker explanation information.

Note that in the first and second embodiments described above, the learning device 300 generates and acquires the learning information by itself based on the information acquired from the sensor 100 provided in the failure prediction system 1 and the mobile terminal 200. , but not limited to. For example, the learning information may be acquired from a learning device of another failure prediction system that performs machine learning using the learning information. For example, the learning device 300 may acquire learning information from multiple failure prediction systems operating in the same area, or acquire learning information from failure prediction systems operating independently in different areas. You may. In this case, the learning device 300 may add or remove another failure prediction system that acquires learning information at any timing.

Note that in the first and second embodiments described above, the learning device 300 provided in advance in the failure prediction system 1 performs machine learning only on learning information based on information acquired from the sensor 100 and the mobile terminal 200 to generate a learned model. generated and outputted, but is not limited to this. For example, the learning device 300 of the failure prediction system 1 is a learning device that acquires learning information generated based on information acquired from sensors and mobile terminals provided in another failure prediction system and performs machine learning. The trained model may be updated and output by performing relearning by generating and acquiring learning information based on information acquired from the mobile terminal 100 and the mobile terminal 200.

In the first and second embodiments described above, the learning device 300 generates a plurality of different learned models for each device, and the failure prediction device 500 uses the plurality of learned models to predict the failure of each device. Forecasts include, but are not limited to: For example, the learning device 300 generates one learned model that indicates the relationship between the state of each device and the failure of each device, and the failure prediction device 500 uses one learned model to predict the failure of each device. You can predict it.

Note that in the first and second embodiments described above, the failure prediction device 500 acquires the trained model generated and output by the learning device 300 provided in the failure prediction system 1 and stored in the storage device 400. , but not limited to. For example, the failure prediction device 500 may acquire a learned model generated and output by another learning device or another failure prediction system. Specifically, the failure prediction device 500 may acquire learned models from other failure prediction systems operating in the same area, or may acquire learned models from other failure prediction systems operating independently in different areas. You may also obtain a trained model from .

In the first and second embodiments described above, the inference result information is displayed on the display unit 55, but the output of the inference result information may be changed as long as it can be confirmed by the administrator or operator of the failure prediction system 1. but not limited to. For example, when the inference result information includes information indicating the probability that the device is malfunctioning, if the probability exceeds a predetermined threshold, the mobile terminal 200 and the failure prediction device 500 owned by the administrator A warning email may be sent to a different management terminal, or an alarm may be sounded in a factory electrically connected to the failure prediction device 500.

Note that in the first and second embodiments described above, the worker explanation information is information that includes audio information and date and time information, but is not limited to this. For example, if the mobile terminal 200 is smart glasses, the worker explanation information may include image information of what is in front of the worker's line of sight in addition to audio information and date and time information. In this case, the learning device 300 may perform machine learning by acquiring learning information including image analysis information obtained by analyzing image information included in the worker explanation information. Further, in this case, the failure prediction device 500 may perform inference by acquiring inference information including image analysis information obtained by analyzing image information included in the worker explanation information.

Note that, as in Embodiments 1 and 2 above, it is preferable that the worker explanation information includes audio information in order to reduce the burden on the worker, but the type of worker explanation information depends on the equipment. The information does not need to include audio information as long as it is information that the worker has explained about the condition. For example, the worker explanation information may be information such as character information input by the worker using a keyboard or image information of handwritten characters drawn using a touch panel.

Note that in the second embodiment described above, the failure record information acquisition unit 380 acquires failure record information from the storage device 400, but is not limited to this, and for example, failure record information is acquired from the terminal owned by the user described above. Performance information may be obtained directly.

Note that in the second embodiment described above, the failure prediction device 500 learns information including numerical analysis information and audio analysis information based on inference information and inference result information based on failure record information as new learning information. Although the output is to the device 300, the present invention is not limited thereto. For example, the failure prediction device 500 outputs information including numerical analysis information and audio analysis information based on the inference information and failure record information on the date and time when the inference information was acquired to the learning device 300 as new learning information. It's okay. For example, the failure prediction device 500 may convert information including numerical analysis information and voice analysis information based on the inference information, failure record information on the date and time when the inference information was acquired, and inference result information into new learning information. It may also be output to the learning device 300 as a.

Note that the core part that performs processing of the learning device 300 and failure prediction device 500, which includes the control section 51, main storage section 52, external storage section 53, operation section 54, transmitting/receiving section 56, internal bus 50, etc., is a dedicated Regardless of the system, it can be realized using a normal computer system. For example, a computer program for executing the above operations is stored and distributed in a computer-readable recording medium, such as a flexible disk or a DVD-ROM (Read-Only Memory), and the computer program is loaded onto a computer. By installing it, the learning device 300 and failure prediction device 500 that execute the above processing may be configured. Further, the learning device 300 and the failure prediction device 500 may be configured by storing the computer program in a storage device included in a server device on a communication network and downloading it by a normal computer system.

In addition, when the functions of the learning device 300 and the failure prediction device 500 are realized by sharing the functions between an OS (Operating System) and an application program, or when realized by cooperation between the OS and an application program, only the application program part may be stored in a recording medium or storage device.

It is also possible to superimpose a computer program on a carrier wave and provide it via a communication network. For example, the computer program may be posted on a bulletin board system (BBS) on a communication network, and the computer program may be provided via the network. Then, the above processing may be executed by starting this computer program and executing it under the control of the OS in the same way as other application programs.

The present disclosure is capable of various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Further, the embodiments described above are for explaining the present disclosure, and do not limit the scope of the present disclosure. In other words, the scope of the present disclosure is indicated by the claims rather than the embodiments. Various modifications made within the scope of the claims and the meaning of the disclosure equivalent thereto are considered to be within the scope of the present disclosure.

Hereinafter, various aspects of the present disclosure will be collectively described as supplementary notes.

(Additional note 1)
A failure prediction system that predicts failures of factory equipment,
a sensor that detects the state of the device;
a mobile terminal carried by a worker who performs work using the device;
a learning device that learns the relationship between the state of the device and a failure of the device;
an inference device that infers a failure of the device based on a state of the device;
Equipped with
The learning device includes:
Obtaining learning information based on sensor detection information that is information that the sensor has detected the state of the equipment, and worker explanation information that is information that the worker has explained about the state of the equipment that has been generated by the mobile terminal. a learning information acquisition unit,
a learned model generation unit that generates a learned model indicating a relationship between the state of the device and a failure of the device by machine learning using the learning information acquired in advance;
a trained model output unit that outputs the trained model;
including;
The inference device includes:
a trained model acquisition unit that acquires the trained model;
a sensor detection information acquisition unit that acquires the sensor detection information from the sensor;
a worker explanation information acquisition unit that acquires the worker explanation information from the mobile terminal;
Inputting inference information based on the newly acquired sensor detection information and the worker explanation information into the trained model to generate inference result information that is information indicating a result of inferring a failure of the equipment. An inference result information generation unit;
an inference result information output unit that outputs the inference result information;
including,
Failure prediction system.
(Additional note 2)
The worker explanation information includes audio information in which the worker explains the state of the equipment by voice.
The failure prediction system described in Appendix 1.
(Additional note 3)
The inference device includes:
a worker explanation information determination unit that determines whether information acquired from the mobile terminal is the worker explanation information based on position information of the worker generated by the mobile terminal;
further including;
The worker explanation information acquisition unit acquires the information obtained from the mobile terminal as the worker explanation information when it is determined that the information acquired from the mobile terminal is the worker explanation information.
The failure prediction system according to appendix 1 or 2.
(Additional note 4)
The learning information acquisition unit acquires the learning information based on the sensor detection information, the worker explanation information, and failure record information that is information indicating a failure record of the equipment,
The learned model generation unit generates the learned model for outputting the inference result information based on the failure record information from the inference information by machine learning using the learning information acquired in advance.
The failure prediction system according to any one of Supplementary Notes 1 to 3.
(Appendix 5)
A failure prediction device that predicts failures of equipment in a factory,
a sensor detection information acquisition unit that acquires sensor detection information that is information on detecting the state of the device from a sensor that detects the state of the device;
a worker explanation information acquisition unit that acquires worker explanation information that is information about the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The newly acquired model is added to a trained model showing the relationship between the state of the device and a failure of the device, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation unit that inputs inference information based on the sensor detection information and the worker explanation information and generates inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output unit that outputs the inference result information;
including,
Failure prediction device.
(Appendix 6)
A learning device for learning the relationship between the state of equipment in a factory and a failure of the equipment, the learning device comprising:
Sensor detection information, which is information obtained by a sensor that detects the state of the equipment, and sensor detection information, which is information about the state of the equipment detected by a sensor that detects the state of the equipment, and sensor detection information, which is information about the state of the equipment, which is generated by a mobile terminal carried by a worker who performs work using the equipment. a learning information acquisition unit that acquires learning information based on worker explanation information that is information explained by the operator;
a learned model generation unit that generates a learned model indicating a relationship between the state of the device and a failure of the device by machine learning using the learning information acquired in advance;
a trained model output unit that outputs the trained model;
A learning device equipped with.
(Appendix 7)
A failure prediction method for predicting failure of equipment in a factory,
a sensor detection information acquisition step in which the computer acquires sensor detection information that is information on detecting the state of the device from a sensor that detects the state of the device;
a worker explanation information acquisition step in which the computer acquires worker explanation information that is information on the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The computer adds a new model to a trained model indicating the relationship between the state of the equipment and a failure of the equipment, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation step of inputting inference information based on the sensor detection information and the worker explanation information acquired in the above, and generating inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output step in which the computer outputs the inference result information;
failure prediction methods including
(Appendix 8)
computer,
a sensor detection information acquisition unit that acquires sensor detection information that is information about the state of the equipment from a sensor that detects the state of the equipment in the factory;
a worker explanation information acquisition unit that acquires worker explanation information that is information about the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The newly acquired model is added to a trained model showing the relationship between the state of the device and a failure of the device, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation unit that inputs inference information based on the sensor detection information and the worker explanation information and generates inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output unit that outputs the inference result information;
A program that functions as

DESCRIPTION OF SYMBOLS 1...Failure prediction system, 50...Internal bus, 51...Control unit, 52...Main storage unit, 53...External storage unit, 54...Operation unit, 55...Display unit, 56...Transmission/reception unit, 59...Control program, 100... Sensor, 110...Sensor detection information generation section, 120...Sensor detection information output section, 200...Mobile terminal, 210...Position information generation section, 220...Worker explanation information generation section, 230...Position information output section, 240...Worker Explanation information output section, 300...Learning device, 310...Sensor detection information acquisition section, 320...Position information acquisition section, 330...Worker explanation information acquisition section, 335...Worker explanation information determination section, 340...Learning information acquisition section , 350... Learned model generation section, 360... Learning information output section, 370... Learned model output section, 380... Failure record information acquisition section, 400... Storage device, 410... Learned model storage section, 420... For learning Information storage unit, 430...Failure record information storage unit, 500...Failure prediction device, 510...Learned model acquisition unit, 520...Sensor detection information acquisition unit, 530...Position information acquisition unit, 535...Worker explanation information determination unit, 540... Worker explanation information acquisition section, 550... Inference information acquisition section, 560... Inference result information generation section, 570... Inference result information output section, 580... Learning information output section.

Claims (8)

A failure prediction system that predicts failures of factory equipment,
a sensor that detects the state of the device;
a mobile terminal carried by a worker who performs work using the device;
a learning device that learns the relationship between the state of the device and a failure of the device;
an inference device that infers a failure of the device based on a state of the device;
Equipped with
The learning device includes:
Obtaining learning information based on sensor detection information that is information that the sensor has detected the state of the equipment, and worker explanation information that is information that the worker has explained about the state of the equipment that has been generated by the mobile terminal. a learning information acquisition unit,
a learned model generation unit that generates a learned model indicating a relationship between the state of the device and a failure of the device by machine learning using the learning information acquired in advance;
a trained model output unit that outputs the trained model;
including;
The inference device includes:
a trained model acquisition unit that acquires the trained model;
a sensor detection information acquisition unit that acquires the sensor detection information from the sensor;
a worker explanation information acquisition unit that acquires the worker explanation information from the mobile terminal;
Inputting inference information based on the newly acquired sensor detection information and the worker explanation information into the trained model to generate inference result information that is information indicating a result of inferring a failure of the equipment. An inference result information generation unit;
an inference result information output unit that outputs the inference result information;
including,
Failure prediction system. The worker explanation information includes audio information in which the worker explains the state of the equipment by voice.
The failure prediction system according to claim 1. The inference device includes:
a worker explanation information determination unit that determines whether information acquired from the mobile terminal is the worker explanation information based on position information of the worker generated by the mobile terminal;
further including;
The worker explanation information acquisition unit acquires the information obtained from the mobile terminal as the worker explanation information when it is determined that the information acquired from the mobile terminal is the worker explanation information.
The failure prediction system according to claim 1. The learning information acquisition unit acquires the learning information based on the sensor detection information, the worker explanation information, and failure record information that is information indicating a failure record of the equipment,
The learned model generation unit generates the learned model for outputting the inference result information based on the failure record information from the inference information by machine learning using the learning information acquired in advance.
The failure prediction system according to claim 1. A failure prediction device that predicts failures of equipment in a factory,
a sensor detection information acquisition unit that acquires sensor detection information that is information on detecting the state of the device from a sensor that detects the state of the device;
a worker explanation information acquisition unit that acquires worker explanation information that is information about the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The newly acquired model is added to a trained model showing the relationship between the state of the device and a failure of the device, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation unit that inputs inference information based on the sensor detection information and the worker explanation information and generates inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output unit that outputs the inference result information;
including,
Failure prediction device. A learning device for learning the relationship between the state of equipment in a factory and a failure of the equipment, the learning device comprising:
Sensor detection information, which is information obtained by a sensor that detects the state of the equipment, and sensor detection information, which is information about the state of the equipment detected by a sensor that detects the state of the equipment, and sensor detection information, which is information about the state of the equipment, which is generated by a mobile terminal carried by a worker who performs work using the equipment. a learning information acquisition unit that acquires learning information based on worker explanation information that is information explained by the operator;
a learned model generation unit that generates a learned model indicating a relationship between the state of the device and a failure of the device by machine learning using the learning information acquired in advance;
a trained model output unit that outputs the trained model;
A learning device equipped with. A failure prediction method for predicting failure of equipment in a factory,
a sensor detection information acquisition step in which the computer acquires sensor detection information that is information on detecting the state of the device from a sensor that detects the state of the device;
a worker explanation information acquisition step in which the computer acquires worker explanation information that is information on the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The computer adds a new model to a trained model indicating the relationship between the state of the equipment and a failure of the equipment, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation step of inputting inference information based on the sensor detection information and the worker explanation information acquired in the above, and generating inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output step in which the computer outputs the inference result information;
failure prediction methods including computer,
a sensor detection information acquisition unit that acquires sensor detection information that is information about the state of the equipment from a sensor that detects the state of the equipment in the factory;
a worker explanation information acquisition unit that acquires worker explanation information that is information about the worker's explanation about the state of the device from a mobile terminal carried by a worker who performs work using the device;
The newly acquired model is added to a trained model showing the relationship between the state of the device and a failure of the device, which is generated by machine learning using learning information based on the sensor detection information and the worker explanation information. an inference result information generation unit that inputs inference information based on the sensor detection information and the worker explanation information and generates inference result information that is information indicating a result of inferring a failure of the equipment;
an inference result information output unit that outputs the inference result information;
A program that functions as

Images