Learning apparatus, learning method, and program

Abstract

A learning apparatus 100 according to the present disclosure comprises: a pseudo label determination unit 121 that determines a pseudo label indicating a remaining life of equipment on the basis of output resulting from inputting non-labeled data comprising time-series operation data indicating the operation state of the equipment, to a first trained model obtained by performing learning by using labeled data comprising a remaining life period and the time-series operation data indicating the operation state of the equipment; a prediction label output unit 122 that uses output resulting from inputting non-labeled data to a second trained model as a prediction label indicating a predicted remaining life predicted of the equipment; and a learning unit 123 that performs learning to obtain the second trained model on the basis of the pseudo label and the prediction label.

Description

Learning device, learning method, and program

[0001] The present disclosure relates to a learning device, a learning method, and a program.

[0002] In order to perform maintenance on various types of equipment, it has been considered to estimate the remaining lifespan of the equipment, which is the period until a failure occurs. For example, Patent Literature 1 discloses a method for predicting the remaining lifespan of a NAND flash memory provided in a numerical control device of a machine tool using a machine learning model. Specifically, Patent Literature 1 generates a machine learning model through supervised learning using supervised data including the actually measured lifespan of the NAND flash memory, i.e., the period until failure.

[0003] Patent No. 6386523

[0004] However, it is difficult to obtain a large amount of supervised data covering the entire lifespan of a device, i.e., data up until a failure occurs. In particular, maintenance may be performed before a failure occurs, making it even more difficult to obtain supervised data. This results in supervised learning using an insufficient amount of supervised data, which poses a problem in that it is not possible to improve the accuracy of the generated machine learning model.

[0005] Therefore, the object of the present disclosure is to provide a learning device that can solve the above-mentioned problem of not being able to improve the accuracy of the machine learning model that is generated when it is difficult to obtain supervised data.

[0006] A learning device according to one embodiment of the present disclosure includes a pseudo label determination unit that determines a pseudo label representing the remaining lifespan of an equipment based on the output of a first learning model trained using labeled data in which the remaining lifespan period for time-series operation data representing the operation status of the equipment is input as unlabeled data, the unlabeled data being the time-series operation data representing the operation status of the equipment; a predicted label output unit that sets the output of a second learning model in which the unlabeled data is input as a predicted label representing the predicted remaining lifespan of the equipment; and a learning unit that trains the second learning model based on the pseudo label and the predicted label.

[0007] Furthermore, a learning method according to one embodiment of the present disclosure is configured as follows: a first learning model is trained using labeled data in which the remaining life span of time-series operation data representing the operation status of the equipment is input with unlabeled data consisting of time-series operation data representing the operation status of the equipment, and a pseudo label representing the remaining life span of the equipment is determined based on the output; a second learning model is trained using the unlabeled data as an input with the output as a predicted label representing the predicted remaining life span of the equipment, and the second learning model is trained based on the pseudo label and the predicted label.

[0008] Furthermore, a program according to one embodiment of the present disclosure has a configuration that causes a computer to execute the following processes: determine a pseudo label representing the remaining lifespan of an equipment based on the output of a first learning model trained using labeled data in which the remaining lifespan period for time-series operating data representing the operating status of the equipment is input as unlabeled data consisting of time-series operating data representing the operating status of the equipment; set the output of a second learning model in which the unlabeled data is input as a predicted label representing the predicted remaining lifespan of the equipment; and train the second learning model based on the pseudo label and the predicted label.

[0009] By being configured as described above, the present disclosure can improve the accuracy of the generated machine learning model even when it is difficult to obtain supervised data.

[0010] FIG. 2 is a block diagram showing the configuration of the information processing device disclosed in FIG. 1. FIG. 3 is a diagram showing the state of processing by the information processing device disclosed in FIG. 1. FIG. 4 is a diagram showing the state of processing by the information processing device disclosed in FIG. 1. FIG. 5 is a diagram showing the state of processing by the information processing device disclosed in FIG. 1. FIG. 6 is a flowchart showing the operation of the information processing device disclosed in FIG. 1. FIG. 7 is a block diagram showing the hardware configuration of an information processing device according to embodiment 2 of the present disclosure. FIG. 8 is a block diagram showing the configuration of an information processing device according to embodiment 2 of the present disclosure.

[0011] First Embodiment A first embodiment of the present disclosure will be described with reference to Fig. 1 to Fig. 7. Fig. 1 is a diagram for explaining the configuration of an information processing device, and Fig. 2 to Fig. 7 are diagrams for explaining the processing operation of the information processing device.

[0012] [Configuration] The information processing device 10 in this embodiment functions as a learning device for generating a model for predicting the remaining lifespan of various devices, which is the period until a failure occurs, and as a prediction device for predicting the remaining lifespan of the devices using the model. For example, devices for which the remaining lifespan is to be predicted include NAND flash memories provided in numerical control devices of machine tools as described in the above-mentioned Patent Document 1, precision devices that can be provided in information processing devices such as hard disks, and rotating machines such as pumps and fans, but any device may be the subject of the remaining lifespan prediction.

[0013] For the equipment whose remaining life is to be predicted, operation data representing actual operating conditions is measured and accumulated in advance. In this embodiment, first, labeled data consisting of operation data and remaining life of equipment that has actually experienced a failure and reached the end of its life is measured. In other words, the labeled data includes time-series operation data including measurement values ​​of multiple types of operating conditions measured over a predetermined period while the equipment is operating, and the remaining life of the equipment from the time the operation data of the equipment was measured until the failure occurred, and the remaining life is associated with the operation data as a label. In addition, in this embodiment, unlabeled data consisting of operation data of equipment for which maintenance was performed before the failure occurred is measured. In other words, the unlabeled data is only time-series operation data including measurement values ​​of multiple types of operating conditions measured over a predetermined period while the equipment is operating, and is not labeled with a label such as remaining life. The unlabeled data is data when no failure has occurred, such as data obtained when a failure is prevented by maintenance.

[0014] Here, if the device is a NAND flash memory, the operational data includes, for example, the number of rewrites, the rewrite interval, the number of reads, the temperature, the manufacturer, the production lot, and ECC performance (ECC: Error Correction Coding). If the device is a hard disk, the operational data includes operational performance information such as S.M.A.R.T. (Self-Monitoring Analysis and Reporting Technology) information and read / write information obtainable via a RAID controller. If the device is a rotating device such as a pump or a fan, the operational data includes acceleration, ultrasonic waves, current, motor torque, distortion, and the like. However, the operational data of the device may be any type of data depending on the device. The labeled data and unlabeled data described above are assumed to be stored in a predetermined storage device.

[0015] The information processing device 10 is composed of one or more information processing devices each having a calculation device and a storage device. As shown in FIG. 1 , the information processing device 10 includes a data collection unit 11, a first learning unit 12, a second learning unit 13, and a prediction unit 14. The functions of the data collection unit 11, the first learning unit 12, the second learning unit 13, and the prediction unit 14 can be realized by the calculation device executing a program for realizing each function stored in the storage device. The information processing device 10 also includes a data storage unit 16 and a model storage unit 17. The data storage unit 16 and the model storage unit 17 are each composed of a storage device. Each component will be described in detail below.

[0016] The data collection unit 11 reads and acquires the labeled data and unlabeled data stored in a predetermined storage device, and stores the labeled data as time-series data and the unlabeled data as maintenance cycle data in the data storage unit 16. Note that the labeled data is data on equipment that has experienced a breakdown and reached the end of its lifespan, and the amount of labeled data acquired may be smaller than that of unlabeled data.

[0017] The first learning unit 12 performs machine learning using the time-series data, which is the labeled data described above, to generate a first learning model called "teacher." At this time, as shown in FIG. 2 , the first learning unit 12 performs supervised learning using the remaining lifespan corresponding to the operation data at a certain time in the time-series data as the teacher data for the operation data. Specifically, the first learning unit 12 calculates a loss function L as shown in the following equation 1 between the predicted value of the remaining lifespan, which is the output when the operation data at a certain time is input to teacher, and the teacher data, which is the remaining lifespan corresponding to the operation data at the input time. ｐｒｅ The first learning unit 12 learns to reduce the remaining life span of a device and updates the teacher's parameters. The first learning unit 12 then stores data such as parameters constituting the first learning model, which is the learned teacher, in the model storage unit 17. As a result, the first learning model, which is the teacher, is configured to input operation data of a device whose remaining life span is unknown and output a predicted value of the remaining life span.

[0018] Bｌ : Index of labeled data

[0019] The first learning unit 12 uses the teacher to calculate the average remaining lifespan μ ｉ and the predictive distribution σ ｉ In this case, for example, a Gaussian distribution is also learned, and the loss function L shown in the following formula 2 is used. ｐｒｅ The teacher may be trained using

[0020] The second learning unit 13 (pseudo label determination unit, predicted label output unit, learning unit) performs machine learning using the maintenance cycle data, which is the unlabeled data described above, to generate a second learning model called "student." At this time, as shown in FIG. 3 , the second learning unit 13 uses the first learning model, which is the teacher described above, to generate pseudo labels based on a predicted remaining lifespan, which is the output when operation data at a certain time in the maintenance cycle data is input to the teacher, and performs supervised learning using the pseudo labels as teaching data. Specifically, the second learning unit 13 first sets the parameters of teacher generated as described above as the initial values ​​of the parameters of student, which is the second learning model to be generated. 3, the second learning unit 13 learns so that the loss function L of the predicted value (predicted label) of the remaining life of the equipment, which is the output when operation data at a certain time in the maintenance cycle data is input to the student, and the pseudo label generated by inputting the operation data at the same time to the teacher as described above, converges, and updates the parameters of the student. As a result, the second learning model, which is the student, is configured to output a predicted value of the remaining life by inputting operation data of the equipment whose remaining life is unknown.

[0021] Here, the function (pseudo label determination unit) of the second learning unit 13 to generate pseudo labels using the maintenance cycle data and teacher will be described. First, the data storage unit 16 stores a plurality of pseudo label candidates (remaining lifespan candidates) that have been set in advance. As shown in FIG. 4 , the pseudo label candidates are data in which the remaining lifespan is set relative to the elapsed time, and each pseudo label candidate has a different terminal remaining lifespan, which is the remaining lifespan that can correspond to the most recent time. For example, FIG. 4 shows pseudo label candidates such as "terminal remaining lifespan 10 years," "terminal remaining lifespan 30 years," and "terminal remaining lifespan 80 years." Furthermore, each pseudo label candidate is configured as data in which the terminal remaining lifespan described above and the remaining lifespan at each time during a predetermined period prior to the most recent time corresponding to the terminal remaining lifespan are set. In this case, the time set on the horizontal axis of the pseudo label candidate corresponds to the time of the maintenance cycle data. For this reason, the remaining lifespan of the pseudo label candidates is set to be long and constant for earlier times further away from the latest time, i.e., for periods when the device has been in operation for a short time, and is set to gradually shorten as the time approaches the latest time for periods closer to the latest time, i.e., for periods when the device has been in operation for a long time. However, the pseudo label candidates are not limited to those shown in Figure 4, and other data may be prepared.

[0022] The second learning unit 13 then selects one of the pseudo label candidates prepared in advance based on the output of the maintenance cycle data input to the teacher, and determines the pseudo label. Specifically, as shown in FIG. 5 , the second learning unit 13 calculates an evaluation value for each pseudo label candidate, representing the likelihood of the pseudo label candidate being a label for the maintenance cycle data. At this time, the second learning unit 13 sets multiple evaluation points P corresponding to each time for the pseudo label candidate for which the evaluation value is to be calculated. As an example, in the example of FIG. 5 , multiple evaluation points P are set for each pseudo label candidate at each time indicated by a black circle, with more evaluation points P being set as the pseudo label candidate approaches the latest time corresponding to the terminal remaining lifespan. Then, for each evaluation point P in the target pseudo label candidate, the second learning unit 13 calculates an evaluation value by comparing the remaining lifespan on the pseudo label candidate at the time corresponding to the evaluation point P with the remaining lifespan that is the output of the maintenance data input to the teacher at the time corresponding to the evaluation point P. For example, the second learning unit 13 calculates the error between the remaining lifespan of the pseudo label candidate and the remaining lifespan that is the output when maintenance data is input to the teacher as an evaluation value. As an example, the second learning unit 13 calculates the root mean squared error (RMSE) of the remaining lifespan for all evaluation points P as the evaluation value. Then, the second learning unit 13 selects the pseudo label candidate with the best evaluation value from the evaluation values ​​calculated for each pseudo label candidate, and determines it as the pseudo label. In the example of FIG. 5, the evaluation value "0.1" is the best, so the pseudo label candidate with a terminal remaining lifespan of 30 years is determined as the pseudo label. In this way, the evaluation value is a value that evaluates the quality of the pseudo label.

[0023] The teacher uses the loss function L shown in Equation 2. ｐｒｅ When the evaluation is performed using the pseudo label candidate, the evaluation value is calculated based on the remaining life y ｐｉ , the average remaining lifespan μ ｐｉ and the predictive distribution σ ｐｉIt may be calculated in the same way as the first term of Equation 2, or may be calculated using the following Equation 3. By calculating the evaluation value in this way, evaluation points that are thought to have high prediction accuracy by the teacher are evaluated more strictly, and evaluation points that are thought to have low prediction accuracy are evaluated more leniently, so that more accurate pseudo label candidates can have better evaluation values. B ｐ : Index of all evaluation points of pseudo label candidates to be evaluated

[0024] In this way, the second learning unit 13 evaluates the pseudo label candidates by comparing the predicted remaining life, which is the output when the unlabeled operation data of the equipment at each evaluation point P corresponding to a plurality of times is input to the teacher, with the remaining life set for the pseudo label candidate at the corresponding time, that is, the evaluation point P. In particular, the more evaluation points are set as the time approaches the latest time corresponding to the terminal remaining life set for the pseudo label candidate, and the output when the maintenance cycle data is input to the teacher is compared and evaluated with the remaining life of the pseudo label candidate, thereby making it possible to determine a more plausible pseudo label for the maintenance cycle data. Note that the method of determining the pseudo label by the second learning unit 13 is not limited to the above-described method. For example, the second learning unit 13 may select pseudo label candidates from a plurality of pseudo label candidates according to the evaluation values ​​as described above, but may not directly determine the selected pseudo label candidates as pseudo labels, but may instead use the selected pseudo label candidates to modify the remaining life of some of the selected pseudo label candidates to generate pseudo labels. For example, the terminal remaining lifespan of the selected pseudo label candidate and multiple terminal remaining lifespans, both large and small, centered on that value, can be fitted to a polynomial with the terminal remaining lifespan value as the explanatory variable and the evaluation value as the objective variable, and the value of the explanatory variable corresponding to the minimum value of the polynomial can be modified to become the terminal remaining lifespan of the selected pseudo label candidate.

[0025] 3, the second learning unit 13 uses the pseudo labels determined as described above as training data to learn the predicted remaining life value, which is the output of inputting the maintenance cycle data to the student, so as to reduce the loss function L as shown in the following equation 4, and updates the parameters of the student. The second learning unit 13 stores data such as parameters that constitute the second learning model, which is the learned student, in the model storage unit 17.

[0026] B ul :, index of unlabeled data

[0027] The second learning unit 13 calculates the average remaining lifespan μ ｓｉ and the predictive distribution σ ｓｉ In this case, for example, a Gaussian distribution is also learned, and the loss function L shown in the following formula 5 may be used to learn the student.

[0028] Furthermore, the second learning unit 13 may add a loss term for labeled data shown in the following equation 6 to the loss function L during learning for the student. This makes it possible to prevent the student from over-fitting to unlabeled data and stabilize the learning results. Here, λ is an adjustment parameter that is determined in advance or adjusted so as to optimize the model evaluation value described below.

[0029] Note that the second learning unit 13 may select a portion of the maintenance cycle data according to the evaluation value when evaluating the pseudo label candidates as described above. For example, if the second learning unit 13 determines that the evaluation value calculated for each evaluation point P in the determined pseudo label is good according to a preset standard (for example, lower than a threshold), it selects the maintenance cycle data for the time corresponding to the evaluation point P. Then, the second learning unit 13 may input only the selected maintenance cycle data to the student, and use the output, that is, the predicted value of remaining life, to learn the student as described above and update the parameters of the student.

[0030] The second learning unit 13 further evaluates the prediction performance of the student after learning in which the parameters have been updated as described above. For example, the second learning unit 13 evaluates the prediction performance of the student using verification data in which the operation data of the equipment and the remaining life span are associated with each other and which are pre-stored in the data storage unit 16. At this time, the second learning unit 13 compares the predicted value of the remaining life span, which is the output when the operation data of the verification data is input to the student, with the remaining life span of the verification data, and stores a model evaluation value, which is the evaluation value of the model. For example, the model evaluation value is the prediction error of the predicted value of the remaining life span, which is the output when the operation data of the verification data is input to the student, relative to the remaining life span of the verification data, and may be the root mean square error. In this way, the model evaluation value is a value that represents the quality of the prediction performance of the second learning model, which is the student.

[0031] Then, the second learning unit 13 evaluates the prediction performance of the student as described above every time the parameters of the student are updated, and changes the parameters of the teacher using the parameters of the student based on the evaluation result. For example, the second learning unit 13 compares a newly calculated model evaluation value for the student with a model evaluation value calculated and stored in the past, and if it is determined that the newly calculated model evaluation value has improved compared to the past, that is, that the model evaluation value has improved according to a preset standard, it changes θ in the following formula (7): ｎｅｗ As shown in the figure, the parameter θ ｓｔｕｄｅｎｔ Using the teacher's parameter θ ｔｅａｃｈｅｒ However, the second learning unit 13 may update the parameters of the teacher by any method. The second learning unit 13 then stores data such as parameters constituting the first learning model, which is the updated teacher, in the model storage unit 17. It is desirable that m is less than 1 and close to 1.

[0032] Furthermore, after the first learning model, which is the teacher, is updated as described above, the second learning unit 13 continues the learning of the student using the teacher, as described above, as shown in Fig. 3. The second learning unit 13 then repeats the above-described learning of the student until a preset number of learning rounds are completed or until the improvement in the model evaluation value of the student converges and the learning is completed, and stores the final student data in the model storage unit 17.

[0033] The prediction unit 14 predicts the remaining life of the equipment using a second learning model that is a learned student stored in the model storage unit 17. That is, the prediction unit 14 inputs operation data of the equipment whose remaining life is unknown to the student, and obtains a predicted value of the remaining life output from the student.

[0034] [Operation] Next, a description will be given of the operation of the above-described information processing device 10. First, the operation of learning the first learning model, which is the teacher, will be described with reference to the flowchart in Fig. 6 and Fig. 2 .

[0035] The information processing device 10 acquires time-series data, which is labeled data (step S1), and inputs operational data at a certain time in the time-series data to a teacher. The information processing device 10 then performs learning to minimize a loss function between the predicted remaining lifespan, which is the output of the teacher, and the teacher data, which is the remaining lifespan corresponding to the operational data at the input time (step S2). After learning, the information processing device 10 updates the teacher's parameters (step S3) and repeats the above-described learning until the learning converges (No in step S4). When the learning converges, the information processing device 10 stores data such as parameters constituting the first learning model, which is the trained teacher.

[0036] Next, the operation of learning the second learning model, which is a student, will be described with reference to the flowchart of FIG. 7 and FIG.

[0037] The information processing device 10 acquires the stored teacher and the maintenance cycle data, which is unlabeled data (step S11), and sets the parameters of the teacher to the initial values ​​of the parameters of the student, which is the second learning model (step S12).

[0038] The information processing device 10 then inputs the maintenance cycle data to the teacher and determines a pseudo label based on the remaining lifespan predicted by the teacher (step S13). The information processing device 10 then compares and evaluates the remaining lifespans of multiple predefined pseudo label candidates with the remaining lifespan predicted by the teacher, thereby determining a pseudo label from the pseudo label candidates. For example, for pseudo label candidates with different terminal remaining lifespans as shown in FIG. 4, evaluation points P are set at multiple times as shown in FIG. 5, and each pseudo label candidate is evaluated by comparing the remaining lifespan of the pseudo label candidate at each evaluation point with the predicted remaining lifespan, which is the output of the maintenance cycle data. The information processing device 10 then determines the pseudo label candidate with the best evaluation value as the pseudo label. In this way, the information processing device 10 determines a pseudo label by calculating an evaluation value representing the quality of the pseudo label using the pseudo label evaluation function (pseudo label evaluation unit) of the second learning unit 13 as described above.

[0039] The information processing device 10 also obtains a predicted value of the remaining life of the equipment, which is the output of inputting the operation data in the maintenance cycle data to the student, and trains the student so that the loss function between this predicted value and the pseudo label determined as described above becomes smaller (step S14). The information processing device 10 then updates the student's parameters after training (step S15). At this time, the information processing device 10 further evaluates the prediction performance of the student after training with the updated parameters. For example, the information processing device 10 compares the predicted value of the remaining life, which is the output of inputting the operation data of the verification data to the student, with the remaining life period of the verification data, and calculates a model evaluation value from the error between these. If the newly calculated model evaluation value is an improvement over the previous value (Yes in step S16), the information processing device 10 updates the teacher's parameters using the student's parameters (step S17). In this way, the information processing device 10 calculates a model evaluation value that represents the quality of the student's predictive performance using the function (model evaluation unit) for evaluating the second learning model, which is the student, possessed by the second learning unit 13 as described above, and determines whether the student is improving.

[0040] Next, the information processing device 10 repeats the above-described learning until the learning converges (No in step S18). When the learning converges, the information processing device 10 stores data such as parameters constituting the second learning model, which is the trained student.

[0041] Thereafter, the information processing device 10 predicts the remaining life of the equipment using the second learning model, which is the student stored by learning as described above. That is, the information processing device 10 inputs the operation data of the equipment whose remaining life is unknown to the student, and obtains the predicted value of the remaining life as the output.

[0042] As described above, according to the information processing device 10 of this embodiment, even in situations where it is difficult to obtain a large amount of labeled data, a first learning model such as "teacher" can be generated from a small amount of labeled data, and then a second learning model such as "student" can be generated from unlabeled data, such as maintenance cycle data, using the first learning model as training data. Specifically, the information processing device 10 can improve the accuracy of the generated model by training the student based on predicted values ​​obtained by inputting unlabeled data into "student" using pseudo labels obtained by inputting unlabeled data into "teacher" as training data. In particular, the information processing device 10 can further improve the accuracy of the student model by evaluating pseudo label candidates using the output of "teacher" as described above and determining pseudo labels, or by updating the parameters of the teacher using the parameters of the trained student model.

[0043] <Embodiment 2> Next, a second embodiment of the present disclosure will be described with reference to Fig. 8 to Fig. 9. Fig. 8 to Fig. 9 are block diagrams showing the configuration of a learning device in embodiment 2. Note that this embodiment shows an outline of the configuration of the learning device described in the above embodiment.

[0044] First, the hardware configuration of the learning device 100 in this embodiment will be described with reference to Figure 8. The learning device 100 is configured as a general information processing device, and is equipped with the following hardware configuration, for example: - CPU (Central Processing Unit) 101 (arithmetic unit) - ROM (Read Only Memory) 102 (storage device) - RAM (Random Access Memory) 103 (storage device) - Programs 104 loaded into RAM 103 - Storage device 105 storing programs 104 - Drive device 106 for reading and writing from / to a storage medium 110 external to the information processing device - Communication interface 107 for connecting to a communication network 111 external to the information processing device - Input / output interface 108 for inputting and outputting data - Bus 109 for connecting each component

[0045] 8 shows an example of the hardware configuration of the information processing device that is the learning device 100, and the hardware configuration of the information processing device is not limited to the above-described case. For example, the information processing device may be configured with a part of the above-described configuration, such as not including the drive device 106. Furthermore, instead of the above-described CPU, the information processing device may use a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating Point Number Processing Unit), a PPU (Physics Processing Unit), a TPU (Tensor Processing Unit), a quantum processor, a microcontroller, or a combination thereof.

[0046] The learning device 100 can be equipped with the pseudo label determination unit 121, predicted label output unit 122, and learning unit 123 shown in FIG. 9 by having the CPU 101 acquire and execute the program group 104. The program group 104 is stored in advance in the storage device 105 or ROM 102, for example, and is loaded into the RAM 103 and executed by the CPU 101 as needed. The program group 104 may be supplied to the CPU 101 via the communication network 111, or may be stored in advance in the storage medium 110, with the drive device 106 reading out the program and supplying it to the CPU 101. However, the pseudo label determination unit 121, predicted label output unit 122, and learning unit 123 described above may be constructed using dedicated electronic circuits for realizing such means.

[0047] The pseudo label determination unit 121 determines a pseudo label representing the remaining life of the equipment based on an output obtained by inputting unlabeled data consisting of time-series operation data representing the operation status of the equipment to a first learning model trained using labeled data consisting of time-series operation data representing the operation status of the equipment and the remaining life span. For example, the pseudo label determination unit 121 determines a pseudo label by evaluating a difference between an output obtained by inputting unlabeled data to the first learning model and a pseudo label candidate prepared in advance.

[0048] The predicted label output unit 122 outputs the unlabeled data input to the second learning model as a predicted label representing the predicted remaining life of the equipment.

[0049] The learning unit 123 learns a second learning model based on the pseudo labels and the predicted labels. At this time, the learning unit 123 may change the parameters of the first learning model using the parameters of the second learning model after learning, and perform further learning.

[0050] By configuring the present disclosure as described above, even in situations where it is difficult to obtain a large amount of labeled data, a first learning model can be generated from labeled data, even if it is a small amount, and by using this, a highly accurate second learning model can be generated even from unlabeled data.

[0051] The above-described program can be stored and supplied to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)). The program may also be supplied to a computer by various types of transitory computer-readable media. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. The transitory computer-readable media can be supplied to a computer via wired communication paths such as electric wires and optical fibers, or via wireless communication paths.

[0052] Although the present disclosure has been described above with reference to the above-described embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that are understandable to those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, at least one or more of the functions of the pseudo label determination unit 121, the predicted label output unit 122, and the learning unit 123 described above may be executed by an information processing device installed and connected anywhere on a network, that is, may be executed by so-called cloud computing.

[0053] <Supplementary Notes> Some or all of the above embodiments may be described as in the following supplementary notes. Below, an outline of the configurations of a learning device, a learning method, and a program according to the present disclosure will be described. However, the present disclosure is not limited to the following configurations. (Supplementary Note 1) A learning device comprising: a pseudo label determination unit that determines a pseudo label that represents a remaining lifespan of an appliance based on an output obtained by inputting unlabeled data consisting of time-series operation data that represents the operation status of the appliance to a first learning model that has been trained using labeled data in which the remaining lifespan periods for time-series operation data that represent the operation status of the appliance are labeled; a predicted label output unit that sets an output obtained by inputting the unlabeled data to a second learning model as a predicted label that represents a predicted remaining lifespan of the appliance; and a learning unit that trains the second learning model based on the pseudo label and the predicted label. (Supplementary Note 2) The learning device according to Supplementary Note 1, wherein the pseudo label determination unit evaluates a plurality of predetermined remaining lifespan candidates based on an output obtained by inputting the unlabeled data to the first learning model, and determines the pseudo label based on the remaining lifespan candidates according to an evaluation result. (Supplementary Note 3) The learning device according to Supplementary Note 2, wherein the plurality of remaining lifespan candidates have different terminal remaining lifespans set, which are remaining lifespans that can correspond to the operational data at the latest time among the unlabeled data, and the pseudo label determination unit evaluates the plurality of remaining lifespan candidates based on an output obtained by inputting the unlabeled data into the first learning model, and determines the pseudo label by selecting from the remaining lifespan candidates according to an evaluation result. (Supplementary Note 4) The learning device according to Supplementary Note 3, wherein the plurality of remaining lifespan candidates have the terminal remaining lifespan and a remaining lifespan at a predetermined period before the latest time corresponding to the terminal remaining lifespan set, and the pseudo label determination unit evaluates the plurality of remaining lifespan candidates based on an output obtained by inputting the unlabeled data at times that represent a plurality of evaluation points into the first learning model, and the remaining lifespans set as the remaining lifespan candidates at the times that represent each of the evaluation points.(Supplementary Note 5) The learning device according to Supplementary Note 4, wherein the pseudo label determination unit sets more evaluation points for each of the remaining life candidates as the remaining life candidate is closer to the latest time, and evaluates the remaining life candidates based on an output obtained by inputting the unlabeled data at the time corresponding to each evaluation point into the first learning model. (Supplementary Note 6) The learning device according to Supplementary Note 4 or 5, wherein the pseudo label determination unit evaluates the remaining life candidates at each evaluation point and selects the unlabeled data based on the evaluation result, and the predicted label output unit inputs the selected unlabeled data into the second learning model and outputs the predicted label. (Supplementary Note 7) The learning device according to any of Supplements 1 to 6, wherein the learning unit evaluates the second learning model based on an output obtained by inputting the operating data, for which a remaining life period has been previously associated, into the second learning model after learning, and the remaining life period associated with the operating data, and changes parameters of the first learning model using parameters set for the second learning model according to the evaluation result. (Supplementary Note 8) The learning device according to Supplementary Note 7, wherein the learning unit changes parameters of the first learning model using parameters set in the second learning model when an evaluation result of the second learning model satisfies a predetermined criterion. (Supplementary Note 9) A learning method comprising: determining a pseudo label representing a remaining lifespan of an equipment based on an output obtained by inputting unlabeled data consisting of time-series operation data representing an operation status of an equipment into a first learning model trained using labeled data in which the remaining lifespan period for time-series operation data representing an operation status of the equipment is labeled; setting an output obtained by inputting the unlabeled data into a second learning model as a predicted label representing a predicted remaining lifespan of the equipment; and training the second learning model based on the pseudo label and the predicted label. (Supplementary Note 10) The learning method according to Supplementary Note 9, wherein a plurality of predetermined remaining lifespan candidates are evaluated based on an output obtained by inputting the unlabeled data into the first learning model, and determining the pseudo label based on the remaining lifespan candidate in accordance with an evaluation result.(Supplementary Note 11) The learning method according to Supplementary Note 10, wherein the plurality of remaining lifespan candidates have different terminal remaining lifespans that are remaining lifespans that can correspond to the operational data at the latest time among the unlabeled data, the plurality of remaining lifespan candidates are evaluated based on an output obtained by inputting the unlabeled data into the first learning model, and the pseudo label is determined by selecting from the remaining lifespan candidates according to the evaluation result. (Supplementary Note 12) The learning method according to Supplementary Note 11, wherein the plurality of remaining lifespan candidates have the terminal remaining lifespan and a remaining lifespan at a predetermined period before the latest time corresponding to the terminal remaining lifespan, and the plurality of remaining lifespan candidates are evaluated based on an output obtained by inputting the unlabeled data at times that represent a plurality of evaluation points into the first learning model and the remaining lifespans set for the remaining lifespan candidates at each of the evaluation points. (Supplementary Note 13) The learning method according to Supplementary Note 12, wherein a larger number of evaluation points are set for each of the plurality of remaining life candidates as the evaluation points become closer to the latest time, and the plurality of remaining life candidates are evaluated based on an output obtained by inputting the unlabeled data at the time corresponding to each evaluation point into the first learning model. (Supplementary Note 14) The learning method according to Supplementary Note 12 or 13, wherein the remaining life candidates are evaluated at each of the evaluation points, and the unlabeled data is selected based on the evaluation result, and the selected unlabeled data is input into the second learning model and the predicted label is output. (Supplementary Note 15) The learning method according to any of Supplements 9 to 14, wherein the second learning model is evaluated based on an output obtained by inputting the operating data, for which a remaining life period has been previously associated, into the second learning model after learning, and the remaining life period associated with the operating data, and parameters of the first learning model are changed using parameters set in the second learning model according to the evaluation result. (Appendix 16) A learning method according to Appendix 15, wherein, when the evaluation result of the second learning model satisfies a predetermined standard, the parameters of the first learning model are changed using the parameters set in the second learning model.(Supplementary Note 17) A computer-readable storage medium storing a program that causes a computer to execute the following processes: determining a pseudo label representing the remaining lifespan of an equipment based on the output of a first learning model trained using labeled data in which the label indicates the remaining lifespan period for time-series operation data representing the operation status of the equipment; determining the output of a second learning model in which the unlabeled data is input as a predicted label representing the predicted remaining lifespan of the equipment; and training the second learning model based on the pseudo label and the predicted label.

[0054] REFERENCE SIGNS LIST 10 Information processing device 11 Data collection unit 12 First learning unit 13 Second learning unit 14 Prediction unit 16 Data storage unit 17 Model storage unit 100 Learning device 101 CPU 102 ROM 103 RAM 104 Program group 105 Storage device 106 Drive device 107 Communication interface 108 Input / output interface 109 Bus 110 Storage medium 111 Communication network 121 Pseudo label determination unit 122 Predicted label output unit 123 Learning unit

Claims

1. a pseudo label determination unit that determines a pseudo label representing the remaining life of the equipment based on the output of a first learning model trained using labeled data in which the remaining life of the time-series operation data representing the operation status of the equipment is input; and a predicted label output unit that outputs the unlabeled data input to a second learning model as a predicted label representing the predicted remaining life of the equipment; and a learning unit that learns the second learning model based on the pseudo label and the predicted label; A learning device equipped with

2. The learning device according to claim 1 , the pseudo label determination unit evaluates a plurality of predetermined remaining life candidates based on an output obtained by inputting the unlabeled data into the first learning model, and determines the pseudo label based on the remaining life candidates in accordance with an evaluation result; Learning device.

3. The learning device according to claim 2, The plurality of remaining lifespan candidates are set to have different terminal remaining lifespans, which are remaining lifespans that can correspond to the latest operational data among the unlabeled data, and the pseudo label determination unit evaluates the plurality of remaining life candidates based on an output of the first learning model when the unlabeled data is input, and selects from the remaining life candidates according to an evaluation result to determine the pseudo label; Learning device.

4. The learning device according to claim 3, The plurality of remaining lifespan candidates are set with the terminal remaining lifespan and a remaining lifespan at a time during a predetermined period before the latest time corresponding to the terminal remaining lifespan, the pseudo label determination unit evaluates the plurality of remaining lifespan candidates based on an output obtained by inputting the unlabeled data at times corresponding to the plurality of evaluation points into the first learning model and a remaining lifespan set as the remaining lifespan candidate at each of the times corresponding to the evaluation point; Learning device.

5. The learning device according to claim 4, the pseudo label determination unit sets a larger number of evaluation points for each of the plurality of remaining life candidates as the evaluation points become closer to the latest time, and evaluates the plurality of remaining life candidates based on an output of the first learning model when the unlabeled data at the time corresponding to each evaluation point is input. Learning device.

6. The learning device according to claim 4, the pseudo label determination unit evaluates the remaining life candidates at each of the evaluation points and selects the unlabeled data based on the evaluation results; the predicted label output unit inputs the selected unlabeled data to the second learning model and outputs the predicted label. Learning device.

7. The learning device according to claim 1 , the learning unit evaluates the second learning model based on an output obtained by inputting the operation data, to which a remaining life span has been previously associated, into the second learning model after learning, and the remaining life span associated with the operation data, and changes parameters of the first learning model using parameters set in the second learning model according to the evaluation result; Learning device.

8. The learning device according to claim 7, the learning unit changes parameters of the first learning model using parameters set in the second learning model when an evaluation result of the second learning model satisfies a preset standard; Learning device.

9. determining a pseudo label representing the remaining life of the equipment based on the output of a first learning model trained using labeled data in which the remaining life of the time-series operation data representing the operation status of the equipment is input; an output obtained by inputting the unlabeled data into a second learning model is used as a predicted label representing the predicted remaining life of the equipment; training the second learning model based on the pseudo labels and the predicted labels; How to learn.

10. determining a pseudo label representing the remaining life of the equipment based on the output of a first learning model trained using labeled data in which the remaining life of the time-series operation data representing the operation status of the equipment is input; an output obtained by inputting the unlabeled data into a second learning model is used as a predicted label representing the predicted remaining life of the equipment; training the second learning model based on the pseudo labels and the predicted labels; A program that causes a computer to perform a process.