Information processing device, inference model generation method, teacher data generation method, inference model generation program, and teacher data generation program

Abstract

According to the invention, the inference model with high universality can be generated. An information processing device (2) is provided with: a data linking unit (204) that links a plurality of pieces of time-series data based on data collected by each of a plurality of facilities and generates pseudo time-series data; a teacher data generation unit (205) that performs standardization processing on the pseudo time series data as teacher data; and a learning unit (206) that generates an inference model by machine learning using the teacher data.

Description

Technical Field

[0001] The present invention relates to an information processing device and the like that generates an inference model. Background Art

[0002] The technology of generating an inference model by machine learning has been known in the past. For example, the following patent document 1 discloses a neural network model that infers the type of event that occurred in a factory based on a plurality of factory data acquired in the factory. In addition, the so-called "plurality of factory data" refers to time series data collected by a factory.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent No. 3002524 Summary of the invention

[0006] Problems to be solved by the invention

[0007] In the prior art described above, there is room for improvement in the versatility of the inference model. That is, in the technology of Patent Document 1, machine learning is performed using factory data collected in the target factory. In the inference model generated by such learning, the type of event generated in the factory can be inferred with high accuracy based on the factory data collected in the factory. However, in the inference model, the type of event generated in other factories cannot be inferred based on the factory data collected from the other factories. Or, even if inference is possible, the accuracy of the inference model becomes low.

[0008] An object of one aspect of the present invention is to provide an information processing device or the like that can generate an inference model with high versatility.

[0009] Technical means of solving problems

[0010] In order to solve the above-mentioned problem, an information processing device of one form of the present invention includes: a data connection unit, which connects multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; a teacher data generating unit, which standardizes or normalizes the pseudo time series data to serve as teacher data; and a learning unit, which generates an inference model through machine learning using the teacher data.

[0011] In addition, in order to solve the above-mentioned problem, another information processing device of one form of the present invention includes: a data connection unit, which connects multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; and a teacher data generating unit, which performs standardization or normalization processing on the pseudo time series data as teacher data.

[0012] In addition, in order to solve the above-mentioned problem, a method for generating an inference model in one form of the present invention is executed by one or more information processing devices, and the method for generating an inference model includes: a data connection step, connecting multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; a teacher data generation step, standardizing or normalizing the pseudo time series data to use as teacher data; and a learning step, generating an inference model through machine learning using the teacher data.

[0013] In addition, in order to solve the above-mentioned problem, a method for generating teacher data in one form of the present invention is executed by one or more information processing devices, and the method for generating teacher data includes: a data connection step, connecting multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; and a teacher data generation step, performing standardization or normalization processing on the pseudo time series data to serve as teacher data.

[0014] Effects of the Invention

[0015] According to one aspect of the present invention, it is possible to generate an inference model with high versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] [ Figure 1 ] is a block diagram showing an example of the main part structure of an information processing device according to one embodiment of the present invention.

[0017] [ Figure 2 ] is a diagram showing the structure of an information processing system including the information processing device.

[0018] [ Figure 3 ] is a diagram showing an example of determining a connection object.

[0019] [ Figure 4 ] is a diagram showing an example of generation of time series data.

[0020] [ Figure 5 ] is a diagram showing an example of generating pseudo time series data and teacher data.

[0021] [ Figure 6 ] is a flowchart showing an example of processing when the information processing device generates an inference model. DETAILED DESCRIPTION

[0022] (System Structure)

[0023] Figure 21 is a diagram showing the structure of an information processing system 5 of the present embodiment. As shown in the figure, the information processing system 5 includes information processing devices 1A to 1D and an information processing device 2. The details will be described below. According to the information processing system 5, an inference model can be generated using a plurality of time series data from different sources. Moreover, the inference model generated by the information processing system 5 becomes an inference model with high versatility compared to an inference model generated using time series data from a single source as teacher data.

[0024] Information processing devices 1A to 1D are devices that collect time series data that are the basis of teacher data for generating the inference model, and are respectively arranged in facilities A to D. Hereinafter, when it is not necessary to distinguish between information processing devices 1A to 1D, they are simply described as "information processing device 1".

[0025] Facility A to Facility D only need to include at least one machine. For example, Facility A to Facility D may also be a factory including a plurality of machines. Here, "factory" is equipment used in industry, including a plurality of machines. The "factory" performs predetermined processing such as production of products or processing of objects through these machines.

[0026] Hereinafter, an example will be described in which facilities A to D are waste treatment facilities that incinerate waste (e.g., combustible garbage) and generate electricity using the exhaust heat. In this case, information processing device 1A collects time series data related to the incineration or power generation of waste in facility A. Similarly, information processing devices 1B to 1D collect time series data related to the incineration or power generation of waste in facilities B to D.

[0027] Time series data can be used as the basis of teacher data for generating an inference model, and can be data corresponding to the inference content, etc. For example, in the case of generating an inference model for predicting the combustion state of an incinerator, it is sufficient to collect various time series data related to the combustion state of the incinerator. If a specific example is given, time series sensor data related to automatic combustion control (ACC) such as time series measurement values ​​of the temperature in the furnace or time series measurement values ​​of the amount of steam generated in the boiler can also be collected as time series data.

[0028] The information processing device 2 connects the time series data collected by the information processing devices 1A to 1D to generate one time series data. Then, the information processing device 2 performs standardization or normalization processing on the connected time series data to obtain teacher data. The information processing device 2 generates an inference model by machine learning using the teacher data.

[0029] The inference model thus generated is more versatile than the case where time series data from a single source is used as teacher data. For example, an inference model generated using teacher data generated from each time series data from facilities A to D can also be used for inference in any one of facilities A to D, and if they are the same type of facilities, they can also be used for inference in facilities other than facilities A to D. In this way, in the generation of the inference model by the information processing device 2, data collected from the facility that is the object of inference is not necessarily required. Therefore, for example, in the case where a new facility is built, the inference model can also be used immediately after its operation begins. In addition, it is not necessarily necessary for all information processing devices 1A to 1D to collect the time series data that serves as the basis of the teacher data. For example, the time series data that serves as the basis of the teacher data is collected from facilities A to C, and on the other hand, the inference model can be generated without collecting the time series data that serves as the basis of the teacher data from facility D. Moreover, the inference model can also be used for inference in facility D.

[0030] In addition, in general, in order to generate an inference model with sufficient inference accuracy, a sufficient amount of teacher data is required, and the data collection period in each facility is also likely to be prolonged. In the above aspect, in the information processing system 5, teacher data is generated based on time series data collected from multiple facilities, so the required amount of teacher data can be prepared in a short period of time and an inference model can be quickly generated.

[0031] In addition, the linked time series data is not limited to data collected by a plurality of different facilities. For example, the information processing device 2 may also link time series data collected for a plurality of different devices or machines in one facility. For example, in the case where the facility A includes two incinerators, the information processing device 2 may link the time series data collected by each incinerator to generate an inference model that can also be used for inference in either of the two incinerators. Therefore, the "facility" in the following description may be replaced by any "object".

[0032] In addition, the method of collecting time series data is also arbitrary. For example, the measured values ​​measured by the sensor installed in the arbitrary object itself can also be collected as time series data. In addition, the measured values ​​measured by the sensor installed around the arbitrary object or the data measured in a wider range such as air temperature and humidity can also be collected as time series data. In addition, for example, the setting value used for the action setting of the arbitrary object or the instruction value when making the arbitrary object perform a specified action can also be collected as time series data.

[0033] (Device Structure)

[0034] based on Figure 1The structures of information processing device 1 and information processing device 2 will be described. Figure 1 1 is a block diagram showing an example of the main structure of the information processing device 1 and the information processing device 2. As shown in the figure, the information processing device 1 includes a control unit 10 that performs overall control of each unit of the information processing device 1, and a storage unit 11 that stores various data used by the information processing device 1. In addition, the information processing device 1 includes: a communication unit 12 for allowing the information processing device 1 to communicate with other devices, an input unit 13 for accepting input of various data for the information processing device 1, and an output unit 14 for causing the information processing device 1 to output various data. In addition, the control unit 10 includes a data acquisition unit 101, a preprocessing unit 102, an inference unit 103, and a control amount determination unit 104.

[0035] The data acquisition unit 101 acquires time series data that is the basis of teacher data for generating an inference model. The time series data may be acquired from a sensor or the like disposed in a facility via the communication unit 12 or the input unit 13, or may be input by a user of the information processing device 1 via the input unit 13. Then, the data acquisition unit 101 transmits the acquired data to the information processing device 2 via the communication unit 12.

[0036] In addition, the data acquisition unit 101 acquires an inference model generated using the data transmitted to the information processing device 2. The inference model may be acquired from the information processing device 2 by communication via the communication unit 12, or may be input by a user of the information processing device 1 via the input unit 13.

[0037] Furthermore, the data acquisition unit 101 acquires data for inference when performing inference using the acquired inference model. Hereinafter, the data is referred to as inference data. The inference data may be input from a sensor or the like disposed in a facility as an inference object via the communication unit 12 or the input unit 13, or may be input by a user of the information processing device 1 via the input unit 13.

[0038] The preprocessing unit 102 preprocesses the inference data acquired by the data acquisition unit 101 to generate input data to the inference model. The preprocessing is a process of standardizing or normalizing the inference data using the same conditions as when generating the teacher data of the inference model in the information processing device 2, which will be described in detail later.

[0039] The inference unit 103 performs inference using the inference model generated by the information processing device 2. More specifically, the inference unit 103 directly uses the output value obtained by inputting the input data generated by the preprocessing unit 102 to the inference model as the inference result, or obtains the inference result based on the output value.

[0040] The inference object of the inference model is not particularly limited, and can be set to predict Figure 2 The combustion state in the incinerator included in the facilities A to D shown. In order to efficiently generate electricity in the facilities A to D, it is necessary to stabilize the amount of steam used to rotate the turbine for power generation. However, the combustion state in the incinerator changes with changes in the quality or amount of the waste to be incinerated, and the amount of steam also changes accordingly. The information processing device 1A to the information processing device 1D predicts the combustion state in the incinerator included in the facilities A to D by using an inference model for predicting the combustion state. The information processing device 1A to the information processing device 1D can appropriately control the equipment in the facilities A to D based on the prediction results, and stably perform waste incineration and power generation.

[0041] The control amount determination unit 104 determines the control amount of the machine installed in the facility as the object based on the inference result of the inference unit 103. The method of determining the control amount varies depending on which inference model is used for inference. For example, it is assumed that an inference result of a reduction in the amount of steam generated is obtained by inference using an inference model that predicts the amount of steam generated by the boiler of the self-incinerator. In this case, the control amount determination unit 104 may also determine the control amount of a machine that affects the amount of steam generated from the boiler (for example, the amount of air supplied to the incinerator or the movement speed of the grate). Of course, in this case, the control amount is determined in such a way as to increase the amount of steam generated. For example, the control amount is determined by increasing the amount of air supplied to the incinerator, or by performing a control to increase the movement speed of the grate.

[0042] On the other hand, the information processing device 2 includes a control unit 20 that performs unified control of each unit of the information processing device 2, and a storage unit 21 that stores various data used by the information processing device 2. In addition, the information processing device 2 includes a communication unit 22 for allowing the information processing device 2 to communicate with other devices, an input unit 23 for accepting input of various data for the information processing device 2, and an output unit 24 for causing the information processing device 2 to output various data. In addition, the control unit 20 includes a data acquisition unit 201, a connection object determination unit 202, a time series data generation unit 203, a data connection unit 204, a teacher data generation unit 205, and a learning unit 206.

[0043] The data acquisition unit 201 acquires data that is the basis of the teacher data. The data acquired by the data acquisition unit 201 may include data that is the explanatory variable of the generated estimation model or data that is the basis of the explanatory variable. For example, when generating an estimation model that uses the time series sensor data collected from each of the facilities A to D as the explanatory variable, the data acquisition unit 201 may communicate with the teacher data via the communication unit 22. Figure 2The information processing devices 1A to 1D shown communicate with each other to acquire the sensing data.

[0044] The connection object determination unit 202 determines the data to be connected by the data connection unit 204 from the data acquired by the data acquisition unit 201. The connection object determination unit 202 is not an essential structure. However, by including the connection object determination unit 202, there is an advantage that even if the data acquired by the data acquisition unit 201 contains data that is not suitable for connection or there is a combination that cannot be connected, appropriate teacher data can be generated.

[0045] The time series data generation unit 203 generates time series data used in the connection by the data connection unit 204. More specifically, the time series data generation unit 203 generates time series data having a difference or ratio between the measured value and the set value as an element, based on the measured value measured by the facility and the set value corresponding to the measured value. The difference or ratio between the measured value and the set value is an explanatory variable in the generated inference model.

[0046] The time series data generation unit 203 is not necessarily a necessary structure. However, by including the time series data generation unit 203, there is an advantage that the deviation of data caused by different facilities can be reduced. Furthermore, linking the data collected by each of multiple facilities has the advantage of reducing the deviation caused by the change of the numerical value due to the time series.

[0047] The data concatenation unit 204 concatenates a plurality of time series data based on data collected by a plurality of facilities to generate one pseudo time series data. The method of generating pseudo time series data will be described in the section "Example of generating pseudo time series data and teacher data" described later.

[0048] The teacher data generation unit 205 performs standardization or normalization processing on the pseudo time series data generated by the data connection unit 204 to obtain teacher data. The method of generating teacher data will also be described in the section "Generation Example of Pseudo Time Series Data and Teacher Data" described later.

[0049] The learning unit 206 generates an inference model by machine learning using the teacher data generated by the teacher data generating unit 205. The machine learning algorithm is not particularly limited, for example, the learning unit 206 may generate an inference model by support vector machine, linear regression, random forest, or neural network.

[0050] As described above, the information processing device 2 includes a data linking unit 204 and a teacher data generating unit 205. The data linking unit 204 links a plurality of time series data based on data collected from each of a plurality of objects to generate a single pseudo time series data. The teacher data generating unit 205 performs standardization or normalization processing on the pseudo time series data generated by the data linking unit 204 to obtain teacher data.

[0051] According to the above structure, it is possible to generate teacher data that can generate an inference model with high versatility. Therefore, according to the above structure, it is possible to provide an inference model with high versatility.

[0052] In addition, as described above, the information processing device 2 includes: a data linking unit 204, a teacher data generating unit 205, and a learning unit 206. The data linking unit 204 links a plurality of time series data based on data collected for each of a plurality of objects to generate a pseudo time series data. The teacher data generating unit 205 performs standardization processing or normalization processing on the pseudo time series data generated by the data linking unit 204 as teacher data. The learning unit 206 generates an inference model by machine learning using the teacher data generated by the teacher data generating unit 205.

[0053] According to the above structure, an inference model can be generated using a plurality of time series data based on data collected for each of a plurality of objects. In addition, according to the above structure, a highly versatile inference model that can be used for inference not only in the facility for which data is collected but also in other facilities can be generated.

[0054] (Example of determining the connection target)

[0055] Figure 3 is a diagram showing an example of determining a connection object. Figure 3 In the example of FIG. 1 , steam quantity sensor data 111A and temperature sensor data 112A are collected in facility A. The collection of these data is performed, for example, by Figure 2 Similarly, steam quantity sensor data 111B and temperature sensor data 112B are collected in facility B, and steam quantity sensor data 111C and temperature sensor data 112C are collected in facility C. These data are collected, for example, by Figure 2 The information processing device 1B and the information processing device 1C shown in the figure perform the above-mentioned data acquisition unit 201 of the information processing device 2. Then, the data acquisition unit 201 of the information processing device 2 acquires the above-mentioned data collected by the information processing devices 1A to 1C.

[0056] The connection target determination unit 202 determines data to be connected by the data connection unit 204 from the data thus acquired. For example, the connection target determination unit 202 may determine data to be connected according to a preset rule and may not connect unrelated data.

[0057] For example, a rule may be set to link data measured by sensors of the same type. Figure 3 As shown, the connection object determination unit 202 determines the steam amount sensor data 111A to 111C including the measurement values ​​measured by the steam amount sensor as the connection object. In addition, the connection object determination unit 202 determines the temperature sensor data 112A to 112C including the measurement values ​​measured by the temperature sensor as the connection object. In addition, the connection object determination unit 202 may also give a common symbol or identification information to the data determined as the connection object.

[0058] In addition, for example, the connection object determination unit 202 may determine the data to be connected according to a rule for connecting data measured by sensors installed at the same position in the facility, based on the rule or in place of the rule. Figure 3 The temperature sensor data 112A and the temperature sensor data 112C in the figure are both measured by a temperature sensor installed near the superheater of the incinerator, while the temperature sensor data 112B is measured by a temperature sensor installed at another location. In this case, the connection object determination unit 202 determines the temperature sensor data 112A and the temperature sensor data 112C measured by the temperature sensors installed at the same location as the connection objects, and does not set the temperature sensor data 112B as the connection object.

[0059] (Example of generating time series data)

[0060] Figure 4 is a diagram showing an example of generation of time series data. More specifically, Figure 4 It is shown in Figure 3 Each of the steam amount sensor data 111A to the steam amount sensor data 111C shown is an example of time series data used for generating teacher data.

[0061] like Figure 4 As shown in FIG. 1 , the steam quantity sensor data 111A to 111C include the measured values ​​(process variables (PV)) at each time. Figure 4The steam quantity sensor data 111A to 111C shown also include set values ​​corresponding to the respective measured values. The set value (set variable (SetVariable, SV)) indicates a value (the steam quantity in the example) that is a target at the time when the set value is established. That is, in the facilities A to C that obtain the steam quantity sensor data 111A to 111C, control is performed so that the steam quantity approaches the set value.

[0062] In this way, the data acquisition unit 201 can also acquire the steam volume sensor data 111A to 111C including the measured values ​​and the set values. In this case, the time series data generation unit 203 can also generate time series data with the difference or ratio between the measured values ​​and the set values ​​as elements, based on the measured values ​​measured by the facilities A to C and the set values ​​corresponding to the measured values ​​contained in the steam volume sensor data 111A to 111C.

[0063] For example, in Figure 4 In the example of , the time series data generation unit 203 generates time series data 113A to 113C using the value of PV / SV at each time, that is, the ratio of the measured value to the set value, as an element based on the steam volume sensor data 111A to the steam volume sensor data 111C. When the value of PV / SV is close to 1, it can be said that the state of the facility is normal. In addition, the time series data generation unit 203 may generate time series data using the difference between PV and SV, that is, the difference between the measured value and the set value, as an element instead of PV / SV.

[0064] In addition, the time series data 113A to 113C include elements of abnormality flags. The value of the abnormality flag at each time point indicates whether the status of the facility A to the facility C is normal. Figure 4 In the example, the value of the abnormal flag is set to 0 in the case of a normal state, and the value of the abnormal flag is set to 1 in the case of an abnormal state. The criterion for determining whether it is normal or not can be determined appropriately. For example, the time series data generation unit 203 may determine that it is abnormal in at least one of the cases where the value of PV / SV is less than a prescribed lower limit value and the value of PV / SV exceeds a prescribed upper limit value, and determine that it is normal in other cases.

[0065] The value of the abnormal flag is the correct answer data in machine learning, in other words, the target variable of the generated inference model. That is, the inference model generated using the time series data 113A to the time series data 113C becomes a model for inferring whether the state of the facility is normal. In addition, the corresponding establishment of the target variable can also be performed by the teacher data generation unit 205.

[0066] Of course, the objective variable of the inference model is arbitrary and is not limited to whether the state of the facility is normal. That is, the time series data generation unit 203 only needs to generate time series data containing any objective variable to be inferred. For example, in the case of generating an inference model for inferring the appropriate control amount for a specified machine in the facility, the time series data generation unit 203 only needs to generate time series data containing the appropriate control amount for the machine. In addition, the objective variable may be automatically determined by the time series data generation unit 203 or input by the user.

[0067] In addition, it is also conceivable that the steam amount sensor data 111A to 111C include a sensor name or a sensor identifier (ID), or the data types of the data included in the steam amount sensor data 111A to 111C are different. In this case, the time series data generation unit 203 may also shape the steam amount sensor data 111A to 111C and convert them into data that can be connected.

[0068] As described above, the time series data generation unit 203 may generate time series data having a difference or ratio between the measured value and the set value as an element for each of a plurality of facilities (objects), based on the measured value measured by the facility and the set value corresponding to the measured value.

[0069] The difference or ratio between the measured value and the set value is an indicator with high universality compared to the measured value measured by the facility. Therefore, according to the structure for generating time series data using such universal indicators as elements, pseudo time series data can be set as natural data. That is, according to the structure, the deviation of data caused by different facilities can be reduced. Furthermore, linking the data collected for each of the multiple facilities can reduce the deviation caused by the change of the numerical value due to the time series. Therefore, according to the structure, an inference model with high inference accuracy can be generated.

[0070] (Example of generating pseudo time series data and teacher data)

[0071] Figure 5 : is a diagram showing an example of generating pseudo time series data and teacher data. More specifically, Figure 5 It shows the Figure 4 The pseudo time series data 114 generated from the time series data 113A to the time series data 113C and the teacher data 115 generated from the pseudo time series data 114 are shown.

[0072] First, the generation of the pseudo time series data 114 by the data concatenation unit 204 will be described. The data concatenation unit 204 sequentially concatenates the time series data 113A to 113C to generate the pseudo time series data 114. However, if the time series data 113A to 113C are simply concatenated directly, the values ​​of "time" may be duplicated or mismatched.

[0073] Therefore, when performing the concatenation, the data concatenation unit 204 replaces the time values ​​in the time series data 113A to the time series data 113C with continuous values ​​of 1 to 15. These values ​​represent the order of each data element contained in the time series data 113A to the time series data 113C, and can be called sequence information. By providing new sequence information instead of the original "time" value, it is possible to avoid mismatching the "time" value while maintaining the continuity of the data.

[0074] Thus, the data concatenation unit 204 may concatenate multiple time series data by assigning sequence information indicating a series of sequences to the multiple time series data. According to the above configuration, multiple time series data can be treated as one pseudo time series data by simply assigning a series of sequence information to the multiple time series data.

[0075] In addition, in general, when an inference model is created using multiple variables with different data distributions, a high-precision inference model can be created by standardizing or normalizing the variables. Here, in order to implement standardization or normalization, statistics of the target data set are required, but the statistics are inherent to the data set. Therefore, in the prior art, it is impossible to connect the time series data after standardization or normalization as a single data, or even if it is possible to connect, learning using such data will significantly reduce the performance of the inference model.

[0076] On the other hand, according to the above configuration, since a plurality of time series data are concatenated in advance as a single pseudo time series data, it is possible to achieve both the normalization process or the normalization process and the concatenation process of the plurality of time series data.

[0077] In addition, the data concatenation unit 204 may concatenate the plurality of time series data after performing processing to reduce the discontinuity of the numerical values ​​of the concatenated parts. Thus, the discontinuity of the numerical values ​​of the concatenated parts can be reduced and the pseudo time series data after concatenation can have continuity, and pseudo time series data with natural numerical changes in the concatenated parts can be generated.

[0078] In addition, according to the above configuration, since the discontinuity of the values ​​of the connected parts in the plurality of time series data can be reduced, general preprocessing applied to time series data such as calculation of moving average can be applied to the pseudo time series data. This can further improve the inference accuracy of the inference model.

[0079] In addition, the process of reducing discontinuity is not particularly limited as long as it can reduce the discontinuity of the numerical value of the connected part in the multiple time series data. For example, as a process of reducing discontinuity, a process of smoothing the numerical value of the connected part in the multiple time series data can also be applied. In addition, for example, as a process of reducing discontinuity, a process of deleting part of the numerical value of each connecting part of the time series data to be connected to reduce the discontinuity can also be applied. In this case, for example, the numerical value can be deleted in a way that the difference between the connecting parts becomes less than a threshold.

[0080] For example, the data concatenation unit 204 may replace the PV / SV value at time = 6 in the pseudo time series data 114 with the average value of the PV / SV value at time = 5. In this case, the PV / SV value at time = 6 is (0.989 + 1.035) / 2 = 1.012, and the change in the PV / SV value in the concatenated portion of the time series data 113A and the time series data 113B can be smoothed.

[0081] Furthermore, for example, the data concatenation unit 204 may delete the PV / SV value at time = 10 in the pseudo time-series data 114. This can smooth the change in the PV / SV value in the concatenated portion of the time-series data 113B and the time-series data 113C.

[0082] Next, the generation of the teacher data 115 by the teacher data generation unit 205 will be described. Figure 5 In the example of , the teacher data generating unit 205 generates the teacher data 115 by performing a normalization process on the pseudo time series data 114. Specifically, the teacher data generating unit 205 uses the data obtained by normalizing each of the three data elements, namely, the measured value, the set value, and the PV / SV, included in the pseudo time series data 114 as the teacher data 115.

[0083] Specifically, the teacher data generating unit 205 generates the teacher data 115 by dividing the difference between the value of the data element contained in the pseudo time series data 114 and the average value by the standard deviation for each data element. For example, in the pseudo time series data 114, the average value of the measured value is 9.387 and the standard deviation is 0.519. Therefore, the teacher data generating unit 205 standardizes the measured value of 9.8 at time = 1 in the pseudo time series data 114 to a value of (9.8-9.387) / 0.519=0.795.

[0084] In addition, the teacher data generating unit 205 may generate teacher data by performing normalization processing instead of standardization processing. In this case, the teacher data generating unit 205 calculates the maximum value and the minimum value of the data elements contained in the pseudo time series data 114. Then, the teacher data generating unit 205 generates teacher data by performing processing on each data element by dividing the difference between the value of the data element and the calculated minimum value by the data range, that is, the difference between the maximum value and the minimum value.

[0085] As described above, the teacher data generation unit 205 generates the teacher data 115 by performing standardization or normalization processing on the pseudo time series data 114. This makes it possible to generate an inference model that absorbs the characteristics of the time series data of each facility, in other words, to generate an inference model that reflects the characteristics of the time series data of each facility.

[0086] In addition, the conditions of the standardization or normalization processing (statistics such as the average, standard deviation, maximum value and minimum value) are also used in the inference using the generated inference model. Therefore, the teacher data generation unit 205 can also notify the conditions of the standardization or normalization processing to the teacher data generation unit 205 through the communication via the communication unit 22. Figure 2 Information processing devices 1A to 1D are shown.

[0087] The learning unit 206 uses the teacher data 115 generated in the above manner to perform machine learning and generate an inference model. The purpose variable of the inference model only needs to be set to the value of the abnormal flag. In addition, the explanatory variable of the inference model only needs to be set to at least any one of the measured value, the set value, and the PV / SV value. In addition, the learning unit 206 may also include values ​​other than these in the explanatory variables. For example, the learning unit 206 may also include at least any one of the measured value, the set value, and the PV / SV value of the time series of the temperature contained in the temperature sensor data 112A in the explanatory variables. In addition, at least any one of the values ​​that have been standardized or normalized may be included in the explanatory variables.

[0088] In addition, Figure 5The teacher data 115 shown does not reflect the data collected by the facility D, but the inference model generated using the teacher data 115 can also be applied to the inference using the inference data collected by the facility D.

[0089] (Flow of Processing Executed by Information Processing Device 2)

[0090] based on Figure 6 The flow of processing executed by the information processing device 2 will be described. Figure 6 1 is a flowchart showing an example of processing when the information processing device 2 generates an inference model. Figure 6 The series of processing shown in the flowchart includes a method for generating teacher data and a method for generating an inference model.

[0091] In S11, the data acquisition unit 201 acquires data obtained in each facility. For example, the data acquisition unit 201 may acquire data obtained in each facility via the information processing devices 1A to 1C. Figure 2 The time-series sensing data collected at facilities A to C are shown. The data collected at each facility may be stored in advance in the storage unit 21 or the like. In this case, the data acquisition unit 201 only needs to acquire the data from the storage unit 21 or the like.

[0092] In S12, the connection target determination unit 202 determines data to be connected by the data connection unit 204 from the data acquired in S11. Figure 3 As described above, the connection target determination unit 202 only needs to determine the data to be connected according to the preset rule and not to set irrelevant data as the connection target.

[0093] In S13, the time series data generation unit 203 generates time series data used in the connection performed by the data connection unit 204 based on the data determined to be the object of connection in S12. Specifically, the time series data generation unit 203 generates time series data with the difference or ratio between the measured value and the set value as an element based on the measured value contained in the data determined to be the object of connection in S12 and the set value corresponding to the measured value. The above processing can also be regarded as a processing of generating an explanatory variable of an inference model using the measured value. In addition, the time series data generation unit 203 can also perform a processing of establishing a correspondence between the value of the target variable of the inference model and the generated time series data in S13.

[0094] In S14 (data concatenation step), the data concatenation unit 204 concatenates a plurality of time series data based on data collected from each of a plurality of facilities to generate one pseudo time series data.

[0095] In S15 (teacher data generation step), the teacher data generation unit 205 performs standardization or normalization processing on the pseudo time series data generated in S14 as teacher data. If the time series data is not associated with the value of the target variable in S13, the teacher data generation unit 205 associates the pseudo time series data with the value in S15.

[0096] In S16 (learning step), the learning unit 206 generates an inference model by machine learning using the teacher data generated in S15, thereby Figure 6 The processing is completed. In addition, the learning unit 206 may also send the generated inference model to the information processing device 1. In addition, at this time, the statistics used in the standardization process or the normalization process in S15 may also be sent.

[0097] In the above description Figure 6 The processing includes a method for generating teacher data. That is, the method for generating teacher data executed by the information processing device 2 includes a data connection step (S14) and a teacher data generation step (S15). The data connection step (S14) connects multiple time series data based on data collected from each of multiple facilities (objects) to generate a pseudo time series data. The teacher data generation step (S15) performs standardization or normalization processing on the pseudo time series data generated in S14 as teacher data. In this way, teacher data that can generate an inference model with high versatility can be generated.

[0098] In addition, Figure 6 The processing also includes a method for generating an inference model. That is, the method for generating an inference model executed by the information processing device 2 includes a data connection step (S14), a teacher data generation step (S15), and a learning step (S16). The data connection step (S14) connects multiple time series data based on data collected from each of multiple facilities (objects) to generate a pseudo time series data. The teacher data generation step (S15) performs standardization or normalization processing on the pseudo time series data generated in S14 as teacher data. The learning step (S16) generates an inference model by machine learning using the teacher data generated in S15. Therefore, an inference model with high versatility can be generated.

[0099] (Flow of Processing Executed by Information Processing Device 1)

[0100] The data acquisition unit 101 of the information processing device 1 acquires the inference model generated in S16 and the statistics used in the normalization process or the regularization process in S15. In addition, the data acquisition unit 101 acquires the inference data collected for the facility to be inferred. For example, if Figure 2The information processing device 1D obtains the inference data collected by the facility D.

[0101] Next, the preprocessing unit 102 uses the statistics to perform standardization or normalization on the inference data to generate input data for the inference model. Then, the inference unit 103 inputs the input data generated by the preprocessing unit 102 into the inference model, and obtains an inference result based on the output value output by the inference model. Then, the control amount determination unit 104 determines the control amount of the equipment installed in the target facility based on the inference result.

[0102] In addition, previously, in Figure 2 In the case where there are multiple facilities, the time series data collected for each facility is standardized or normalized for each facility, teacher data is generated for each facility, and an inference model dedicated to each facility is generated. Therefore, in each facility, when inference is performed using inference data, it is necessary to perform standardization or normalization using the statistics used in the facility.

[0103] In contrast, Figure 2 In the information processing system 5, the information processing device 2 connects the time series data collected for each facility as a pseudo time series data and then performs standardization processing or normalization processing. Therefore, the same statistics can be used to perform standardization processing or normalization processing on the inference data collected at any facility.

[0104] [Variation]

[0105] The execution subject of each process described in each embodiment is arbitrary and is not limited to the above-described examples. That is, as long as each process described in each embodiment can be executed, the devices constituting the information processing system 5 can be changed as appropriate.

[0106] For example, in Figure 2 In the information processing system 5, the information processing device 2 which is different from the information processing devices 1A to 1D generates the teacher data and the inference model, but any one of the information processing devices 1A to 1D can generate the teacher data and the inference model. Figure 2 In the information processing system 5, the information processing apparatuses 1A to 1D collect data in the facilities A to D and perform inference using an inference model, but these processes may be performed by different information processing apparatuses.

[0107] In addition, Figure 2In the information processing system 5, the generation of teacher data and the generation of inference models are performed by one information processing device 2, but the generation of teacher data and the generation of inference models may be performed by different information processing devices, respectively.

[0108] in addition, Figure 6 The execution subject of each of the processes shown does not necessarily need to be a single device, and multiple arbitrary information processing devices (computers) may be made to share and execute these processes. For example, the information processing device 2 may be made to execute Figure 6 The processing of S11 to S15 causes other information processing devices to execute the processing of S16.

[0109] 〔Software-based implementation example〕

[0110] The functions of the information processing device 1 and the information processing device 2 (hereinafter referred to as "devices") are programs for causing a computer to function as the devices, and can be realized by programs for causing each control block (particularly each unit included in the control unit 10 and the control unit 20) of the computer to function as the devices. For example, the function of generating an inference model in the information processing device 2 can be realized by an inference model generation program, and the function of generating teacher data in the information processing device 2 can be realized by a teacher data generation program.

[0111] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The functions described in each embodiment are implemented by executing the program using the control device and the storage device.

[0112] The program may also be recorded in one or more non-temporary and computer-readable recording media. The recording medium may or may not be included in the device. In the latter case, the program may also be supplied to the device via any transmission medium, whether wired or wireless.

[0113] In addition, part or all of the functions of each control block can also be realized by a logic circuit. For example, an integrated circuit having a logic circuit that functions as each control block is also included in the scope of the present invention. In addition, for example, the functions of each control block can also be realized by a quantum computer.

[0114] 〔Summarize〕

[0115] The information processing device of form 1 of the present invention includes: a data connection unit, which connects multiple time series data based on data collected for each of multiple objects to generate a pseudo time series data; a teacher data generating unit, which performs standardization or normalization processing on the pseudo time series data as teacher data; and a learning unit, which generates an inference model through machine learning using the teacher data.

[0116] In the information processing device according to the second aspect of the present invention, in the first aspect, the data concatenation unit may concatenate the plurality of time-series data by providing sequence information indicating a series order to the plurality of time-series data.

[0117] In the information processing device of aspect 3 of the present invention, in aspect 1 or 2, the data concatenation unit may perform concatenation after performing processing to reduce discontinuity of numerical values ​​of concatenated parts in the plurality of time series data.

[0118] In the information processing device of form 4 of the present invention, in any one of forms 1 to 3, it includes a time series data generating unit, which generates the time series data with the difference or ratio between the measured value and the set value as an element for each of the multiple objects based on the measured value measured for the object and the set value corresponding to the measured value.

[0119] The information processing device of form 5 of the present invention includes: a data connection unit, which connects multiple time series data based on data collected from multiple objects to generate a pseudo time series data; and a teacher data generation unit, which performs standardization or normalization processing on the pseudo time series data as teacher data.

[0120] The generation method of form 6 of the present invention is a method for generating an inference model executed by one or more information processing devices, comprising: a data connection step, connecting multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; a teacher data generation step, standardizing or normalizing the pseudo time series data to use as teacher data; and a learning step, generating an inference model through machine learning using the teacher data.

[0121] The generation method of form 7 of the present invention is a method for generating teacher data executed by one or more information processing devices, comprising: a data connection step, connecting multiple time series data based on data collected from each of multiple objects to generate a pseudo time series data; and a teacher data generation step, performing standardization or normalization processing on the pseudo time series data to serve as teacher data.

[0122] The inference model generation program of form 8 of the present invention is used to enable a computer to function as the information processing device described in form 1, and the inference model generation program enables the computer to function as the data connection unit, the teacher data generation unit and the learning unit.

[0123] The teacher data generating program of form 9 of the present invention is used to make a computer function as the information processing device described in form 5, and the teacher data generating program makes the computer function as the data connection unit and the teacher data generating unit.

[0124] The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope of the claims. Embodiments obtained by appropriately combining technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0125] Explanation of symbols

[0126] 2 Information processing device

[0127] 203 Time series data generation unit

[0128] 204 Data Connection Department

[0129] 205 Teacher Data Generation Department

[0130] 206 Learning Department

[0131] 111A, 111B, 111C Steam volume sensor data (measured value, set value)

[0132] 112A, 112B, 112C Temperature sensor data (measured value, set value)

[0133] 113A, 113B Time Series Data

[0134] 114 Pseudo time series data

[0135] 115 Teacher Data

Claims

1. An information processing device, comprising: a data concatenation unit that concatenates a plurality of time series data based on data collected for each of a plurality of objects to generate a single pseudo time series data; a teacher data generating unit that performs standardization processing or normalization processing on the pseudo time series data to obtain teacher data; and The learning unit generates an inference model through machine learning using the teacher data.

2. The information processing device according to claim 1, wherein: The data concatenation unit concatenates the plurality of time-series data by providing sequence information indicating a series of sequences to the plurality of time-series data.

3. The information processing device according to claim 1 or 2, wherein: The data concatenation unit concatenates the plurality of time-series data after performing processing to reduce discontinuity of numerical values ​​of concatenated portions.

4. The information processing device according to claim 1 or 2 includes a time series data generating unit, which generates the time series data with the difference or ratio between the measured value and the set value as an element for each of the multiple objects based on the measured value measured for the object and the set value corresponding to the measured value.

5. An information processing device, comprising: a data concatenation unit that concatenates a plurality of time series data based on data collected for each of a plurality of objects to generate a single pseudo time series data; as well as The teacher data generating unit performs standardization processing or normalization processing on the pseudo time series data to obtain teacher data.

6. A method for generating an inference model, executed by one or more information processing devices, the method for generating an inference model comprising: A data concatenation step of concatenating a plurality of time series data based on data collected from each of a plurality of subjects to generate a pseudo time series data; a teacher data generating step of performing standardization or normalization processing on the pseudo time series data to obtain teacher data; and A learning step generates an inference model through machine learning using the teacher data.

7. A method for generating teacher data, executed by one or more information processing devices, the method for generating teacher data comprising: a data concatenation step of concatenating a plurality of time series data based on data collected for each of a plurality of subjects to generate a pseudo time series data; and The teacher data generating step performs standardization processing or normalization processing on the pseudo time series data to obtain teacher data.

8. An inference model generation program for causing a computer to function as the information processing device as described in claim 1, and the inference model generation program is used to cause the computer to function as the data connection unit, the teacher data generation unit and the learning unit.

9. A teacher data generating program for causing a computer to function as the information processing device as described in claim 5, and the teacher data generating program is used to cause the computer to function as the data linking unit and the teacher data generating unit.

Images