JP2023089671A - Prediction system, prediction method and program - Google Patents

Abstract

To provide a prediction system allowed to predict the replacement time of a volcanic ash filter before a start of use.SOLUTION: A prediction system comprises an acquiring section that acquires a predetermined evaluation condition concerning a volcanic ash filter of an object of evaluation and a predicting section that predicts the replacement time of the volcanic ash filter of the object of evaluation based on a prediction model that, if inputting the evaluation condition, outputs a prediction value of a lifetime of the volcanic ash filter related to the evaluation condition and the evaluation condition the acquiring section acquired.SELECTED DRAWING: Figure 1

Description

The present disclosure relates to prediction systems, prediction methods, and programs. Specifically, the present invention relates to a life prediction system, prediction method and program for volcanic ash filters.

When a volcano erupts, ash fall affects buildings and equipment within the volcanic ash fall area. As one measure against volcanic ash, nuclear power plants install volcanic ash filters in the volcanic ash inflow path to prevent volcanic ash from entering equipment. In order to suppress the impact on the plant when an eruption actually occurs, it is desirable to clean or replace the filter before its performance reaches its limit. For this purpose, it is necessary to know in advance the time until the performance of the filter reaches its limit.

As a related technique, Patent Document 1 describes a filter installed in general air conditioning equipment, based on the differential pressure of the filter, the ambient temperature, the humidity, and the actual usage period of the filter, predicting the life of the filter, and predicting the filter replacement time A diagnostic method is disclosed that provides notification of. The diagnostic method of Patent Document 1 collects online information such as differential pressure and ambient temperature for the filter after the start of use, corrects the period of use of the filter using these parameters, and predicts the life of the filter. It is not intended to predict the life of the filter before starting to use it.

JP 2017-217616 A

There is a need for a technology that predicts the life of volcanic ash filters in advance.

The present disclosure provides a prediction system, prediction method, and program that can solve the above problems.

The prediction system of the present disclosure includes an acquisition unit that acquires predetermined evaluation conditions regarding a volcanic ash filter to be evaluated, and a prediction model that outputs a predicted value of the life of an unused volcanic ash filter according to the evaluation conditions when the evaluation conditions are input. and a prediction unit that predicts the life of the volcanic ash filter to be evaluated based on the evaluation conditions acquired by the acquisition unit.

The prediction method of the present disclosure is executed by a prediction system, acquires a predetermined evaluation condition regarding a volcanic ash filter to be evaluated, and inputs the evaluation condition, a prediction model that outputs a predicted value of the life of the volcanic ash filter according to the evaluation condition. , the life of the volcanic ash filter to be evaluated is predicted by inputting the acquired evaluation conditions.

The program of the present disclosure acquires a predetermined evaluation condition for a volcanic ash filter to be evaluated in a computer, and inputs the evaluation condition to a prediction model that outputs a predicted value of the life of the volcanic ash filter according to the evaluation condition. By inputting the obtained evaluation conditions, a process of predicting the service life of the volcanic ash filter to be evaluated is executed.

According to the prediction system, prediction method, and program described above, it is possible to predict the life of the filter before starting to use it.

It is a block diagram showing an example of a prediction system concerning an embodiment. It is a figure which shows an example of the learning data which concerns on embodiment. It is a figure which shows an example of the principal part of the prediction system which concerns on embodiment. It is a figure which shows an example of the search result of the parameter which concerns on embodiment. It is a figure which shows an example of the prediction model construction log which concerns on embodiment. FIG. 1 shows an example of a scatter diagram of predicted values and measured values according to the embodiment; FIG. 2 is a second diagram showing an example of a scatter diagram of predicted values and measured values according to the embodiment; It is a figure which shows an example of the graph showing transition of the prediction error which concerns on embodiment. It is a flow chart which shows an example of operation of a prediction system concerning an embodiment. It is a figure showing an example of hardware constitutions of a prediction system concerning an embodiment.

A prediction system according to the present disclosure will be described below with reference to FIGS. 1 to 9. FIG.
<Embodiment>
(composition)
FIG. 1 is a block diagram showing an example of a prediction system according to an embodiment. The prediction system 10 is a system for predicting the life of volcanic ash filters. Volcanic ash filter life is when the filter pressure drop reaches a threshold value. Hereinafter, the service life of the volcanic ash filter is also referred to as the filter replacement time. The prediction system 10 includes an input reception unit 11 , a control unit 12 , a storage unit 13 and a display unit 14 .

The input reception unit 11 is configured using input devices such as a keyboard, mouse, touch panel, and buttons. The input reception unit 11 acquires various types of information input using an input device, and outputs the information to the control unit 12 . For example, the input reception unit 11 includes evaluation conditions used for life prediction, such as the particle size distribution pattern of the volcanic ash and the mesh roughness of the volcanic ash filter, and the test conditions and test results of filter tests conducted on the volcanic ash filter in the past. Get information about The test conditions and test results of past filter tests serve as learning data for constructing the prediction model.

The control unit 12 controls the process of predicting the life of the volcanic ash filter. In the learning phase, the control unit 12 builds a prediction model used for life prediction, and in the prediction phase, uses the built prediction model to predict the life of the volcanic ash filter. The control unit 12 includes a prediction model construction unit 121 , a prediction unit 122 and a determination unit 123 . The predictive model construction unit 121 constructs a predictive model that outputs the service life of a volcanic ash filter when inputting an evaluation condition such as a particle size distribution pattern of volcanic ash using test results of past filter tests as learning data. The prediction unit 122 uses a prediction model used for unused life prediction related to evaluation conditions based on learning data accumulated to some extent, and life The service life of the volcanic ash filter to be evaluated is predicted from information such as the evaluation conditions used for prediction and the test conditions of filter tests conducted on past volcanic ash filters. Furthermore, using the prediction model constructed by the prediction model construction unit 121, the life of the volcanic ash filter to be evaluated is predicted. When the learning data is accumulated by performing the filter test, the determination unit 123 reconstructs (or newly constructs) the prediction model in order to reflect the newly accumulated learning data in the prediction model. determine what Also, the control unit 12 outputs the prediction results of the prediction unit 122 to a display device or an electronic file.

The storage unit 13 stores various information necessary for life prediction of the volcanic ash filter. For example, the storage unit 13 stores learning data and constructed prediction models.
The display unit 14 is configured using a display device such as a liquid crystal display.

FIG. 2 shows an example of learning data. In testing the volcanic ash filter, the test data shown in FIG. 2 are recorded. The test data includes "item number" which is the identification number of the test, "customer information" such as customer name and nuclear plant site name, "test date" when the volcanic ash filter was tested, and "test time". , test implementation information such as "temperature" and "humidity" at the time of test implementation, and "particle size 1", "particle size 2", which indicates the particle size (particle size, particle size) of the volcanic ash used in the test, ...・, "particle size N", "concentration" indicating the weight per unit volume of volcanic ash used in the test, whether or not an "upstream labyrinth" is installed on the upstream side of the volcanic ash filter used in the test, and "downstream Presence or absence of installation of "labyrinth", roughness of "mesh" of the volcanic ash filter, and pressure loss which is the value obtained by subtracting the pressure on the downstream side of the filter from the pressure on the upstream side of the filter when sending volcanic ash to the volcanic ash filter at the rated air flow. , and when the volcanic ash specified by the test conditions is supplied to the volcanic ash filter at a predetermined flow rate, the pressure loss of the unused volcanic ash filter is "allowable Includes "allowable pressure drop reaching time" that indicates the time until reaching "pressure drop". "Permissible pressure drop reaching time" is the test result. The particle size of the volcanic ash is associated in advance with the test conditions "particle size 1", "particle size 2", ... "particle size N", and In each item of , the ratio of ash of that particle size contained in volcanic ash is set (particle size distribution pattern). For example, in the example of FIG. 2, in the test of item number 1, 70% of the ash with a particle size of “particle size 1”, 20% of ash with a particle size of “particle size 2”, and 10% of ash with a particle size of “particle size N” This indicates that volcanic ash with a particle size distribution pattern containing a ratio of The test conditions and test results of the test data illustrated in FIG. 2 are used as learning data when constructing the prediction model, the test conditions being explanatory variables and the test results being objective variables. Note that the test conditions illustrated in FIG. 2 are an example. For example, test conditions (explanatory variables) may include volcanic ash flow velocity. When predicting the service life of the volcanic ash filter, the value of each item of the test conditions is input to the prediction model, and the allowable pressure drop reaching time output by the prediction model is obtained as the predicted value of the service life of the volcanic ash filter. The input reception unit 11 acquires the test data collected when the test was performed in the past at the nuclear power plant and stored in the storage unit 13 . The test data may be edited using, for example, general spreadsheet software and stored in the storage unit 13 in a data format managed by the spreadsheet software. By using spreadsheet software, for example, when adding explanatory variables, explanatory variables can be added by adding columns, and when adding learning data, test data in the row direction Learning data can be added by adding

(prediction model construction function)
FIG. 3 is a diagram showing an example of a main part of the prediction system of FIG. 1; A prediction model construction function and a volcanic ash filter life prediction function of the prediction system 10 will be described with reference to FIG. 3 . The prediction model construction unit 121 includes a search condition setting unit 1211, a search unit 1212, a search result display unit 1213, a construction setting unit 1214, a construction unit 1215, a construction result display unit 1216, and a prediction error confirmation unit 1217. , and a prediction error display section 1218 . The storage unit 13 also stores the past test data 131 described with reference to FIG.

The prediction model construction unit 121 searches for how to set the setting values for the machine learning parameters used for building the prediction model, and presents the search results to the user when constructing the prediction model. , a function of constructing a prediction model using a combination of setting values for parameters selected by the user from search results, and a function of presenting the prediction accuracy (prediction error) of the constructed prediction model to the user.

The search condition setting unit 1211 sets a search range for parameter setting values. For example, the prediction model construction unit 121 uses a random forest to construct a prediction model, and performs oversampling by SMOTE (Synthetic Minority Oversampling Technique) as preprocessing, for example. In this case, the search condition setting unit 1211 includes, for example, the number of decision trees as a random forest parameter, the number of k-nearest neighbor data samples as a SMOTE parameter, a sampling threshold, and a threshold for outlier removal as a parameter for prediction results. Set the search range for each parameter. Here, the number of decision trees, the number of k-nearest neighbor data samples, and the sampling threshold are parameters required for each algorithm. The threshold value for outlier removal is to use the learning data once, perform regression learning by random forest, and as a result, obtain the difference between the predicted value of the created trained model and the actual value of the learning data. It is used for the purpose of removing the prediction result when it exceeds the "threshold for removing outliers". For example, the user can specify a value ranging from 20 to 50 for the number of decision trees, a value ranging from 2 to 5 for the number of samples of k-nearest neighbor data, a value ranging from 0 to 60 for the sampling threshold, and a value ranging from 0 to 60 for the outlier removal. Set a value in the range of 0 to 180 for the threshold. It should be noted that the parameters used for constructing the prediction model mentioned here are just an example. For example, regarding a random forest, in addition to the number of decision trees, the depth of decision trees and the number of leaves may be included in the search parameters. In the following, the number of decision trees, the number of k-nearest neighbor data samples, the threshold for sampling, and the threshold for removing outliers are described as parameters 1 to 4, and the values set for each parameter are described as set values 1 to 4. sometimes. For example, the search condition setting unit 1211 sets {15, 25, 35, 45} as the search range for the setting value 1, and {2, 3, 4} as the search range for the setting value 2, according to the user's setting. , the search range for setting value 3 is set to {none, 30, 60}, and the search range for setting value 4 is set to {none, 120, 180}.

The search unit 1212 searches for appropriate setting values 1 to 4 for the search range set by the search condition setting unit 1211 . Specifically, search unit 1212 creates a combination of setting values 1 to 4 set by search condition setting unit 1211 . In the case of the above example, 4×3×3×3 combinations of setting values are created. For each combination, the search unit 1212 performs preprocessing and machine learning using the test conditions and test results of the past test data 131 as learning data, and the prediction accuracy when a prediction model is constructed with the combination of the set values 1 to 4. Evaluate. For example, in the case of a combination of setting value 1 = 25, setting value 2 = 2, setting value 3 = 30, and setting value 4 = 120, the search unit 1212 sets the number of samples of the k nearest neighbor data to 2 and the sampling threshold to 30. to increase the learning data (oversampling) by SMOTE, remove outliers from the learning data after oversampling, and use that data to set the number of decision trees to 25 and perform regression analysis by random forest and calculate the prediction error of each parameter of the trained model. Similarly, search section 1212 calculates the prediction error of each parameter for each combination.

Search result display section 1213 displays the search result by search section 1212 . For example, the search result display unit 1213 displays a predetermined number of combinations of setting values 1 to 4 when the prediction error is small among the search results by the search unit 1212, arranging and displaying them in ascending order of the prediction error. FIG. 4 shows a display example of search results. In the example of FIG. 4, combinations of set values 1 to 4 are displayed in descending order of prediction accuracy. The combination of setting values 1 to 4 with the highest prediction accuracy is 25 for setting value 1, 2 for setting value 2, 30 for setting value 3, 120 for setting value 4, and 120 for setting value 4. It shows that the prediction error when learning with these values set to ˜4 was "26". The lower limit value and upper limit value of the search result are the lower limit value and upper limit value of the "allowable pressure loss reaching time" (objective variable) included in the learning data used for the search. This value indicates the coverage of the prediction model. For example, if the lower limit is 30 (minutes) and the upper limit is 500 (minutes), the prediction result of the filter life predicted using this learning data will be a value in the range of 30 to 500 minutes.

The construction setting unit 1214 receives settings of setting values 1 to 4 used for constructing a prediction model. For example, the user refers to the search results displayed by the search result display unit 1213 and determines combinations of setting values 1 to 4 to be used for construction of the prediction model. For example, the user selects the search No. with the smallest prediction error among the search results illustrated in FIG. The same setting as 1 may be performed. Alternatively, the user adopts a combination (not shown) of setting values 1 to 4 when the range of the lower limit value and the upper limit value is wide (the range of the prediction result is wide) even if the prediction error value is somewhat large. good too. The construction setting unit 1214 sets upper and lower limits of parameters and learning data used for constructing a prediction model as set by the user. For example, the construction setting unit 1214 sets “25”, “2”, “30”, and “120” to the setting values 1 to 4, respectively. In general, it is considered that a more accurate prediction model can be constructed if the amount of learning data is large. Therefore, if it is desired to construct a prediction model with high prediction accuracy even if the prediction range is narrow, adjustment may be made to narrow down the input data to a range in which a large amount of learning data is obtained.

The construction unit 1215 constructs a prediction model (unconstrained) 133 and/or a prediction model (with constraints) 134 using the past test data 131 as learning data using the setting values 1 to 4 set by the construction setting unit 1214. . Here, the prediction model (without constraints) indicates the case where the model is constructed using all data without checking the value of the allowable pressure drop arrival time, and (with constraints) indicates that the value is confirmed and the model A case where a model is constructed by narrowing down the data to be input at the time of construction (for example, 600 (minutes) or less) is shown. For example, when the construction setting unit 1214 makes the above settings, the construction unit 1215 sets the number of samples of the k-nearest neighbor data to 2, sets the sampling threshold to 30, increases the learning data by SMOTE, and after the increase, Data corresponding to outliers are removed from the data, the data is used to perform random forest regression analysis with the number of decision trees set to 25, a prediction model is created, and a prediction error is calculated. The constructing unit 1215 stores the constructed prediction model (unconstrained) 133 in the storage unit 13 . Also, the construction unit 1215 records a log (prediction model construction log 132 ) at the time of constructing the prediction model (no constraint) 133 in the storage unit 13 . FIG. 5 shows an example of data recorded as the prediction model construction log 132. As shown in FIG. As illustrated, the prediction model construction log 132 includes items of No, model name, setting value 1, setting value 2, setting value 3, setting value 4, lower limit value, upper limit value, and prediction error. No is set with the identification number of the constructed prediction model. The model name and date are respectively set with the name of the constructed prediction model and the date when the prediction model was constructed. Other items are the same as the items with the same names in FIG. In addition, the construction unit 1215 records the item number (the item number of the test data in FIG. 2) of the learning data used to construct the prediction model in the storage unit 13, for example, in association with the identification number of the prediction model. . In addition, the construction unit 1215 confirms the value of the "allowable pressure loss arrival time" and narrows down the data to be input when constructing the model (for example, 600 (minutes) or less), and with the same parameter settings, the prediction model (with restrictions ) 134 (first time) and calculate the prediction error of the constructed prediction model (with constraints) 134 . In the case of the above setting example, using only the learning data in the range of 33 to 395 for the "allowable pressure drop arrival time", setting the same values for the setting values 1 to 4, constructing the prediction model (with restrictions) 134. Also for the prediction model (with restrictions) 134, the prediction model construction log 132 is recorded. FIG. 5 illustrates the log for the predictive model (constrained) 134 . The constructing unit 1215 may construct both the prediction model (without constraint) 133 and the prediction model (with constraint) 134, or may construct only one of them.

Furthermore, the building of predictive models can be done iteratively. When building a prediction model by setting the same values to the setting values 1 to 4 for the second and subsequent times, in the second construction, the construction unit 1215 uses the added data as an input, oversampling as in the first time, Outlier removal and machine learning are performed to build a prediction model (unconstrained) 133 (second time) and a prediction model (constrained) 134 (second time), and a prediction model construction log 132 is recorded for each. Henceforth, a prediction model can be similarly constructed. (When the range of learning data is restricted/not required, and the number of times of learning is not required, it is simply referred to as "prediction model.")

The construction result display unit 1216 displays the prediction model construction result on the display unit 14 based on the prediction model construction log 132 recorded by the construction unit 1215 . For example, the construction result display unit 1216 displays on the display unit 14 the log information illustrated in FIG. Also, the construction result display unit 1216 creates a scatter diagram of the predicted values and the measured values by the prediction model, and displays the created scatter diagram on the display unit 14 . FIG. 6A shows an example of a scatter diagram for the prediction model (unconstrained) 133 (nth time), and FIG. 6B shows an example of a scatter diagram for the prediction model (with constraints) 134 (nth time). The vertical axis of the graphs in FIGS. 6A and 6B is the measured value of the "allowable pressure drop arrival time" included in the learning data, and the horizontal axis is the explanatory variable included in the learning data and the "allowable pressure drop arrival time" predicted by the prediction model. Show predicted values. Each point corresponds to a measured value and a predicted value obtained from one piece of training data. The more points that exist near L1, the higher the performance of the prediction model. (L2 shows an example of the predicted value output by the prediction model when the evaluation conditions are input to the constructed prediction model in the prediction phase, and L2 is not displayed in the learning phase.)

The prediction error confirmation unit 1217 reads and acquires the history of prediction errors of the prediction models recorded in the prediction model construction log 132 . For example, the prediction error confirmation unit 1217 reads and acquires the first to fifth prediction errors (58, 34, 38, 30, 28) from the prediction model construction log 132. FIG.

The prediction error display unit 1218 creates a graph representing the transition of the prediction error according to the number of times of learning from the prediction error history information acquired by the prediction error confirmation unit 1217, and displays this graph. FIG. 7 shows an example of a graph representing changes in prediction errors. Thus, even if the set values 1 to 4 are the same, the prediction accuracy can be improved by increasing the learning data each time the construction of the prediction model is repeated.

(Service life prediction function)
Next, the life prediction function of the volcanic ash filter will be explained. The prediction unit 122 includes an evaluation condition setting unit 1221 , a prediction model setting unit 1222 , a filter replacement time prediction unit 1223 and a prediction error display unit 1224 . For example, the storage unit 13 stores N prediction models (without constraints) 133 and/or N prediction models (with constraints) 134 created by building N times, and N prediction models recorded when building The prediction model building log 132 is stored.

The evaluation condition setting unit 1221 records the evaluation conditions input by the user in the storage unit 13 . As a result, the evaluation condition data 135 is saved in the storage unit 13 . Evaluation conditions are explanatory variables illustrated in FIG. For example, when a user creates a test plan for a volcanic ash filter, the user inputs test conditions related to a test to be performed as evaluation conditions. Or, not limited to filter tests, when a volcano near a nuclear power plant erupts and a volcanic ash filter is to be installed, the user can predict the particle size distribution pattern of the volcanic ash and enter evaluation conditions. do. The evaluation condition setting unit 1221 saves the input evaluation conditions in the storage unit 13 .

The prediction model setting unit 1222 displays a list of prediction models constructed by the prediction model construction unit 121 on the display unit 14 . For example, the prediction model setting unit 1222 displays the log information illustrated in FIG. 5 and the graph representing the transition of the prediction error illustrated in FIG. 7 . Then, prediction model setting section 1222 outputs the identification number (model No.) of the prediction model selected by the user from the list to filter replacement time prediction section 1223 .

The filter replacement time prediction unit 1223 predicts the filter replacement time of the volcanic ash filter to be evaluated. The filter replacement time prediction unit 1223 inputs the evaluation condition set by the evaluation condition setting unit 1221 to the prediction model specified by the prediction model setting unit 1222 (prediction model selected by the user from the list), and outputs the prediction model. Obtain the time to reach the allowable pressure drop, and use this value as the filter replacement time (filter life).

The prediction error display section 1224 displays the filter replacement time predicted by the filter replacement time prediction section 1223 on the display section 14 . For example, the prediction error display unit 1224 displays the predicted value of the filter replacement time ( L2 in FIG. 6A and L2 in FIG. 6B) are superimposed and displayed.

(motion)
Next, the flow of prediction model construction processing and filter replacement time prediction processing according to the present embodiment will be described.
FIG. 8 is a flow chart showing an example of the operation of the prediction system.
As a premise, it is assumed that the life of the volcanic ash filter is predicted. First, the control unit 12 determines whether or not to construct a prediction model (step S1). For example, when a prediction model has not been constructed, when learning data has been accumulated to some extent, when the number of explanatory variables has increased, when test data from different types of plants has increased, etc., the user can use the prediction model. It judges to construct and instructs the prediction system 10 to construct a prediction model. The input reception unit 11 notifies the control unit 12 that construction of a prediction model has been instructed. In this case, the control unit 12 determines to build a prediction model. In addition, if the prediction model has already been built, and there is no addition of learning data or changes in the explanatory variables after building the prediction model, or if there is no addition of test data conducted at a different type of plant, the user can The prediction system 10 is instructed to perform filter life prediction. The input reception unit 11 notifies the control unit 12 that the life prediction of the volcanic ash filter has been instructed. In this case, the control unit 12 determines not to build a prediction model.

If it is determined to construct a prediction model (step S1; Yes), the user edits past test data using spreadsheet software (step S2). For example, the user checks whether there are any additions to the explanatory variables and whether there are any additions to the learning data, and adds explanatory variables and learning data. The input reception unit 11 acquires data input by the user, and makes changes or additions to the past test data 131 .

When the editing of the test data (learning data) is completed, the user next sets the search range of parameter setting values used for machine learning and preprocessing (step S3). For example, the user sets {15, 25, 35, 45} etc. for the number of decision trees in the random forest. The user also sets the search range for other parameters. The input reception unit 11 acquires the search range set by the user and outputs it to the search condition setting unit 1211 . The search condition setting unit 1211 records the search range of setting values set by the user in the storage unit 13 .

The user then directs the prediction system 10 to search for setpoints. The input accepting unit 11 accepts the instruction operation of the user and notifies the searching unit 1212 that the search for the setting value has been instructed. Search unit 1212 searches for a setting value (step S4). The search unit 1212 searches for a combination of setting values that provide a good prediction error from within the search range of setting values set for each parameter. Next, the search result display unit 1213 displays a list of search results (step S5). For example, the search result display unit 1213 displays the search results illustrated in FIG. 4 on the display unit 14. FIG.

Next, the user refers to the list of search results and determines setting values to be set for each parameter (step S6). For example, the user confirms the set values 1 to 4 determined for the parameters 1 to 4 and the value of the allowable pressure drop arrival time, and inputs settings to the prediction system 10 for narrowing down the data to be input when constructing the model. The input receiving unit 11 outputs the setting values 1 to 4 input by the user to the construction setting unit 1214 . The construction setting unit 1214 sets the setting value 1 to the parameter 1, sets the setting value 2 to the parameter 2, sets the setting value 3 to the parameter 3, and sets the setting value 4 to the parameter 4. FIG.

The user then instructs the prediction system 10 to build a prediction model. The input reception unit 11 receives the user's instruction operation and notifies the construction unit 1215 that construction of a prediction model has been instructed. Based on this instruction, the construction unit 1215 constructs a prediction model (step S7). The constructing unit 1215 constructs a prediction model (no constraint) 133 by performing preprocessing (oversampling) and regression analysis by random forest based on the setting values 1 to 4. FIG. Also, the construction unit 1215 constructs the prediction model (with constraints) 134 using the data subjected to oversampling and outlier removal. The building unit 1215 also calculates the prediction errors of the prediction model (without constraint) 133 and the prediction model (with constraint) 134 . The constructing unit 1215 stores the constructed prediction model in the storage unit 13 . The constructing unit 1215 records a predictive model construction log 132 for the predictive model (without constraint) 133 and the predictive model (with constraint) 134 . In addition, the constructing unit 1215 records the item number of the learning data used for constructing the prediction model in the storage unit 13 in association with the identification number of the prediction model.

Next, the construction result display unit 1216 displays a scatter diagram of predicted values and measured values (step S8). The construction result display unit 1216 displays the scatter diagrams (excluding L2) illustrated in FIGS. 6A and 6B. The user confirms the prediction errors of the prediction model (without constraint) 133 and the prediction model (with constraint) 134 . If the prediction accuracy is not sufficient, the user can repeat the processes after step S6. For example, as described above, a prediction model may be constructed by setting the same set values 1 to 4 five times in succession until the prediction error becomes equal to or less than a predetermined value.

Next, the prediction error confirmation unit 1217 acquires the history of prediction errors from the prediction model construction log 132 . Next, the prediction error display unit 1218 displays log information based on the prediction model construction log 132 and changes in prediction errors (step S9). For example, the prediction error display unit 1218 displays, on the display unit 14, log information of prediction model construction illustrated in FIG. 5 and a graph showing changes in prediction errors illustrated in FIG. The user can confirm that the prediction error is small and a prediction model with high prediction accuracy is being constructed. Once the prediction model is built, the learning is completed and the life prediction of the volcanic ash filter can be performed.

In step S1, when it determines with not constructing a prediction model (step S1; No), the prediction system 10 performs life prediction of a volcanic ash filter. First, the user sets evaluation conditions using spreadsheet software (step S10). For example, the user sets test conditions for a filter performance test as evaluation conditions. The input reception unit 11 acquires the evaluation conditions input by the user and stores the evaluation condition data 135 in the storage unit 13 .

The user then instructs prediction system 10 to predict the life of the volcanic ash filter under the evaluated conditions. The input accepting unit 11 accepts the user's instruction operation and notifies the prediction model setting unit 1222 that the prediction of the filter life has been instructed. Then, the prediction model setting unit 1222 displays a list of prediction models on the display unit 14 (step S11). For example, the prediction model setting unit 1222 displays a list (log information) of constructed prediction models illustrated in FIG.

Next, the user selects a prediction model to be used for prediction (step S12). The input reception unit 11 outputs the model number of the prediction model selected by the user to the prediction model setting unit 1222 . The prediction model setting unit 1222 outputs the model number of the prediction model selected by the user to the filter replacement time prediction unit 1223 . Next, the filter replacement time prediction unit 1223 predicts the filter replacement time (step S13). The filter replacement time prediction unit 1223 inputs the evaluation condition set in step S10 to the prediction model of the model number output from the prediction model setting unit 1222, and predicts the replacement time of the volcanic ash filter. Next, the prediction error display unit 1224 displays the prediction result (step S14). For example, the prediction error display unit 1224 displays a graph (L2) in which the currently predicted “allowable pressure drop reaching time” is superimposed on the scatter diagram of the measured values and predicted values illustrated in FIGS. 6A and 6B. . The user can grasp the prediction accuracy as well as the predicted value. For example, when the prediction result illustrated in FIG. 6B, that is, when P2 is obtained as a predicted value, it can be expected that the prediction accuracy is high in the vicinity of P2 because the measured values and predicted values are concentrated in the vicinity of L1.

(effect)
As described above, according to this embodiment, the life of the volcanic ash filter can be predicted before use. As a result, for example, the filter replacement time can be predicted in advance at the stage of planning the test of the volcanic ash filter, and the test results for various test condition patterns can be predicted in advance. can be useful for In addition to the filter test, it is possible to predict and understand the life of the filter in advance in the actual volcanic eruption situation, so it is possible to replace the volcanic ash filter at the appropriate time, and the plant can receive it. The impact can be reduced. Further, according to the present embodiment, it is possible to search for parameter setting values that can maintain good prediction accuracy simply by setting test data of past filter tests and a search range for parameters related to machine learning. Since a prediction model can be constructed using the searched setting values, a prediction model with high prediction accuracy can be constructed. Also, the accuracy of the prediction model can be confirmed by referring to the scatter diagrams illustrated in FIGS. 6A and 6B. Also, by limiting the range of learning data with upper and lower limits, it is possible to construct a prediction model according to the prediction range. In addition, it is possible to predict the filter life by selecting an appropriate prediction model from the prediction models created in the past, and to check whether the prediction error learning of the created prediction model is progressing. For example, referring to the graph showing the transition of the prediction error illustrated in FIG. 7, if the prediction error is reduced by repeating the construction of the prediction model, the last constructed prediction model can be selected. Further, for example, when selecting a prediction model with reference to the log information illustrated in FIG. Various prediction results can be obtained for unknown volcanic ash filters by selecting multiple prediction models from prediction models built with training data with a narrow range of upper and lower limits and predicting the filter life. . In addition, when learning data is added, the prediction model can be reconstructed by adjusting the parameters 1 to 4, so a model with a smaller prediction error can be constructed.

<Example 1>
In the description of the above embodiment, the prediction model to be used for prediction is selected from the list of built learning models, and the life prediction of the volcanic ash filter is performed (steps S11 to S13). At the stage of setting (step S10), which prediction model should be used may be proposed. For example, as a design policy for a volcanic ash filter, it is assumed that "flow velocity is XX or less, mesh size is XX or more, and there is no labyrinth upstream or downstream". In this case, when testing this volcanic ash filter, it is preferable to perform life prediction using a prediction model that has learned a lot of learning data that satisfies this condition. In such a case, it is very difficult for the user to manually check the learning data used to construct each prediction model and to identify the prediction model constructed using the learning data containing much data that satisfies the above conditions. is. Therefore, the prediction model setting unit 1222 displays a list of prediction models according to the conditions set by the user. The user sets the priority of each condition. For example, priority 1 is set to "flow velocity is XX or less", priority 2 is set to "mesh XX or more", and priority 3 is set to "no labyrinth". Then, in step S10, the evaluation condition setting unit 1221 records the set priorities 1 to 3 together with the evaluation conditions in the storage unit 13. FIG. Next, in step S12, the predictive model setting unit 1222 selects and displays a list of constructed predictive models, using a learning data group containing many learning data satisfying the conditions set with high priority. Select a predictive model built using For example, the prediction model setting unit 1222 examines the learning data used for building the prediction model based on the item number of the learning data recorded in the storage unit 13 when building the prediction model, and builds each prediction model. For the learning data group used in , the number of learning data that satisfies the highest priority condition of "flow velocity is XX or less" and the percentage of learning data that satisfies the condition are calculated. The predictive model setting unit 1222 extracts a predictive model constructed using a learning data group including many learning data that satisfy the condition of priority 1 . For example, the prediction model setting unit 1222 selects a predetermined number of pieces of learning data satisfying the priority 1 condition in descending order. Alternatively, the prediction model setting unit 1222 selects a predetermined number of learning data in descending order of ratio of learning data satisfying the condition of priority 1 . Similarly, the predictive model setting unit 1222 extracts a predictive model constructed by learning a learning data group containing many learning data that satisfies the condition of “mesh XX or more” given the second highest priority. . The same is true for "no labyrinth", which has the third highest priority. Then, the predictive model setting unit 1222 displays a predetermined number of predictive models constructed using many of the extracted learning data satisfying priority 1 in descending order of the number and/or ratio of the data, and similarly prioritizes 2. A predetermined number of prediction models extracted for priority 3 are also displayed. The predictive model setting unit 1222 displays, together with the extracted predictive model, the number and ratio of data satisfying the conditions included in the learning data used to construct the predictive model. As a result, the user can predict the life of the volcanic ash filter to be evaluated by utilizing the prediction model constructed using many learning data that match the test conditions of the filter test to be performed from now on.

<Example 2>
In the above embodiment, the user determines whether or not to construct (reconstruct) the prediction model in step S1, but the determination unit 123 may automatically perform this determination. The user sets priorities for conditions that trigger model building. For example, the user sets “the number of explanatory variables has changed” as priority 1, sets “the number of test data has increased by a predetermined number or more” as priority 2, and sets “learning has not been performed” as priority 3. "The number of types of nuclear power plants that have not been learned has increased by a predetermined number or more" is set, and as priority 4, "the number of test data in the rainy season that has not been learned has increased by a predetermined number or more" is set. Then, the determination unit 123 acquires the setting information through the input reception unit 11 and sets the priority of the condition that triggers model building in the storage unit 13 . In step S<b>1 , the determination unit 123 determines whether or not to construct a model, and displays the determination result on the display unit 14 . For example, the determination unit 123 determines whether it is necessary to build a model according to the priority. First, for priority 1, the determination unit 123 compares the test conditions, that is, the explanatory variables of the latest test data stored in the storage unit 13 with the explanatory variables of the learning data used to construct the constructed prediction model. to see if the number of explanatory variables has increased. When the number of explanatory variable items increases, the determination unit 123 determines to build a prediction model. If there is no change in the number of explanatory variables, the determination unit 123 determines whether or not to build a prediction model based on priority 2, which has the next highest priority. For example, the determination unit 123 subtracts the number of learning data used to construct the constructed prediction model from the number of all learning data including the latest test data stored in the storage unit 13, and the difference is a predetermined number. If it is the above, it determines with building a prediction model. For priority 3, the determination unit 123 determines the customer information of the test information of the test data corresponding to the item number of the learning data that is not used for building the prediction model, and the customer information for building the prediction model. Compare the customer information of the test data corresponding to the item number of the used learning data, and among the unused test data, customer information (for example, the site name of the nuclear power plant) that is not used for building the prediction model When the number of pieces of test data to be held reaches a predetermined number or more, it is determined to construct a prediction model. If the site name of the nuclear power plant is different, the nearby volcano will be different and the particle size distribution pattern of the volcanic ash will change. Because. Next, the determination unit 123 determines, for priority 4, the customer information of the test information of the test data corresponding to the item number of the learning data that is not used for constructing the prediction model, and the customer information of the prediction model. Compare the customer information of the test data corresponding to the item number of the learning data used for construction, and when the number of data corresponding to the rainy season among the unused test data reaches a predetermined number or more, the prediction model is started. Determine to build. This eliminates the need for the user to confirm the contents of the test data and determine whether or not to construct (reconstruct) the prediction model. In addition, when determining to build a prediction model, the determination unit 123 does not automatically build a prediction model, for example, if it is determined to build a prediction model based on priority 1, "Description Test data in which the number of variables has been changed has been accumulated. Please consider constructing a prediction model." Note that the determination as to whether or not to construct the prediction model of the second embodiment is not limited to step S1 in the flowchart of FIG. such as once).

9 is a diagram illustrating an example of a hardware configuration of a prediction system according to the embodiment; FIG.
A computer 900 includes a CPU 901 , a main memory device 902 , an auxiliary memory device 903 , an input/output interface 904 and a communication interface 905 . The prediction system 10 described above is implemented on a computer 900 . Each function described above is stored in the auxiliary storage device 903 in the form of a program. The CPU 901 reads out the program from the auxiliary storage device 903, develops it in the main storage device 902, and executes the above processing according to the program. Also, the CPU 901 secures a storage area in the main storage device 902 according to the program. In addition, the CPU 901 secures a storage area for storing data being processed in the auxiliary storage device 903 according to the program.

A program for realizing all or part of the functions of the prediction system 10 is recorded on a computer-readable recording medium, and the program recorded on this recording medium is read and executed by a computer system. may be processed by The "computer system" here includes hardware such as an OS and peripheral devices. The "computer system" also includes the home page providing environment (or display environment) if the WWW system is used. The term "computer-readable recording medium" refers to portable media such as CDs, DVDs, and USBs, and storage devices such as hard disks incorporated in computer systems. Moreover, when this program is delivered to the computer 900 via a communication line, the computer 900 receiving the delivery may develop the program in the main storage device 902 and execute the above process. Further, the program may be for realizing part of the functions described above, or may be capable of realizing the functions described above in combination with a program already recorded in the computer system. .

Although several embodiments of the present disclosure have been described above, all these embodiments are provided by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, as well as the scope of the invention described in the claims and equivalents thereof.

<Appendix>
The prediction system, prediction method, and program described in each embodiment are understood as follows, for example.

(1) The prediction system according to the first aspect includes an acquisition unit (input reception unit 11) that acquires a predetermined evaluation condition related to the volcanic ash filter to be evaluated, and an unused value related to the evaluation condition when the evaluation condition is input. A prediction unit 122 that predicts the life of the volcanic ash filter to be evaluated based on a prediction model that outputs a predicted value of the life of the volcanic ash filter and the evaluation conditions acquired by the acquisition unit.
This makes it possible to predict the life of the volcanic ash filter before starting to use it.

(2) The prediction system according to the second aspect is the prediction system of (1), wherein the evaluation conditions include the particle size distribution pattern of the volcanic ash captured by the volcanic ash filter, and the mesh roughness of the volcanic ash filter. , a threshold pressure drop of the volcanic ash filter, and the predictive model predicts the time until the pressure drop of an unused volcanic ash filter reaches the threshold.
This makes it possible to predict the life of the volcanic ash filter before starting to use it.

(3) A prediction system according to a third aspect is the prediction system of (2), wherein the evaluation condition is an explanatory variable, and the pressure loss of the unused volcanic ash filter under the evaluation condition is the threshold The relationship between the evaluation condition and the time until the pressure loss reaches the threshold is determined by machine learning based on learning data with the time measured in a filter test that measures the time to reach the time as an objective variable. It further comprises a prediction model building unit that learns and builds the prediction model.
This makes it possible to build a predictive model that predicts the filter life before the volcanic ash filter is put into use.

(4) A prediction system according to a fourth aspect is the prediction system of (3), wherein the prediction model construction unit searches a predetermined search range for setting values for parameters used in the machine learning. specifying a predetermined number of the setting values in descending order of the prediction error of the prediction model, displaying the prediction error associated with the setting value together with the specified setting value; The setting value selected from is set as the parameter to construct the prediction model.
This makes it possible to grasp the setting values for the parameters with high prediction accuracy. By referring to the set values that can improve the prediction accuracy, the set values that are actually used for constructing the prediction model are determined, so that the prediction model with high prediction accuracy can be constructed.

(5) The prediction system according to the fifth aspect is the prediction system of (3) to (4), wherein when a filter test for measuring the life of the volcanic ash filter is performed, based on the results of the filter test the learning data is generated, the number of evaluation items included in the evaluation conditions in the learning data is changed, the number of learning data increases, the number of plants that have performed the filter test related to the learning data increases, and and a judgment unit that judges whether or not to execute construction of the prediction model based on any one of the conditions of when the test related to the learned data was performed.
As a result, it is possible to automatically determine at what timing a prediction model should be constructed when learning data is accumulated.

(6) The prediction system according to the sixth aspect is the prediction system of (5), wherein the determination unit constructs the latest learning data and the prediction model for the conditions in descending order of priority. Any of the learning data not used in the prediction model is compared with the learning data used in building the prediction model, and the determination is made based on the result of this comparison.
Thereby, it is possible to automatically determine whether or not to construct a prediction model.

(7) The prediction system according to the seventh aspect is the prediction system of (1) to (6), wherein the prediction unit uses the evaluation condition as an explanatory variable, and the unused volcanic ash under the evaluation condition A plurality of the prediction models and the prediction models constructed by varying the parameters of the machine learning in performing machine learning on the learning data with the life measured in the filter test for measuring the life of the filter as the objective variable. is displayed, and the life of the volcanic ash filter is predicted using the prediction model selected from the list.
As a result, the prediction model used for predicting the life of the volcanic ash filter can be selected with reference to the prediction accuracy.

(8) The prediction system according to the eighth aspect is the prediction system of (7), wherein, in displaying the list, the prediction unit uses the learning data used to construct the prediction model for the evaluation A predetermined number of learning data that satisfy a predetermined condition are selected in descending order of the number and displayed in the list.
This makes it easier to select a prediction model suitable for the evaluation conditions related to prediction.

(9) In the prediction method according to the ninth aspect, the prediction system obtains a predetermined evaluation condition regarding the volcanic ash filter to be evaluated, and when the evaluation condition is input, the predicted value of the life of the volcanic ash filter according to the evaluation condition is obtained. The service life of the volcanic ash filter to be evaluated is predicted by inputting the obtained evaluation conditions into the prediction model to be output.

(10) A program according to a tenth aspect acquires a predetermined evaluation condition regarding a volcanic ash filter to be evaluated in a computer 900, and when the evaluation condition is input, outputs a predicted value of the life of the volcanic ash filter according to the evaluation condition. By inputting the obtained evaluation conditions into the prediction model, a process of predicting the service life of the volcanic ash filter to be evaluated is executed.

10 prediction system 11 (acquisition unit) input reception unit 12 control unit 121 prediction model construction unit 1211 search condition setting unit 1212 search unit 1213 search Result display unit 1214 Construction setting unit 1215 Construction unit 1216 Construction result display unit 1217 Prediction error confirmation unit 1218 Prediction error display unit 122 Prediction unit 1221 Evaluation condition setting unit 1222 Prediction model setting unit 1223 Filter replacement time prediction unit 1224 Prediction error display unit 123 Judgment unit 13 Storage unit 14 Display unit 900・Computer 901 CPU
902 Main storage device 903 Auxiliary storage device 904 Input/output interface 905 Communication interface

Claims (10)

an acquisition unit that acquires a predetermined evaluation condition regarding a volcanic ash filter to be evaluated;
Based on a prediction model that outputs a predicted value of the life of an unused volcanic ash filter according to the evaluation condition when the evaluation condition is input, and the evaluation condition acquired by the acquisition unit, the volcanic ash filter to be evaluated. a prediction unit that predicts the life;
A forecasting system with The evaluation conditions include the particle size distribution pattern of the volcanic ash captured by the volcanic ash filter, the mesh roughness of the volcanic ash filter, and the pressure loss threshold of the volcanic ash filter. predicting the time it takes for the filter pressure drop to reach the threshold;
The prediction system of claim 1. Learning using the evaluation condition as an explanatory variable and the time measured in a filter test that measures the time until the pressure loss of the unused volcanic ash filter reaches the threshold based on the evaluation condition as an objective variable A prediction model building unit that builds the prediction model by learning the relationship between the evaluation condition and the time until the pressure loss reaches the threshold based on the data by machine learning;
3. The prediction system of claim 2, further comprising: The prediction model construction unit
Searching a predetermined search range for setting values for the parameters used in the machine learning, and identifying a predetermined number of the setting values in order from the one that minimizes the prediction error of the prediction model,
displaying the prediction error associated with the specified setting value together with the specified setting value;
setting the setting value selected from the identified setting values to the parameter to construct the prediction model;
The prediction system according to claim 3. When a filter test for measuring the life of the volcanic ash filter is performed, the learning data is generated based on the results of the filter test,
Change the number of evaluation items included in the evaluation conditions in the learning data, increase the learning data, increase the number of plants that have performed the filter test related to the learning data, and carry out the filter test related to the increased learning data. a determination unit that determines whether or not to execute the construction of the prediction model based on any condition of time,
5. The prediction system of claim 3 or claim 4, further comprising: The determination unit is
For each of the conditions, one of the latest learning data and the learning data not used for building the prediction model is compared with the learning data used for building the prediction model in descending order of priority. , making said determination based on the result of this comparison;
A prediction system according to claim 5 . The prediction unit
In performing machine learning on learning data with the evaluation condition as an explanatory variable and the life measured in a filter test for measuring the life of the unused volcanic ash filter based on the evaluation condition as the objective variable, the machine learning Displaying a list of the plurality of prediction models constructed with different parameters and prediction errors of the prediction models, and using the prediction model selected from the list to predict the life of the volcanic ash filter do,
The prediction system according to any one of claims 1 to 6. The prediction unit
In displaying the list, for the learning data used to construct the prediction model, a predetermined number of the learning data that satisfies a predetermined evaluation condition are selected in descending order and displayed in the list.
The prediction system of claim 7. the prediction system
Acquiring predetermined evaluation conditions for the volcanic ash filter to be evaluated,
Predicting the life of the volcanic ash filter to be evaluated by inputting the obtained evaluation condition into a prediction model that outputs a predicted value of the life of the volcanic ash filter according to the evaluation condition when the evaluation condition is input,
Forecast method. to the computer,
Acquiring predetermined evaluation conditions for the volcanic ash filter to be evaluated,
A process of predicting the life of the volcanic ash filter to be evaluated by inputting the obtained evaluation conditions into a prediction model that, when the evaluation conditions are input, outputs a predicted value of the life of the volcanic ash filter according to the evaluation conditions. ,
program to run.

Images