A modeling system and method for predicting the remaining life of equipment, and a prediction system

Abstract

The present invention discloses a modeling system for a remaining life prediction model of an equipment, including a basic algorithm library, a database and a general platform. The basic algorithm library is used to store the basic algorithms required in the modeling process of the remaining life prediction model. The database is used to store the data sets required in the modeling process of the remaining life prediction model. The general platform is connected to the basic algorithm library and the database respectively, and is used to call the basic algorithm and the data set to build, train and test the basic model, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type, and the remaining life prediction model is used to implement the remaining life prediction of the equipment. A corresponding modeling method and a remaining life prediction system for nuclear power equipment are also provided. The modeling system can help researchers in the field of nuclear power to quickly and easily establish a remaining life prediction model for specific equipment, and reduce the requirements for the IT technical capabilities of modeling personnel.

Description

Technical Field

[0001] The present invention belongs to the technical field of nuclear power, and in particular relates to a residual life prediction model building system and method for equipment, and a residual life prediction system for nuclear power equipment. Background Art

[0002] Predictive maintenance of nuclear power equipment refers to the prediction and accurate decision-making of maintenance timing and maintenance objects through continuous monitoring and prediction of the status and failure of nuclear power plant systems / equipment, avoiding unplanned downtime caused by sudden equipment failures, maximizing the use of equipment service life, and reducing the storage scale of spare parts required for maintenance. Remaining life prediction is the core key technology for implementing equipment predictive maintenance. Among them, data-driven prediction methods mainly rely on information mined from data for life prediction. For large and complex equipment or systems whose failure mechanisms are difficult to accurately describe, as long as relevant data can be obtained, data-driven models can be used to obtain life prediction results with application value. Therefore, applying data-driven prediction methods to the nuclear industry will greatly improve the safety and economy of nuclear power equipment.

[0003] However, data-driven remaining life prediction requires the ability to analyze and process massive amounts of data. Therefore, it is highly dependent on big data processing technology and intelligent algorithms. As a result, the current research on remaining life prediction technology in the nuclear industry is largely limited by researchers' understanding of intelligent algorithms and programming development capabilities, which limits the development and application of this technology in the nuclear industry. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a remaining life prediction model modeling system and method for equipment, and a remaining life prediction system for nuclear power equipment in view of the above-mentioned deficiencies in the prior art, so that researchers in the nuclear power field can easily and quickly establish a remaining life prediction model for specific nuclear power equipment, thereby reducing the requirements for the IT technical capabilities of modeling personnel.

[0005] An embodiment of the present invention provides a system for modeling a remaining life prediction model of an equipment, including a basic algorithm library, a database, and a general platform. The basic algorithm library is used to store the basic algorithms required in the process of modeling the remaining life prediction model. The database is used to store the data sets required in the process of modeling the remaining life prediction model. The general platform is connected to the basic algorithm library and the database respectively, and is used to call the basic algorithm and the data set to build, train, and test the basic model, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type.

[0006] Preferably, the types of the basic algorithms include data preprocessing algorithms, feature extraction algorithms, data driven algorithms and optimization algorithms.

[0007] The general platform includes a modeling module, a training module, and a test and evaluation module. The modeling module is connected to the basic algorithm library and the database, and is used to divide the called data set into a training set and a test set according to the division ratio, and is used to call the data preprocessing algorithm to preprocess the training set, call the feature extraction algorithm to extract features from the training set after data preprocessing and generate a feature set, and is used to call the data-driven algorithm and the optimization algorithm to obtain the basic model. The training module is connected to the modeling module and is used to apply the basic model to the generated feature set for model training. The test and evaluation module is connected to the modeling module and is used to call the test set to test and evaluate the trained basic model, and to compare and analyze the evaluation results. The modeling module is also used to encapsulate the basic model whose evaluation result exceeds the preset threshold into a remaining life prediction model of a preset file type to meet the integration requirements of the equipment health management platform of the nuclear power plant.

[0008] Preferably, the modeling module includes an algorithm association unit, which is used to define interfaces between various basic algorithms to achieve associations between data preprocessing algorithms, feature extraction algorithms, data-driven algorithms, and optimization algorithms.

[0009] Preferably, the universal platform also includes an algorithm management module. The algorithm management module is connected to the basic algorithm library and is used to delete, modify, or view the basic algorithms in the basic algorithm library, and to add new algorithms to the basic algorithm library by programming or script file import.

[0010] Preferably, the equipment remaining life prediction model modeling system further includes a typical use case library. The typical use case library is connected to the general platform and is used to store the remaining life prediction models of typical failure modes of typical equipment encapsulated by the general platform for users to view or integrate and call.

[0011] Preferably, the general platform also includes a use case management module. The use case management module is connected to the modeling module and is used to view, call, add, delete or modify the basic model obtained by the modeling module.

[0012] Preferably, the modeling module further includes a parameter adjustment unit, which is used to receive parameters input by the user when calling data preprocessing algorithms, feature extraction algorithms, data driven algorithms and optimization algorithms.

[0013] In a second aspect, an embodiment of the present invention further provides a remaining life prediction system for nuclear power equipment, comprising an input module and a remaining life prediction model modeling system for the equipment described in the first aspect. The input module is used to receive a data set of nuclear power equipment input by a user. The remaining life prediction model modeling system for the equipment is connected to the input module and is used to output a remaining life prediction model and a remaining life prediction result of the nuclear power equipment according to the data set of nuclear power equipment input by the user.

[0014] In a third aspect, an embodiment of the present invention also provides a method for modeling a remaining life prediction model for an equipment, comprising: obtaining a basic algorithm required in the process of modeling a remaining life prediction model, and obtaining a data set of an equipment for which a remaining life prediction model is to be established; dividing the data set into a training set and a test set according to a division ratio; constructing and training a basic model according to the basic algorithm and the training set, and testing and evaluating the training basic model according to the test set, and encapsulating the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type.

[0015] The equipment remaining life prediction modeling system of the present invention includes a basic algorithm library, a database and a general platform. The basic algorithm library is used to store the basic algorithms required in the remaining life prediction modeling process. The database is used to store the data sets required in the remaining life prediction modeling process. The general platform is connected to the basic algorithm library and the database respectively, and is used to call the basic algorithms and data sets to build, train, and test the basic model, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type. Nuclear power researchers can use the modeling system to carry out a process-standardized, building block-type remaining life prediction modeling method, and can quickly and easily establish a remaining life prediction model for specific nuclear power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 : This is a structural diagram of a remaining life prediction model building system for a device in Example 1 of the present invention. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0018] Embodiment 1:

[0019] like Figure 1As shown, this embodiment provides a modeling system for the remaining life prediction model of an equipment, which is suitable for establishing a remaining life prediction model, training, testing and packaging for a specific equipment, and is particularly suitable for researchers in the nuclear power field to model, train, test and package a data-driven remaining life prediction model for the nuclear power equipment after obtaining a large amount of operating data of a certain nuclear power equipment and fault data of the same type of nuclear power equipment.

[0020] The equipment remaining life prediction model modeling system of this embodiment includes a basic algorithm library 1, a database 2 and a general platform 3.

[0021] Basic algorithm library 1 is used to store the basic algorithms required in the process of modeling the remaining life prediction model, wherein the types of basic algorithms include data preprocessing algorithms, feature extraction algorithms, data-driven algorithms, and optimization algorithms, and the number of basic algorithms of each type is multiple. Database 2 is used to store the data sets required in the process of modeling the remaining life prediction model. General platform 3 is connected to basic algorithm library 1 and database 2 respectively, and is used to call basic algorithms and data sets to build, train, and test basic models, and encapsulate the basic models that have passed the test evaluation into a remaining life prediction model of a preset file type.

[0022] In this embodiment, the basic algorithm library 1 is used to store commonly used basic algorithms. The basic algorithms in the basic algorithm library 1 can be directly called during the modeling process of the remaining life prediction model. The types of basic algorithms include at least but not limited to the following types, and the following algorithms are all algorithms well known to those skilled in the art:

[0023] (1) Data preprocessing algorithms, including missing data filling algorithms, abnormal data detection algorithms, noise data processing algorithms, and data normalization algorithms.

[0024] (2) Feature extraction algorithms, including Wingner-Ville analysis, wavelet analysis, spectral analysis, blind source separation and other algorithms, to extract feature parameters and establish a feature parameter set.

[0025] (3) Sensitive feature parameter selection algorithm, which reduces the dimension of the feature parameter set and automatically selects the most optimized sensitive feature parameters, including principal component analysis algorithm and hierarchical clustering algorithm.

[0026] (4) Data-driven algorithms for life prediction based on feature data, including support vector machines, artificial neural networks, and recurrent neural networks.

[0027] (5) Optimization algorithms that automatically optimize model parameters, such as genetic algorithms.

[0028] (6) Model uncertainty analysis algorithms, such as Bootstrap and Gaussian regression.

[0029] The data types in database 2 include: life data, status data, and environmental parameters. Among them, life data is the duration from the start of operation of the equipment to failure; status data is a parameter that characterizes the operating status and performance of the equipment, such as the vibration frequency of the pump shaft; environmental parameters are used to describe the use conditions of the equipment, such as the temperature, pressure, and humidity of the system in which the equipment is located. These data can be real data of equipment operation (including historical data of the equipment operation and data of similar equipment operation), or they can be data generated by simulating operation through experiments or modeling simulation. The life data, status data, and environmental parameters constitute the data set of device A. Usually, the user plans to model the remaining life prediction model for device A, and must first import the acquired data set of device A into database 2. In the table of the data set to be imported, there should be a header describing the attributes of each column of data so that the storage module can store the data in a structured manner.

[0030] Optionally, the general platform 3 includes a modeling module 31 , a training module 32 , and a test and evaluation module 33 .

[0031] The modeling module 31 is connected to the basic algorithm library 1 and the database 2, and is used to divide the called data set into a training set and a test set according to the division ratio, and is used to call the data preprocessing algorithm to perform data preprocessing on the training set, call the feature extraction algorithm to extract features from the training set after data preprocessing and generate a feature set, and is used to call the data driven algorithm and the optimization algorithm to obtain the basic model. The training module 32 is connected to the modeling module 31, and is used to apply the basic model to perform model training on the generated feature set. The modeling module 31 also includes a parameter adjustment unit. The parameter adjustment unit is used to receive parameters input by the user when calling the data preprocessing algorithm, the feature extraction algorithm, the data driven algorithm and the optimization algorithm.

[0032] In this embodiment, the modeling module 31 is used to provide the user with a selection interface for the path where the data set to be called is stored, so that the user can select the required data set, and divide the called data set into a training set and a test set according to the division ratio set by the user. For example, for the data set of the called device A, the proportion of the training set is set to 70%, and the proportion of the test set is set to 30%. Then, the modeling module automatically divides and generates a temporary training set and a test set, wherein the amount of data in the training set accounts for 70% of the data set of the device A, and the amount of data in the test set accounts for 30% of the data set of the device A. The ratio of the training set and the test set can be set according to user needs. The modeling module 31 is also used to provide the user with a selection interface for a data preprocessing algorithm, so that the user can select a specific data preprocessing algorithm to be called, and receive the algorithm parameters input by the user. The modeling training module 31 is used to automatically call the data preprocessing algorithm selected by the user to perform data preprocessing on the generated training set. The modeling module 31 is also used to provide the user with a selection interface for a feature extraction algorithm, and can receive the parameters input by the user for the specific feature extraction algorithm selected. For example, the modeling module 31 is used to select feature parameters for the training set after data preprocessing according to the feature parameter extraction algorithm and sensitive parameter selection algorithm selected by the user and generate multiple sensitive feature sets. The modeling module 31 is also used to provide the user with a selection interface for data-driven algorithms and optimization algorithms, and can receive the parameters input by the user for the specific data-driven algorithm and optimization algorithm selected to obtain the basic model. For example, if the user selects a data-driven algorithm based on historical failure records and exponential distribution, it is necessary to input parameters such as the failure rate parameter and initial value in the exponential distribution. The user also selects a genetic algorithm to optimize the basic model parameters. The basic model in this embodiment includes the specific data preprocessing algorithm, feature extraction algorithm, data-driven algorithm, and optimization algorithm selected by the user. The training module 32 is used to automatically perform model training on the generated feature set using the basic model. Before model training, the training module 32 is also used to receive the training cutoff conditions set by the user. The training cutoff conditions include the number of training times, performance indicators, etc. When the model training reaches the training cutoff condition set by the user, the training module 32 stops the model training and outputs the result. There are two situations at this time: First, the training number is reached, the training is stopped, and the output result is training failure, prompting the user to re-select the data preprocessing algorithm, feature extraction algorithm, data-driven algorithm, and optimization algorithm and re-train the model. Second, the performance index is reached, the output result is model training success, the performance index of the model is displayed, and the user is prompted whether to store the successfully trained basic model. The modeling module 31 is used to save the successfully trained basic model.

[0033] The test evaluation module 33 is connected to the modeling module 31 and is used to call the test set to test and evaluate the trained basic model, and to compare and analyze the evaluation results. The modeling module 31 is also used to encapsulate the basic model whose evaluation results meet the preset index requirements into a remaining life prediction model of a preset file type to meet the integration requirements of the equipment health management platform of the nuclear power plant.

[0034] In this embodiment, the test evaluation module 33 is used to provide the user with a selection interface for whether to perform a test. If the user chooses to perform a test, the test evaluation module 33 is used to call the test set to test and evaluate the trained basic model. When the basic model is tested, the prediction results are dynamically displayed, and the prediction results include: remaining service life and failure probability distribution. When evaluating the basic model, the evaluation indicators include but are not limited to prediction accuracy, uncertainty, credibility, model complexity and response speed (specific evaluation indicators are customized according to user needs and algorithms). The test evaluation module 33 is also used to compare and analyze the evaluation results of multiple basic models. If the evaluation result does not meet the performance indicators set by the user, the user is prompted to re-select the data preprocessing algorithm, feature extraction algorithm, data-driven algorithm, and optimization algorithm, model training and test evaluation. If the evaluation result meets the performance indicators set by the user, the user is prompted that the remaining life prediction model modeling is completed, and the user is further prompted whether to encapsulate the remaining life prediction model. If the user selects the encapsulation type (such as C language), the modeling module 31 is used to convert the completed remaining life prediction model into a program file of the file type specified by the user, completing the model encapsulation to meet the integration requirements of the nuclear power plant's equipment health management platform or facilitate integration with other superior nuclear power platforms.

[0035] Optionally, the modeling module 31 includes an algorithm association unit. The algorithm association unit is used to define interfaces between various basic algorithms to achieve associations between data preprocessing algorithms, feature extraction algorithms, data driven algorithms, and optimization algorithms.

[0036] In this embodiment, there are many types of basic algorithms. The business processes, data, input and output, and core algorithms between different algorithms and models may be different. Therefore, how to ensure the versatility of the remaining life prediction model modeling system of the equipment in this embodiment is a key technical difficulty. First, for different algorithm categories, sort out their business processes and the applicable algorithms at each stage in the business process; then, for algorithms with the same function, establish a basic algorithm library (container), support the addition, deletion, modification, and query functions for the elements (single algorithm) in the basic algorithm library, and develop corresponding algorithm call functions and interfaces based on the business process of the prediction model to guide users to model the remaining life prediction; finally, provide a packaging function for the model established and tested by the user, so as to facilitate integration into the existing equipment health management platform of the nuclear power plant and perform the life prediction function. In addition, an algorithm association unit is set to define and standardize the input and output interface form of each type of algorithm, and visually describe the meaning of the interface, so as to guide users to accurately define the input and output interfaces (such as data source, model storage location, parameter storage location, etc.) of the selected algorithm during the use of the remaining life prediction model modeling system of the equipment, and ensure that the multiple algorithms selected by the user can be correctly connected, thereby building a complete remaining life prediction model. As shown in Table 1, this is an example of interface standardization for data-driven algorithms.

[0037] Table 1

[0038]

[0039]

[0040] Optionally, the general platform 3 further comprises an algorithm management module 34. The algorithm management module 34 is connected to the basic algorithm library 1 and is used to delete, modify, or view the basic algorithms in the basic algorithm library, and to add new algorithms to the basic algorithm library by programming or script file import.

[0041] In this embodiment, a "new algorithm" function is provided, and the user can load the newly written algorithm file (.m file format of matlab) into the equipment remaining life prediction model system to expand the content of the basic algorithm library to meet the application requirements of a wider range of modeling scenarios. Similarly, the user can standardize and customize the new algorithm interface according to the system guidance so that it can be associated with the existing algorithms in the basic algorithm library.

[0042] Optionally, the equipment's remaining life prediction model modeling system further includes a typical use case library 4. The typical use case library 4 is connected to the general platform 3 and is used to store the remaining life prediction models of typical failure modes of typical equipment encapsulated by the general platform 3 (specifically the modeling module 31) for users to view or integrate and call.

[0043] Optionally, the general platform 3 further includes a use case management module 35. The use case management module 35 is connected to the modeling module 31 and is used to view, call, add, delete or modify the trained basic model stored in the modeling module 31.

[0044] The remaining life prediction modeling system of the equipment of this embodiment provides a standardized process and building block-type remaining life prediction modeling method for equipment, which can help nuclear power researchers to conveniently and quickly select appropriate data preprocessing technology, feature extraction technology, data-driven algorithm, parameter optimization technology, and establish a remaining life prediction model for specific equipment, so as to avoid the need for nuclear power researchers to program to realize algorithm functions and the combination of various algorithms in the manual modeling process, thereby reducing the requirements for modeling personnel's understanding of intelligent algorithms and programming technical capabilities, and making general technologies such as information technology and intelligent algorithms become general tools that can be directly applied in professional and technical fields or applied after a small amount of development; at the same time, it simplifies the interface of cross-professional technology (such as nuclear power technology and data processing technology) integration in the process of data-driven remaining life prediction model development, and reduces communication difficulty and cost. In addition, in the manual modeling process, the modeling personnel learn, optimize, code implementation, test evaluation, and algorithm optimization of prediction methods and algorithms until a satisfactory prediction model is obtained, and the cycle is at least about two months. If the remaining life prediction modeling system of the equipment of this embodiment is used, the cycle of building, testing and packaging the prediction model for the same equipment is about ten days. Therefore, the use of the modeling system of this embodiment can save labor costs and improve modeling efficiency.

[0045] Embodiment 2:

[0046] This embodiment provides a remaining life prediction system for nuclear power equipment, including an input module 21 and a remaining life prediction model modeling system for the equipment described in Example 1.

[0047] The input module 21 is used to receive the data set of nuclear power equipment input by the user.

[0048] The equipment's remaining life prediction model modeling system is connected to the input module 21 and is used to output a remaining life prediction model and remaining life prediction results of the nuclear power equipment according to a data set of the nuclear power equipment input by a user.

[0049] Embodiment 3:

[0050] This embodiment provides a method for modeling a remaining life prediction model of equipment, including:

[0051] Step 301, obtain the basic algorithm required in the process of building the remaining life prediction model, and obtain the data set of the equipment for which the remaining life prediction model is to be established. The types of basic algorithms include data preprocessing algorithms, feature extraction algorithms, data driven algorithms and optimization algorithms.

[0052] Step 302: divide the data set into a training set and a test set according to a division ratio.

[0053] Step 303, construct and train a basic model according to the basic algorithm and the training set, test and evaluate the trained basic model according to the test set, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type.

[0054] Optionally, step 303 specifically includes:

[0055] Step 31, calling a data preprocessing algorithm to perform data preprocessing on the training set;

[0056] Step 32, calling a feature extraction algorithm to extract features from the training set after data preprocessing and generate a feature set;

[0057] Step 33, calling data-driven algorithms and optimization algorithms to obtain a basic model;

[0058] Step 34, apply the basic model to perform model training on the generated feature set. When the training is successful, store the trained basic model; when the training fails, re-execute steps 31 to 34 to build a new basic model until the new basic model is successfully trained.

[0059] Step 35, calling the test set to test and evaluate the trained basic model, and performing comparative analysis on the evaluation results. When the evaluation results do not meet the performance indicators, re-execute steps 31 to 35 until the evaluation results of the new basic model meet the performance indicators.

[0060] Step 36, encapsulate the basic model whose evaluation results meet the preset index requirements into a remaining life prediction model of a preset file type to meet the integration requirements of the equipment health management platform of the nuclear power plant.

[0061] Optionally, step 303 also includes defining interfaces between various basic algorithms to achieve associations between data preprocessing algorithms, feature extraction algorithms, data-driven algorithms, and optimization algorithms.

[0062] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A system for predicting the remaining life of equipment, characterized in that: Including basic algorithm library, database and general platform, The basic algorithm library is used to store the basic algorithms required in the modeling process of the remaining life prediction model. Database, used to store the data sets required in the remaining life prediction model building process, The general platform is connected to the basic algorithm library and database respectively, and is used to call the basic algorithm and data set to build, train, and test the basic model, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of the preset file type. The types of basic algorithms include data preprocessing algorithms, feature extraction algorithms, data driven algorithms and optimization algorithms. The general platform includes modeling module, training module, and test evaluation module. The modeling module is connected to the basic algorithm library and database, and is used to divide the called data set into a training set and a test set according to the division ratio, and is used to call the data preprocessing algorithm to preprocess the training set, call the feature extraction algorithm to extract features from the preprocessed training set and generate a feature set, and is used to call the data-driven algorithm and the optimization algorithm to obtain the basic model. The training module is connected to the modeling module and is used to apply the basic model to the generated feature set for model training. The test evaluation module is connected to the modeling module and is used to call the test set to test and evaluate the trained basic model, and to compare and analyze the evaluation results. The modeling module is also used to encapsulate the basic model whose evaluation results exceed the preset threshold into a remaining life prediction model of a preset file type to meet the integration requirements of the equipment health management platform of the nuclear power plant. The modeling module includes an algorithm association unit, The algorithm association unit is used to define the interfaces between various basic algorithms to realize the association between data preprocessing algorithms, feature extraction algorithms, data-driven algorithms and optimization algorithms.

2. The equipment remaining life prediction model modeling system according to claim 1, characterized in that: The general platform also includes an algorithm management module. The algorithm management module is connected to the basic algorithm library and is used to delete, modify, or view the basic algorithms in the basic algorithm library, as well as add new algorithms to the basic algorithm library by programming or script file import.

3. The equipment remaining life prediction model modeling system according to claim 2, characterized in that: It also includes a typical use case library. The typical use case library is connected to the general platform and is used to store the remaining life prediction models of typical failure modes of typical equipment packaged in the general platform for users to view or integrate and call.

4. The equipment remaining life prediction modeling system according to claim 3, characterized in that: The general platform also includes a use case management module. The use case management module is connected to the modeling module and is used to view, call, add, delete or modify the basic model obtained by the modeling module.

5. The equipment remaining life prediction modeling system according to claim 4, characterized in that: The modeling module also includes a parameter adjustment unit. The parameter adjustment unit is used to receive parameters input by the user when calling data preprocessing algorithms, feature extraction algorithms, data-driven algorithms and optimization algorithms.

6. A remaining life prediction system for nuclear power equipment, characterized in that: A modeling system comprising an input module and a remaining life prediction model of the equipment according to any one of claims 1 to 5, The input module is used to receive a data set of nuclear power equipment input by a user. The equipment's remaining life prediction model modeling system is connected to the input module and is used to output a remaining life prediction model and remaining life prediction results of the nuclear power equipment according to a data set of the nuclear power equipment input by a user.

7. A method for modeling a remaining life prediction model of equipment, characterized in that: include: Obtain the basic algorithms required in the process of modeling the remaining life prediction model, and obtain the data set of the equipment for which the remaining life prediction model is to be established; Divide the data set into training set and test set according to the division ratio; Construct and train a basic model based on the basic algorithm and training set, test and evaluate the basic model based on the test set, and encapsulate the basic model that has passed the test evaluation into a remaining life prediction model of a preset file type. Among them, the basic model is constructed and trained according to the basic algorithm and the training set, the basic model is tested and evaluated according to the test set, and the basic model that has passed the test evaluation is packaged into a remaining life prediction model of a preset file type, specifically including: Step 31, calling a data preprocessing algorithm to perform data preprocessing on the training set; Step 32, calling a feature extraction algorithm to extract features from the training set after data preprocessing and generate a feature set; Step 33, calling data-driven algorithms and optimization algorithms to obtain a basic model; Step 34, applying the basic model to perform model training on the generated feature set. When the training is successful, the trained basic model is stored; when the training fails, steps 31 to 34 are re-executed to construct a new basic model until the new basic model is successfully trained; Step 35, calling the test set to test and evaluate the trained basic model, and performing comparative analysis on the evaluation results. When the evaluation results do not meet the performance indicators, re-execute steps 31 to 35 until the evaluation results of the new basic model meet the performance indicators; Step 36, encapsulate the basic model whose evaluation results meet the preset index requirements into a remaining life prediction model of a preset file type to meet the integration requirements of the equipment health management platform of the nuclear power plant, The basic model is constructed and trained according to the basic algorithm and the training set, and the basic model is tested and evaluated according to the test set, and the basic model that has passed the test and evaluation is packaged into a remaining life prediction model of a preset file type, which also includes: Define interfaces between various basic algorithms to realize the association between data preprocessing algorithms, feature extraction algorithms, data-driven algorithms, and optimization algorithms.

Images