System and method for evaluating corrosion state of nuclear power pipeline

Abstract

The invention relates to the field of nondestructive testing, in particular to a nuclear power pipeline corrosion state evaluation system and method, and the system comprises a plurality of fiber grating sensors which are arranged at a plurality of preset positions of a nuclear power pipeline and are used for converting the states of the nuclear power pipeline at the preset positions into optical signals; the fiber bragg grating demodulator is connected with the plurality of fiber bragg grating sensors and is used for analyzing the optical signals to obtain pipeline parameters; and the evaluation equipment is used for evaluating the corrosion state of the nuclear power pipeline based on a preset pipeline corrosion prediction model according to the pipeline parameters. Various fiber grating sensors are used, pipeline parameters are represented through optical signals, and the optical signals are processed through a fiber grating demodulator. The dependence of optical signals on pipeline materials is small, and the method can adapt to pipelines of different materials and structures.

Description

Technical Field

[0001] The present invention relates to the field of nondestructive testing, and in particular to a system and method for evaluating the corrosion state of a nuclear power pipeline. Background Art

[0002] Nuclear power pipelines are channels for transporting key fluids in nuclear power plants. Pipeline safety is directly related to the safe operation of nuclear power plants and the safety of personnel. Corrosion is one of the main threats faced during pipeline operation. Corrosion can cause the thickness of pipeline walls to decrease and the strength to decrease, which in turn can cause serious accidents such as leakage and explosion. These accidents will not only cause huge economic losses, but may also pose a serious threat to the environment and public safety. Through corrosion monitoring, pipeline corrosion problems can be discovered in a timely manner and corresponding repair measures can be taken to prevent accidents.

[0003] At present, the main method for pipeline corrosion detection is electromagnetic detection. The electromagnetic detection method generates an alternating magnetic field around the nuclear power pipeline. When there are defects caused by corrosion inside or on the surface of the pipeline, these defects will change the magnetic permeability and electrical conductivity of the material, thereby affecting the distribution of the magnetic field. The detection instrument identifies and quantifies the extent and location of corrosion by analyzing the changes in the magnetic field. However, this method has a narrow scope of application and cannot detect pipelines made of non-ferromagnetic materials. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a system and method for evaluating the corrosion status of nuclear power pipelines.

[0005] The first aspect of the present invention discloses a nuclear power pipeline corrosion state assessment system, the system comprising:

[0006] A plurality of fiber grating sensors are arranged at a plurality of preset positions of the nuclear power pipeline, and are used to convert the nuclear power pipeline status at the preset positions into optical signals;

[0007] A fiber Bragg grating demodulator connected to a plurality of fiber Bragg grating sensors for analyzing the optical signal to obtain pipeline parameters;

[0008] The evaluation equipment is used to evaluate the corrosion status of the nuclear power pipeline according to the pipeline parameters and based on a preset pipeline corrosion prediction model.

[0009] Furthermore, the fiber grating sensor comprises:

[0010] The optical fiber temperature and humidity sensor is installed outside the nuclear power pipeline and is used to convert the temperature and humidity of the nuclear power pipeline where it is installed into optical signals;

[0011] The optical fiber strain sensor is arranged outside the nuclear power pipeline and is used to convert the strain generated by the inner wall of the nuclear power pipeline where it is arranged into an optical signal;

[0012] The optical fiber pH sensor is installed inside the nuclear power pipeline and is used to convert the pH value of the fluid in the nuclear power pipeline into an optical signal.

[0013] The optical fiber flow velocity sensor is arranged inside the nuclear power pipeline and is used for converting the flow velocity of the fluid inside the nuclear power pipeline where it is arranged into an optical signal.

[0014] Furthermore, the pipeline corrosion prediction model is a Stacking ensemble learning model;

[0015] The evaluation device is used to input the pipeline parameters into multiple different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculate the initial prediction result according to the prediction result of each prediction model; and according to the initial prediction result, the corrosion state of the nuclear power pipeline is obtained through the secondary learner of the Stacking ensemble learning model.

[0016] The second aspect of the present invention discloses a method for evaluating the corrosion state of a nuclear power pipeline, which is applied to the evaluation system for the corrosion state of a nuclear power pipeline as described in any one of the first aspects of the present invention. The evaluation method comprises:

[0017] The pipeline status of the nuclear power pipeline is collected through the fiber grating sensor and the corresponding optical signal is generated;

[0018] The optical signal is analyzed by a fiber Bragg grating demodulator to obtain the pipeline parameters of the nuclear power pipeline;

[0019] The pipeline parameters are collected regularly through the evaluation equipment, and all the collected pipeline parameters are used to predict the corrosion status of the nuclear power pipeline based on the preset pipeline corrosion prediction model.

[0020] Furthermore, the pipeline corrosion prediction model is a Stacking ensemble learning model, and the steps of predicting the corrosion state of the nuclear power pipeline based on the preset pipeline corrosion prediction model using all collected pipeline parameters include:

[0021] Inputting the pipeline parameters into a plurality of different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculating an initial prediction result according to the prediction result of each prediction model;

[0022] The initial prediction result is input into the secondary learner of the Stacking ensemble learning model to obtain the corrosion status of the nuclear power pipeline.

[0023] Furthermore, the primary learner includes a random forest model, an adaptive boosting model, a gradient boosting decision tree model and an extreme gradient boosting model, and the secondary learner is a BP neural network model.

[0024] Furthermore, the training method of the pipeline corrosion prediction model includes:

[0025] Acquire multiple samples; wherein the samples include sample data and corresponding sample labels; the sample data include the temperature of the nuclear power pipeline, the humidity of the nuclear power pipeline, the strain generated by the inner wall of the nuclear power pipeline, the pH value of the fluid and the flow rate of the fluid; the sample label is the corrosion state of the nuclear power pipeline;

[0026] According to the K-fold cross-validation algorithm, the plurality of samples are grouped into a plurality of training sets including training subsets and validation subsets; wherein the number of samples in each of the training subsets is the same, the number of samples in each of the validation subsets is the same; and each of the validation subsets in the training set is different from each other;

[0027] Training the primary learner based on the training set to obtain the trained primary learner; and obtaining an initial prediction result through the trained primary learner;

[0028] Based on the initial prediction result, the secondary learner is trained to obtain the trained secondary learner;

[0029] The trained primary learner and the trained secondary learner are combined into a pipeline corrosion prediction model.

[0030] Furthermore, the step of training the primary learner based on the training set to obtain the trained primary learner includes:

[0031] Using the training subset in the training set to train each prediction model in the primary learner, and obtaining a prediction model of the primary learner that has been trained;

[0032] Each prediction model in the initial learner is validated using the validation subset in the training set:

[0033] The model that passes the verification is the prediction model of the trained primary learner;

[0034] For models that fail the verification, the prediction model is trained again using the training subset;

[0035] When all prediction models are trained, all the trained prediction models are combined into the primary learner.

[0036] Furthermore, each prediction model in the primary learner is trained using a training subset in the training set, and the step of obtaining the prediction model of the primary learner after training includes:

[0037] For each prediction model in the primary learner:

[0038] Input the sample data of the training subset in the training set into the prediction model to obtain the corresponding prediction results;

[0039] Based on the genetic algorithm, the value of the preset fitness function is calculated according to the prediction results and the sample labels in the training subset; the model parameters are adjusted according to the calculation results.

[0040] Furthermore, the fitness function includes a goodness of fit evaluation and a root mean square error; wherein the goodness of fit evaluation represents the variance percentage of the prediction result of a prediction model in the primary learner, and the root mean square error represents the square root of the average of the sum of squares of the deviations between the prediction result of a prediction model in the primary learner and the sample label.

[0041] The present invention uses a variety of fiber grating sensors to represent pipeline parameters through optical signals, and uses a fiber grating demodulator to process the optical signals. The optical signals are less dependent on the pipeline material and can adapt to pipelines of different materials and structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 It is a structural schematic diagram of a nuclear power pipeline corrosion state assessment system disclosed in an embodiment of the present invention;

[0044] Figure 2 It is a flow chart of a method for evaluating the corrosion status of a nuclear power pipeline disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, or product end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0047] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0048] See also Figure 1 As shown, Figure 1 Schematic diagram of a nuclear power pipeline corrosion status assessment system disclosed in an embodiment of the present invention. Figure 1 As shown, the nuclear power pipeline corrosion status assessment system method may include the following modules:

[0049] A plurality of fiber grating sensors 101 are arranged at a plurality of preset positions of the nuclear power pipeline, and are used to convert the nuclear power pipeline state at the preset positions into an optical signal;

[0050] In this optional embodiment, nuclear power pipelines refer to channels in nuclear power plants that transport critical fluids, including but not limited to coolants, steam, water, fuel, and waste. For example, the pipelines of the main coolant circulation system are responsible for carrying high-temperature, high-pressure coolants generated from the reactor, sending them to the steam generator for heat exchange, and returning the cooled coolant to the reactor. These fluids are critical to the operation of nuclear power plants, and any leakage or contamination may lead to serious consequences.

[0051] In an optional embodiment, the fiber grating sensor comprises:

[0052] The optical fiber temperature and humidity sensor is installed outside the nuclear power pipeline and is used to convert the temperature and humidity of the nuclear power pipeline where it is installed into optical signals;

[0053] The optical fiber strain sensor is arranged outside the nuclear power pipeline and is used to convert the strain generated by the inner wall of the nuclear power pipeline where it is arranged into an optical signal;

[0054] The optical fiber pH sensor is installed inside the nuclear power pipeline and is used to convert the pH value of the fluid in the nuclear power pipeline into an optical signal.

[0055] The optical fiber flow velocity sensor is arranged inside the nuclear power pipeline and is used for converting the flow velocity of the fluid inside the nuclear power pipeline where it is arranged into an optical signal.

[0056] In this optional embodiment, the optical fiber temperature and humidity sensor can convert the temperature and humidity information of the environment into optical signals. This sensor is installed on the outside of the nuclear power pipeline and can accurately measure the temperature and humidity of the pipeline surface by monitoring and analyzing the reflected optical signals.

[0057] A fiber optic strain sensor is a sensor that can convert physical deformation (such as stretching or compression) into an optical signal. This sensor is usually installed on the outside of a nuclear power pipeline. By measuring the change in the reflected light signal, the strain magnitude of the inner wall of the pipeline can be accurately obtained. Strain is a physical quantity used to describe the degree of change in the shape of an object, including stretching and compression. When the inner wall of the pipeline is strained, the fiber optic strain sensor installed on the outside of the pipeline will deform accordingly. This deformation will change the reflection spectrum of the fiber grating. The fiber grating demodulator can obtain the magnitude of the strain by analyzing this change.

[0058] The fiber optic pH sensor is a sensor that can measure the pH of liquids by converting the pH value into an optical signal and then transmitting it to the data processing center through optical fiber. This sensor is usually installed on the inside of nuclear power pipelines to monitor the pH value of the fluid in the pipeline. Accurate measurement of pH helps prevent corrosion problems caused by changes in pH. pH is a parameter that measures the acidity and alkalinity of a solution. Its value range is between 0 and 14, with 7 indicating neutrality, less than 7 indicating acidity, and greater than 7 indicating alkalinity. The pH value of the fluid in nuclear power pipelines is very important for predicting and preventing pipeline corrosion. Generally speaking, fluids with strong acidity and alkalinity are more likely to cause pipeline corrosion. By monitoring and adjusting the pH value of the fluid, pipeline corrosion can be effectively prevented or slowed down, thereby extending the service life of the pipeline and reducing safety risks.

[0059] Fiber optic flow sensor is a sensor that can measure the flow rate of fluid. It works by converting the flow rate into an optical signal. This sensor is usually installed on the inside of nuclear power pipelines to monitor the flow rate of fluid in the pipeline. Monitoring the flow rate helps prevent problems caused by too fast or too slow flow, such as corrosion, wear, etc.

[0060] A fiber Bragg grating demodulator 102, connected to a plurality of fiber Bragg grating sensors, is used to analyze the optical signal to obtain pipeline parameters;

[0061] In this optional embodiment, the fiber Bragg grating interrogator is a specialized device used to analyze the optical signal transmitted by the fiber Bragg grating sensor. The fiber Bragg grating is a special optical fiber with tiny gratings engraved at specific intervals along its length. When light passes through these gratings, different wavelengths of light will produce different mutual interferences, forming unique reflection spectra. By analyzing these reflection spectra, the fiber Bragg grating interrogator can obtain physical or chemical parameters such as temperature, humidity, strain, pH value and flow rate at the location of the fiber Bragg grating sensor.

[0062] The evaluation device 103 is used to evaluate the corrosion state of the nuclear power pipeline according to the pipeline parameters and based on a preset pipeline corrosion prediction model.

[0063] In an optional embodiment, the pipeline corrosion prediction model is a Stacking ensemble learning model;

[0064] The evaluation device is used to input the pipeline parameters into multiple different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculate the initial prediction result according to the prediction result of each prediction model; and according to the initial prediction result, the corrosion state of the nuclear power pipeline is obtained through the secondary learner of the Stacking ensemble learning model.

[0065] The Stacking ensemble learning model is an ensemble learning model that uses the outputs of multiple basic models as new training sets and then trains a meta-model. By combining the prediction results of different models, the accuracy and robustness of the overall prediction can be improved, while the risk of overfitting can be reduced. It is a very effective ensemble learning strategy.

[0066] See also Figure 2 As shown, Figure 2 1 is a flow chart of a method for evaluating the corrosion status of a nuclear power pipeline disclosed in an embodiment of the present invention. Figure 2 As shown, the method for evaluating the corrosion status of nuclear power pipelines may include the following steps:

[0067] S201, collecting the pipeline status of the nuclear power pipeline through a fiber grating sensor and generating a corresponding optical signal;

[0068] Regarding the type and configuration of the fiber Bragg grating sensor in this embodiment, please refer to the relevant introduction in the evaluation system for the corrosion status of nuclear power pipelines in the present invention.

[0069] S202, analyzing the optical signal by using a fiber grating demodulator to obtain pipeline parameters of the nuclear power pipeline;

[0070] For the introduction of the fiber Bragg grating demodulator in this embodiment, please refer to the relevant introduction of the evaluation system for the corrosion status of nuclear power pipelines in the present invention.

[0071] S203, regularly collecting pipeline parameters through evaluation equipment, and predicting the corrosion state of nuclear power pipelines based on a preset pipeline corrosion prediction model using all collected pipeline parameters.

[0072] In an optional embodiment, the pipeline corrosion prediction model is a Stacking ensemble learning model, and the steps of predicting the corrosion state of the nuclear power pipeline based on the preset pipeline corrosion prediction model using all collected pipeline parameters include:

[0073] Inputting the pipeline parameters into a plurality of different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculating the initial prediction result according to the prediction result of each prediction model;

[0074] The initial prediction result is input into the secondary learner of the Stacking ensemble learning model to obtain the corrosion status of the nuclear power pipeline.

[0075] In a further optional embodiment, the primary learner includes a random forest model, an adaptive boosting model, a gradient boosting decision tree model and an extreme gradient boosting model, and the secondary learner is a BP neural network model.

[0076] In this optional embodiment, random forest is an ensemble learning algorithm based on decision trees. It makes a final prediction by constructing multiple decision trees and combining their prediction results. During the training process, random forest uses a self-service sampling method to randomly select samples and feature subsets to construct each decision tree, which can improve the diversity and robustness of the model. Random forest has good classification and regression performance, can effectively handle high-dimensional data and noise, and has strong robustness to outliers and missing values.

[0077] The adaptive boosting model is an iterative boosting algorithm that builds a strong learner by combining multiple weak learners. In each iteration, the adaptive boosting model adjusts the weights of samples based on the performance of the weak learners in the previous round, increasing the weights of samples that are misclassified and decreasing the weights of samples that are correctly classified. In this way, subsequent weak learners will pay more attention to samples that are difficult to classify. Finally, the final prediction is made by weighted combination of the predictions of all weak learners. The adaptive boosting model is sensitive to noise and outliers, but performs well in many practical applications.

[0078] The gradient boosting decision tree model is a decision tree-based boosting algorithm that gradually optimizes the model by iteratively building a series of decision trees. In each iteration, the gradient boosting decision tree model fits a new decision tree based on the residual of the current model and adds it to the model. The residual represents the difference between the predicted value of the current model and the true value. By continuously fitting the residual, the gradient boosting decision tree model gradually reduces the error of the model and improves the prediction performance. The gradient boosting decision tree model can handle various types of data, including continuous values ​​and categorical values, and has a certain degree of robustness to outliers and noise.

[0079] The extreme gradient boosting model is an optimized implementation of the gradient boosting decision tree model. It introduces some improvements and optimization strategies based on the gradient boosting decision tree model. The extreme gradient boosting model uses a second-order Taylor expansion to approximate the loss function, taking into account the first-order derivative and the second-order derivative, so as to fit the residual more accurately. In addition, the extreme gradient boosting model also introduces a regularization term to control the complexity of the model and prevent overfitting. The extreme gradient boosting model has significant improvements in training speed and prediction performance, and supports parallel computing, which can effectively process large-scale data sets. The extreme gradient boosting model has achieved excellent results in many data mining competitions.

[0080] BP neural network is a supervised learning algorithm based on gradient descent, which trains multi-layer perceptrons through back propagation algorithm. BP neural network consists of input layer, hidden layer and output layer, each layer consists of multiple neurons. In the forward propagation process, the input signal is transmitted from the input layer to the output layer through the hidden layer to generate the prediction result. In the back propagation process, according to the error between the prediction result and the true value, the error term of each neuron is calculated by the chain rule, and the weight and bias of the network are updated using the gradient descent method to minimize the loss function. BP neural network has strong nonlinear fitting ability and can learn complex patterns and relationships.

[0081] In this optional embodiment, the selected primary learner includes different types of models, such as tree-based models (random forest, gradient boosting decision tree model, extreme gradient boosting model) and boosting-based models (adaptive boosting model). These models are different in algorithm principles, feature processing and decision boundaries, and can capture the characteristics and patterns of data from different angles. The diversity of models helps to improve the performance of the Stacking ensemble learning model, because the prediction errors of different models may be unrelated, and by combining their prediction results, the overall error rate can be reduced. As a secondary learner, the BP neural network has a strong nonlinear fitting ability. It can learn the complex nonlinear relationship between the prediction results of the primary learner and the true value, thereby further improving the overall prediction performance. The addition of the BP neural network enables the Stacking ensemble learning model to better handle complex data patterns and nonlinear problems.

[0082] In an optional embodiment, the pipeline corrosion prediction model training method includes:

[0083] Acquire multiple samples; wherein the samples include sample data and corresponding sample labels; the sample data include the temperature of the nuclear power pipeline, the humidity of the nuclear power pipeline, the strain generated by the inner wall of the nuclear power pipeline, the pH value of the fluid and the flow rate of the fluid; the sample label is the corrosion state of the nuclear power pipeline;

[0084] According to the K-fold cross-validation algorithm, the plurality of samples are grouped into a plurality of training sets including training subsets and validation subsets; wherein the number of samples in each of the training subsets is the same, the number of samples in each of the validation subsets is the same; and each of the validation subsets in the training set is different from each other;

[0085] Training the primary learner based on the training set to obtain the trained primary learner; and obtaining an initial prediction result through the trained primary learner;

[0086] Based on the initial prediction result, the secondary learner is trained to obtain the trained secondary learner;

[0087] The trained primary learner and the trained secondary learner are combined into a pipeline corrosion prediction model.

[0088] Taking the K-fold cross-validation algorithm as an example, the multiple samples are divided into multiple training sets including training subsets and validation subsets, the multiple samples are equally divided into K parts, each of which is used as a validation subset, and the remaining K-1 parts are used as training subsets corresponding to the validation subset, so that a total of K validation subsets and corresponding K training subsets are obtained.

[0089] In this optional embodiment, each prediction model in the initial learner is trained K times, and each time a validation subset prediction result {V1, V2, ..., V K} and a training subset prediction result {P1, P2, ..., P K}, after training m prediction models, we can get m sets of validation subset prediction results {V 11 , V 12 , …, V 1K , V 21 , V 22 , …, V 2K , …, V m1 , …, V mK}, the prediction results of the m group training subset are

[0090] {P 11 , P 12 ,…,P 1K , P 21 , P 22 ,…,P 2K ,…,P m1 ,…,P mK}. The average of the training subset prediction results obtained from K trainings is used to obtain the m groups of training subset prediction results: P =

[0091] {P1, P2, ..., P m}, the average of the validation subset prediction results obtained by K trainings is used to obtain the m groups of validation subset prediction results V = {V1, V2, ..., V m}. P and V are used as inputs of the secondary learner.

[0092] In an optional embodiment, the step of training the primary learner based on the training set to obtain the trained primary learner includes:

[0093] Using the training subset in the training set to train each prediction model in the primary learner, and obtaining a prediction model of the primary learner that has been trained;

[0094] Each prediction model in the initial learner is validated using the validation subset in the training set:

[0095] The model that passes the verification is the prediction model of the trained primary learner;

[0096] For models that fail the verification, the prediction model is trained again using the training subset;

[0097] When all prediction models are trained, all the trained prediction models are combined into the primary learner.

[0098] In an optional embodiment, each prediction model in the primary learner is trained using a training subset in the training set, and the step of obtaining the prediction model of the primary learner after training includes:

[0099] For each prediction model in the primary learner:

[0100] Input the sample data of the training subset in the training set into the prediction model to obtain the corresponding prediction results;

[0101] Based on the genetic algorithm, the value of the preset fitness function is calculated according to the prediction results and the sample labels in the training subset; the model parameters are adjusted according to the calculation results.

[0102] In this optional embodiment, the genetic algorithm searches according to the principles of natural selection and inheritance, and finds the optimal solution to the problem by repeatedly modifying and improving the current solution. The initial value of the genetic algorithm is a possible combination of model parameters of the prediction model. This initial value can be a set of parameter combinations generated randomly, or it can be pre-set based on experience.

[0103] In genetic algorithms, the fitness function is used to evaluate the performance of each candidate solution (i.e., model parameter combination). Solutions with good fitness will have a higher probability of being selected as the parent of the next generation. Then, new solutions are generated through the two main operations of the genetic algorithm, namely crossover (or pairing) and mutation. Crossover is to exchange some parameters of two solutions to generate a new solution; mutation is to randomly modify some parameters of the solution. These two steps add randomness to ensure that the algorithm can search globally and avoid falling into local optimal solutions. After many iterations (i.e., the evolution of multiple generations of populations), the algorithm will eventually converge to one, or a group of solutions, which are optimal in terms of the fitness function, that is, the parameter configuration of the model will make the model's prediction performance optimal.

[0104] In an optional embodiment, the fitness function includes a goodness of fit evaluation and a root mean square error; wherein the goodness of fit evaluation represents the variance percentage of the prediction result of a prediction model in the primary learner, and the root mean square error represents the square root of the average of the sum of squares of the deviations between the prediction result of a prediction model in the primary learner and the sample label.

[0105] In this alternative embodiment, R-squared (R 2 , determination coefficient) to measure the goodness of fit. R 2The value of is between 0 and 1. The closer it is to 1, the more consistent the model's prediction results are with the actual values, and the better the model's fitting effect is. The root mean square error is a commonly used prediction error measurement method, which represents the square root of the average of the squares of the deviations between the predicted values ​​and the actual values. The smaller the root mean square error, the higher the accuracy of the prediction model.

[0106] In summary, the present invention discloses a nuclear power pipeline corrosion status assessment system and method, which uses a variety of fiber grating sensors to represent pipeline parameters through optical signals, and uses a fiber grating demodulator to process the optical signals. The optical signal is less dependent on the pipeline material and can adapt to pipelines of different materials and structures. Therefore, the present invention effectively overcomes various shortcomings in the prior art and has a high industrial utilization value.

[0107] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A nuclear power pipeline corrosion status assessment system, characterized in that: The system comprises: A plurality of fiber grating sensors are arranged at a plurality of preset positions of the nuclear power pipeline, and are used to convert the nuclear power pipeline status at the preset positions into optical signals; A fiber Bragg grating demodulator connected to a plurality of fiber Bragg grating sensors for analyzing the optical signal to obtain pipeline parameters; The evaluation equipment is used to evaluate the corrosion status of the nuclear power pipeline according to the pipeline parameters and based on a preset pipeline corrosion prediction model.

2. A nuclear power pipeline corrosion status assessment system according to claim 1, characterized in that: The fiber grating sensor comprises: The optical fiber temperature and humidity sensor is installed outside the nuclear power pipeline and is used to convert the temperature and humidity of the nuclear power pipeline where it is installed into optical signals; The optical fiber strain sensor is arranged outside the nuclear power pipeline and is used to convert the strain generated by the inner wall of the nuclear power pipeline where it is arranged into an optical signal; The optical fiber pH sensor is installed inside the nuclear power pipeline and is used to convert the pH value of the fluid in the nuclear power pipeline into an optical signal. The optical fiber flow velocity sensor is arranged inside the nuclear power pipeline and is used for converting the flow velocity of the fluid inside the nuclear power pipeline where it is arranged into an optical signal.

3. A nuclear power pipeline corrosion status assessment system according to claim 1, characterized in that: The pipeline corrosion prediction model is a Stacking ensemble learning model; The evaluation device is used to input the pipeline parameters into multiple different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculate the initial prediction result according to the prediction result of each prediction model; and according to the initial prediction result, the corrosion state of the nuclear power pipeline is obtained through the secondary learner of the Stacking ensemble learning model.

4. A method for evaluating the corrosion status of nuclear power pipelines, characterized in that: The evaluation system for the corrosion state of a nuclear power pipeline as claimed in any one of claims 1 to 3, wherein the evaluation method comprises: The pipeline status of the nuclear power pipeline is collected through the fiber grating sensor and the corresponding optical signal is generated; The optical signal is analyzed by a fiber Bragg grating demodulator to obtain the pipeline parameters of the nuclear power pipeline; The pipeline parameters are collected regularly through the evaluation equipment, and all the collected pipeline parameters are used to predict the corrosion status of the nuclear power pipeline based on the preset pipeline corrosion prediction model.

5. A method for evaluating the corrosion status of nuclear power pipelines according to claim 4, characterized in that: The pipeline corrosion prediction model is a Stacking ensemble learning model. The steps of predicting the corrosion state of nuclear power pipelines based on the preset pipeline corrosion prediction model with all collected pipeline parameters include: Inputting the pipeline parameters into a plurality of different prediction models in the initial learner of the Stacking ensemble learning model respectively, and calculating the initial prediction result according to the prediction result of each prediction model; The initial prediction result is input into the secondary learner of the Stacking ensemble learning model to obtain the corrosion status of the nuclear power pipeline.

6. A method for evaluating the corrosion status of nuclear power pipelines according to claim 5, characterized in that: The primary learner includes a random forest model, an adaptive boosting model, a gradient boosting decision tree model and an extreme gradient boosting model, and the secondary learner is a BP neural network model.

7. A method for evaluating the corrosion status of nuclear power pipelines according to claim 5, characterized in that: The training method of the pipeline corrosion prediction model includes: Acquire multiple samples; wherein the samples include sample data and corresponding sample labels; the sample data include the temperature of the nuclear power pipeline, the humidity of the nuclear power pipeline, the strain generated by the inner wall of the nuclear power pipeline, the pH value of the fluid and the flow rate of the fluid; the sample label is the corrosion state of the nuclear power pipeline; According to the K-fold cross-validation algorithm, the plurality of samples are grouped into a plurality of training sets including training subsets and validation subsets; wherein the number of samples in each of the training subsets is the same, the number of samples in each of the validation subsets is the same; and each of the validation subsets in the training set is different from each other; Training the primary learner based on the training set to obtain the trained primary learner; and obtaining an initial prediction result through the trained primary learner; Based on the initial prediction result, the secondary learner is trained to obtain the trained secondary learner; The trained primary learner and the trained secondary learner are combined into a pipeline corrosion prediction model.

8. A method for evaluating the corrosion status of nuclear power pipelines according to claim 7, characterized in that: The step of training the primary learner based on the training set to obtain the trained primary learner comprises: Using the training subset in the training set to train each prediction model in the primary learner, and obtaining a prediction model of the primary learner that has been trained; Each prediction model in the initial learner is validated using the validation subset in the training set: The model that passes the verification is the prediction model of the trained primary learner; For models that fail the verification, the prediction model is trained again using the training subset; When all prediction models are trained, all the trained prediction models are combined into the primary learner.

9. A method for evaluating the corrosion status of nuclear power pipelines according to claim 8, characterized in that: The steps of training each prediction model in the primary learner using a training subset in the training set and obtaining the prediction model of the primary learner after training include: For each prediction model in the primary learner: Input the sample data of the training subset in the training set into the prediction model to obtain the corresponding prediction results; Based on the genetic algorithm, the value of the preset fitness function is calculated according to the prediction results and the sample labels in the training subset; the model parameters are adjusted according to the calculation results.

10. A method for evaluating the corrosion status of nuclear power pipelines according to claim 9, characterized in that: The fitness function includes a goodness of fit evaluation and a root mean square error; wherein the goodness of fit evaluation represents the variance percentage of the prediction result of a prediction model in the primary learner, and the root mean square error represents the square root of the average of the sum of squares of the deviations between the prediction result of a prediction model in the primary learner and the sample label.

Images