A nuclear metal aging grade diagnosis system and method

Abstract

The application discloses a nuclear metal aging grade diagnosis system and method, and relates to the field of nuclear power station safety monitoring.The system comprises a fiber laser, a collimating light receiving lens group, a spectrometer and a data analysis module; wherein the fiber laser is used for gathering laser on the surface of a nuclear metal sample and generating plasma; the collimating light receiving lens group is used for outputting the gathered plasma into the fiber coupler; the fiber coupler transmits the collected plasma into the spectrometer through an optical fiber; and the data analysis module is used for reading the spectrum of the spectrometer and obtaining a spectrum capable of representing the element composition of the nuclear metal sample.Compared with the detection method of a traditional LIBS system, the application has the advantages of in-situ detection, high efficiency, good economy and high robustness, and the like.Therefore, the system and method of the application can provide new technical support for the rapid in-situ diagnosis of the aging state of key metal components of a nuclear power station.

Description

TECHNICAL FIELD

[0001] The application relates to a nuclear metal aging grade diagnosis system and method. BACKGROUND

[0002] Whether the aging state of key metal components of a nuclear island can be quickly diagnosed during the in-service period of a nuclear power plant directly affects the safe operation of the nuclear power plant. The main nuclear power plants in operation in China are light water reactor type pressurized water reactor nuclear power plants. The safety of the pressurized water reactor unit mainly relies on three barriers, namely, fuel and cladding, pressure vessel and primary circuit pipeline, and safety shell. Among them, the pressure vessel and the primary circuit pipeline are exposed to extreme conditions such as high temperature, high pressure and radiation for a long time, and there is a great safety hazard due to material failure. Therefore, the aging state diagnosis of such key metal components of the nuclear island has always been the focus and difficulty of the operation and maintenance of the nuclear power unit. In addition, the special service environment of the nuclear power plant makes the aging mechanism very complex, increasing the difficulty of the aging state diagnosis.

[0003] At present, the nuclear metal material aging state diagnosis methods can be divided into non-in-situ detection and in-situ detection. As a laboratory analysis method, non-in-situ detection needs to cut the pipe for sampling. Although the results obtained are more accurate, the operation is tedious and time-consuming, and also causes irreversible damage to the original metal components. The in-situ detection method, including ultrasonic detection, ray detection, magnetic powder detection and eddy current detection, although does not need to cut the pipe for sampling, can only analyze the existing macroscopic defects, and is difficult to meet the demand for early warning of the failure of the service metal components. Therefore, in recent years, a nuclear metal component aging state diagnosis method based on laser-induced breakdown spectroscopy (LIBS) has been developed.

[0004] Due to the complex equipment layout and harsh working environment of the nuclear power plant, the existing LIBS diagnosis system of the nuclear metal component takes the Nd:YAG laser as the plasma excitation light source, and mainly cooperates with fiber optic transmission (Fiber Optic-LIBS, FO-LIBS). FO-LIBS transmits the laser beam to the surface of the metal to be detected through an optical fiber at a long distance, effectively improving the flexibility of the LIBS diagnosis system. However, the optical fiber transmission is extremely sensitive to the laser energy and coupling mode, the output laser pulse energy is low, which leads to poor sensitivity of trace metal detection, and the focusing effect of the laser after transmission through the optical fiber is significantly deteriorated, which further affects the detection effect. At the same time, the traditional Nd:YAG laser is large in size and heavy in structure, and is difficult to be integrated into a portable device, and the low laser repetition frequency (Hz level) limits the detection speed, and cannot realize short-time and efficient scanning measurement.

[0005] Therefore, the existing LIBS diagnosis system for nuclear metal components has problems of low detection efficiency, poor sensitivity and inability to meet the in-situ detection requirements of nuclear power plants. SUMMARY

[0006] The present application aims to overcome the deficiencies of the prior art, and provide a nuclear metal aging grade diagnosis system and method to improve the efficiency and accuracy of diagnosis.

[0007] To achieve the above-mentioned purpose, the technical solution of the present application is:

[0008] In a first aspect, the present application provides a nuclear metal aging grade diagnosis system, comprising a fiber laser, a focusing lens, a collimating light receiving lens group, a spectrometer and a data analysis module; wherein,

[0009] The fiber laser is used to generate laser, and the generated laser is focused on the surface of the nuclear metal sample after passing through the focusing lens and generates plasma;

[0010] The collimating light receiving lens group is used to output the collected plasma to the fiber coupler;

[0011] The fiber coupler transmits the collected plasma to the spectrometer through an optical fiber;

[0012] The data analysis module is used to read the spectrum of the spectrometer and obtain the spectrum that can represent the element composition of the nuclear metal sample.

[0013] Optionally, the nuclear metal aging grade diagnosis system further comprises:

[0014] A measurement platform is used to place the nuclear metal sample and control the movement process of the nuclear metal sample.

[0015] Optionally, the fiber laser comprises a fiber laser host, a laser transmission optical fiber and a laser probe; the fiber laser host is used to turn on the laser, the laser transmission optical fiber is used to transmit the laser at a long distance, and the laser probe is used to output the laser; the laser output by the laser probe is focused on the surface of the nuclear metal sample after passing through the focusing lens.

[0016] In a second aspect, the present application provides a nuclear metal aging grade diagnosis method, comprising:

[0017] Obtaining nuclear metal samples of different aging grades;

[0018] Detecting the nuclear metal samples of different aging grades based on the above-mentioned nuclear metal aging grade diagnosis system, and obtaining the spectrum of the nuclear metal samples of different aging grades;

[0019] Pretreating the spectrum to obtain an effective spectrum;

[0020] Scores between the characteristic variables of the effective spectrum and the aging grade labels are calculated by using mutual information, and features whose scores reach a threshold are stored as a new spectrum data matrix;

[0021] The spectrum data matrix is input into a random forest model to establish an aging grade classification model;

[0022] Based on the above nuclear metal aging grade diagnosis system, the spectrum of the to-be-tested nuclear metal is obtained;

[0023] The spectrum of the to-be-tested nuclear metal is input into the aging grade classification model to obtain the aging grade of the to-be-tested nuclear metal.

[0024] Optionally, the nuclear metal samples of different aging grades include:

[0025] The nuclear metal is subjected to an artificial aging experiment to obtain nuclear metal samples in different aging states;

[0026] The aging grade of the nuclear metal sample is comprehensively determined according to the aging time length and the mechanical performance index.

[0027] Optionally, the process of mutual information calculation is:

[0028]

[0029] wherein p(x, y) represents the joint probability distribution of the characteristic variable X and the classification label Y, p(x) and p(y) represent the marginal probability distribution of the characteristic variable X and the classification label Y, respectively.

[0030] Optionally, the optimization range of the variable tree of the random forest is 50-500, and the optimization range of the random variable is 1-7.

[0031] Optionally, the variable tree of the random forest is 150, and the random variable is 3.

[0032] Optionally, the aging grade classification model is:

[0033]

[0034] wherein f(x) is the prediction result of the random forest, B is the number of decision trees, f b (x) is the prediction result of the bth decision tree, k is the category label, and I(f b (x)=k) is an indicator function.

[0035] For the label k, the proportion of the prediction as k in all decision trees is calculated, and then the category with the highest proportion is selected as the final prediction result.

[0036] The aging grade of the measured main pipeline steel material is calculated according to f(x), so that the aging grade measurement is realized.

[0037] Optionally, the preprocessing of the spectrum includes: preprocessing the spectrum by using the methods of effectiveness screening, channel total intensity normalization and spectrum averaging.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] The system and method disclosed by the present application do not need to cut the pipe or mechanically process the sample surface before diagnosing the aging grade of the main pipeline steel material for nuclear power. The fiber laser can quickly remove the sample surface oxide by virtue of the high-repetition-rate ablation feature, and directly collect the plasma spectrum reflecting the sample matrix characteristics. Based on the method combining the mutual information feature extraction and the random forest, an aging grade classification model is established to realize the in-situ rapid diagnosis of the aging state of the metal component for nuclear power. Compared with the detection method of the traditional LIBS system, the system established by the present application has the advantages of in-situ detection, high efficiency and good economy. Compared with the traditional aging grade detection method, the method proposed by the present application has the advantages of accurate result, short operation time and high robustness. Therefore, the system and method of the present application can provide new technical support for the rapid in-situ diagnosis of the aging state of the key metal component of the nuclear power station. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The composition schematic diagram of the nuclear metal aging grade diagnosis system provided by the embodiment of the present application is shown in the figure;

[0041] Figure 2 The flowchart of the nuclear metal aging grade diagnosis method provided by the embodiment of the present application is shown in the figure;

[0042] Figure 3 The spectrum diagram of the sample with different aging grades is shown in the figure;

[0043] Figure 4 The prediction result diagram of the nuclear metal aging grade diagnosis method provided by the embodiment of the present application is shown in the figure;

[0044] In the figure: 1, fiber laser; 2, fiber laser host; 3, laser transmission optical fiber; 4, laser probe; 5, focusing lens; 6, measurement platform; 7, collimating light receiving lens group; 8, fiber coupler; 9, multi-channel optical fiber; 10, multi-channel spectrometer; 11, computer. DETAILED DESCRIPTION

[0045] Embodiment:

[0046] The technical solutions of the present application will be further described in combination with the drawings and embodiments.

[0047] Referring to Figure 1 The nuclear metal aging grade diagnosis system provided in the embodiment shown mainly comprises a fiber laser 1, a focusing lens 5, a collimating light receiving lens group 7, a multi-channel spectrometer 10 and a data analysis module 11.

[0048] The fiber laser 1 is used to generate laser, and the generated laser is focused on the surface of the nuclear metal sample through the focusing lens 5 and generates plasma; the conventional LIBS system mainly applies an Nd:YAG laser as a plasma excitation source, and has the disadvantages of bulky structure, slow output power, etc. The fiber laser is adopted in the present application, which has the outstanding advantages of good beam quality, compact structure, long-term stable output power, high repetition frequency (kHz level) and adaptability to harsh environments such as high temperature and high dust, greatly increases the detection efficiency, effectively improves the portability and flexibility of the measurement system, can realize in-situ rapid scanning of the surface of the nuclear metal component in a large area, and further obtain more complete information representing the aging characteristics of the material, thereby solving the fundamental defects of the conventional LIBS system.

[0049] The collimating light receiving lens group 7 is used to output the plasma after being collected to a fiber coupler 8, the fiber coupler 8 collects the plasma to the multi-channel spectrometer 10 through a multi-channel optical fiber 9; and the data analysis module is used to read the spectrum of the spectrometer and obtain the spectrum representing the element composition of the nuclear metal sample.

[0050] As can be seen, the fiber laser of the nuclear metal aging grade diagnosis system provided in the embodiment can quickly remove the oxides on the surface of the sample by virtue of the high repetition frequency ablation characteristics, and form a high-temperature plasma, and the wavelength and intensity of the emission spectrum of the plasma in the cooling process are analyzed to obtain the sample element type and concentration information, so that the plasma spectrum reflecting the characteristics of the sample matrix can be directly collected, and the detection efficiency can be greatly improved.

[0051] In addition, in order to facilitate the laser emitted by the fiber laser to be focused on different positions on the surface of the nuclear metal sample, the nuclear metal aging grade diagnosis system provided in the embodiment further comprises a measurement platform for placing the nuclear metal sample and controlling the movement process of the nuclear metal sample.

[0052] In one specific embodiment, the fiber laser 1 includes a fiber laser main unit 2, a laser transmission fiber 3, and a laser probe 4. The fiber laser main unit 2 is used to turn on the laser, the laser transmission fiber 3 is used to transmit the laser over a long distance, and the laser probe 4 is used to output the laser. The laser output by the laser probe 4 is focused onto the sample surface after passing through a focusing lens 5. Since the fiber laser mainly consists of three parts—the fiber laser main unit 2, the laser transmission fiber 3, and the laser probe 4—it has a compact and small structure, stable output power, high detection frequency, and can also reduce costs, showing great potential in practical applications.

[0053] The data analysis module mainly includes a computer 11. The collimating and receiving lens group 7 mainly consists of two focusing lenses 5.

[0054] Accordingly, such as Figure 2 As shown, this embodiment also provides a method for diagnosing the aging level of nuclear metals, which mainly includes the following steps:

[0055] Step 110: Obtain nuclear metal samples with different aging levels;

[0056] In practical implementation, taking the main pipeline steel as an example, artificial aging experiments were conducted on Z3CN20-09M material according to the national standard "T / CNS80-2022 Test Method for Hot Aging of Stainless Steel Materials for Nuclear Power Plants" to obtain nuclear-grade metal material samples in different aging states. The aging level of the main pipeline steel samples was comprehensively determined based on the aging time and Charpy V-shaped impact energy, thus comprehensively determining the aging level of the nuclear-grade metal samples.

[0057] Step 120: Based on the aforementioned nuclear metal aging level diagnostic system, nuclear metal samples of different aging levels are tested to obtain the spectra of nuclear metal samples of different aging levels. In this application, the spectrum is also known as plasma spectrum.

[0058] Step 130: Preprocess the spectrum to obtain an effective spectrum;

[0059] In this step, the spectrum is preprocessed to reduce interference such as invalid spectra and background noise, thereby improving the spectral quality and obtaining the effective spectra of nuclear metal samples with different aging levels.

[0060] In a specific implementation, the spectrum is preprocessed by using the methods of effectiveness screening, channel total intensity normalization, and spectral averaging to reduce the interference of invalid spectrum and noise on subsequent analysis. At the same time, considering that the spectral characteristics will be affected by the fluctuation of spectral intensity at different measurement points, the obtained spectrum is averaged by 10 groups, that is, a total of 50 spectra are used to represent the same material. The standard deviation (SD) of the characteristic spectral line Fe I 424.11 nm is selected for effectiveness screening, and the set SD value is 4500.

[0061] In step 140, the scores between the feature variables of the effective spectrum and the aging grade labels are calculated by using mutual information, and the features with high scores are stored as a new spectrum data matrix.

[0062] In this step, the feature variables of the plasma spectrum of different aging grade samples are extracted by using mutual information, which can effectively identify the data features with high correlation with the aging characteristics in the nuclear metal components.

[0063] In a specific implementation, the scores between the feature variables of the effective spectrum and the aging grade labels are calculated and arranged in descending order. Taking the classification accuracy of the prediction set as an indicator, the variable with the highest score is added one by one to the random forest classification model until all variables are re-inputted. The variable combination corresponding to the best classification accuracy is used as the final variable for modeling, and the mutual information calculation process is as follows:

[0064]

[0065] Wherein, p(x, y) represents the joint probability distribution of the feature variable X and the classification label Y, and p(x) and p(y) represent the marginal probability distribution of the feature variable X and the classification label Y, respectively.

[0066] In step 150, the spectrum data matrix after mutual information feature extraction is input into the random forest model to establish an aging grade classification model.

[0067] In this step, the random forest is used to establish an aging grade classification model for the spectrum data matrix, which has good robustness for the characteristics of alloy steel substrate (element concentration and uneven distribution of metallographic structure), and can quickly and accurately predict the aging grade of the sample.

[0068] In a preferred embodiment, in order to optimize the classification accuracy of the model, the key parameters of the random forest are optimized by grid search, and the optimization range of the variable tree is 50-500; the optimization range of the random variable is 1-7.

[0069] In step 160, the nuclear metal to be tested is detected based on the above nuclear metal aging grade diagnosis system, and the spectrum of the nuclear metal to be tested is obtained.

[0070] Step 170, input the spectrum of the to-be-tested nuclear metal into the aging grade classification model to obtain the aging grade of the to-be-tested nuclear metal.

[0071] In specific embodiments, the aging grade classification model is:

[0072]

[0073] In the formula, f(x) is the prediction result of the random forest, B is the number of decision trees, f b (x) is the prediction result of the bth decision tree, k is the category label, and I(f b (x)=k) is an indicator function.

[0074] For label k, the proportion of all decision trees that predict k is calculated, and then the category with the highest proportion is selected as the final prediction result. The aging grade of the tested main pipeline steel material is calculated according to f(x) to achieve aging grade measurement.

[0075] Therefore, the nuclear metal aging grade diagnosis method provided in this embodiment first performs artificial aging treatment on the nuclear metal, and comprehensively determines the aging grade of the sample according to the aging time and charpy V-type impact energy. Then, the FL-LIBS system is used to obtain the spectral data of samples with different aging grades, and the spectral data is preprocessed by effectiveness screening, channel total intensity normalization and averaging. 50 representative spectra are obtained for each aging grade sample. Subsequently, the mutual information algorithm is used to extract features from the full spectrum variables, and a random forest aging grade classification model is established. Finally, the FL-LIBS system is used to measure the spectrum of the to-be-tested nuclear metal, and input into the aging grade classification model to obtain the aging grade of the tested material. Compared with the traditional aging grade detection method, this method has the advantages of accurate prediction result, short operation time and high robustness.

[0076] The nuclear metal aging grade diagnosis method of the present application will be further described below in conjunction with an application scenario embodiment:

[0077] Step 1: Design and build a LIBS detection system (Fiber Laser Based-LIBS, FL-LIBS) with a fiber laser as a plasma excitation source, which specifically includes a 1064nm wavelength single pulse fiber laser, a three-dimensional electric translation stage, a multi-channel grating spectrometer, a computer and optical elements. Among them, the fiber laser frequency, power, spectrometer integration time and electric translation stage moving speed are adjustable. In this embodiment, the fiber laser frequency is set to 25kHz, the power is set to 20W, the spectrometer integration time is set to 30ms, and the electric translation stage moving speed is set to 3mm / s.

[0078] Step 2: Obtain the main pipe steel material Z3CN20-09M samples of different aging states. Based on the aging mechanism during the service process of the main pipe, thermal aging caused by ferrite phase amplitude decomposition and G phase precipitation is the main reason for material failure. Therefore, according to the national standard “T / CNS80-2022 Nuclear Power Plant Stainless Steel Material Thermal Aging Test Method”, the main pipe material Z3CN20-09M is subjected to artificial aging experiment. Under the condition of 400℃, 0, 500h, 1000h, 3000h, 6000h and 12000h of thermal aging treatment are carried out respectively. The microstructure and mechanical properties of the samples in different aging states are analyzed. According to the aging time and charpy V-type impact energy, the aging grade of the sample is determined, i.e. 1-6.

[0079] Step 3: Obtain the plasma spectrum of the main pipe steel material Z3CN20-09M for nuclear power by using the FL-LIBS system designed and built in step 1, as shown in Figure 3 Each aging grade sample obtains 500 spectra. The SD value of the characteristic spectral line Fe I 424.11nm is selected as the basis for judging the effectiveness of the spectrum, and the SD value is set to 4500. Finally, the spectra of each aging grade sample are all judged as effective spectra. Then, the effective spectral data is subjected to channel total intensity normalization processing. Finally, 50 representative spectra are obtained for each aging grade sample.

[0080] Step 4: Calculate the score between the spectral feature variables and the aging grade label by using the mutual information algorithm, and arrange them in descending order. Extract the first 1476 feature variables, i.e. 1476x300-dimensional spectral data matrix is obtained for the samples of 1-6 aging grades.

[0081] Step 5: Input the spectral data matrix into the random forest algorithm to establish an aging grade classification model. The key parameters of the random forest are optimized by grid search, i.e. the best parameters of the tree are 150 and the best parameters of the random variable are 3.

[0082] Step 6: Measure the plasma emission spectrum of the main pipe steel sample to be tested by using the established FL-LIBS system, and input the spectral matrix into the aging grade classification model:

[0083]

[0084] In the formula, f(x) is the prediction result of the random forest, B is the number of decision trees, f b (x) is the prediction result of the bth decision tree, k is the class label, and I(f b (x)=k) is an indicator function. For label k, calculate the proportion of all decision trees that predict k, and then select the class with the highest proportion as the final prediction result.

[0085] The aging grade of the measured main pipeline steel material is calculated according to f(x), so as to realize aging grade prediction. Figure 4 As shown in the figure, the accuracy of the prediction result is as high as 99%.

[0086] It can be seen that the system and method disclosed by the application do not need to cut the pipe and do not need to mechanically process the surface of the sample before diagnosing the aging grade of the main pipeline steel material for nuclear power. The fiber laser can quickly remove the oxide on the surface of the sample by virtue of the ablation characteristics of high repetition frequency, and directly collect the plasma spectrum reflecting the characteristics of the sample matrix. Based on the method combining mutual information feature extraction and random forest, an aging grade classification model is established, and the in-situ rapid diagnosis of the aging state of the metal component for nuclear power is realized. Compared with the traditional LIBS system, the system established by the application has the advantages of in-situ detection, high efficiency and good economy. Compared with the traditional aging grade detection method, the method proposed by the application has the advantages of accurate result, short operation time and high robustness. Therefore, the system and method of the application can provide new technical support for the in-situ rapid diagnosis of the aging state of the key metal component of the nuclear power station.

[0087] The above examples are only for illustrating the technical concept and characteristics of the application, and the purpose is to enable those skilled in the art to understand the content of the application and implement it, and cannot limit the protection scope of the application. Any equivalent changes or modifications made according to the essence of the application should be covered within the protection scope of the application.

Claims

1. A method of diagnosing a nuclear metal aging grade, characterized by, include: Obtain nuclear metal samples with different aging levels; The nuclear metal aging level diagnostic system is used to test nuclear metal samples with different aging levels and obtain the spectra of nuclear metal samples with different aging levels. The spectrum is preprocessed to obtain an effective spectrum; The scores between the aging level labels of the feature variables of the effective spectrum are calculated using mutual information, and the features that reach the threshold scores are extracted and stored as a new spectral data matrix. Input the spectral data matrix into the random forest model to establish an aging level classification model; The nuclear metal aging level diagnostic system is used to detect the nuclear metal to be tested and obtain the spectrum of the nuclear metal to be tested. The spectrum of the metal to be tested for nuclear use is input into the aging level classification model to obtain the aging level of the metal to be tested for nuclear use. The aging level classification model is as follows: where f(x) is the prediction result of the random forest, B is the number of decision trees, f b (x) is the prediction result of the bth decision tree, k is the class label, I(f b (x) = k) is an indicator function. For label k, calculate the proportion of all decision trees that predict k, and then select the category with the highest proportion as the final prediction result; The aging grade of the main pipeline steel under test is calculated based on f(x) to achieve aging grade measurement. The nuclear metal aging level diagnostic system includes a fiber laser, a focusing lens, a collimating and receiving lens group, a spectrometer, and a data analysis module; wherein... The fiber laser is used to generate laser light, which is focused onto the surface of the nuclear metal sample by the focusing lens and generates plasma. The collimating receiving lens group is used to focus the plasma and output it to the fiber coupler; The fiber optic coupler transmits the collected plasma to the spectrometer via optical fiber; The data analysis module is used to read the spectrum of the spectrometer and obtain the spectrum that can characterize the elemental composition of the nuclear metal sample; The fiber laser includes a fiber laser main unit, a laser transmission fiber, and a laser probe; the fiber laser main unit is used to turn on the laser, the laser transmission fiber is used to transmit the laser over long distances, and the laser probe is used to output the laser.

2. The nuclear metal aging grade diagnostic method according to Claim 1, characterized by, The process of obtaining nuclear metal samples with different aging levels includes: Artificial aging experiments were conducted on nuclear metals to obtain samples of nuclear metals in different aging states; The aging level of nuclear metal samples is determined by comprehensively considering aging time and mechanical performance indicators.

3. The nuclear metal aging grade diagnostic method according to Claim 1, characterized by, The process of calculating mutual information is as follows: Where p(x,y) represents the joint probability distribution of feature variable X and classification label Y, and p(x) and p(y) represent the marginal probability distributions of feature variable X and classification label Y, respectively.

4. The nuclear metal aging grade diagnostic method according to Claim 1, characterized by, The optimization range of the variable trees in the random forest is 50 to 500, and the optimization range of the random variables is 1 to 7.

5. The nuclear metal aging grade diagnostic method according to claim 4, characterized by, The random forest has 150 variable trees and 3 random variables.

6. The method for diagnosing the aging level of nuclear metals as described in claim 1, characterized in that, The preprocessing of the spectrum includes: preprocessing the spectrum using methods such as effectiveness screening, channel total intensity normalization, and spectral averaging.

7. The method for diagnosing the aging level of nuclear metals as described in claim 1, characterized in that, The nuclear metal aging level diagnostic system also includes: A measurement platform for placing nuclear metal samples and controlling the movement of the nuclear metal samples.

Images