Nondestructive testing method for flame retardant property of nonmetal shell of electric energy metering box

Abstract

The invention relates to the technical field of material performance detection, and particularly discloses a nondestructive testing method for flame retardance of a nonmetal shell of an electric energy metering box. According to the method, a spectrum technology, machine learning and a prediction technology are combined, a plurality of groups of samples are collected for flame combustion tests to obtain sample flame retardant performance data as labels, a spectrometer is used for nondestructive in-situ detection to obtain material surface spectrum data, after pretreatment and feature extraction, various machine learning algorithms are adopted for training and optimization, and the flame retardant performance of the samples is obtained. And finally, a BO-XGBoost model of Bayesian optimization is selected as a prediction algorithm. The non-destructive testing of the flame retardant property of the non-metal shell of the electric energy metering box is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of material performance detection, and more specifically, to a non-destructive testing method for flame retardant performance of a non-metallic shell of an electric energy metering box. Background Art

[0002] The combustion performance test of polymer materials is usually a destructive test process. The test principle of existing combustion performance test equipment, such as cone calorimeter, horizontal vertical combustion tester and various combustion test machines, is to place the test sample in a flame environment and observe the combustion state of the sample, flame extinguishing speed and other phenomena as the basis for judging the combustion performance of the material. Since the test sample needs to be made into a shape suitable for the combustion test and then placed on the flame for the test, after the flame retardant performance test, the sample cannot be used anymore regardless of whether it burns or not.

[0003] As the terminal of electric power equipment that directly serves users, the flame retardant performance of its non-metallic shell is crucial to protecting the safety of users' lives and property. According to current testing technology, the flame retardant performance of the non-metallic shell of the electric power meter box can only be tested by sampling, and it is impossible to test one by one. This poses a certain hidden danger to the safety of power equipment.

[0004] According to the purchasing requirements of the power company, the chemical composition of the non-metallic shell of the energy meter box is relatively certain. Its main body is a composite of two polymer materials, polycarbonate (PC) and ABS resin. Each manufacturer adds relevant chemical additives such as flame retardants, plasticizers and reinforcing materials according to its own production process. Therefore, the flame retardant performance of the non-metallic shell mainly depends on the type and content of the added flame retardant. By measuring the relationship between the composition and content of the flame retardant in the polymer composite system and the flame retardant performance, the flame retardant performance of the shell sample can be judged and analyzed. However, due to the large number of manufacturers of energy meter boxes, different processes and raw materials, and different stability of process control, the flame retardant performance of even the same batch of products may be different. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box, which is achieved by combining spectral detection technology and machine learning prediction technology, collecting samples for flame combustion tests to obtain accurate flame retardant performance data as labels, and using spectral in-situ non-destructive testing technology to obtain sample spectral data. After preprocessing and feature extraction, a variety of machine learning algorithms are used for training and optimization, and finally a Bayesian optimized BO-XGBoost model is selected for predictive analysis, thereby realizing non-destructive testing of the flame retardant properties of the non-metallic shell of the electric energy meter box to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box combines spectral technology with machine learning and prediction technology, uses spectral in-situ measurement technology that does not damage the sample to obtain the chemical composition spectrum of the test object, and realizes the prediction of the flame retardant properties of the material based on in-situ spectral characteristics through large-scale test data learning and training of machine learning models, including the following steps:

[0008] Step S1, collecting non-metallic shell samples of the electric energy meter box, and performing in-situ spectroscopy detection on the collected non-metallic shell samples using a portable infrared spectrometer, a portable Raman spectrometer, or an infrared / Raman spectrometer equipped with a movable optical fiber probe to collect spectral data;

[0009] Step S2, performing a flame combustion test on the collected non-metallic shell samples of the electric energy meter box;

[0010] Step S3, spectral data preprocessing and establishment of machine learning model;

[0011] Step S4, using the Bayesian optimization method to search for the hyperparameters of the XGBoost model, including the maximum depth of the tree (max_depth), the learning rate (learning_rate), the minimum split loss (gamma), and the subsample ratio (subsample), and evaluating the model performance through five-fold cross validation;

[0012] Step S5, using the optimized BO-XGBoost model to predict and identify the flame retardant properties of the non-metallic shell material of the electric energy meter box.

[0013] As a further solution of the present invention, step S1 comprises:

[0014] Step S11, widely collect samples of electric energy meter boxes from various brands and manufacturers on the market, obtain non-metallic shells, and ensure the diversity and representativeness of test data;

[0015] Step S12, using clean water and anhydrous ethanol to alternately clean the sample to remove dust, oil and other pollutants on the surface to ensure the accuracy of the flame combustion test.

[0016] As a further solution of the present invention, in step S2, the flame retardant performance test is carried out in accordance with the national standard GB / T2408-2021. Specifically, the non-metallic shell sample after cleaning and drying is cut into test strips with a length of 125.0mm±5.0mm, a width of 13.0mm±0.5mm, and a thickness of <13.0mm to ensure that its size meets the standard requirements. The test adopts horizontal and vertical combustion test equipment to carry out combustion tests under specified flame height, ignition time (10 seconds or 30 seconds) and test environment (23±2℃, relative humidity 50±10%). In the test, the flame propagation speed, self-extinguishing time after flame removal, dripping phenomenon and whether cotton is ignited after the sample is burned are observed. According to the comprehensive performance of the combustion test, the flame retardant grade is divided into V0 (the highest grade, indicating short burning time, no droplets and strong self-extinguishing ability), V1 (slightly longer self-extinguishing time, but no burning droplets), V2 (burning droplets and igniting cotton) and other grades (unqualified samples). Test data, as the key label for model training, provides a basis for the construction of machine learning models, while improving data reliability through repeated experiments on different samples.

[0017] As a further solution of the present invention, step S3 comprises:

[0018] Step S31, performing Savitzky-Golay filtering, baseline correction and normalization processing on the collected spectral data in sequence;

[0019] Step S32, extracting key band features from the spectral data, including information such as peak position and intensity directly related to the composition and content of the flame retardant;

[0020] In step S33, support vector machine (SVM), inverse neural network, CatBoost, BO-CatBoost (Bayesian optimized CatBoost), XGBoost and BO-XGBoost (Bayesian optimized XGBoost) are used for preliminary model training, and the BO-XGBoost model is selected as the final prediction model.

[0021] As a further solution of the present invention, step S5 comprises:

[0022] Step S51, using a portable infrared or Raman spectrometer, or an infrared or Raman spectrometer capable of moving an optical fiber probe, to in-situ test the non-metallic housing of the electric energy meter box to obtain spectral data of the test sample;

[0023] Step S52, using the optimized BO-XGBoost model to read the spectral data of the test sample, predict the flame retardant properties of the test sample, and output the result as "qualified sample" or "unqualified sample";

[0024] In step S53, the samples predicted by the optimized BO-XGBoost model are further verified by the combustion test in step S13.

[0025] As a further solution of the present invention, in step S31, a Savitzky-Golay filter is used to smooth the collected spectral data, and by setting a 7-point window width and a second-order polynomial order, random noise in the spectral signal is removed while retaining the key spectral characteristic peak shape; a least squares baseline fitting algorithm is used to perform baseline correction to eliminate spectral baseline drift caused by the test environment or instrument errors, ensuring that the spectral data is based on an accurate zero point; and maximum-minimum normalization is used to normalize the spectral data, mapping the spectral intensity value to the range of [0,1] to eliminate the influence of external factors such as sample thickness and light source intensity on the characteristic intensity.

[0026] As a further solution of the present invention, in step S32, key band features are extracted from the original spectrum. Since flame retardants usually have unique characteristic absorption peaks in infrared spectra and Raman spectra, key bands related to these chemical components and contents are extracted from the spectrum. In infrared spectra, different chemical groups will produce absorption peaks in a specific wavenumber range. For example, phosphate groups usually have significant characteristic absorption peaks in the range of 1200-1300cm-1, bromides will show characteristic peaks in the range of 500-600cm-1, and chlorides will show characteristic peaks in the range of 600-800cm-1. -1 There are weak characteristic absorption peaks in the range. By analyzing these peaks in the sample spectrum, information related to the type and content of flame retardants can be extracted; in Raman spectroscopy, the vibration mode of chemical bonds will produce unique scattering peaks in specific wavelength regions. For example, nitrogen- or phosphorus-containing chemical bonds usually show different frequency shifts in Raman spectra, and this information can be used to further identify and quantify the content of flame retardants. Extracting the peak position and intensity directly related to the content includes the following specific steps:

[0027] Step Y1, using a peak detection algorithm to identify significant peaks in the spectrum, wherein the peak detection algorithm is based on the first-order derivative to find the inflection point of the spectrum and determine its wavelength or wave value;

[0028] Step Y2, the intensity of the peak reflects the strength of the light absorbed in this band, which can be obtained by calculating the maximum absorption value of each peak in the spectrum. The intensity of the peak is proportional to the content of the substance. Therefore, by measuring the intensity of these peaks, the concentration or content of the flame retardant can be inferred.

[0029] As a further solution of the present invention, in step S33, the combustion experiment result of step S2 is used as a label, and the spectral data collected in step S1 is used as a feature input. In order to obtain the best prediction model, a variety of classic machine learning algorithms are used for preliminary training and comparison. The machine learning algorithms include support vector machine, inverse neural network, CatBoost, BO-CatBoost, XGBoost and BO-XGBoost. The BO-XGBoost model with the highest accuracy, precision, recall and F1 score is selected as the final machine prediction model.

[0030] As a further solution of the present invention, in step S4, the Bayesian optimization method is used to search the hyperparameters of the XGBoost model, and the initial hyperparameter setting range is: max_depth is between 3 and 10, learning_rate is between 0.01 and 0.3, gamma is between 0 and 5, and subsample is between 0.5 and 1.0; the optimal hyperparameter results are: max_depth=7, learning_rate=0.15, gamma=1.2, subsample=0.8.

[0031] The technical effects and advantages of the non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box of the present invention: The present invention combines spectral technology with machine learning and prediction technology to achieve non-destructive testing of the flame retardant properties of the non-metallic shell of the electric energy meter box. Accurate data is obtained as labels by collecting multiple samples for flame combustion tests, and non-destructive spectral data is obtained by in-situ spectral detection. After preprocessing and feature extraction, a variety of machine learning algorithms are used for training and optimization, and finally the Bayesian optimized BO-XGBoost model prediction is selected. The method overcomes the limitations of existing destructive testing and sampling detection, can detect products one by one, and ensure the safety of power equipment. The testing process is non-destructive and does not affect the subsequent use of samples. It provides an efficient, accurate, non-destructive and widely applicable solution for the flame retardant performance test of the non-metallic shell of the electric energy meter box. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention is a flow chart of a non-destructive testing method for the flame retardant performance of a non-metallic shell of an electric energy meter box.

[0033] Figure 2 This is a combustion test result diagram of the first sample using a horizontal vertical combustion instrument in the present invention.

[0034] Figure 3 This is a combustion test result diagram of the second sample using a horizontal vertical combustion instrument of the present invention.

[0035] Figure 4This is a combustion test result diagram of the third sample using a horizontal vertical combustion instrument of the present invention. DETAILED DESCRIPTION

[0036] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0037] Example 1

[0038] See also Figure 1 The method flow chart shown in the figure is a non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box of the present invention. The spectral technology is combined with machine learning and prediction technology. The chemical composition spectrum of the test object is obtained by using the in-situ spectral measurement technology that does not cause damage to the sample. After large-scale test data learning and training of the machine learning model, the flame retardant properties of the material based on the in-situ spectral characteristics are predicted. The method comprises the following steps:

[0039] Step S1, collecting non-metallic shell samples of the electric energy meter box, and performing in-situ spectroscopy detection on the collected non-metallic shell samples using a portable infrared spectrometer, a portable Raman spectrometer, or an infrared / Raman spectrometer equipped with a movable optical fiber probe to collect spectral data;

[0040] Step S11, widely collect samples of electric energy meter boxes from various brands and manufacturers on the market to ensure the diversity and representativeness of test data and obtain non-metallic shells;

[0041] Step S12, using clean water and anhydrous ethanol to alternately clean the sample to remove dust, oil and other pollutants on the surface to ensure the accuracy of the flame combustion test;

[0042] Step S2, performing a flame combustion test on the collected non-metallic shell samples of the electric energy meter box;

[0043] Step S3, spectral data preprocessing and establishment of machine learning model;

[0044] Step S31, performing Savitzky-Golay filtering, baseline correction and normalization processing on the collected spectral data in sequence;

[0045] Step S32, extracting key band features from the original spectrum, including information such as peak position and intensity directly related to the composition and content of the flame retardant;

[0046] Step S33, using support vector machine (SVM), inverse neural network, CatBoost, BO-CatBoost, XGBoost and BO-XGBoost (Bayesian optimized XGBoost) for preliminary model training, and selecting BO-XGBoost model as the final prediction model;

[0047] Step S4, using the Bayesian optimization method to search for the hyperparameters of the XGBoost model, including the maximum depth of the tree (max_depth), the learning rate (learning_rate), the minimum split loss (gamma), and the subsample ratio (subsample), and evaluating the model performance through five-fold cross validation;

[0048] Step S5, using the optimized BO-XGBoost model to predict and identify the flame retardant properties of the non-metallic shell material of the electric energy meter box;

[0049] Step S51, using an infrared or Raman spectrometer capable of moving an optical fiber probe to in-situ test the non-metallic shell of the electric energy meter box to obtain spectral data of the test sample;

[0050] Step S52, using the optimized BO-XGBoost model to read the spectral data of the test sample, predict the flame retardant properties of the test sample, and output the result as "qualified sample" or "unqualified sample";

[0051] In step S53, the samples predicted by the optimized BO-XGBoost model are further verified by the combustion test in step S13.

[0052] Furthermore, in step S2, the flame retardant performance test is carried out in accordance with the national standard GB / T 2408-2021. Specifically, the non-metallic shell sample after cleaning and drying is cut into test strips with a length of 125.0mm±5.0mm, a width of 13.0mm±0.5mm, and a thickness of <13.0mm to ensure that its size meets the standard requirements. The test adopts horizontal and vertical combustion test equipment to carry out combustion tests under specified flame height, ignition time (10 seconds or 30 seconds) and test environment (23±2℃, relative humidity 50±10%). During the test, the flame propagation speed, self-extinguishing time after flame removal, dripping phenomenon and whether cotton is ignited after the sample is burned are observed. According to the comprehensive performance of the combustion test, the flame retardant grade is divided into V0 (the highest grade, indicating short burning time, no droplets and strong self-extinguishing ability), V1 (slightly longer self-extinguishing time, but no burning droplets), V2 (burning droplets and igniting cotton) and other grades (unqualified samples). Test data, as the key label for model training, provides a basis for the construction of machine learning models, while improving data reliability through repeated experiments on different samples.

[0053] Furthermore, in step S31, a Savitzky-Golay filter is used to smooth the collected spectral data, and by setting a 7-point window width and a second-order polynomial order, random noise in the spectral signal is removed while retaining the key spectral characteristic peak shape; a least squares baseline fitting algorithm is used to perform baseline correction to eliminate spectral baseline drift caused by the test environment or instrument errors, ensuring that the spectral data is based on an accurate zero point; and maximum-minimum normalization is used to normalize the spectral data, mapping the spectral intensity value to the range of [0,1] to eliminate the influence of external factors such as sample thickness and light source intensity on the characteristic intensity.

[0054] Furthermore, in step S32, key band features are extracted from the original spectrum. Since flame retardants usually have unique characteristic absorption peaks in infrared spectra and Raman spectra, key bands related to these chemical components and contents are extracted from the spectrum. In infrared spectra, different chemical groups will produce absorption peaks in a specific wavenumber range. For example, phosphate groups usually have significant characteristic absorption peaks in the infrared spectrum range of 1200-1300cm-1, bromide will show characteristic peaks in the infrared spectrum range of 500-600cm-1, and chloride will show characteristic peaks in the infrared spectrum range of 600-800cm-1. -1 The infrared spectrum has weak characteristic absorption peaks. By analyzing these peaks in the sample spectrum, information related to the type and content of flame retardants can be extracted; in Raman spectroscopy, the vibration mode of chemical bonds will produce unique scattering peaks in specific wavelength regions. For example, chemical bonds containing nitrogen or phosphorus usually show different frequency shifts in Raman spectra, and this information can be used to further identify and quantify the content of flame retardants. Extracting the peak position and intensity directly related to the content includes the following specific steps:

[0055] Step Y1, using a peak detection algorithm to identify significant peaks in the spectrum, wherein the peak detection algorithm is based on the first-order derivative to find the inflection point of the spectrum and determine its wavelength or wave value;

[0056] Step Y2, the intensity of the peak reflects the strength of the light absorbed in this band, which can be obtained by calculating the maximum absorption value of each peak in the spectrum. The intensity of the peak is proportional to the content of the substance. Therefore, by measuring the intensity of these peaks, the concentration or content of the flame retardant can be inferred.

[0057] Furthermore, in step S33, the combustion experiment result of step S2 is used as a label, and the spectral data collected in step S1 is used as a feature input. In order to obtain the best prediction model, a variety of classic machine learning algorithms are used for preliminary training and comparison. The machine learning algorithms include support vector machine, inverse neural network, CatBoost, BO-CatBoost, XGBoost and BO-XGBoost. The following table shows the accuracy, precision, recall and F1 score indicators of support vector machine, inverse neural network, CatBoost, BO-CatBoost, XGBoost and BO-XGBoost on the training set.

[0058] Table 1 Accuracy, precision, recall and F1 score indicators of support vector machine, reverse neural network, CatBoost, XGBoost and BO-XGBoost on the training set

[0059] Model Accuracy (%) Accuracy (%) Recall rate (%) F1 score (%) Support Vector Machine 92.5 91.8 90.2 91.0 Inverse Neural Network 89.8 88.7 87.4 88.0 CatBoost 94.2 93.5 92.8 93.1 BO-CatBoost 95.0 94.5 93.7 94.1 XGBoost 93.6 92.3 91.7 92.0 BO-XGBoost 96.1 95.4 94.8 95.1

[0060] According to the accuracy, precision, recall and F1 score indicators of different models on the training set, the BO-XGBoost model was finally selected as the final machine prediction model.

[0061] Furthermore, in step S4, the Bayesian optimization method is used to search the hyperparameters of the XGBoost model, and the initial hyperparameter setting range is: max_depth is between 3 and 10, learning_rate is between 0.01 and 0.3, gamma is between 0 and 5, and subsample is between 0.5 and 1.0; the optimal hyperparameter results are: max_depth=7, learning_rate=0.15, gamma=1.2, subsample=0.8.

[0062] In this embodiment, 239 non-metallic shell samples of electric energy meter boxes from different manufacturers and different models were collected on the market, and the flame retardant properties of the shell samples were tested to obtain 206 qualified samples (flame retardant level V0) and 33 unqualified samples (flame retardant level below V0); the 239 collected samples were tested using a portable Raman spectrometer to obtain spectral data; the Raman spectral data obtained from the test were subjected to supervised machine learning training based on the flame retardant test results (qualified - flame retardant level V0, unqualified - flame retardant level lower than V0); according to step S3, step S4 and step S5 of the scheme of the present invention, the BO-XGBoost model was used for learning and training to obtain a prediction model; the prediction model obtained by training was used to evaluate the in-situ Raman spectral data of three untested flame retardant samples to predict the flame retardant properties of each sample; the three samples were subjected to flame retardant performance test using a horizontal and vertical combustion instrument to obtain the flame retardant properties of the three samples. The combustion test results of the three samples are as follows: Figure 2 , Figure 3 , Figure 4 The flame combustion test photos are shown in Table 2. The degree of conformity between the model prediction results and the combustion test detection of the three samples - the identification accuracy rate is shown in Table 2.

[0063] Example 2

[0064] The BO-XGBoost model trained in Example 1 was used to predict the Raman spectral data of 8 non-metallic electric energy meter box shell samples; the flame retardant properties of the samples were tested using a horizontal and vertical combustion instrument. The degree of conformity between the model prediction results and the sample combustion test results - the identification accuracy is shown in Table 2.

[0065] Table 2 The degree of conformity between the model prediction results and the combustion test detection - identification accuracy

[0066]

[0067]

[0068] The present invention combines spectral technology with machine learning and prediction technology to achieve non-destructive testing of the flame retardant properties of the non-metallic shell of the electric energy meter box. The flame retardant performance data is obtained by collecting multiple samples for flame combustion tests as labels, and the non-destructive spectral data is obtained by using spectral in-situ detection samples. After preprocessing and feature extraction, a variety of machine learning algorithms are used for training and optimization, and finally the Bayesian optimized BO-XGBoost model prediction is selected. The method overcomes the limitations of existing destructive testing and sampling detection, can detect products one by one, and ensure the safety of power equipment. The test process is non-destructive and does not affect the subsequent use of samples. It provides an efficient, accurate, non-destructive and widely applicable solution for the flame retardant performance test of the non-metallic shell of the electric energy meter box.

[0069] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0070] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A non-destructive testing method for the flame retardant properties of a non-metallic shell of an electric energy meter box, which uses an in-situ spectral measurement technique that does not damage the sample to obtain the chemical composition spectrum of the test object, and realizes the prediction of the flame retardant properties of the material based on the in-situ spectral characteristics, characterized in that: The steps include: Step S1, collecting non-metallic shell samples of the electric energy meter box, and performing in-situ spectroscopy detection on the collected non-metallic shell samples using a portable infrared spectrometer, a portable Raman spectrometer, or an infrared / Raman spectrometer equipped with a movable optical fiber probe to collect spectral data; Step S2, performing a flame combustion test on the collected non-metallic shell samples of the electric energy meter box; Step S3, spectral data preprocessing and establishment of machine learning model; Step S4, using the Bayesian optimization method to search for the hyperparameters of the XGBoost model, including the maximum depth of the tree, the learning rate, the minimum split loss, and the subsample ratio, and evaluating the model performance through five-fold cross validation; Step S5, using the optimized BO-XGBoost model to predict and identify the flame retardant properties of the non-metallic shell material of the electric energy meter box.

2. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In the step S1, collecting the non-metallic shell samples of the electric energy meter box specifically includes: widely collecting samples of electric energy meter boxes of various brands and manufacturers on the market to ensure the diversity and representativeness of the test data and obtain the non-metallic shell; using clean water and anhydrous ethanol to alternately clean the samples to remove dust, oil and other pollutants on the surface to ensure the accuracy of the flame combustion test.

3. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In step S2, the flame combustion test specifically includes: according to the GB / T2408-2021 standard, cutting the sample into a standard test strip and conducting a combustion test to obtain flame retardant performance data and provide labels for subsequent model training.

4. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In the step S3, the collected spectral data is smoothed by using Savitzky-Golay filtering, and the random noise in the spectral signal is removed by setting a 7-point window width and a second-order polynomial order, while retaining the key spectral characteristic peak shape; the baseline correction is performed by using a least squares baseline fitting algorithm to eliminate the spectral baseline drift caused by the test environment or instrument error, ensuring that the spectral data is based on an accurate zero point; The spectral data were normalized using maximum-minimum normalization to map the spectral intensity values ​​to the range of [0,1] to eliminate the influence of external factors such as sample surface finish and scattered light intensity on the characteristic intensity.

5. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In step S3, key band features are extracted from the original spectrum, including peak position and intensity information directly related to the flame retardant composition and content. Extracting the peak position and intensity directly related to the content includes the following specific steps: Step Y1, using a peak detection algorithm to identify significant peaks in the spectrum, wherein the peak detection algorithm is based on the first-order derivative to find the inflection point of the spectrum and determine its wavelength or wave value; Step Y2, the peak intensity reflects the strength of the test substance's absorption capacity for the spectrum in this band, which is obtained by calculating the maximum absorption value of the spectral peak. The peak intensity reflects the content of the substance. Therefore, by measuring the intensity of these peaks, the concentration or content of the flame retardant of a specific substance in the test sample is calculated.

6. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In step S3, support vector machine, inverse neural network, CatBoost, BO-CatBoost, XGBoost and BO-XGBoost are used for preliminary model training, and finally the BO-XGBoost model with the highest accuracy, precision, recall and F1 score on the training set is selected as the final machine prediction model.

7. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In step S4, the Bayesian optimization method is used to search the hyperparameters of the XGBoost model, and the initial hyperparameter setting range is: max_depth is between 3 and 10, learning_rate is between 0.01 and 0.3, gamma is between 0 and 5, and subsample is between 0.5 and 1.0; the optimal hyperparameter results are: max_depth=7, learning_rate=0.15, gamma=1.2, subsample=0.

8.

8. The non-destructive testing method for the flame retardant properties of the non-metallic shell of an electric energy meter box according to claim 1 is characterized in that: In the step S5, a portable infrared or Raman spectrometer that can be moved, or an infrared or Raman spectrometer with a movable optical fiber probe, is used to test the non-metallic shell of the electric energy meter box in situ to obtain spectral data of the test sample; the spectral data of the test sample is read in using the optimized BO-XGBoost model to predict the flame retardant properties of the test sample, and the output result is "qualified sample" or "unqualified sample". The sample predicted by the optimized BO-XGBoost model is further verified by the combustion test in step S1.

Images