A multi-physical environment electromagnetic sensitivity prediction and evaluation method

Abstract

The application discloses a kind of multi-physical environment electromagnetic sensitivity prediction and evaluation method, comprising the following steps: step S1. Constructing the electromagnetic sensitive characteristic map of target product under multi-physical environment, selecting multi-physical stress parameter for sampling;Step S2. Data preprocessing is carried out based on the data obtained by sampling, and the obtained sample set is divided;Step S3. Establishing LSTM-SVR hybrid model, and realizing multi-physical environment electromagnetic sensitivity prediction and evaluation.The application constructs electromagnetic sensitive characteristic map, and based on electromagnetic sensitive characteristic map, carries out product electromagnetic sensitive characteristic test under the action of multi-physical stress in limited time, establishes product electromagnetic anti-interference ability degradation prediction model in whole life cycle, and provides strong support for product design and development.

Description

Technical Field

[0001] This invention relates to electromagnetic susceptibility assessment, and in particular to a method for predicting and assessing electromagnetic susceptibility in multi-physical environments. Background Technology

[0002] With the continuous development of materials technology, component technology, and electronic design technology, electronic and electrical products or systems are becoming increasingly complex and highly integrated, placing higher demands on product reliability. The main factors affecting product reliability include the natural environment, electromagnetic environment, mechanical structure, and manufacturing processes. In the natural environment, temperature, humidity, and air pressure can all affect the normal operation of electronic components, degrading their electrical performance or even damaging them, thus causing malfunctions. Electromagnetic waves are ubiquitous in the environment; under the influence of electromagnetic signals, the noise of electronic circuits increases, and their stability deteriorates. Severe interference can even lead to equipment malfunctions or endanger personal safety.

[0003] Currently, product reliability assessments and predictions are based on product requirements or the environment in which they operate, involving single environmental tests or electromagnetic susceptibility tests, with the goal of meeting the respective requirements. However, electronic and electrical products or systems in real-world environments are not only affected by natural factors such as temperature and humidity, but also by electromagnetic interference. The combined effect of these two factors, coupled with insufficient preventative measures, can lead to product instability or performance degradation during use. Furthermore, if the reliability requirements of the natural and electromagnetic environments are met separately during the product design phase, it may result in design redundancy or deficiencies, making it difficult to achieve the optimal cost-effectiveness ratio of the system. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for predicting and evaluating electromagnetic susceptibility in multi-physical environments. By constructing an electromagnetic susceptibility characteristic spectrum and conducting electromagnetic susceptibility characteristic tests of products under multi-physical stress within a limited time based on the electromagnetic susceptibility characteristic spectrum, a prediction model for the degradation of electromagnetic interference resistance of products throughout their entire life cycle is established, providing strong support for product design and development.

[0005] The objective of this invention is achieved through the following technical solution: a method for predicting and assessing electromagnetic susceptibility in multi-physics environments, comprising the following steps:

[0006] Step S1. Conduct electromagnetic susceptibility threshold tests on the target product under different multi-physical environments and frequency parameters, and construct an electromagnetic susceptibility characteristic spectrum; the multi-physical environment is also called environmental parameters, including temperature, humidity and air pressure;

[0007] Environmental parameters and frequency parameters were selected as multi-physical stress parameters for sampling to obtain data on temperature, humidity, air pressure, frequency, running time, and product electromagnetic sensitivity threshold.

[0008] Step S2. Perform data preprocessing based on the sampled data, including:

[0009] The sampled data is normalized, split into N groups of two-dimensional sequences, and converted into sliding window samples. The normalized electromagnetic sensitivity threshold of each sliding window sample at the next time step is used as the sample label, and the samples are divided into training set and test set.

[0010] Step S3. Establish an LSTM-SVR hybrid model and realize the prediction and assessment of electromagnetic susceptibility in multi-physics environments, including:

[0011] An LSTM-SVR hybrid model consisting of an LSTM network and an SVR model is constructed. After training on the training set and testing on the test set, the trained LSTM-SVR model is obtained. The multi-physical environment aging time is input, the predicted value is output, and inverse normalization is performed to obtain the predicted electromagnetic susceptibility threshold. The electromagnetic susceptibility is evaluated by comparing it with the electromagnetic susceptibility limit of the target product.

[0012] In step S1, constructing the electromagnetic susceptibility characteristic spectrum of the target product under multiple physical environments includes:

[0013] S101. Conduct electromagnetic susceptibility tests on the target product under various physical environments to obtain the electromagnetic susceptibility thresholds of the target product under different physical environments.

[0014] The multi-physical environment, also known as environmental parameters, includes temperature, humidity, and air pressure.

[0015] The different multi-physical environments include standard conditions and non-standard conditions; the standard conditions refer to normal temperature, normal humidity, and standard atmospheric pressure.

[0016] Step S101 includes:

[0017] A1. Given multiple different multi-physical environments;

[0018] A2. For any multi-physical environment, the interference signal frequency and interference signal level are first preset to given initial values; whereby the interference signal frequency is also called the frequency parameter.

[0019] A201. Fix the frequency of the interference signal, start from the initial value of the interference signal level, and gradually increase the interference signal level. When the target product cannot work, record the interference signal level at this time, and record it as the electromagnetic sensitivity threshold corresponding to the current interference signal frequency in the current multi-physical environment.

[0020] A202. Gradually increase the frequency of the interference signal. At each interference signal frequency, repeat step A201 to obtain the electromagnetic sensitivity threshold corresponding to different interference signal frequencies in the current multi-physical environment. Stop when the interference signal exceeds the preset maximum frequency.

[0021] A2. For each multi-physical environment, repeat step A3 to obtain the test results of electromagnetic susceptibility thresholds for different multi-physical environments;

[0022] Preferably, the testing methods include conduction sensitivity testing and radiation sensitivity testing;

[0023] S102. Select the electromagnetic susceptibility threshold of the target product under standard conditions in a multi-physical environment, and plot the electromagnetic susceptibility threshold curve under standard conditions, i.e., the electromagnetic susceptibility response diagram under standard conditions.

[0024] In the electromagnetic susceptibility response diagram under standard conditions, the horizontal axis represents the frequency of the interference signal, and the vertical axis represents the level of the interference signal.

[0025] S103. Based on the actual environmental conditions faced by the target product during use, select multiple sets of target multi-physical environments consisting of temperature, humidity, and air pressure from step S101.

[0026] For each interference signal frequency, the maximum and minimum electromagnetic susceptibility threshold values ​​are selected from the test results of multiple targets in multiple physical environments.

[0027] Since there are multiple interference signal frequencies, the maximum electromagnetic sensitivity threshold of all interference signal frequencies constitutes the upper boundary curve of the electromagnetic sensitivity of the target product, with the vertical axis representing the maximum electromagnetic sensitivity threshold and the horizontal axis representing the interference signal frequency.

[0028] The minimum electromagnetic susceptibility threshold of all interference signal frequencies constitutes the lower boundary curve of the electromagnetic susceptibility of the target product, with the vertical axis representing the minimum electromagnetic susceptibility threshold and the horizontal axis representing the interference signal frequency.

[0029] The upper and lower boundary curves of electromagnetic sensitivity constitute the sensitive boundary of multi-physical stress variation.

[0030] S104. Subtract the maximum and minimum values ​​of the electromagnetic sensitivity thresholds of different interference signal frequencies under multiple target multi-physical environments to obtain the fluctuation difference value. Obtain the fluctuation difference value of the electromagnetic sensitivity threshold of the target product under different physical environments, and plot the electromagnetic sensitivity fluctuation curve of the product under multi-physical stress changes as the sensitivity fluctuation characteristic.

[0031] The vertical axis represents the fluctuation difference, and the horizontal axis represents the frequency of the interference signal;

[0032] S105. Construct an electromagnetic susceptibility profile of the product under multi-physical environment from three dimensions: electromagnetic susceptibility response under standard conditions, sensitive boundary of multi-physical stress change, and sensitive fluctuation characteristics.

[0033] The beneficial effects of this invention are as follows: By constructing an electromagnetic susceptibility characteristic spectrum, this invention can identify electromagnetic susceptibility failure points of products under different regional climate conditions, thereby enabling targeted electromagnetic protection design; at the same time, based on the electromagnetic susceptibility characteristic spectrum, electromagnetic susceptibility characteristic tests of products under multi-physical stress are conducted within a limited time, establishing a predictive model for the degradation of electromagnetic interference resistance of products throughout their entire life cycle, providing strong support for product design and development, avoiding excessive or insufficient preventive reserves, achieving the best cost-effectiveness ratio of the system, and ensuring the electromagnetic reliability of products throughout their entire life cycle. Attached Figure Description

[0034] Figure 1 Electromagnetic susceptibility characteristics of the product under multi-physical environments;

[0035] Figure 2 This is a schematic diagram of the electromagnetic sensitivity response of a sensor under standard conditions.

[0036] Figure 3 A schematic diagram for evaluating whether the sensor meets the requirements of the Chinese military standard CS114;

[0037] Figure 4 This is a schematic diagram of the electromagnetically sensitive boundary of a sensor under multiple physical stresses.

[0038] Figure 5 This is a schematic diagram of the electromagnetic sensitivity wave behavior of a certain sensor under multiple physical stresses.

[0039] Figure 6 This is a flowchart illustrating a method for predicting electromagnetic susceptibility under long-term multi-physical stress. Detailed Implementation

[0040] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0041] This invention constructs an electromagnetic susceptibility characteristic map of a product under multiple physical environments, thereby achieving a comprehensive assessment of the product's electromagnetic susceptibility characteristics. This provides strong support for the electromagnetic protection design of the product and enables effective early warning of electromagnetic safety risks during operation in different regional climates. By combining the electromagnetic susceptibility characteristic map with regression modeling based on Gaussian process regression, the relevant characteristics of multiple physical stresses are fully extracted. From short-term, limited test data, a nonlinear mapping relationship is accurately established between the long-term action parameters of multiple physical stresses and the product's electromagnetic susceptibility response, thus achieving high-precision prediction of the sensitivity threshold curve under product aging. This method can strongly support the electromagnetic protection design of products, improve the electromagnetic reliability of products throughout their entire life cycle, and has practical reference value. Specifically:

[0042] A method for predicting and assessing electromagnetic susceptibility in multi-physics environments includes the following steps:

[0043] Step S1. Construct an electromagnetic susceptibility characteristic spectrum of the target product under multi-physical environment, and select multi-physical stress parameters for sampling;

[0044] 1. Multi-physical environment electromagnetic susceptibility assessment method

[0045] This invention provides a method for evaluating the electromagnetic susceptibility of products in multiple physical environments. It constructs an electromagnetic susceptibility characteristic spectrum of the product under these environments, addressing the large and diverse electromagnetic susceptibility test data generated in various physical environments. Figure 1 As shown, this method enables the identification of electromagnetically sensitive failure points of products under different regional climate conditions. The following example illustrates this using a certain model of sensor.

[0046] 1) Multi-physical environment electromagnetic susceptibility testing was conducted on a certain model of sensor to obtain its electromagnetic susceptibility threshold under different temperature, humidity, and air pressure conditions. Electromagnetic interference response data were selected under normal temperature, normal humidity, and standard atmospheric pressure conditions to plot the electromagnetic susceptibility threshold curve under standard conditions (temperature 20℃, humidity 50%, air pressure 101KPa), i.e., the electromagnetic susceptibility response diagram under standard conditions. The electromagnetic susceptibility threshold curve under standard conditions is divided into conducted immunity sensitivity curves (e.g., ...) according to the test scenario. Figure 2 The graph shows the radiated immunity sensitivity curve and the conducted immunity sensitivity response graph. The vertical axis of the graph represents the conducted interference signal level, with units such as dBm and dBmA depending on the specific test scenario. The horizontal axis represents the frequency range of conducted immunity, which is from 9kHz to 80MHz, and can be extended to 1GHz in some scenarios. Figure 2 The legend in the upper right corner indicates the standard physical environment conditions under which the product operates. The vertical axis of the radiated immunity sensitivity response graph represents the radiated interference signal level, with units such as dBpT and dBmV / m depending on the specific test scenario. The horizontal axis represents the radiated immunity frequency, typically ranging from 80MHz to 6GHz, but may extend down to 20MHz for specific products.

[0047] In electromagnetic susceptibility testing, the sensitivity scan frequency step size is set to obtain the product sensitivity threshold test data. The sensitivity scan step size is shown in Table 1.

[0048] surface Sensitive scan step size in relevant standards

[0049]

[0050] Based on the application scenario of a certain sensor, determine the electromagnetic immunity standard that the sensor must meet. If the sensor is to be pre-equipped on a certain type of naval vessel, it needs to meet the limit requirements of curve two in CS114 of the National Military Standard GJB151B-2013. Convert the unit metric of the sensor's electromagnetic susceptibility response under standard conditions to dBuA and compare it with the limit curve two of CS114. Figure 3 As shown, it can be found that the 150-250MHz range is difficult to meet the requirements of the national military standard for conducted immunity, and targeted rectification measures need to be taken to meet the factory requirements of the sensor.

[0051] Based on the extreme environmental conditions the product faces during use, a range of temperature, humidity, and air pressure levels is selected. The electromagnetic susceptibility threshold of the product is then traversed within this multi-physical stress range to obtain the maximum and minimum values. An electromagnetic susceptibility boundary curve under varying multi-physical stresses is then plotted, yielding the electromagnetic susceptibility boundary of the product under these stresses. The electromagnetic susceptibility boundary under multi-physical stresses is categorized into conducted immunity electromagnetic susceptibility boundary and radiated immunity electromagnetic susceptibility boundary, depending on the test scenario. Figure 4 The electromagnetic susceptibility boundary of a certain sensor under various physical stress conditions (-50℃~100℃, humidity 0~100%, air pressure 330~1040 hPa) was compared with the CS114 limit curve II (yellow line in the figure). Within the frequency range of 25~100MHz, the electromagnetic susceptibility boundary of this sensor overlaps with the limit curve. This indicates a risk of electromagnetic safety failure when the sensor operates in environments with these conditions. Therefore, it is necessary to improve the conducted immunity capability within this frequency band to meet the electromagnetic safety requirements of the sensor operating in different environments.

[0052] 2) Subtract the maximum and minimum values ​​of the product's electromagnetic susceptibility threshold under different physical stresses to obtain the fluctuation difference of the product's electromagnetic susceptibility threshold under different physical environments, and plot the electromagnetic susceptibility fluctuation characteristics of the product under the above-mentioned multi-physical stress changes. The electromagnetic susceptibility fluctuation under multi-physical stress is divided into conducted immunity electromagnetic susceptibility fluctuation and radiated immunity electromagnetic susceptibility fluctuation according to the test scenario, where the vertical axis is measured in dB and the horizontal axis is frequency. Figure 5This study examines the conducted immunity and electromagnetic susceptibility fluctuations of a certain sensor model under various physical stress conditions, including -50℃ to 100℃, humidity 0 to 100%, and air pressure 330 to 1040 hPa. It indicates that the sensor exhibits poor conducted immunity in frequency bands below 100MHz and above 200MHz when these environmental parameters change. Therefore, the robustness of its conducted immunity should be specifically improved during the electromagnetic protection design phase.

[0053] In summary, an electromagnetic susceptibility profile of a product under multi-physical environments is constructed from three dimensions: sensitive response under standard conditions, sensitive boundary under multi-physical stress changes, and sensitive fluctuation characteristics. This enables a comprehensive assessment of the product's electromagnetic susceptibility characteristics, providing strong support for the product's electromagnetic protection design and enabling effective early warning of electromagnetic safety risks when the product operates under different regional climates.

[0054] Step S2. Perform data preprocessing based on the sampled data and divide the resulting sample set;

[0055] Electronic and electrical equipment / systems subjected to prolonged multi-physical stresses face the risk of electromagnetic interference degradation. This invention, based on a combination of electromagnetic susceptibility characteristic maps and a hybrid model (LSTM-SVR), fully extracts the relevant characteristics of multi-physical stresses and accurately establishes a nonlinear mapping relationship between long-term multi-physical stress parameters and the product's electromagnetic susceptibility response from limited test data. This enables the prediction and assessment of the sensitivity threshold of products under time-related aging in multi-physical environments. Figure 6 As shown.

[0056] First, based on the multi-physical environment electromagnetic susceptibility assessment method, an electromagnetic susceptibility characteristic spectrum of a product under multi-physical environments is constructed. Based on this spectrum, environmental parameters such as temperature, humidity, and air pressure corresponding to the lower boundary curve of electromagnetic susceptibility under multi-physical stress are selected. and frequency parameters Temperature, humidity, air pressure, and frequency parameters are fixed at several values, i.e. ; ,in For the set of frequency points, The elements in the frequency point set are read out from the lower boundary of electromagnetic susceptibility under multi-physical stress and correlated with environmental parameters. Correspondingly.

[0057] The sampling process requires m sets of tests. In each set of tests, after 8 hours of stable operation, the environment must be restored to standard conditions for electromagnetic susceptibility testing. In each of the m sets of electromagnetic susceptibility tests, [the following parameters are selected]. Corresponding frequency set Mid-frequency scanning sampling is performed to obtain product sensitivity threshold test data, completing the sampling of the product's electromagnetic susceptibility. The test time is set to t hours, and the range of time t is determined based on the product's lifespan. It is expected that N=t / 8 sets of sampling data will be obtained.

[0058] The absolute value fluctuation of the sensitivity threshold obtained during the sampling process should not exceed 3dB and must fall within the electromagnetic sensitivity boundary under multiple physical stresses of the product. If this condition is not met, the sampling data at that frequency point will be considered as bad value, and bad values ​​will be removed before resampling.

[0059] After sampling, six complete data points were obtained, including temperature, humidity, air pressure, frequency, running time, and product electromagnetic sensitivity threshold. The data consisted of N sets of six-dimensional discrete points, as shown below.

[0060]

[0061] In the formula It is the total number of data points; Represents temperature; Represents humidity; Represents dry air pressure; Representing interference frequencies, with a total quantity of k; The total operating time of the representative product in the multi-physical environment chamber during the i-th measurement, in hours; It is the electromagnetic sensitivity threshold when the product fails;

[0062] The difference between the maximum and minimum electromagnetic susceptibility threshold values ​​(Diff) in the complete test data is used as a ratio. All electromagnetic susceptibility threshold values ​​(X) in the data are divided by this ratio for normalization. This compresses all electromagnetic susceptibility threshold values ​​in the complete test data to [0,1]. The compressed electromagnetic susceptibility threshold is denoted as... ;

[0063] The normalized N sets of six-dimensional discrete points are divided into N sets of two-dimensional sequences according to frequency points. Then, convert the N sets of two-dimensional sequences into sliding window samples in a supervised learning format:

[0064] , ,..., ;

[0065] Sample the next time step for each sliding window. As a sample label;

[0066] Since each sample is a small sample, the sample data is divided into training set and test set according to the ratio of 80% / 20%.

[0067] For different frequency points, repeat the above process of splitting; and model and train them separately during modeling.

[0068] Step S3. Establish an LSTM-SVR hybrid model and realize the prediction and assessment of electromagnetic susceptibility in multi-physics environments.

[0069] Establishing an LSTM-SVR hybrid model:

[0070] An LSTM network is used to model the training set of single-frequency electromagnetic susceptibility thresholds to obtain a hidden state that contains the time-series characteristics of the product's electromagnetic susceptibility aging. ,in, Indicates that the products in the training set have undergone The single-frequency electromagnetic susceptibility threshold during aging time, i.e., the single-frequency electromagnetic susceptibility threshold sample. These represent the weight matrices of the input layer and the hidden layer, respectively. This represents the bias vector. Indicates the activation function;

[0071] Then hide the state Input into the SVR model to obtain Among them, the SVR model is the support vector regression model. This represents the training set for single-frequency electromagnetic susceptibility thresholds. K represents the kernel function; represents the coefficients of the support vector machine; b represents the bias term; Indicates that the product has undergone Predicted electromagnetic susceptibility threshold at a single frequency point during aging time;

[0072] For the established LSTM-SVR hybrid model, grid search cross-validation is set up, and the LSTM-SVR hybrid model is trained and its parameters are tuned using the training set;

[0073] The trained model is tested on the test set. When the test accuracy reaches the set threshold, a product electromagnetic sensitivity threshold aging prediction model based on the collaborative LSTM-SVR model is obtained.

[0074] The above-obtained collaborative LSTM-SVR model is used to predict the electromagnetic susceptibility threshold of a product after aging in multiple physical environments. The aging time in multiple physical environments is input, and the predicted value is output. The same scaler used during training is used for inverse normalization to obtain the predicted electromagnetic susceptibility threshold.

[0075] In the embodiments of this application, the mean square error and mean absolute error between the predicted value and the actual electromagnetic susceptibility threshold are calculated to evaluate the effectiveness of the model. Compared with three commonly used regression algorithms—cubic spline interpolation, support vector machine regression, and generalized regression neural network—the absolute prediction error of this model is generally smaller than that of other methods.

[0076] In the embodiments of this application, a validated hybrid model is used to predict the electromagnetic susceptibility threshold after 1000 hours of aging, and the predicted value is compared with the electromagnetic susceptibility limit of the product, such as the CS114 limit curve. If the predicted value falls below the limit, it indicates that the product has a risk of electromagnetic safety failure during long-term use, and its immunity needs to be improved so that the target product meets electromagnetic safety requirements throughout its entire life cycle.

[0077] This invention proposes a multi-physical environment electromagnetic susceptibility prediction and assessment method, which can assess and predict the electromagnetic susceptibility of electronic and electrical products and systems under various physical environments such as different temperatures, humidity levels, and air pressures, obtaining a more realistic baseline for electromagnetic susceptibility throughout the entire lifecycle. Compared to existing methods that focus on the electromagnetic susceptibility of a single product, this method offers the following advantages:

[0078] This technology enables the assessment of regional climate sensitivity and the prediction of electromagnetic interference immunity degradation under the effects of aging over time. It diagnoses the electromagnetic safety vulnerabilities of products throughout their entire life cycle, providing strong support for product design and development. It also enables targeted electromagnetic protection design, avoiding excessive or insufficient preventative reserves, shortening the system development cycle, achieving the best cost-effectiveness ratio, and ensuring the electromagnetic reliability of products throughout their entire life cycle.

[0079] The foregoing description illustrates and describes a preferred embodiment of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for predicting and assessing electromagnetic susceptibility in multi-physical environments, characterized in that: Includes the following steps: Step S1. Conduct electromagnetic susceptibility threshold tests on the target product under different multi-physical environments and frequency parameters, and construct an electromagnetic susceptibility characteristic spectrum; the multi-physical environment is also called environmental parameters, including temperature, humidity and air pressure; Environmental parameters and frequency parameters were selected as multi-physical stress parameters for sampling to obtain temperature, humidity, air pressure, frequency, running time, and electromagnetic sensitivity threshold. Step S2. Perform data preprocessing based on the sampled data, including: The sampled data is normalized, split into N groups of two-dimensional sequences, and converted into sliding window samples. The normalized electromagnetic sensitivity threshold of each sliding window sample at the next time step is used as the sample label, and the samples are divided into training set and test set. Step S2 includes: S201. After sampling, six-dimensional complete data of temperature, humidity, air pressure, frequency, running time, and product electromagnetic sensitivity threshold are obtained. The data consists of N sets of six-dimensional discrete points, as shown below. ； In the formula It is the total number of data points; Represents temperature; Represents humidity; Represents air pressure; Representing interference frequencies, the total number is k; The time the product spent operating in the multi-physical environment chamber during the i-th measurement, expressed in hours, is considered the aging time. It is the electromagnetic sensitivity threshold when the product fails. S202. Perform data preprocessing: The complete test data is normalized, and all data points are compressed to [0,1]. The difference between the maximum and minimum electromagnetic susceptibility threshold values ​​(Diff) in the complete test data is used as a ratio. All electromagnetic susceptibility threshold values ​​(X) in the data are divided by this ratio for normalization. This compresses all electromagnetic susceptibility threshold values ​​in the complete test data to [0,1]. The compressed electromagnetic susceptibility threshold is denoted as... ; The normalized N sets of six-dimensional discrete points are divided into N sets of two-dimensional sequences according to frequency points. Then, convert the N sets of two-dimensional sequences into sliding window samples in a supervised learning format: , ,..., ； Sample the next time step for each sliding window. As a sample label; S203. The sample data is used to form a sample set, which is then divided into a training set and a test set for electromagnetic susceptibility thresholds at a single frequency point. For different frequency points, repeat the above process of splitting; and model and train them separately during modeling. Step S3. Establish an LSTM-SVR hybrid model and realize the prediction and assessment of electromagnetic susceptibility in multi-physics environments, including: An LSTM-SVR hybrid model consisting of an LSTM network and an SVR model is constructed. After training on the training set and testing on the test set, the trained LSTM-SVR model is obtained. The multi-physical environment aging time is input, the predicted value is output, and inverse normalization is performed to obtain the predicted electromagnetic susceptibility threshold. The electromagnetic susceptibility is evaluated by comparing it with the electromagnetic susceptibility limit of the target product.

2. The method for predicting and assessing electromagnetic susceptibility in multi-physics environments according to claim 1, characterized in that: In step S1, constructing the electromagnetic susceptibility characteristic spectrum of the target product under multiple physical environments includes: S101. Conduct electromagnetic susceptibility tests on the target product under various physical environments to obtain the electromagnetic susceptibility thresholds of the target product under different physical environments. The multi-physical environment, also known as environmental parameters, includes temperature, humidity, and air pressure. The different multi-physical environments include standard conditions and non-standard conditions; the standard conditions refer to normal temperature, normal humidity, and standard atmospheric pressure. Step S101 includes: A1. Given multiple different multi-physical environments; A2. For any multi-physical environment, the interference signal frequency and interference signal level are first preset to given initial values; whereby the interference signal frequency is also called the frequency parameter. A201. Fix the frequency of the interference signal, start from the initial value of the interference signal level, and gradually increase the interference signal level. When the target product cannot work, record the interference signal level at this time, and record it as the electromagnetic sensitivity threshold corresponding to the current interference signal frequency in the current multi-physical environment. A202. Gradually increase the frequency of the interference signal. At each interference signal frequency, repeat step A201 to obtain the electromagnetic sensitivity threshold corresponding to different interference signal frequencies in the current multi-physical environment. Stop when the interference signal exceeds the preset maximum frequency. A3. For each multi-physical environment, repeat step A2 to obtain the test results of electromagnetic susceptibility thresholds for different multi-physical environments; S102. Select the electromagnetic susceptibility threshold of the target product under standard conditions in a multi-physical environment, and plot the electromagnetic susceptibility threshold curve under standard conditions, i.e., the electromagnetic susceptibility response diagram under standard conditions. In the electromagnetic susceptibility response diagram under standard conditions, the horizontal axis represents the frequency of the interference signal, and the vertical axis represents the level of the interference signal. S103. Based on the actual environmental conditions faced by the target product during use, select multiple sets of target multi-physical environments consisting of temperature, humidity, and air pressure from step S101. For each interference signal frequency, the maximum and minimum electromagnetic susceptibility threshold values ​​are selected from the test results of multiple targets in multiple physical environments. Since there are multiple interference signal frequencies, the maximum electromagnetic susceptibility threshold of all interference signal frequencies constitutes the upper boundary curve of the electromagnetic susceptibility of the target product; the minimum electromagnetic susceptibility threshold of all interference signal frequencies constitutes the lower boundary curve of the electromagnetic susceptibility of the target product. The upper and lower boundary curves of electromagnetic sensitivity constitute the sensitive boundary of multi-physical stress variation. S104. Subtract the maximum and minimum values ​​of the electromagnetic sensitivity thresholds of different interference signal frequencies under multiple target multi-physical environments to obtain the fluctuation difference value. Obtain the fluctuation difference value of the electromagnetic sensitivity threshold of the target product under different physical environments, and plot the electromagnetic sensitivity fluctuation curve or bar chart of the product under multi-physical stress changes as the sensitivity fluctuation characteristic. The vertical axis represents the fluctuation difference, and the horizontal axis represents the frequency of the interference signal; S105. Construct an electromagnetic susceptibility profile of the product under multi-physical environment from three dimensions: electromagnetic susceptibility response under standard conditions, sensitive boundary of multi-physical stress change, and sensitive fluctuation characteristics.

3. The method for predicting and assessing electromagnetic susceptibility in multi-physics environments according to claim 2, characterized in that: In step S1, selecting multiple physical stress parameters and sampling includes: Based on the electromagnetic susceptibility characteristic spectrum of the product under multi-physical environment, select the environmental parameters corresponding to the lower boundary curve of electromagnetic susceptibility under multi-physical stress of the product. and frequency parameters , Temperature, humidity, air pressure, and frequency parameters are fixed at Above, that is ; ,in For the set of frequency points, The elements in the frequency point set are read from the lower boundary of electromagnetic susceptibility under multi-physical stress and correlated with environmental parameters. Correspondingly; environmental parameters Including temperature, humidity, and air pressure; During the sampling process, m sets of tests are required. In each set of tests, after running stably for p hours, the environment must be restored to standard conditions for electromagnetic susceptibility testing to obtain the electromagnetic susceptibility threshold when the product fails. Selecting from m groups of electromagnetic susceptibility tests respectively Corresponding frequency set Mid-frequency scanning and sampling are performed to obtain product sensitivity threshold test data, thus completing the sampling of the product's electromagnetic susceptibility; The test time is set to t hours, and the range of time t is determined based on the product's lifespan, resulting in N = t / p sets of sampled data; The absolute value fluctuation of the sensitivity threshold obtained during the sampling process should not exceed 3dB and must fall within the electromagnetic sensitivity boundary under multiple physical stresses of the product. If this condition is not met, the sampling data at that frequency point will be considered as bad value, and bad values ​​will be removed before resampling.

4. The method for predicting and assessing electromagnetic susceptibility in multi-physics environments according to claim 1, characterized in that: Step S3 includes: Establishing an LSTM-SVR hybrid model: An LSTM network is used to model the training set of single-frequency electromagnetic susceptibility thresholds to obtain a hidden state that contains the time-series characteristics of the product's electromagnetic susceptibility aging. ,in, Indicates that the products in the training set have undergone The single-frequency electromagnetic susceptibility threshold during aging time, i.e., the single-frequency electromagnetic susceptibility threshold sample. These represent the weight matrices of the input layer and the hidden layer, respectively. This represents the bias vector. Indicates the activation function; Then hide the state Input into the SVR model to obtain Among them, the SVR model is the support vector regression model. This represents the training set for single-frequency electromagnetic susceptibility thresholds. K represents the kernel function; represents the coefficients of the support vector machine; b represents the bias term; Indicates that the product has undergone Predicted electromagnetic susceptibility threshold at a single frequency point during aging time; For the established LSTM-SVR hybrid model, grid search cross-validation is set up, and the LSTM-SVR hybrid model is trained and its parameters are tuned using the training set; The trained model is tested on the test set. When the test accuracy reaches the set threshold, a product electromagnetic sensitivity threshold aging prediction model based on the collaborative LSTM-SVR model is obtained. The above-obtained collaborative LSTM-SVR model is used to predict the electromagnetic susceptibility threshold of a product after aging in multiple physical environments. The aging time in multiple physical environments is input, and the predicted value is output. The same scaler used during training is used for inverse normalization to obtain the predicted electromagnetic susceptibility threshold.

5. The method for predicting and assessing electromagnetic susceptibility in multi-physics environments according to claim 1, characterized in that: In step S3, after obtaining the predicted electromagnetic susceptibility threshold, an evaluation step is also included: The predicted electromagnetic susceptibility threshold is compared with the electromagnetic susceptibility limit of the target product: If the predicted electromagnetic susceptibility threshold falls below the electromagnetic susceptibility limit, it indicates that the target product has a risk of electromagnetic safety failure during long-term use and its immunity needs to be improved.

Images