Electronic component service life intelligent prediction method and system

Abstract

The invention relates to the technical field of electronic components, in particular to an intelligent prediction method and system for the service life of an electronic component, and the method comprises the following steps: S1, data collection: carrying out the real-time monitoring of the electronic component through a plurality of sensors, and collecting the multi-dimensional data; s2, data preprocessing: preprocessing the collected multi-modal data; s3, data fusion and feature enhancement: fusing the preprocessed multi-modal data to form comprehensive features, and performing feature enhancement on the comprehensive features; s4, life prediction: constructing a life prediction model, and performing life prediction on the electronic component; and S5, dynamically adjusting the prediction result: dynamically adjusting the life prediction result of the electronic component according to the actual use condition. According to the method, the influence of complex working conditions on the service life of the component can be reflected in real time, and the prediction result after dynamic adjustment better fits an actual application scene.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic components, and in particular to a method and system for intelligently predicting the life of electronic components. Background Art

[0002] With the rapid development of modern industry and electronic technology, electronic components play a vital role in various equipment and systems. Their performance and lifespan directly affect the operating efficiency and reliability of the entire system. In a complex working environment, electronic components are affected by multiple factors such as temperature changes, vibrations, mechanical strains, and workloads. Their health status may change dynamically. In order to ensure the normal operation of equipment and reduce the risk of downtime caused by component failure, a method that can accurately predict the life of components is urgently needed to detect potential problems in advance, optimize equipment maintenance strategies, and improve the efficiency of component use.

[0003] Existing methods for predicting the life of electronic components usually rely on a single data source, such as temperature or vibration signals. However, due to the complexity of the working environment and the diversity of component failure modes, a single data source often cannot fully reflect the health status of the components. In addition, existing methods have limited fusion processing capabilities for multimodal data, and the feature extraction and prediction models are not accurate enough. They are easily disturbed by noise and abnormal data, resulting in instability and inaccuracy in the prediction results. At the same time, most traditional prediction methods ignore the dynamic changes in actual operating conditions, which makes the prediction results less adaptable in practical applications and difficult to meet the needs of modern complex industrial scenarios.

[0004] In view of the shortcomings of the prior art, the present invention provides a method and system for intelligent prediction of the life of electronic components, which improves the accuracy and robustness of electronic component life prediction, solves problems such as single data source and insufficient model adaptability, and provides more reliable technical support for the intelligent management and maintenance of electronic components. Summary of the invention

[0005] The present invention provides a method and system for intelligently predicting the life of electronic components.

[0006] An intelligent prediction method for the life of electronic components comprises the following steps:

[0007] S1, data collection: real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency, and strain value;

[0008] S2, data preprocessing: preprocessing the collected multimodal data, including data denoising and outlier removal;

[0009] S3, data fusion and feature enhancement: fuse the preprocessed multimodal data to form comprehensive features, and enhance the comprehensive features;

[0010] S4, life prediction: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components;

[0011] S5, dynamic adjustment of prediction results: dynamically adjust the life prediction results of electronic components according to actual usage conditions (workload, environmental changes).

[0012] Optionally, the data collection in S1 includes:

[0013] S11, temperature data acquisition: use temperature sensors to monitor the operating temperature of components in real time;

[0014] S12, vibration intensity data collection: using a vibration sensor to measure the vibration intensity of components and collect vibration frequency and amplitude data;

[0015] S13, acoustic wave frequency data collection: using an acoustic wave sensor to capture the acoustic wave frequency related to the operation of the component;

[0016] S14, strain data acquisition: use strain sensors to obtain the strain values ​​of components during use to monitor the physical deformation of components.

[0017] Optionally, the data preprocessing in S2 includes:

[0018] S21, data denoising: using the moving average method to denoise the collected multimodal data;

[0019] S22, outlier removal: Use the Z-Score algorithm to detect and remove outliers.

[0020] Optionally, the data fusion and feature enhancement in S3 include:

[0021] S31, feature extraction: extracting features from the preprocessed multimodal data, including temperature variation amplitude, vibration intensity, frequency distribution, and strain peak value;

[0022] S32, data fusion: integrating the extracted features to generate comprehensive features;

[0023] S33, feature enhancement and redundancy removal: The self-attention mechanism is introduced to enhance the fused comprehensive features and remove redundant information.

[0024] Optionally, the feature extraction in S31 includes:

[0025] S311, temperature change feature extraction: The temperature change feature reflects the thermal state of electronic components under different working conditions. The temperature change feature is extracted by calculating the maximum temperature difference;

[0026] S312, vibration intensity feature extraction: the vibration intensity reflects the degree to which the components are affected by external forces. The characteristics of the vibration intensity are extracted by calculating the standard deviation of the vibration signal;

[0027] S313, frequency distribution feature extraction: The frequency distribution reflects the frequency components in the vibration or sound wave signal. The time domain signal is converted into a frequency domain signal through Fourier transform to extract the frequency feature;

[0028] S314, strain peak feature extraction: The strain peak feature is used to describe the maximum deformation of the material during operation, indicating whether the material has structural damage during use. The strain peak is extracted by analyzing the maximum value of the strain signal.

[0029] Optionally, the data fusion in S32 includes:

[0030] S321, feature normalization: normalizing the extracted features using a minimum-maximum normalization method;

[0031] S322, weighted fusion: Set the weight coefficient w according to the importance difference of multimodal data features. i Perform weighted summation on each feature to generate a comprehensive feature.

[0032] Optionally, the feature enhancement and redundancy removal in S33 includes:

[0033] S331, weight calculation of self-attention mechanism: The correlation weights between the features within the comprehensive features are calculated through the self-attention mechanism to extract the dependency relationship of key information in the comprehensive features;

[0034] S332, feature enhancement: performing weighted summation on each feature in the comprehensive feature according to the calculated attention weight to generate an enhanced feature;

[0035] S333, redundancy removal: By setting the attention weight threshold τ, the enhanced features are redundancy removed to generate the final features (the feature set after redundancy removal).

[0036] Optionally, the life prediction model in S4 adopts a Gaussian process regression (GPR) model, and the Gaussian process regression (GPR) model includes:

[0037] S41, feature input preparation: using the generated final features as input data, setting the final feature matrix of the training sample to X;

[0038] S42, kernel function definition: select radial basis function kernel (RBF kernel) as the kernel function of life prediction task;

[0039] S43, kernel matrix calculation: using the final feature matrix of the training samples as X, calculate the similarity between the training samples and construct the kernel matrix K;

[0040] S44, kernel function calculation: Given a new test sample F 最终,* , calculate its similarity with the training sample and use it to predict the life span value;

[0041] S45, life prediction: according to the kernel matrix K of the training sample and the kernel vector of the test sample, the life prediction value and uncertainty of the test sample are calculated;

[0042] S46, result output: output the prediction results of the Gaussian process regression (GPR) model, including the life prediction value and uncertainty estimation.

[0043] Optionally, the dynamic adjustment of the prediction result in S5 includes:

[0044] S51, workload adjustment factor calculation: Calculate the workload adjustment factor Δ based on the actual workload (operating frequency, current) of the components. load ;

[0045] S52, calculation of environmental change adjustment factor: Calculate the environmental change adjustment factor Δ based on the environmental conditions (temperature) env ;

[0046] Dynamically adjusted life prediction calculation: combined with workload adjustment factor Δ load and environmental change adjustment factor Δ env , dynamically adjust the life prediction value to obtain the corrected final life prediction value.

[0047] An electronic component life intelligent prediction system, used to implement the above-mentioned electronic component life intelligent prediction method, includes the following modules:

[0048] Data acquisition module: Real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency and strain value;

[0049] Data preprocessing module: preprocess the collected multimodal data, including data denoising and outlier removal;

[0050] Data fusion and feature enhancement module: fuses the preprocessed multimodal data to form comprehensive features, and enhances the comprehensive features;

[0051] Life prediction module: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components;

[0052] Dynamic adjustment module: dynamically adjust the life prediction results according to the actual use conditions of electronic components to optimize the life prediction results.

[0053] Beneficial effects of the present invention:

[0054] The present invention uses a variety of sensors to monitor the operating status of electronic components in real time, comprehensively collects multi-dimensional data such as temperature, vibration intensity, sound wave frequency and strain value, and uses data denoising and outlier removal technology to pre-process the collected data, thereby effectively improving the accuracy and consistency of the data, solving the limitation that a single data source cannot fully reflect the health status of components, and providing high-quality basic data support for subsequent feature extraction and life prediction.

[0055] The present invention accurately captures the key change patterns in various sensor data through feature extraction, data fusion and feature enhancement. The fused comprehensive features can comprehensively reflect the dynamic operation status of electronic components, and enhance the expression ability of the comprehensive features through the self-attention mechanism. At the same time, redundant information is removed, and the weight distribution of key features is further optimized to avoid the interference of redundant information, which significantly improves the accuracy and robustness of the prediction model. Through the Gaussian process regression model, a nonlinear mapping relationship based on the kernel function is established to accurately capture the complex interaction between the fused features, which can not only provide high-precision life prediction results, but also quantify the uncertainty of the prediction results, providing a scientific basis for risk assessment.

[0056] The present invention corrects the preliminary prediction results by combining workload and environmental changes. The introduction of workload adjustment factors and environmental change adjustment factors can reflect the impact of complex working conditions on the life of components in real time. The dynamically adjusted prediction results are more in line with actual application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 A schematic diagram of a prediction method flow chart of an embodiment of the present invention;

[0059] Figure 2 Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0061] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0062] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0063] like Figure 1 As shown, a method for intelligently predicting the life of electronic components comprises the following steps:

[0064] S1, data collection: real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency, and strain value;

[0065] S2, data preprocessing: preprocessing the collected multimodal data, including data denoising and outlier removal;

[0066] S3, data fusion and feature enhancement: fuse the preprocessed multimodal data to form comprehensive features, and enhance the comprehensive features;

[0067] S4, life prediction: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components;

[0068] S5, dynamic adjustment of prediction results: dynamically adjust the life prediction results of electronic components according to actual usage conditions (workload, environmental changes);

[0069] Through the above content, the accuracy and robustness of electronic component life prediction are improved. It can effectively integrate multi-dimensional data from different sensors, overcome the limitation that a single data source cannot fully reflect the status of components, and enhance feature expression and model prediction capabilities through deep learning technology. At the same time, it can be dynamically adjusted according to real-time workload and environmental changes to ensure the accuracy and adaptability of prediction results in actual applications, thereby improving the intelligence level and application value of component management.

[0070] Data collection in S1 includes:

[0071] S11, temperature data collection: Use temperature sensors to monitor the operating temperature of components in real time to ensure that temperature changes can be accurately recorded and reflect the thermal status of components in a timely manner;

[0072] S12, vibration intensity data collection: using a vibration sensor to measure the vibration intensity of components and collect vibration frequency and amplitude data;

[0073] S13, sound wave frequency data collection: using sound wave sensors to capture sound wave frequencies related to the operation of components and detect abnormal sound waves caused by changes in the working state;

[0074] S14, strain data acquisition: using strain sensors to obtain the strain values ​​of components during use to monitor the physical deformation of components;

[0075] Through the above content, the operating status of components can be fully and accurately reflected. Different types of sensors work together to provide richer diagnostic information, which is helpful for early detection of potential faults and abnormal phenomena, and avoids misjudgment or missed judgment due to the limitations of a single data source. This multi-dimensional data collection method can effectively improve the accuracy of electronic component life prediction and provide a solid data foundation for subsequent predictive analysis, thereby improving the intelligence level and reliability of component management.

[0076] Data preprocessing in S2 includes:

[0077] S21, data denoising: The collected multimodal data is denoised using the moving average method, expressed as:

[0078]

[0079] in, is the denoised data value, y i is the original data, N is the sliding window size, and t is the current time point;

[0080] S22, outlier removal: Use the Z-Score algorithm to detect and remove outliers, expressed as:

[0081]

[0082] Among them, y i is the data point, μ is the mean of the data, σ is the standard deviation of the data, and Z is the data value after removing outliers. When the Z-Score value is greater than the set threshold (set to 3);

[0083] Through the above content, the accuracy and consistency of the data are improved. Data denoising can effectively remove high-frequency noise, retain the main trend of the signal, and reduce interference. Outlier removal can avoid erroneous analysis and prediction due to abnormal data, ensuring the representativeness and reliability of the data.

[0084] Data fusion and feature enhancement in S3 include:

[0085] S31, feature extraction: extracting features from the preprocessed multimodal data, including temperature variation amplitude, vibration intensity, frequency distribution, and strain peak value;

[0086] S32, data fusion: integrating the extracted features to generate comprehensive features;

[0087] S33, feature enhancement and redundancy removal: introduce the self-attention mechanism to enhance the integrated features after fusion and remove redundant information;

[0088] Through the above content, it is possible to accurately capture the key change patterns in various sensor data, integrate the features from different sources through data fusion, generate comprehensive features, and comprehensively reflect the health status of components. The self-attention mechanism is used to enhance the fused comprehensive features, optimize the weights of key information, and remove redundant information. It can effectively avoid information redundancy and feature loss, and can comprehensively improve the expression ability of features, further enhance the performance of the prediction model, and thus more accurately predict the life of components.

[0089] Feature extraction in S31 includes:

[0090] S311, temperature change feature extraction: The temperature change feature reflects the thermal state of electronic components under different working conditions. By calculating the maximum temperature difference, the temperature change feature is extracted, which is expressed as:

[0091] ΔT=T max -T min ;

[0092] Among them, T max is the maximum temperature in the collected data, T min is the minimum temperature in the collected data, ΔT is the temperature change amplitude;

[0093] S312, vibration intensity feature extraction: The vibration intensity reflects the degree to which the components are affected by external forces. By calculating the standard deviation of the vibration signal, the characteristics of the vibration intensity are extracted, which is expressed as:

[0094]

[0095] Among them, x i is the i-th vibration sampling point, is the mean value of the vibration signal, M is the number of sampling points, σ vib is the discrete degree of the vibration signal, that is, the vibration intensity;

[0096] S313, frequency distribution feature extraction: The frequency distribution reflects the frequency components in the vibration or sound wave signal. The time domain signal is converted into a frequency domain signal through Fourier transform, and the frequency feature is extracted, which is expressed as:

[0097]

[0098] Where X(f) is the Fourier transform result of the signal at frequency f, x(t) is the time domain signal, and T is the duration of the signal;

[0099] S314, strain peak feature extraction: The strain peak feature is used to describe the maximum deformation of the material during operation, indicating whether the material has structural damage during use. The strain peak is extracted by analyzing the maximum value of the strain signal, which is expressed as:

[0100]

[0101] Among them, ∈(t) is the strain value at time t, ∈ max is the maximum value of the strain signal, T is the duration of the signal;

[0102] Through the above content, combined with the extraction methods of time domain and frequency domain features, the key information of multimodal data can be fully captured, which not only reflects the dynamic operating status of electronic components, but also reveals potential problems that may cause failures, improves the accuracy and diversity of feature extraction, and provides high-quality data support for subsequent life prediction and health assessment, thereby significantly enhancing the robustness and accuracy of the prediction.

[0103] Data fusion in S32 includes:

[0104] S321, feature normalization: The extracted features are normalized using the minimum-maximum normalization method, which is expressed as:

[0105]

[0106] Among them, x is the original eigenvalue, min(x) and max(x) are the minimum and maximum values ​​of the feature respectively, and x′ is the normalized eigenvalue;

[0107] S322, weighted fusion: Set the weight coefficient w according to the importance difference of multimodal data features. i Perform weighted summation on each feature to generate a comprehensive feature, expressed as:

[0108]

[0109] Among them, F 融合 is the comprehensive feature after fusion, w i is the weight of the i-th feature, satisfying x i ′ is the normalized ith eigenvalue, and n is the number of features;

[0110] Through the above content, the problems of multimodal data in dimensional differences, feature importance differences and redundant information have been solved, and the efficiency and quality of data processing have been significantly improved. Feature normalization eliminates the dimensional influence of different features and ensures that features are fused on a unified scale. Weighted fusion assigns weights according to the importance of features, highlighting the contribution of key features and making data fusion more accurate.

[0111] Feature enhancement and redundancy removal in S33 include:

[0112] S331, weight calculation of self-attention mechanism: The correlation weights between the features within the comprehensive features are calculated through the self-attention mechanism, and the dependency relationship of key information in the comprehensive features is extracted, which is expressed as:

[0113]

[0114] Among them, F 融合 =[F1, F2, ... F n ] represents the integrated features after fusion, including n single features, α ij is the attention weight of the i-th feature and the j-th feature in the comprehensive feature, indicating the correlation between the features, q i =W q ·F i , k j =W k ·F j Represents the feature F i and F j The query vector and key vector of is the similarity between the query vector and the key vector, W q , W k is the weight matrix used to generate query and key vectors;

[0115] S332, feature enhancement: According to the calculated attention weight, each feature in the comprehensive feature is weighted and summed to generate enhanced features, which are expressed as:

[0116]

[0117] Among them, F 增强，i is the enhanced i-th feature, Value(F j )=W v ·F j For feature F j The value vector, W v is the weight matrix of the value vector;

[0118] S333, redundancy removal: By setting the attention weight threshold τ, the enhanced features are redundancy removed to generate the final features (the feature set after redundancy removal), which is expressed as:

[0119]

[0120] Among them, F 最终 is the feature set after removing redundancy, is the comprehensive feature finally used by the model, max(α ij ) is the feature F i The maximum attention weight between other features, τ is the attention weight threshold;

[0121] The setting of the attention weight threshold τ specifically includes:

[0122] Calculate the mean and standard deviation of the weights: All attention weights α for the comprehensive features ij Perform statistical analysis and calculate the mean and standard deviation of the weights, expressed as:

[0123]

[0124] Among them, μ α is the mean of all attention weights, σ α is the standard deviation of all attention weights, α ij is the attention weight between comprehensive features, s is the number of comprehensive features;

[0125] Set dynamic weight threshold: Combine the mean and standard deviation to set a dynamic weight threshold, expressed as:

[0126] τ=μ α +k·σ α ;

[0127] Among them, k is the adjustment factor (the value range is 0 to 2);

[0128] Through the above content, it is possible to effectively highlight the features that play a key role in life prediction, while weakening or removing irrelevant features, and setting dynamic thresholds based on the mean and standard deviation of the weights to ensure that the screening process is adaptable and flexible, thereby avoiding the interference of redundant features on the model, improving the quality of data input and the efficiency of the model, and being able to fully explore the key information of the fused features, optimize the feature expression, and significantly improve the prediction accuracy and robustness of the model.

[0129] The life prediction model in S4 adopts the Gaussian process regression (GPR) model, which includes:

[0130] S41, feature input preparation: take the generated final features as input data, set the final feature matrix of the training sample to X, and set X = {F 最终，1 , F 最终，2 , …, F 最终，S} is the final feature matrix of the training sample, Y = {y1, y2, ..., y S} is the target lifespan value corresponding to the training sample, where F 最终，1 , F 最终，2 , ..., F 最终，S are the final feature vectors of the 1st, 2nd, ..., Sth samples, y1, y2, ..., y S are the true lifespan values ​​of the 1st, 2nd, ..., Sth samples respectively, and S is the number of training samples;

[0131] S42, kernel function definition: The radial basis function kernel (RBF kernel) is selected as the kernel function of the life prediction task to measure the similarity between different samples, which is expressed as:

[0132]

[0133] Among them, F 最终 , F′ 最终 are two sets of final eigenvectors, ||F 最终 -F′ 最终 || is the Euclidean distance between two sets of features, σ 2 is the length scale parameter of the kernel function, indicating the rate at which feature similarity decays;

[0134] S43, kernel matrix calculation: Using the final feature matrix of the training samples as X, calculate the similarity between the training samples and construct the kernel matrix K, which is expressed as:

[0135] K ij = k(F 最终，i 'F 最终，j );

[0136] Among them, K is the kernel matrix, each element K ijrepresents the kernel function value of the i-th sample and the j-th sample, F 最终，i 、F 最终，j are the final features of the i-th and j-th samples respectively;

[0137] S44, kernel function calculation: Given a new test sample F 最终，* , calculate its similarity with the training sample and use it to predict the life span value, expressed as:

[0138] K * ={k(F 最终，* , F 最终，1 ), k(F 最终，* , F 最终，2 ),…,k(F 最终，* , F 最终，S )};

[0139] Among them, K * is the kernel value vector of the test sample and all training samples;

[0140]

[0141] S45, life prediction: According to the kernel matrix K of the training sample and the kernel vector of the test sample, the life prediction value and uncertainty of the test sample are calculated, which is expressed as:

[0142] Predicted mean:

[0143] Among them, μ(F 最终，* ) is the life prediction value of the test sample, K -1 is the inverse of the kernel matrix, and Y is the target lifespan value of the training sample;

[0144] Prediction Variance:

[0145] Among them, σ 2 (F 最终，* ) is the uncertainty (variance) of the predicted value;

[0146] S46, result output: output the prediction results of the Gaussian process regression (GPR) model, including the life prediction value and uncertainty estimation;

[0147] Through the above content, we can make full use of the nonlinear relationship and complex interaction of the fused comprehensive features to accurately predict the life of electronic components. By calculating the similarity between samples through the kernel function, we can not only provide high-precision prediction results, but also quantify the uncertainty of the prediction results, and provide a basis for reliability evaluation and decision support of life prediction. Its non-parametric modeling method does not require explicit assumptions about data distribution and is suitable for processing multimodal features and small sample data. At the same time, the output of GPR includes life prediction values ​​and confidence intervals, which can help identify potential risks and significantly improve the robustness and practicality of the model.

[0148] The dynamic adjustment of prediction results in S5 includes:

[0149] S51, workload adjustment factor calculation: Calculate the workload adjustment factor Δ based on the actual workload (operating frequency, current) of the components. load , used to correct the initial prediction value, expressed as:

[0150]

[0151] Among them, Δ load is the workload adjustment factor, which indicates the impact of load changes on life. w1 is the load sensitivity coefficient. current Load is the current workload. baseline is the baseline workload;

[0152] S52, calculation of environmental change adjustment factor: Calculate the environmental change adjustment factor Δ based on the environmental conditions (temperature) env , used to correct the initial prediction value, expressed as:

[0153]

[0154] Among them, Δ env is the environmental adjustment factor, which indicates the impact of environmental changes on life, w2 is the environmental sensitivity coefficient, Temp current is the current ambient temperature, Temp optimal is the optimal operating temperature of the components;

[0155] Dynamically adjusted life prediction calculation: combined with workload adjustment factor Δ load and environmental change adjustment factor Δ env , dynamically adjust the life prediction value to obtain the corrected final life prediction value, expressed as:

[0156] μ′(F 最终，* )=μ(F 最终，* )+Δ load +Δ env ;

[0157] Among them, μ′(F 最终，* ) is the life prediction value after dynamic adjustment, μ(F 最终，* ) is the preliminary predicted value;

[0158] Through the above content, not only can the life of electronic components be predicted with high precision, but it can also effectively adapt to complex and changeable working conditions. The initial prediction results are corrected by workload adjustment factors and environmental change adjustment factors to dynamically reflect the actual life status of components under different conditions, thereby improving the reliability and practicality of the prediction.

[0159] like Figure 2 As shown, an electronic component life intelligent prediction system is used to implement the above-mentioned electronic component life intelligent prediction method, including the following modules:

[0160] Data acquisition module: Real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency and strain value;

[0161] Data preprocessing module: preprocess the collected multimodal data, including data denoising and outlier removal;

[0162] Data fusion and feature enhancement module: fuses the preprocessed multimodal data to form comprehensive features, and enhances the comprehensive features;

[0163] Life prediction module: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components;

[0164] Dynamic adjustment module: dynamically adjust the life prediction results according to the actual use conditions of electronic components to optimize the life prediction results.

[0165] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0166] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent prediction method for the life of electronic components, characterized in that: The following steps are involved: S1, data collection: real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency, and strain value; S2, data preprocessing: preprocessing the collected multimodal data, including data denoising and outlier removal; S3, data fusion and feature enhancement: fuse the preprocessed multimodal data to form comprehensive features, and enhance the comprehensive features; S4, life prediction: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components; S5, dynamic adjustment of prediction results: dynamically adjust the life prediction results of electronic components according to actual usage conditions.

2. The method for intelligent prediction of electronic component life according to claim 1, characterized in that: The data collection in S1 includes: S11, temperature data acquisition: use temperature sensors to monitor the operating temperature of components in real time; S12, vibration intensity data collection: using a vibration sensor to measure the vibration intensity of components and collect vibration frequency and amplitude data; S13, acoustic wave frequency data collection: using an acoustic wave sensor to capture the acoustic wave frequency related to the operation of the component; S14, strain data acquisition: use strain sensors to obtain the strain values ​​of components during use to monitor the physical deformation of components.

3. The method for intelligent prediction of electronic component life according to claim 1, characterized in that: The data preprocessing in S2 includes: S21, data denoising: using the moving average method to denoise the collected multimodal data; S22, outlier removal: Use the Z-Score algorithm to detect and remove outliers.

4. The method for intelligently predicting the life of electronic components according to claim 1, characterized in that: The data fusion and feature enhancement in S3 include: S31, feature extraction: extracting features from the preprocessed multimodal data, including temperature variation amplitude, vibration intensity, frequency distribution, and strain peak value; S32, data fusion: integrating the extracted features to generate comprehensive features; S33, feature enhancement and redundancy removal: The self-attention mechanism is introduced to enhance the fused comprehensive features and remove redundant information.

5. The method for intelligently predicting the life of electronic components according to claim 4, characterized in that: The feature extraction in S31 includes: S311, temperature change feature extraction: The temperature change feature reflects the thermal state of electronic components under different working conditions. The temperature change feature is extracted by calculating the maximum temperature difference; S312, vibration intensity feature extraction: the vibration intensity reflects the degree to which the components are affected by external forces. The characteristics of the vibration intensity are extracted by calculating the standard deviation of the vibration signal; S313, frequency distribution feature extraction: The frequency distribution reflects the frequency components in the vibration or sound wave signal. The time domain signal is converted into a frequency domain signal through Fourier transform to extract the frequency feature; S314, strain peak feature extraction: The strain peak feature is used to describe the maximum deformation of the material during operation, indicating whether the material has structural damage during use. The strain peak is extracted by analyzing the maximum value of the strain signal.

6. The method for intelligently predicting the life of electronic components according to claim 5, characterized in that: The data fusion in S32 includes: S321, feature normalization: normalizing the extracted features using a minimum-maximum normalization method; S322, weighted fusion: Set the weight coefficient w according to the importance difference of multimodal data features. i Perform weighted summation on each feature to generate a comprehensive feature.

7. The method for intelligently predicting the life of electronic components according to claim 6, characterized in that: The feature enhancement and redundancy removal in S33 include: S331, weight calculation of self-attention mechanism: The correlation weights between the features within the comprehensive features are calculated through the self-attention mechanism to extract the dependency relationship of key information in the comprehensive features; S332, feature enhancement: performing weighted summation on each feature in the comprehensive feature according to the calculated attention weight to generate an enhanced feature; S333, redundancy removal: By setting the attention weight threshold τ, the enhanced features are redundancy removed to generate the final features.

8. The method for intelligently predicting the life of electronic components according to claim 7, characterized in that: The life prediction model in S4 adopts a Gaussian process regression model, and the Gaussian process regression model includes: S41, feature input preparation: using the generated final features as input data, setting the final feature matrix of the training sample to X; S42, kernel function definition: select radial basis function kernel as the kernel function for life prediction task; S43, kernel matrix calculation: using the final feature matrix of the training samples as X, calculate the similarity between the training samples and construct the kernel matrix X; S44, kernel function calculation: Given a new test sample F 最终,* , calculate its similarity with the training sample and use it to predict the life span value; S45, life prediction: according to the kernel matrix K of the training sample and the kernel vector of the test sample, the life prediction value and uncertainty of the test sample are calculated; S46, result output: output the prediction results of the Gaussian process regression model, including the life prediction value and uncertainty estimation.

9. The method for intelligently predicting the life of electronic components according to claim 8, characterized in that: The dynamic adjustment of the prediction result in S5 includes: S51, workload adjustment factor calculation: Calculate the workload adjustment factor Δ based on the actual workload of the components. load ; S52, calculation of environmental change adjustment factor: Calculate the environmental change adjustment factor Δ based on environmental conditions env ; Dynamically adjusted life prediction calculation: combined with workload adjustment factor Δ load and environmental change adjustment factor Δ env , dynamically adjust the life prediction value to obtain the corrected final life prediction value.

10. An electronic component life intelligent prediction system, used to implement an electronic component life intelligent prediction method as claimed in any one of claims 1 to 9, characterized in that: Includes the following modules: Data acquisition module: Real-time monitoring of electronic components through a variety of sensors, collecting multi-dimensional data, including temperature, vibration intensity, sound wave frequency and strain value; Data preprocessing module: preprocess the collected multimodal data, including data denoising and outlier removal; Data fusion and feature enhancement module: fuses the preprocessed multimodal data to form comprehensive features, and enhances the comprehensive features; Life prediction module: Based on the comprehensive features after feature enhancement, a life prediction model is constructed to predict the life of electronic components; Dynamic adjustment module: dynamically adjust the life prediction results according to the actual use conditions of electronic components to optimize the life prediction results.

Images