Multi-representation degradation prediction method for electronic system and related equipment
By designing a multi-characterized degradation prediction method in electronic systems, using characterization trend decomposition and deep fusion technology, the problems of complexity and resource demand for multi-characterized parameters in the existing technology are solved, and efficient and accurate multi-characterized degradation prediction effect is achieved.
Patent Information
- Application Number
- CN202411840403.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-13
- Publication Date
- 2025-04-29
AI Technical Summary
When predicting the degradation trend of multiple key characterization parameters in electronic systems, the prior art requires the deployment of multiple prediction models separately, resulting in large demand for computing resources, complex system construction, maintenance and optimization, and the difficulty of a single prediction model to capture the complexity of diversified data, limiting the accuracy and reliability of prediction results.
Design a multi-characterization degradation prediction method for electronic systems, and achieve deep mining and efficient processing of complex timing characterization data through characterization trend decomposition and characterization perception fusion. The method includes obtaining time series and sequence categories, performing characterization trend decomposition to obtain indicator vectors, deep fusion based on sequence categories and indicator vectors, generating final outputs, and inputting them into a timing prediction model to complete multi-characterization degradation prediction.
Through the design of the multi-character compatible feature fusion module, efficient processing and accurate prediction of multiple time series characterization data is achieved, which significantly improves the adaptability, discrimination and robustness of the model, and improves the accuracy and reliability of the prediction results.
Smart Images
Figure CN119903388A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of long-term degradation trend prediction of key parameters, and relates to a multi-characterization degradation prediction method for electronic systems and related equipment. Background Art
[0002] When discussing the reliability of the entire computer system, the key characterization parameters in its degradation process play a vital role. These parameters can accurately and scientifically reflect the performance stability and reliability requirements of the computer in various application scenarios. For example, as the time reference of the entire system, any parameter degradation of the crystal oscillator will directly affect the operating timing of the entire system, which may cause a series of problems such as timing disorder. Similarly, the parameter changes of the optocoupler and the degradation of the interface crystal oscillator parameters will also have a significant impact on the baud rate of the communication waveform, resulting in deviations or out-of-tolerances in the communication data, which in turn affects the communication quality and stability of the entire system. However, in the current long-term prediction task of computer key parameter degradation, we face a severe challenge: for each different key characterization in the electronic system, a prediction model needs to be deployed separately for prediction. This approach not only greatly increases the demand for computing resources, but also makes the construction, maintenance and optimization of the entire prediction system extremely complex and cumbersome. In addition, with the continuous improvement of system complexity and the increasing diversity of key characterization parameters, a single prediction model has been unable to fully capture and distinguish the complexity of these different characterizations, thereby limiting the accuracy and reliability of the prediction results.
[0003] In order to reflect the degradation trend of the entire computer as comprehensively and reasonably as possible, we need to accurately predict the changing trends of multiple characterization parameters. However, the existing single prediction model is unable to cope with such diverse data and cannot meet the needs of accurate prediction of different characterization parameters at the same time. If the method of designing multiple independent models is adopted, although it may improve the prediction effect to a certain extent, it will also bring about a significant increase in development and maintenance costs, and may also lead to a waste of computing resources and an increase in the complexity of the entire prediction system. Therefore, how to design and implement a prediction model that can handle multiple representations and changing trends at the same time to improve the model's ability to process diverse data has become a key issue that needs to be solved urgently. Summary of the invention
[0004] The purpose of the present invention is to solve the technical problem in the prior art that a separate prediction model needs to be deployed to predict the changing trends of multiple characterization parameters, and to provide a multi-characterization degradation prediction method and related equipment for electronic systems.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a multi-characterization degradation prediction method for an electronic system, comprising: Obtaining a time series and a series category, performing a characterization trend decomposition on the time series, and obtaining an indicator vector as a characterization representation; Based on the time series, the sequence category and the indicator vector are deeply fused to obtain a final output; The final output is input into a time series prediction model to obtain a final predicted time series, thereby completing multi-representation degradation prediction.
[0006] Further improvements are: Decomposing the time series to characterize the trend to obtain the indicator vector specifically includes: Performing global trend decomposition on the time series to obtain global trend features; Performing local trend decomposition on the time series to obtain local trend features; The global trend feature and the local trend feature are concatenated to obtain an indicator vector.
[0007] The time series is decomposed into global trends to obtain global trend features, including: The time series is Sub-convolution and nonlinear activation processing are performed to obtain global trend features.
[0008] The local trend characteristics obtained by performing local trend decomposition on the time series include: Fast Fourier transform is applied to the time series to map the features to the frequency domain, and frequency domain operations are performed through a learnable linear transformation; then, the features are restored to the time domain through an inverse Fourier transform to obtain local trend features.
[0009] Based on the time series, the sequence category and the indicator vector are deeply fused to obtain the final output, which specifically includes: Performing word embedding processing on the sequence category to obtain an embedding vector, and converting the embedding vector into a modulation parameter through high-dimensional mapping; generating a dynamic weighted mask based on the indicator vector; Multiplying the modulation parameter and the dynamic weighted mask element by element to generate a modulation feature; After downsampling the modulation features, they are added to the time series through residual connection to generate the final output.
[0010] Generating a dynamic weighted mask based on the indicator vector specifically includes: The indicator vector is upsampled and flattened, and then passes through two fully connected layers and activated using a nonlinear activation function to generate a dynamic weighted mask.
[0011] The nonlinear activation function includes a ReLU function and a sigmoid function.
[0012] In a second aspect, the present invention provides a multi-representation degradation prediction system for electronic systems, comprising: A representation trend decomposition module is used to obtain a time series and a sequence category, perform a representation trend decomposition on the time series, and obtain an indicator vector as a representation; A representation perception fusion module, used for deeply fusing the sequence category and the indicator vector based on the time series to obtain a final output; The final prediction module is used to input the final output into the time series prediction model to obtain the final predicted time series and complete the multi-representation degradation prediction.
[0013] In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned multi-characterization degradation prediction method for electronic systems when executing the computer program.
[0014] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned multi-characterization degradation prediction method for electronic systems.
[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a multi-representation degradation prediction method for electronic systems, the core of which is to design an efficient multi-representation compatible feature fusion module. The module realizes deep mining and efficient processing of complex time series representation data through two key steps, namely, representation trend decomposition and representation perception fusion. First, the representation trend decomposition module can accurately capture the long-term trend and short-term dynamic changes of input data. This feature enables the model to understand the inherent laws of the data more comprehensively, whether it is the long-term stable trend or the short-term fluctuation and change, it can be accurately captured and reflected in the indicator vector. This in-depth insight into the data provides a solid foundation for subsequent deep fusion and prediction. Secondly, the representation perception fusion block uses label information to generate modulation parameters, and combines it with a dynamic weighted mask to realize the deep fusion of representation features. This process not only improves the sensitivity and discrimination of the model to different representation features, but also significantly enhances the robustness of the model. Through the dynamic weighted mask, the model can intelligently adjust the importance of different representation features, thereby realizing the accurate identification and utilization of key features. In summary, the multi-representation degradation prediction method proposed in the present invention realizes efficient processing and accurate prediction of multiple time series representation data through the design of a multi-representation compatible feature fusion module. This method not only significantly improves the adaptability, discrimination and robustness of the model, but also provides reliable technical support for the processing of complex time series data. Therefore, the present invention has a wide range of application potential and is expected to promote the further development of electronic system degradation prediction technology in the future. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0017] Figure 1 It is a flow chart of a multi-characterization degradation prediction method for electronic systems in the present invention; Figure 2 A step diagram of a multi-characterization degradation prediction method for an electronic system in an embodiment of the present invention; Figure 3 Schematic diagram of the structure of a multi-representation compatible feature fusion module in an embodiment of the present invention; Figure 4 A schematic diagram of a characterization trend decomposition process of a multi-characterization degradation prediction method for electronic systems in the present invention; Figure 5 It is a schematic diagram of a representation fusion perception process of a multi-representation degradation prediction method for electronic systems in the present invention; Figure 6 A schematic diagram of prediction results of a multi-characterization degradation prediction method for electronic systems in an embodiment of the present invention; Figure 6 (a) is a schematic diagram of the prediction results when characterized as s2; Figure 6 (b) is a schematic diagram of the prediction results when characterized as s7; Figure 6 (c) is a schematic diagram of the prediction results when the representation is s8; Figure 6 (d) is a schematic diagram of the prediction results when characterized as s13; Figure 6 (e) is a schematic diagram of the prediction results when characterized as s17; Figure 6 (f) is a schematic diagram of the prediction results when characterized as s21. Figure 7 A diagram of a multi-characterization degradation prediction system for electronic systems in the present invention; Figure 8 This is a module diagram of an electronic device in the present invention. DETAILED DESCRIPTION
[0018] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0019] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0021] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 1 The embodiment of the present invention discloses a multi-characterization degradation prediction method for an electronic system, comprising: S1, obtaining a time series and a series category, performing a characterization trend decomposition on the time series, and obtaining an indicator vector as a characterization representation; S2, performing deep fusion of the sequence category and the indicator vector based on the time series to obtain a final output; S3, inputting the final output into a time series prediction model to obtain a final predicted time series, thereby completing multi-representation degradation prediction.
[0022] The present invention discloses a multi-representation degradation prediction method for electronic systems, the core of which is to design an efficient multi-representation compatible feature fusion module. The module realizes deep mining and efficient processing of complex time series representation data through two key steps, namely, representation trend decomposition and representation perception fusion. First, the representation trend decomposition module can accurately capture the long-term trend and short-term dynamic changes of input data. This feature enables the model to understand the inherent laws of the data more comprehensively, whether it is the long-term stable trend or the short-term fluctuation and change, it can be accurately captured and reflected in the indicator vector. This in-depth insight into the data provides a solid foundation for subsequent deep fusion and prediction. Secondly, the representation perception fusion block uses label information to generate modulation parameters, and combines it with a dynamic weighted mask to realize the deep fusion of representation features. This process not only improves the sensitivity and discrimination of the model to different representation features, but also significantly enhances the robustness of the model. Through the dynamic weighted mask, the model can intelligently adjust the importance of different representation features, thereby realizing the accurate identification and utilization of key features. In summary, the multi-representation degradation prediction method proposed in the present invention realizes efficient processing and accurate prediction of multiple time series representation data through the design of a multi-representation compatible feature fusion module. This method not only significantly improves the adaptability, discrimination and robustness of the model, but also provides reliable technical support for the processing of complex time series data. Therefore, the present invention has a wide range of application potential and is expected to promote the further development of electronic system degradation prediction technology in the future.
[0023] The content of the present invention is described in detail below in conjunction with specific embodiments: See also Figure 2 and Figure 3 The present invention firstly performs global and local trend decomposition on the time series data through the representation trend decomposition encoder in the multi-representation compatible feature fusion module, and extracts the indicator vector As a key representation, it can capture the subtle differences and temporal correlations between different representations. The representation-aware fusion block then uses this representation With time series data labels Deep fusion is performed to generate feature representations with higher specificity and robustness, allowing the model to make better use of representation trend information and adapt to specific key degraded representation inputs.
[0024] The time series features processed by the multi-representation compatible feature fusion module can be directly used as the input of the subsequent time series prediction model network (such as LSTM, RNN, etc.), enhancing the model's ability to capture complex features and improving the prediction accuracy in multi-representation tasks. The entire technical solution is developed according to the following process: (1) Overall architecture of multi-representation compatible feature fusion module This module consists of a representation trend decomposition encoder and a representation perception fusion block, such as Figure 3 First, the time series is input into the representation trend decomposition encoder to extract the representation , thereby obtaining key feature information. Next, the time series , representation The label information of the time series data is input into the representation perception fusion block for deep fusion processing, and finally generates a time series feature result that can characterize the key degradation features. This network structure effectively improves the compatibility and processing accuracy of multiple representations.
[0025] (2) Input timing sequence Input time series data and sequence categories Passed into the model, ready for feature extraction and processing.
[0026] (3) Characterization trend decomposition The design goal of the characterization trend decomposition module is to adapt to and handle multiple key degradation characterizations, such as crystal oscillators, AD and DA conversions, optocouplers, baud rates, and overall voltage and current. In practical applications, these different characterization categories need to be accurately distinguished and captured without explicit category labels. Therefore, the characterization trend decomposition module generates a characterization trend decomposition indicator vector To characterize and distinguish different inputs, and enhance the robustness and adaptability of the model to various representations. Specifically, the representation trend decomposition module The following steps were performed: 1) Global trend decomposer: such as Figure 4 As shown in the global trend decomposer of The model is composed of two-dimensional convolutional layers, which enhances the adaptability of the model by capturing the long-term trend characteristics of the input sequence. Sub-convolution and nonlinear activation processing to obtain global trend features The design of multi-layer convolution can effectively capture the overall change trend of the input sequence, enabling the model to understand the evolution of time series data on a global scale.
[0027] 2) Local trend decomposer: such as Figure 4As shown in the local trend decomposer of Fig. 1, the fast Fourier transform (FFT) is applied to map the features to the frequency domain, and the frequency domain operation is performed through a learnable linear transformation. Then, the inverse Fourier transform (iFFT) is used to restore it to the time domain to obtain the local change trend. The frequency domain-based decomposition method can effectively identify the high-frequency and low-frequency components of the input data and improve the model's ability to respond to different rates of change.
[0028] 3) Global / local mapping merging: global trends and local trends Splicing as a representation ,Through the fusion of global and local features, it is possible to capture both the long-term and short-term dynamics of the input data, and provide a multi-dimensional and fine-grained representation for the subsequent deep feature processing of the representation perception fusion layer.
[0029] (4) Representation-Perception Fusion The module as a whole Figure 5 As shown, the representation obtained by decomposing the representation trend is expressed as With sequence category Deep fusion is performed to adapt to multiple representation features and enhance the discrimination ability. Specifically, the representation perception fusion layer performs deep fusion on the input time series X and the representation With sequence category Do the following: 1) Label modulation parameter generation: sequence category First, the word is embedded to obtain its embedding vector. Then, the embedding vector is converted into modulation parameters through high-dimensional mapping. , which provides support for the extensive expression of label features in high-dimensional space.
[0030] 2) Dynamic Weighted Mask Generation: Representation It is further processed in this step to generate a dynamic weighted mask , thereby dynamically adjusting the importance of different representation features. Specifically, Upsample and flatten to get , Generate dynamic weighted masks through two fully connected layers, ReLU and sigmoid activation .
[0031] 3) Feature modulation and fusion: Modulation parameters and dynamic weighted mask Perform element-by-element multiplication to generate modulated features This modulation mechanism can generate more accurate weight distribution between different representation features, improve the model's adaptability to specific feature categories, and thus enhance the model's discriminative effect in multi-representation tasks.
[0032] 4) Downsampling and residual connection: modulation features After downsampling, the residual connection is connected with the original input time series Add together to generate the final output This process improves the prediction accuracy in multi-representation tasks while preserving the original information.
[0033] Example 1 Evaluation platform: (1) Hardware Platform Use an embedded computer system configured with 32GB of memory, an Intel Core i7 processor with a main frequency of 2.9GHz; equipped with a high-performance GPU that supports multi-tasking (such as NVIDIA RTX 3060) to accelerate model training and testing.
[0034] (2) Software Platform Based on the Linux operating system, the algorithm is implemented in Python programming language, and the deep learning framework PyTorch is used for model construction and training. The main libraries used include NumPy and SciPy for data processing, and Matplotlib for result visualization.
[0035] Methods and steps: (1) Data preparation Collect time series data of common characteristics (voltage, current, temperature, etc.) in embedded systems, perform data preprocessing, including normalization and denoising, and mark degradation labels for various characteristics.
[0036] (2) Model training The preprocessed data is input into the multi-representation compatible degradation prediction framework, where the time series prediction model adopts LSTM, the training process adopts the Adam optimization algorithm, the learning rate is 0.001, and the training rounds are 100 rounds.
[0037] (3) Test method Independent tests were performed on different degradation characterization categories (voltage, current, temperature, etc.) to evaluate the model's prediction performance for single and multiple degradation characterizations.
[0038] See also Figure 6, different representations (such as s2, s7, s8, s13, s17 and s216) are predicted. The orange curve represents the predicted value and the green curve represents the true value. By comparison, it can be seen that under multiple representations, the fitting effect between the predicted value and the true value is good, showing that the proposed method can effectively capture the degradation trend. The prediction effect of using and not using the multi-representation compatible feature fusion module is compared by the two indicators of root mean square error (RMSE) and mean absolute error (MAE). As shown in Table 1: Table 1
[0039] The results show that the model using the multi-representation compatible feature fusion module is significantly better than the model without this module in terms of RMSE and MAE indicators. The RMSE value of this method is 0.051, which is nearly 80% lower than the LSTM model (0.256) without this module, indicating that the prediction accuracy has been significantly improved. The MAE value of this method is 0.041, which is significantly lower than 0.203 of the LSTM model without the module, showing that the prediction error of this method is smaller under different representations and has higher stability.
[0040] The present invention provides a multi-characterization degradation prediction method for electronic systems, which has the ability to be compatible with at least 5 or more characterizations and can accurately predict the degradation trends of multiple key parameters. For multiple characterizations (crystal oscillator, optocoupler, baud rate, whole machine voltage and current, etc.) in embedded computer systems, this method can comprehensively process the degradation trends of multiple characterization features and provide consistent predictions by constructing a multi-characterization compatible degradation prediction framework.
[0041] See also Figure 7 The present invention also discloses a multi-characterization degradation prediction system for electronic systems, comprising: A representation trend decomposition module is used to obtain a time series and a sequence category, perform a representation trend decomposition on the time series, and obtain an indicator vector as a representation; A representation perception fusion module, used for deeply fusing the sequence category and the indicator vector based on the time series to obtain a final output; The final prediction module is used to input the final output into the time series prediction model to obtain the final predicted time series and complete the multi-representation degradation prediction.
[0042] See also Figure 8 The third object of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-characterization degradation prediction method for electronic systems when executing the computer program.
[0043] The multi-characterization degradation prediction method for electronic systems comprises the following steps: Obtaining a time series and a series category, performing a characterization trend decomposition on the time series, and obtaining an indicator vector as a characterization representation; Based on the time series, the sequence category and the indicator vector are deeply fused to obtain a final output; The final output is input into a time series prediction model to obtain a final predicted time series, thereby completing multi-representation degradation prediction.
[0044] A fourth object of the present invention is to provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the multi-characterization degradation prediction method for electronic systems.
[0045] The multi-characterization degradation prediction method for electronic systems comprises the following steps: Obtaining a time series and a series category, performing a characterization trend decomposition on the time series, and obtaining an indicator vector as a characterization representation; Based on the time series, the sequence category and the indicator vector are deeply fused to obtain a final output; The final output is input into a time series prediction model to obtain a final predicted time series, thereby completing multi-representation degradation prediction.
[0046] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0047] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0048] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0050] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A multi-characterization degradation prediction method for electronic systems, characterized in that: include: Obtaining a time series and a series category, performing a characterization trend decomposition on the time series, and obtaining an indicator vector as a characterization representation; Based on the time series, the sequence category and the indicator vector are deeply fused to obtain a final output; The final output is input into a time series prediction model to obtain a final predicted time series, thereby completing multi-representation degradation prediction.
2. The multi-representation degradation prediction method for electronic systems according to claim 1, characterized in that: Decomposing the time series to characterize the trend to obtain the indicator vector specifically includes: Performing global trend decomposition on the time series to obtain global trend features; Performing local trend decomposition on the time series to obtain local trend features; The global trend feature and the local trend feature are concatenated to obtain an indicator vector.
3. The multi-representation degradation prediction method for electronic systems according to claim 2, characterized in that: The time series is decomposed into global trends to obtain global trend features, including: The time series is Sub-convolution and nonlinear activation processing are performed to obtain global trend features.
4. The multi-representation degradation prediction method for electronic systems according to claim 2, characterized in that: The local trend characteristics obtained by performing local trend decomposition on the time series include: Fast Fourier transform is applied to the time series to map the features to the frequency domain, and frequency domain operations are performed through a learnable linear transformation; then, the features are restored to the time domain through an inverse Fourier transform to obtain local trend features.
5. The multi-characterization degradation prediction method for electronic systems according to claim 1, characterized in that: Based on the time series, the sequence category and the indicator vector are deeply fused to obtain the final output, which specifically includes: Performing word embedding processing on the sequence category to obtain an embedding vector, and converting the embedding vector into a modulation parameter through high-dimensional mapping; generating a dynamic weighted mask based on the indicator vector; Multiplying the modulation parameter and the dynamic weighted mask element by element to generate a modulation feature; After downsampling the modulation features, they are added to the time series through residual connection to generate the final output.
6. The multi-representation degradation prediction method for electronic systems according to claim 1, characterized in that: Generating a dynamic weighted mask based on the indicator vector specifically includes: The indicator vector is upsampled and flattened, and then passes through two fully connected layers and activated using a nonlinear activation function to generate a dynamic weighted mask.
7. The multi-representation degradation prediction method for electronic systems according to claim 6, characterized in that: The nonlinear activation function includes a ReLU function and a sigmoid function.
8. A multi-characterization degradation prediction system for electronic systems, characterized in that: include: A representation trend decomposition module is used to obtain a time series and a sequence category, perform a representation trend decomposition on the time series, and obtain an indicator vector as a representation; A representation perception fusion module, used for deeply fusing the sequence category and the indicator vector based on the time series to obtain a final output; The final prediction module is used to input the final output into the time series prediction model to obtain the final predicted time series and complete the multi-representation degradation prediction.
9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-characterization degradation prediction method for electronic systems according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the multi-characterization degradation prediction method for electronic systems according to any one of claims 1 to 7.