High-voltage connector fault prediction method and system based on thermal management

Abstract

The invention relates to the technical field of high-voltage connectors, and discloses a high-voltage connector fault prediction method and system based on thermal management, and the method comprises the steps: collecting the temperature distribution data and electric signal transmission characteristic data of a high-voltage connector, and building a thermal-electric coupling network model comprising a temperature node and a signal transmission node; separating and optimizing the transmission characteristic function to obtain a temperature-transmission characteristic mapping relation; gaussian process regression calculation is carried out on the temperature time sequence data to obtain temperature dynamic characteristic statistical characteristics; inputting the statistical characteristics of the dynamic temperature characteristics into a specified time delay thermal management neural network for signal transmission delay prediction to obtain delay prediction characteristic parameters; according to the method, the weighted error value in the transmission feature space is calculated, and the signal transmission delay is dynamically adjusted to obtain the compensated signal transmission delay parameter, so that the accurate prediction of the signal transmission delay is realized, and the accuracy and robustness of the compensation effect are improved.

Description

Technical Field

[0001] The present application relates to the technical field of high-voltage connectors, and in particular to a method and system for predicting high-voltage connector failures based on thermal management. Background Art

[0002] As a key signal transmission component, high-voltage connectors are widely used in various industrial equipment. However, in actual working environments, high-voltage connectors are often subjected to complex thermal stress and electromagnetic interference. This thermal-electric coupling effect will seriously affect the reliability and stability of signal transmission.

[0003] Traditional high-voltage connector fault prediction methods mainly rely on single electrical parameter monitoring or temperature monitoring, which cannot effectively capture the comprehensive impact of thermal-electrical coupling effects on signal transmission, resulting in low accuracy and reliability of fault prediction. Especially under conditions of drastic temperature changes, the dynamic changes of signal transmission delay are more difficult to accurately predict and compensate. Summary of the invention

[0004] The present application provides a high-voltage connector fault prediction method and system based on thermal management, thereby achieving accurate prediction of signal transmission delay and improving the accuracy and robustness of the compensation effect.

[0005] In a first aspect, the present application provides a high-voltage connector fault prediction method based on thermal management, and the high-voltage connector fault prediction method based on thermal management includes: Collect temperature distribution data and electrical signal transmission characteristic data of high-voltage connectors, and establish a thermal-electric coupling network model including temperature nodes and signal transmission nodes; According to the thermal-electric coupling network model, the transmission characteristic function is separated and optimized to obtain a temperature-transmission characteristic mapping relationship; Based on the temperature-transmission characteristic mapping relationship, Gaussian process regression calculation is performed on the temperature time series data to obtain the statistical characteristics of the temperature dynamic characteristics; Inputting the temperature dynamic characteristic statistical features into a specified time delay thermal management neural network to predict signal transmission delay, and obtaining delay prediction characteristic parameters; The weighted error value in the transmission feature space is calculated according to the delay prediction characteristic parameter, and the signal transmission delay is dynamically adjusted to obtain the compensated signal transmission delay parameter.

[0006] A second aspect of the present application provides a high-voltage connector fault prediction system based on thermal management, the high-voltage connector fault prediction system based on thermal management comprising: An acquisition module is used to acquire temperature distribution data and electrical signal transmission characteristic data of the high-voltage connector, and to establish a thermal-electric coupling network model including temperature nodes and signal transmission nodes; A separation optimization module, used to separate and optimize the transmission characteristic function according to the thermal-electric coupling network model to obtain a temperature-transmission characteristic mapping relationship; A regression calculation module, used to perform Gaussian process regression calculation on the temperature time series data based on the temperature-transmission characteristic mapping relationship to obtain statistical characteristics of temperature dynamic characteristics; A delay prediction module, used for inputting the statistical characteristics of the temperature dynamic characteristics into a predetermined time delay thermal management neural network to predict the signal transmission delay and obtain delay prediction characteristic parameters; The dynamic adjustment module is used to calculate the weighted error value in the transmission feature space according to the delay prediction characteristic parameter, and dynamically adjust the signal transmission delay to obtain the compensated signal transmission delay parameter.

[0007] Compared with the prior art, the present application has the following beneficial effects: by establishing a thermal-electric coupling network model including temperature nodes and signal transmission nodes, an accurate description of the thermal-electric coupling characteristics of the high-voltage connector is achieved, and a nonlinear thermal-electric separation network optimization method is adopted to effectively extract the mapping relationship between the temperature field distribution and the signal transmission characteristics, thereby reducing the complexity of the model and improving the computational efficiency. The temperature dynamic characteristic identification method based on Gaussian process regression accurately captures the temperature variation law and provides a reliable temperature feature input for delay prediction. A specified time delay thermal management neural network structure is proposed, and through multi-level feature extraction and fusion, an accurate prediction of the signal transmission delay is achieved. A three-layer Bayesian inference algorithm is used to optimize the delay compensation strategy, which significantly improves the accuracy and robustness of the compensation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0009] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0010] Figure 1It is a flow chart of a high-voltage connector fault prediction method based on thermal management provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of a high-voltage connector fault prediction system based on thermal management provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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. 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.

[0012] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0013] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0014] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 In the embodiment of the present application, an embodiment of a high-voltage connector fault prediction method based on thermal management includes: Step 100: collecting temperature distribution data and electrical signal transmission characteristic data of the high-voltage connector, and establishing a thermal-electric coupling network model including temperature nodes and signal transmission nodes; It is understandable that the execution subject of the present application may be a high-voltage connector fault prediction system based on thermal management, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0015] Specifically, the physical structure of the high-voltage connector is divided into equally spaced grids, and the complex physical structure of the connector is discretized into several standard units, each of which contains temperature monitoring points and signal monitoring points. In the process of equally spaced grid division, the finite element method is used to discretize the units to obtain a grid structure with higher consistency and computability. Temperature sampling is performed on the temperature monitoring points in the standard units. The temperature distribution information of each unit is obtained by relying on high-precision temperature sensors, while ensuring that the temporal and spatial resolution of the data meets the requirements of subsequent modeling. After collecting enough temperature data, these sampled data are processed based on the heat conduction equation to calculate the temperature gradient of each unit. The temperature gradient reflects the direction and rate of heat transfer in the unit and is an important basic parameter for describing the heat conduction behavior. Through this step, a basic data set containing the temperature distribution of all units is formed. The temperature distribution data is input into the heat conduction differential equation group for heat conduction characteristic analysis. By solving the heat conduction equation group, the heat exchange characteristics between adjacent units are described, and the heat conduction characteristic function is obtained to quantify the flow rate and intensity of heat energy between units. In this process, the influence of the thermal physical parameters of the material and the environmental conditions on heat transfer is fully considered to ensure the accuracy and physical significance of the calculation results. At the same time, for the signal monitoring points in the standard unit, the electrical signal characteristics are collected and analyzed in parallel. The amplitude, phase and delay of the signal are sampled through high-precision electrical signal measurement equipment to form a basic data set that reflects the dynamic characteristics of the signal. After the collection is completed, these data are input into the transmission line model, and the signal characteristics are calculated by solving the transmission line equation to obtain the characteristic data of the signal transmission in each unit, including key indicators such as signal attenuation, delay and distortion. The electrical signal transmission characteristic data is input into the transmission line equation group for electromagnetic field analysis to obtain the transmission characteristic function that describes the electromagnetic coupling behavior between adjacent units. This function is the core description of the transmission law of electrical signals in space, reflecting the signal interference or coupling phenomenon caused by the electromagnetic field between different units. The coupling relationship between the heat conduction characteristic function and the transmission characteristic function is modeled to form a unit-level thermal-electric coupling equation group. By constructing this equation group, the characteristics of heat transfer and electrical signal transmission are combined and integrated to describe the complex behavior of high-voltage connectors in dynamic environments. The unit-level thermal-electric coupling equation group is substituted into the energy conservation constraint condition for numerical solution. The energy conservation constraint ensures that the solution of the model can truly reflect the energy flow and conversion inside the high-voltage connector. In the process of numerical solution, the matrix calculation method is used to obtain the mapping matrix between the temperature node and the signal transmission node. Based on the mapping matrix, a thermal-electric coupling network model is constructed, which uniformly expresses the temperature distribution and signal transmission characteristics in the form of nodes and connection relationships.

[0016] Step 200: Separate and optimize the transmission characteristic function according to the thermal-electric coupling network model to obtain a temperature-transmission characteristic mapping relationship; Specifically, the heat conduction and electromagnetic field coupling relationship in the mathematical model of the thermal-electric coupling network is linearly separated, and the thermal coupling relationship is decoupled from the electromagnetic coupling relationship through matrix decomposition or other numerical methods, and the complex coupling network is split into independent thermal network subsystems and electrical network subsystems. For the thermal network subsystem, the temperature node data contained in it is substituted into the heat conduction equation group for solution, the distribution characteristics of the temperature field at different nodes are predicted, and a prediction function describing the temperature field distribution is obtained. This prediction function not only contains the spatial variation information of the temperature, but also can reflect its dynamic characteristics over time. At the same time, for the electrical network subsystem, the gain characteristic matrix and the phase response characteristic matrix are constructed based on the signal transmission parameters. These two matrices are used to quantify the degree of amplification or attenuation of the signal between different nodes and the phase change of the signal. Through this step, the initial transmission characteristic function is obtained. The transmission characteristic function is substituted into the nonlinear optimizer for signal distortion minimization processing to obtain the optimization constraint condition. By analyzing the gain change and phase offset of the signal, the main source of distortion is clarified, and the optimization constraint condition is set based on this. According to the optimization constraints, the objective function is constructed, and the boundary parameters are set in combination with the stability threshold of the system to ensure the physical meaning of the system during the optimization process and prevent the solution results from deviating from the actual achievable range. Based on the objective function and boundary parameters, an optimization iteration sequence is generated. In order to ensure that the optimization iteration sequence can quickly converge to the global optimal solution, the step size of the optimization algorithm is adaptively adjusted, and the convergence is dynamically judged. During the optimization process, the adaptive adjustment of the step size can adjust the search range according to the speed of iterative convergence, thereby avoiding slow convergence or falling into the local optimum. By monitoring the error change of each iteration, it is judged whether the optimization goal of minimizing distortion has been achieved. When the optimization algorithm converges to the optimal solution, the optimal transmission characteristic function that satisfies the distortion minimization is obtained. The correlation between the temperature field distribution prediction function and the optimal transmission characteristic function is calculated to reveal the influence of temperature change on the signal transmission characteristics. In the correlation calculation, a method based on statistical methods or physical models is used to map the dynamic changes of the temperature prediction function to the spatial and temporal distribution of the transmission characteristic function to obtain the temperature-transmission characteristic mapping relationship.

[0017] Step 300: Based on the temperature-transmission characteristic mapping relationship, Gaussian process regression calculation is performed on the temperature time series data to obtain the statistical characteristics of the temperature dynamic characteristics; It should be noted that a large amount of temperature change data collected by the high-voltage connector in each working cycle is screened and classified. The temperature time series data under normal working state and fault state are distinguished. During the screening process, according to the predefined temperature change pattern and historical operation records, the abnormal or incomplete data is eliminated through rule matching or machine learning classification algorithm to ensure the quality and representativeness of the input data. The screened temperature time series data is preprocessed. The temperature data is denoised by Gaussian filtering method to effectively remove the interference of random noise in the data while retaining key trend information. The denoised data is centralized and standardized, and the temperature data is normalized to a unified scale to eliminate the influence between data of different dimensions and enhance the adaptability of the model to different data distributions. After completing the data preprocessing, the temperature-transmission feature mapping relationship is input into the normalized temperature data sequence as prior knowledge. This mapping relationship can be used to link the temperature change with the signal transmission characteristics, construct a high-dimensional feature space of temperature change, and capture the complex nonlinear relationship between temperature and transmission characteristics. In the high-dimensional feature space, the temperature data is transformed by radial basis function, and the temperature time series data is mapped into the kernel space to enhance the model's ability to capture nonlinear patterns. After the transformation, the kernel function probability density matrix is ​​obtained to describe the distribution state of the temperature data in the feature space. In order to optimize the performance of the model, the kernel function probability density matrix is ​​optimized for hyperparameters. By inputting it into the maximum likelihood estimation function, the optimal parameter set of the Gaussian process kernel function is determined based on the actual data distribution. Based on the optimal parameter set, the temperature change trajectory is modeled by Gaussian process, and a multivariate Gaussian distribution function describing the dynamic characteristics of temperature is constructed. The multivariate Gaussian distribution function describes the trend and uncertainty of temperature change in the form of probability, and captures the distribution characteristics of temperature at different time points. The multivariate Gaussian distribution function is substituted into the temperature prediction model to calculate the mean and variance of the temperature change. The mean represents the main trend of temperature change, while the variance reflects the uncertainty or fluctuation range of temperature change. These statistics together constitute the statistical characteristics of temperature change. The temperature change statistics are coupled with the temperature-transmission feature mapping relationship to reveal the dynamic influence of temperature change on signal transmission characteristics and obtain the statistical characteristics of temperature dynamic characteristics.

[0018] Step 400, inputting the statistical characteristics of the temperature dynamic characteristics into the specified time delay thermal management neural network to predict the signal transmission delay, and obtaining the delay prediction characteristic parameters; Specifically, the statistical characteristics of temperature dynamic characteristics are reconstructed, and various related information is integrated to fuse the temperature change rate, temperature distribution characteristics, signal integrity indicators and initial delay values ​​into a multi-dimensional input data matrix. The multi-dimensional input data matrix is ​​input into the thermal feature extraction layer of the thermal management neural network with a specified time delay. The thermal feature extraction layer is the first layer of the neural network, which extracts key features related to temperature from the input data. This layer contains three core units: temperature dynamic encoding unit, heat conduction feature analysis unit and temperature fluctuation processing unit. The temperature dynamic encoding unit extracts key dynamic information related to time by encoding the temperature change rate and distribution characteristics; the heat conduction feature analysis unit focuses on modeling the spatial characteristics of temperature distribution and capturing the laws of heat conduction behavior; the temperature fluctuation processing unit extracts stability and disturbance information by analyzing short-term and long-term temperature fluctuations. After being processed by the thermal feature extraction layer, a high-dimensional thermal management feature vector is generated. Time series feature learning is performed on the thermal management feature vector. The time series feature learning layer contains long short-term memory units, time attention mechanism units and time series dependency modeling units. The long short-term memory unit captures feature changes over a long time span through its gating mechanism, and can effectively extract the long-term dependency between temperature dynamics and signal delay; the time attention mechanism unit identifies the most important time points for delay prediction by assigning weights to different time steps, and enhances the network's sensitivity to key time information; the time series dependency modeling unit integrates the features of the time series and establishes global associations between each time step. After being processed by the time series feature learning layer, a feature matrix containing time series associations is generated. The time series association feature matrix is ​​input into the delay prediction backbone network layer. The delay prediction backbone network layer consists of a multi-head self-attention mechanism unit, a residual connection unit, and a specified time constraint unit. The multi-head self-attention mechanism unit can perform multi-perspective correlation analysis on the input data from different feature dimensions, and enhance the network's modeling ability for complex feature relationships; the residual connection unit avoids the gradient vanishing problem and improves the training efficiency of the model by adding jump connections in the network structure; the specified time constraint unit ensures that the prediction result meets the time range of the system's physical characteristics by adding time constraints. After being processed by the delay prediction backbone network layer, the initial delay prediction result is obtained. Temperature compensation is performed on the initial delay prediction result. The temperature compensation layer includes a temperature disturbance analysis unit, a nonlinear compensation unit, and a dynamic correction unit. The temperature disturbance analysis unit analyzes short-term fluctuations in temperature changes to identify temperature change factors that have an important impact on delay prediction; the nonlinear compensation unit introduces nonlinear functions to correct complex temperature effects that cannot be captured by simple linear models; the dynamic correction unit adjusts the prediction results based on real-time temperature data to improve the accuracy of delay prediction. After this layer of processing, the compensated delay prediction value is obtained. The compensated delay prediction value is input into the multi-task learning layer to simultaneously achieve delay prediction, temperature prediction, and system stability evaluation.The multi-task learning layer completes the comprehensive analysis of transmission delay, temperature change trend and system operation status through the delay prediction branch, temperature prediction branch and stability assessment branch. Each branch independently optimizes its task objectives while sharing the underlying feature representation to improve the overall performance of the model. Through this step, a feature set containing multi-dimensional prediction features is generated. The multi-dimensional prediction feature set is subjected to feature fusion processing. The feature fusion layer includes a cross-modal feature fusion unit, an adaptive weight allocation unit and a global information aggregation unit. The cross-modal feature fusion unit captures the correlation information between delay, temperature and stability by fusing the features of different branches; the adaptive weight allocation unit dynamically adjusts the weight of each feature according to its contribution to the final prediction; the global information aggregation unit generates the final delay prediction feature parameters by integrating all feature information.

[0019] Step 500: Calculate the weighted error value in the transmission feature space according to the delay prediction characteristic parameter, and dynamically adjust the signal transmission delay to obtain the compensated signal transmission delay parameter.

[0020] Specifically, the delay prediction feature parameters are mapped to the transmission feature space. The delay prediction feature parameters include delay error, temperature influence and signal integrity characteristics, which together describe the multiple influences on the signal transmission process of the high-voltage connector. By constructing the transmission feature space, these influence characteristics are mapped into a unified multi-dimensional space. In the transmission feature space, a weighted error calculation model is established to quantify the relationship between the delay error and other key features. The weighted error calculation model reflects the relative importance of each feature to the transmission delay by assigning weights to different feature dimensions. The model can capture the main sources of delay error and quantify the specific impact of temperature change and signal integrity on transmission performance. Through this model, the transmission feature space is analyzed for errors. The weighted coefficients of each feature dimension are calculated by the least squares method to obtain the weighted error value. The least squares method ensures that the weighted error value can accurately reflect the actual error situation in the transmission feature space by minimizing the deviation between the weighted error and the real data. The weighted error value and the delay prediction feature parameters are subjected to three-layer Bayesian inference. The Bayesian inference introduces prior knowledge and incorporates uncertainty into the calculation to better predict the compensation strategy parameters. The first layer of the three-layer Bayesian inference constructs a preliminary parameter distribution model by analyzing the direct relationship between the weighted error value and the delay prediction feature parameters; the second layer updates the parameter distribution model to be closer to the actual situation by considering the mutual influence between different features; the third layer forms an optimized compensation strategy parameter set based on a global perspective and combines the multi-layer inference results, which contains a dynamic adjustment scheme for the delay error and reflects the compensation priority under the system operation state. Gibbs sampling and variational inference calculations are performed based on the compensation strategy parameter set. The probability distribution of the delay compensation amount is optimized through sampling and inference methods. Gibbs sampling, as a Markov chain Monte Carlo method, can generate the global optimal parameter distribution by sampling the conditional distribution of each parameter one by one; variational inference accelerates the convergence process of parameter optimization by maximizing the approximate form of the posterior distribution. The two methods are used in combination to efficiently optimize the delay compensation amount so that it can adapt to the actual needs of the complex dynamic environment of the high-voltage connector. Through the above steps, the compensated signal transmission delay parameters are obtained. These parameters reflect the current state of signal transmission and effectively reduce the impact of delay error on system performance through dynamic adjustment.

[0021] The weighted error value and delay prediction feature parameters are input into the first-layer Bayesian inferencer to calculate the distribution parameters. The inferencer uses the input error value and feature parameters to establish a probability distribution model of the delay error. This probability distribution characterizes the possibility distribution characteristics of the delay error under different conditions through comprehensive analysis of historical data and real-time measurement data, providing a preliminary statistical description for the compensation strategy. The probability distribution of the delay error is subjected to variational inference operation, and the distribution parameters are optimized by maximizing the posterior probability to obtain the initial distribution parameters of the delay compensation, including the mean, variance and potential uncertainty factors of the delay error. The initial distribution parameters of the delay compensation and the temperature influence characteristics are input into the second-layer Bayesian inferencer to calculate the joint distribution. Through joint modeling, the interactive relationship between the delay compensation parameters and the temperature influence characteristics is captured. The influence of temperature on delay is nonlinear and has dynamic change characteristics. Through Bayesian inference, the regulatory effect of temperature change on delay error is incorporated into the joint distribution to generate a temperature-delay probability function, which describes the possibility distribution of delay compensation under different temperature conditions and provides a theoretical basis for the adjustment of the system in a dynamic environment. The transition probability matrix of the temperature-delay probability function is constructed. The transition probability matrix is ​​an important tool for describing the state changes of the system. By analyzing the transition law of the delay state due to temperature changes, the dynamic impact of temperature on delay compensation can be quantified. The state transition matrix and the signal integrity index are input into the third-layer Bayesian inference to construct the probability field. The Bayesian inference generates a Markov random field structure by combining the state transition matrix with the signal integrity characteristics. The random field describes the interdependence between states in the form of a graphical model, so that the system can capture the comprehensive impact of signal integrity on the delay state transition from a global perspective. The Markov random field is subjected to conditional probability sampling to obtain the state transition probability sequence. Conditional probability sampling gradually generates a series of possible state transition paths by calculating the probability of the current state under the condition of fixing other states. These paths reflect the dynamic behavior of the system under different input conditions. The maximum expectation calculation is performed based on the state transition probability sequence to obtain the final compensation strategy parameter set. The maximum expectation calculation selects the strategy parameter set that maximizes the expectation by weighted summing the probabilities of the state transition paths to ensure that the compensation strategy is optimal in the global scope. The compensation strategy parameter set contains the optimal compensation scheme for different temperature changes and signal states, and can dynamically adjust the delay compensation to adapt to the complex working environment of the high-voltage connector.

[0022] In the embodiment of the present application, by establishing a thermal-electric coupling network model including temperature nodes and signal transmission nodes, an accurate description of the thermal-electric coupling characteristics of the high-voltage connector is achieved, and a nonlinear thermal-electric separation network optimization method is adopted to effectively extract the mapping relationship between the temperature field distribution and the signal transmission characteristics, thereby reducing the complexity of the model and improving the computational efficiency. A temperature dynamic characteristic identification method based on Gaussian process regression accurately captures the temperature variation law and provides a reliable temperature feature input for delay prediction. A neural network structure for thermal management with a specified time delay is proposed, and accurate prediction of signal transmission delay is achieved through multi-level feature extraction and fusion. A three-layer Bayesian inference algorithm is used to optimize the delay compensation strategy, which significantly improves the accuracy and robustness of the compensation effect.

[0023] In a specific embodiment, the process of executing step 100 may specifically include the following steps: The physical structure of the high-voltage connector is divided into equally spaced grids to obtain several standard units including temperature monitoring points and signal monitoring points; The temperature monitoring points in the standard unit are sampled, and the temperature gradient of the sampled data is calculated based on the heat conduction equation to obtain the temperature distribution data of each standard unit; The temperature distribution data is input into the heat conduction differential equation group to analyze the heat conduction characteristics, and the heat conduction characteristic function describing the heat exchange between adjacent units is obtained; The signal amplitude, phase and delay are sampled at the signal monitoring points in the standard unit, and the signal characteristics of the sampled data are calculated based on the transmission line equation to obtain the electrical signal transmission characteristic data of each standard unit; The electric signal transmission characteristic data is input into the transmission line equation group for electromagnetic field analysis to obtain the transmission characteristic function describing the electromagnetic coupling between adjacent units, and the coupling relationship between the heat conduction characteristic function and the transmission characteristic function is modeled to obtain the unit-level thermal-electric coupling equation group; The unit-level thermal-electric coupling equations are substituted into the energy conservation constraints for numerical solution to obtain the mapping matrix between temperature nodes and signal transmission nodes, and a thermal-electric coupling network model is constructed based on the mapping matrix.

[0024] Specifically, the physical structure of the high-voltage connector is divided into equally spaced grids using the finite element analysis method. Assuming that the length of the connector is , width is , height is , which is divided into A three-dimensional grid, each grid is called a standard unit, the size is ,in , , In each unit, temperature monitoring points and signal monitoring points are arranged. The temperature monitoring points are used to collect thermal characteristic data, and the signal monitoring points are used to obtain electrical signal transmission characteristics. Temperature sampling is performed on the temperature monitoring points in the standard unit to obtain the temperature distribution data during the operation of the high-voltage connector. Assume that in a certain unit The collected temperature is , indicating time Based on the heat conduction equation, calculate the temperature gradient of each unit The basic form of the heat conduction equation is: ; in, is the material density, is the specific heat capacity, is the thermal conductivity, is the temperature, is the heat source term per unit volume. By performing numerical calculations on the sampling data in space and time, the temperature gradient is obtained. , thus describing the distribution and change of heat in the unit. The calculated temperature gradient is input into the heat conduction differential equations to perform heat conduction characteristic analysis to quantify the heat exchange between adjacent units. and , heat exchange using heat conduction characteristic function It means that its expression is: ; in, is the temperature difference between the two cells, is the distance between the centers of the two units, is the heat conduction area, is the thermal conductivity. By performing similar calculations on all units, the thermal conductivity function matrix of the entire system is obtained: , describing the heat transfer relationship between each unit and its neighboring units. At the same time, the signal amplitude, phase and delay are sampled at the signal monitoring points in the standard unit. Assuming that The sampled signal amplitude is , the phase is , the delay is Based on the transmission line equation, the signal characteristics within the cell are calculated. The transmission line equation is in the form of: ; in, is the voltage, is the current, is the inductance per unit length, is the capacitance per unit length. By solving these equations, the characteristic data of the signal in the unit, such as propagation speed, reflection coefficient and attenuation rate, are obtained. The signal transmission characteristic data is input into the transmission line equation group for electromagnetic field analysis to calculate the electromagnetic coupling characteristics between adjacent units. For two adjacent units and , electromagnetic coupling using the transmission characteristic function It is expressed in the form of: ; in, is the voltage difference, is the signal propagation time, is the characteristic impedance. By calculating the transmission characteristic function matrix of the entire system , describing the electromagnetic coupling relationship between all units. After obtaining the heat conduction characteristic function matrix And the transfer characteristic function matrix After that, they are coupled and modeled to form a unit-level thermal-electric coupling equation system. The coupling equation system is in the form of: ; in, is the coupling matrix, is the coupling coefficient, which is used to balance the influence of thermal effect and electromagnetic effect. Substitute the unit-level thermal-electric coupling equations into the energy conservation constraint conditions for numerical solution to ensure that the system satisfies heat and energy conservation. The energy conservation constraint condition is expressed as: ; By solving this set of equations, we can obtain the mapping matrix between temperature nodes and signal transmission nodes: Each column of the matrix represents the coupling relationship between the temperature of a node and the signal characteristics. Based on the mapping matrix , construct a thermal-electric coupling network model. The model uses temperature nodes and signal transmission nodes as basic units, and describes the thermal-electric characteristics of the entire high-voltage connector through mapping relationships.

[0025] In a specific embodiment, the process of executing step 200 may specifically include the following steps: Perform linear separation operation on the heat conduction and electromagnetic field coupling relationship of the thermal-electric coupling network mathematical model to obtain the thermal network subsystem and the electrical network subsystem; Substitute the temperature node data in the thermal network subsystem into the heat conduction equations to solve them, and obtain the prediction function describing the temperature field distribution. Construct the gain characteristic matrix and phase response characteristic matrix for the signal transmission parameters in the electrical network subsystem to obtain the transmission characteristic function. Substitute the transmission characteristic function into the nonlinear optimizer to minimize the signal distortion, obtain the optimization constraints, construct the objective function for the optimization constraints, and set the boundary parameters based on the system stability threshold to obtain the optimization iteration sequence; The step size of the optimization iteration sequence is adaptively adjusted and the convergence is judged to obtain the optimal transmission characteristic function that satisfies the distortion minimization. The temperature field distribution prediction function and the optimal transmission characteristic function are correlated to obtain the temperature-transmission characteristic mapping relationship.

[0026] Specifically, the mathematical model of the thermal-electric coupling network is obtained through physical modeling, which includes the relationship between heat conduction and electromagnetic field interaction. Its basic form is expressed as the following coupling equations: ; in, represents the coupling matrix, represents the thermal network characteristic matrix, represents the electrical network characteristic matrix, is the thermal-electric coupling coefficient, which indicates the degree of influence of heat conduction on signal characteristics. The goal of linear separation operation is to separate and solve these two subsystems independently. Through matrix decomposition methods, such as eigenvalue decomposition or singular value decomposition, the coupling matrix is ​​decomposed into thermal network subsystem and electrical network subsystem, thus obtaining: ; in, and Describe the behavior of heat conduction and electrical signal characteristics when they exist independently. For the thermal network subsystem, substitute its temperature node data into the heat conduction equations for solution. The basic form of the heat conduction equations is: ; in, is the density, is the specific heat capacity, is the thermal conductivity, is the temperature, is the heat source term per unit volume. Assume that the temperature node data Over time By numerically solving this partial differential equation through finite difference method or finite element method, the prediction function describing the temperature field distribution is obtained. ,in Represents the spatial coordinates. At the same time, the gain characteristic matrix and phase response characteristic matrix are constructed for the signal transmission parameters of the electrical network subsystem. Gain characteristic matrix Describes the signal at the node and nodes The amplification or attenuation between them is in the basic form: ; in, and Respectively represent nodes and nodes The voltage of the phase response characteristic matrix It represents the phase change of the signal during transmission and is defined as: ; By combining the gain characteristic matrix and the phase response characteristic matrix, the transmission characteristic function is constructed. ,in represents the signal frequency. The transmission characteristic function Substitute into the nonlinear optimizer to minimize signal distortion. The main sources of distortion include amplitude distortion and phase distortion. The goal is to minimize the following objective function: ; in, and are the measured values, and are calculated values ​​respectively. By setting optimization constraints (such as system stability threshold) and boundary parameters (such as signal power range), an optimization iteration sequence is constructed. During the optimization iteration process, the step size is adjusted and updated according to the current gradient change through the step size adaptive adjustment method (such as Adam optimization algorithm), thereby accelerating convergence. At the same time, through the convergence judgment condition (such as the objective function change is less than the preset threshold), it is ensured that the iteration result reaches the optimal solution. The optimization process obtains the optimal transmission characteristic function that satisfies the distortion minimization . The temperature field distribution prediction function and the optimal transmission characteristic function The correlation calculation is performed to establish the temperature-transmission characteristic mapping relationship. This process is achieved by calculating the correlation coefficient matrix between the two: ; in, Represents the temperature node and The covariance of and Represents the temperature nodes and The standard deviation of . Correlation coefficient matrix The coupling strength between temperature change and signal characteristics is described, thus forming a complete temperature-transmission characteristic mapping relationship.

[0027] In a specific embodiment, the process of executing step 300 may specifically include the following steps: The temperature change data collected by the high-voltage connector in each working cycle is screened and classified to obtain the temperature time series data including the normal working state and the fault state; The temperature time series data is subjected to Gaussian filtering, denoising and centralization standardization to obtain a normalized temperature data sequence, and the temperature-transmission feature mapping relationship is input into the normalized temperature data sequence as prior knowledge to obtain a high-dimensional feature space of temperature change; The temperature data in the high-dimensional feature space is transformed by radial basis function and decomposed by eigenvalues ​​to obtain the kernel function probability density matrix, which is then input into the maximum likelihood estimation function for hyperparameter optimization to obtain the optimal parameter set of the Gaussian process kernel function. Based on the optimal parameter set, the temperature change trajectory is modeled by Gaussian process, and the multivariate Gaussian distribution function describing the dynamic characteristics of temperature is obtained. The multivariate Gaussian distribution function is substituted into the temperature prediction model to calculate the mean and variance, and the temperature change statistics are obtained. The coupling analysis of the temperature change statistics and the temperature-transmission characteristic mapping relationship is carried out to obtain the statistical characteristics of the temperature dynamic characteristics.

[0028] Specifically, the temperature change data collected from the high-voltage connector in multiple working cycles are screened and classified. Assume that the collected temperature time series data is ,in Indicates Monitoring points at time The temperature value, is the total number of monitoring points. By analyzing the stability and change trend of the data, it is divided into normal working state and fault state. The temperature data under normal working state is relatively stable with a small fluctuation range, while the fault state is usually accompanied by drastic abnormal temperature changes. Using rule definition or supervised learning methods, the data is labeled as and , representing the time series under normal and fault conditions, respectively. The filtered temperature time series data is subjected to Gaussian filtering to eliminate random noise in the acquisition process. The Gaussian filtering process is implemented through convolution operations. Represents the smoothed temperature time series. After filtering, the data is centered and standardized to obtain the normalized temperature data series The normalization formula is: ; in, and Respectively represent the sequence The role of centralization and standardization is to eliminate the differences in temperature range and scale of different monitoring points, making the data more suitable for subsequent modeling and analysis. On this basis, the temperature-transmission feature mapping relationship is input into the normalized temperature data sequence as prior knowledge to construct a high-dimensional feature space. Describes temperature changes Signal characteristics The impact is expressed as: ; in Represents the mapping function. By embedding this relationship into the normalized temperature sequence, a high-dimensional feature space containing temperature change characteristics and signal influence characteristics is generated. The temperature data in the high-dimensional feature space is transformed by radial basis function and feature decomposition. The mathematical form of the radial basis function is: ; in represents the feature vectors of two data points, is the scale parameter of the kernel function. Through radial basis function transformation, the original features are mapped to a higher-dimensional feature space to capture nonlinear relationships. Feature decomposition is performed to extract the main features and generate the kernel function probability density matrix , whose elements Represents data points and The kernel function probability density matrix is ​​input into the maximum likelihood estimation function for hyperparameter optimization to determine the optimal parameter set of the Gaussian process kernel function. The goal of maximum likelihood estimation is to maximize the following log-likelihood function: ; in, is the temperature data vector, is the position vector in the feature space, is the hyperparameter of the kernel function, is the kernel function matrix, is the number of data points. , we get the optimal kernel function parameter set to describe the temperature change characteristics. Based on the optimal parameter set, we perform Gaussian process modeling on the temperature change trajectory. The multivariate Gaussian distribution function generated by Gaussian process modeling is: ; in, is the mean function of temperature, is the kernel matrix. By calculating the mean and variance of the distribution function, we can get the statistics of temperature change: and , respectively represent the expectation and uncertainty of temperature. The temperature change statistics are coupled with the temperature-transmission characteristic mapping relationship. By calculating the correlation index between the two, such as the Pearson correlation coefficient, the dynamic impact of temperature change on signal characteristics is quantitatively described, and the statistical characteristics of temperature dynamic characteristics are obtained. In a specific embodiment, the process of executing step 400 may specifically include the following steps: Reconstruct the statistical characteristics of temperature dynamic characteristics to obtain a multi-dimensional input data matrix including temperature change rate, temperature distribution characteristics, signal integrity indicators and initial delay values; Inputting the multi-dimensional input data matrix into the thermal feature extraction layer of the thermal management neural network with a specified time delay, the thermal feature extraction layer comprising a temperature dynamic encoding unit, a heat conduction feature analysis unit and a temperature fluctuation processing unit, to obtain a thermal management feature vector; The thermal management feature vector is processed by the time series feature learning layer, which includes a long short-term memory unit, a time attention mechanism unit, and a time series dependency modeling unit to obtain a time series correlation feature matrix; The time series correlation feature matrix is ​​input into the delay prediction backbone network layer, which includes a multi-head self-attention mechanism unit, a residual connection unit, and a specified time constraint unit to obtain the initial delay prediction result; The initial delay prediction result is processed by a temperature compensation layer, wherein the temperature compensation layer includes a temperature disturbance analysis unit, a nonlinear compensation unit and a dynamic correction unit to obtain a compensated delay prediction value; The compensated delay prediction value is input into a multi-task learning layer, which includes a delay prediction branch, a temperature prediction branch, and a stability evaluation branch, to obtain a multi-dimensional prediction feature set; The multi-dimensional prediction feature set is processed by a feature fusion layer, which includes a cross-modal feature fusion unit, an adaptive weight allocation unit and a global information aggregation unit to obtain delay prediction feature parameters.

[0029] Specifically, the statistical characteristics of temperature dynamic characteristics are reconstructed and converted into a multi-dimensional input data matrix suitable for neural network processing. Assume that the statistical characteristics of temperature dynamic characteristics include the temperature change rate (Describes the rate of change of temperature over time), temperature distribution characteristics (Describes the distribution of temperature in space), signal integrity index (a measure of the quality of signal transmission, such as bit error rate or signal strength), and initial delay value (representing the initial delay of the signal under the current conditions). These features are combined to form a multi-dimensional input data matrix ,in: ; in is the number of sampling points, and each row of the matrix represents the multidimensional features of a sampling point. The constructed multidimensional input data matrix is ​​input into the thermal feature extraction layer of the thermal management neural network with a specified time delay. In the thermal feature extraction layer, the temperature dynamic encoding unit encodes the input features. and Encode and generate characteristic vectors of temperature changes in time and space; the heat conduction characteristic analysis unit simulates the heat conduction process and converts the temperature distribution characteristics into The temperature fluctuation processing unit is used to process the temperature change rate. The short-term temperature fluctuation characteristics are extracted by analyzing the fluctuation amplitude. Assuming that the thermal management feature vector is , which is generated from the output of the extraction layer: ; in Represents the nonlinear mapping function of the thermal feature extraction layer. The thermal management feature vector is input to the time series feature learning layer. This layer contains long short-term memory units (LSTM), time attention mechanism units, and time series dependency modeling units to learn the time series relationship between temperature dynamics and signal delay. LSTM remembers important features over a long time span through a gating mechanism and generates a time series output , the formula is: ; in and is the weight matrix and bias of LSTM, and © represents the element-by-element product. The temporal attention mechanism unit assigns weights to different time steps to strengthen the information of key time points. The temporal dependency modeling unit generates a temporal correlation feature matrix by constructing a global temporal dependency relationship. The time series correlation feature matrix is ​​input into the delay prediction backbone network layer, which contains a multi-head self-attention mechanism unit, a residual connection unit, and a prescribed time constraint unit. The multi-head self-attention mechanism is calculated by the following formula: ; in are query, key, and value matrices, is the dimension of the key. The residual connection unit prevents the gradient from disappearing and enhances the stability of training by adding jump connections, while the time constraint unit ensures that the prediction results meet the physical constraints of the system. The output of the delay prediction backbone network is the initial delay prediction result, which is input into the temperature compensation layer for processing. The temperature compensation layer includes a temperature disturbance analysis unit (analyzing the short-term impact of temperature on delay), a nonlinear compensation unit (introducing nonlinear functions to correct deviations), and a dynamic correction unit (updating the prediction results based on real-time temperature data). After this layer of processing, the compensated delay prediction value is obtained. .Will Input the multi-task learning layer, which contains a delay prediction branch, a temperature prediction branch, and a stability assessment branch. The delay prediction branch generates the final delay prediction result, the temperature prediction branch is used to predict future temperature change trends, and the stability assessment branch evaluates the operating status of the system. The output of the multi-task learning layer is a multi-dimensional prediction feature set. The multi-dimensional prediction feature set is processed by the feature fusion layer, which contains a cross-modal feature fusion unit (integrating features of different modes), an adaptive weight allocation unit (dynamically adjusting feature importance weights), and a global information aggregation unit (extracting global features). The fused output is the delay prediction feature parameter, which is used for dynamic adjustment and fault prediction of high-voltage connectors.

[0030] In a specific embodiment, the process of executing step 500 may specifically include the following steps: The delay prediction characteristic parameters are mapped into transmission characteristic space, and a weighted error calculation model is constructed by combining delay error, temperature influence and signal integrity characteristics; Based on the weighted error calculation model, the error analysis of the transmission feature space is performed, and the weighted coefficients of each feature dimension are calculated by the least square method to obtain the weighted error value in the transmission feature space; Perform three-layer Bayesian inference on the weighted error value and delay prediction feature parameters to obtain the compensation strategy parameter set; Based on the compensation strategy parameter set, Gibbs sampling and variational inference calculations are performed, and the probability distribution of the delay compensation amount is optimized to obtain the compensated signal transmission delay parameters.

[0031] Specifically, the delay prediction feature parameters are mapped to the transmission feature space. Assume that the delay prediction feature parameters include the delay error , Temperature influence , Signal Integrity Characteristics , these parameters together describe the characteristic changes of signal transmission. By defining a mapping function , convert the above parameters into position vectors in the transmission feature space , which is of the form: ; in is a point in the feature space, which is used for subsequent calculation and analysis. After obtaining the transmission feature space representation, a weighted error calculation model is constructed to quantify the error value. Assume that the error function is expressed as: ; in Represents the total error value, is the weight coefficient of the corresponding feature dimension. These weight coefficients are calculated by the least squares method to minimize the deviation between the error model and the actual data. The least squares objective function is: ; in is the sample size, It is The actual error value of samples. Derivative the above objective function and set the derivative to zero to solve the weighting coefficient: ; in , is the input feature matrix, is the error vector. The weighted error value obtained by calculation is combined with the delay prediction feature parameters to perform three-layer Bayesian inference to determine the compensation strategy parameter set. In the first layer of Bayesian inference, the weighted error value is used to construct the probability distribution model of the delay error Assuming that the delay error follows a normal distribution, its probability density function is: ; in and is the mean and variance of the distribution. By maximizing the posterior probability, and Optimize and obtain the initial distribution parameters of delay compensation. In the second layer of Bayesian inference, the initial distribution parameters of delay compensation are jointly modeled with the temperature impact characteristics to construct the temperature-delay joint probability distribution: ; By introducing the transition probability matrix , describing the dynamic impact of different temperature states on delay compensation: ; in Indicates that the temperature state Transfer to state In the third layer of Bayesian inference, the signal integrity features are combined , construct a Markov random field to describe the dependencies between global features. The joint probability distribution of the random field is expressed as: ; in and They are edge potential function and node potential function respectively. Through conditional probability sampling, the state transition sequence of the delay compensation strategy is generated. Using the compensation strategy parameter set, Gibbs sampling and variational inference calculation are performed on the delay compensation amount to optimize the probability distribution of the compensation amount. Gibbs sampling gradually updates the conditional distribution of each parameter through the following iterative formula: ; Variational inference is performed by maximizing the variational lower bound To optimize the distribution parameters: ; in is an approximate distribution. By combining Gibbs sampling and variational inference, the compensated signal transmission delay parameters are obtained.

[0032] In a specific embodiment, the execution step performs three-layer Bayesian inference on the weighted error value and the delay prediction characteristic parameter to obtain the compensation strategy parameter set, which may specifically include the following steps: The weighted error value and the delay prediction feature parameter are input into the first-layer Bayesian inference device to calculate the distribution parameters, and the delay error probability distribution is obtained. The delay error probability distribution is then subjected to variational inference to obtain the initial distribution parameters of the delay compensation. The initial distribution parameters of delay compensation and the temperature influence characteristics are input into the second-layer Bayesian inference machine for joint distribution calculation to obtain the temperature-delay probability function, and the transition probability matrix of the temperature-delay probability function is constructed to obtain the state transfer matrix; The state transfer matrix and the signal integrity index are input into the third-layer Bayesian inference device to construct the probability field, and the Markov random field structure is obtained. The Markov random field structure is then subjected to conditional probability sampling to obtain the state transfer probability sequence. The maximum expectation calculation is performed based on the state transition probability sequence to obtain the compensation strategy parameter set.

[0033] Specifically, the weighted error value and delay prediction characteristic parameters (including delay error , Temperature influence characteristics and signal integrity indicators ) Input the first layer Bayesian inference to calculate the distribution parameters. Assuming the delay error Satisfies the normal distribution, and its probability density function is expressed as: ; in and are the distribution parameters that need to be inferred, representing the mean and variance of the delay error. By maximizing the posterior probability , and obtain the initial estimate of the delay error probability distribution. In the calculation of the posterior probability, the prior assumption of normal distribution is added, such as and Inv-Gamma , the posterior form is: ; Based on the initially calculated probability distribution of delay errors, the distribution parameters are optimized by variational inference. The goal of variational inference is to maximize the variational lower bound (ELBO): ; in is an approximate distribution, by optimizing The initial distribution parameters of delay compensation are obtained by combining the initial distribution parameters of delay compensation with the temperature influence characteristic Input to the second layer Bayesian inference to calculate the joint distribution. Joint distribution function Represents the relationship between delay compensation and temperature characteristics, decomposed into the product of conditional probability and marginal probability: ; In order to characterize the dynamic behavior of the system under different temperature conditions, a temperature-delay transition probability matrix is ​​constructed. , where each element Indicates that the temperature state Transfer to state The probability of is defined as: ; By analyzing the state change pattern in historical data, the transition probability is estimated. In the third level of Bayesian inference, the state transition matrix Signal integrity indicators Combined, a Markov random field (MRF) is constructed. The joint probability distribution of MRF is: ; in, is the potential function between nodes, is the potential function of the node itself. Through conditional probability sampling, the state transition probability sequence is generated , its recursive form is: ; Based on the state transition probability sequence, the maximum expectation algorithm (EM) is used to calculate the final compensation strategy parameter set. The E step of the EM algorithm updates the objective function by calculating the expected value under the current parameters: ; The M step is achieved by maximizing Function update parameters: ; After iterative optimization, the optimized compensation strategy parameter set is finally obtained.

[0034] The above describes the high-voltage connector fault prediction method based on thermal management in the embodiment of the present application. The following describes the high-voltage connector fault prediction system 10 based on thermal management in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of a high-voltage connector fault prediction system 10 based on thermal management includes: The acquisition module 11 is used to acquire the temperature distribution data and the electrical signal transmission characteristic data of the high-voltage connector, and to establish a thermal-electric coupling network model including temperature nodes and signal transmission nodes; A separation optimization module 12 is used to separate and optimize the transmission characteristic function according to the thermal-electric coupling network model to obtain a temperature-transmission characteristic mapping relationship; A regression calculation module 13 is used to perform Gaussian process regression calculation on the temperature time series data based on the temperature-transmission characteristic mapping relationship to obtain the statistical characteristics of the temperature dynamic characteristics; The delay prediction module 14 is used to input the statistical characteristics of the temperature dynamic characteristics into the specified time delay thermal management neural network to predict the signal transmission delay and obtain the delay prediction characteristic parameters; The dynamic adjustment module 15 is used to calculate the weighted error value in the transmission feature space according to the delay prediction characteristic parameter, and dynamically adjust the signal transmission delay to obtain the compensated signal transmission delay parameter.

[0035] Through the collaborative cooperation of the above components, by establishing a thermal-electric coupling network model including temperature nodes and signal transmission nodes, the thermal-electric coupling characteristics of the high-voltage connector are accurately described. The nonlinear thermal-electric separation network optimization method is used to effectively extract the mapping relationship between the temperature field distribution and the signal transmission characteristics, reducing the complexity of the model and improving the computational efficiency. The temperature dynamic characteristics identification method based on Gaussian process regression accurately captures the temperature change law and provides a reliable temperature feature input for delay prediction. A neural network structure for thermal management with a specified time delay is proposed, and the accurate prediction of signal transmission delay is achieved through multi-level feature extraction and fusion. The three-layer Bayesian inference algorithm is used to optimize the delay compensation strategy, which significantly improves the accuracy and robustness of the compensation effect.

[0036] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0037] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0038] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A high voltage connector fault prediction method based on thermal management, characterized in that: The method comprises: Collect temperature distribution data and electrical signal transmission characteristic data of high-voltage connectors, and establish a thermal-electric coupling network model including temperature nodes and signal transmission nodes; According to the thermal-electric coupling network model, the transmission characteristic function is separated and optimized to obtain a temperature-transmission characteristic mapping relationship; Based on the temperature-transmission characteristic mapping relationship, Gaussian process regression calculation is performed on the temperature time series data to obtain the statistical characteristics of the temperature dynamic characteristics; Inputting the temperature dynamic characteristic statistical features into a specified time delay thermal management neural network to predict signal transmission delay, and obtaining delay prediction characteristic parameters; The weighted error value in the transmission feature space is calculated according to the delay prediction characteristic parameter, and the signal transmission delay is dynamically adjusted to obtain the compensated signal transmission delay parameter.

2. The high voltage connector fault prediction method based on thermal management according to claim 1, characterized in that: The method collects temperature distribution data and electrical signal transmission characteristic data of the high-voltage connector and establishes a thermal-electric coupling network model including temperature nodes and signal transmission nodes, including: The physical structure of the high-voltage connector is divided into equally spaced grids to obtain several standard units including temperature monitoring points and signal monitoring points; Performing temperature sampling on the temperature monitoring points in the standard unit, and calculating the temperature gradient of the sampled data based on the heat conduction equation to obtain the temperature distribution data of each standard unit; Inputting the temperature distribution data into a heat conduction differential equation group to perform heat conduction characteristic analysis to obtain a heat conduction characteristic function describing heat exchange between adjacent units; Sampling the signal amplitude, phase and delay of the signal monitoring points in the standard unit, and calculating the signal characteristics of the sampled data based on the transmission line equation to obtain the electrical signal transmission characteristic data of each standard unit; Inputting the electrical signal transmission characteristic data into the transmission line equation group to perform electromagnetic field analysis, obtaining a transmission characteristic function describing electromagnetic coupling between adjacent units, and modeling the coupling relationship between the heat conduction characteristic function and the transmission characteristic function to obtain a unit-level thermal-electric coupling equation group; The unit-level thermal-electric coupling equations are substituted into the energy conservation constraint conditions for numerical solution to obtain a mapping matrix between temperature nodes and signal transmission nodes, and a thermal-electric coupling network model is constructed based on the mapping matrix.

3. The high voltage connector fault prediction method based on thermal management according to claim 2, characterized in that: The separation and optimization of the transmission characteristic function according to the thermal-electric coupling network model to obtain a temperature-transmission characteristic mapping relationship includes: Performing linear separation operation on the heat conduction and electromagnetic field coupling relationship of the thermal-electric coupling network mathematical model to obtain a thermal network subsystem and an electrical network subsystem; Substituting the temperature node data in the thermal network subsystem into the heat conduction equation group for solving, obtaining a prediction function describing the temperature field distribution, and constructing a gain characteristic matrix and a phase response characteristic matrix for the signal transmission parameters in the electrical network subsystem to obtain a transmission characteristic function; Substituting the transmission characteristic function into a nonlinear optimizer to perform signal distortion minimization processing to obtain optimization constraints, constructing an objective function for the optimization constraints, and setting boundary parameters based on a system stability threshold to obtain an optimization iteration sequence; The step size of the optimization iteration sequence is adaptively adjusted and the convergence is judged to obtain the optimal transmission characteristic function that satisfies the distortion minimization, and the correlation between the temperature field distribution prediction function and the optimal transmission characteristic function is calculated to obtain the temperature-transmission characteristic mapping relationship.

4. The high-voltage connector fault prediction method based on thermal management according to claim 3, characterized in that: Based on the temperature-transmission characteristic mapping relationship, Gaussian process regression calculation is performed on the temperature time series data to obtain the statistical characteristics of the temperature dynamic characteristics, including: The temperature change data collected by the high-voltage connector in each working cycle is screened and classified to obtain the temperature time series data including the normal working state and the fault state; Performing Gaussian filtering denoising and centralization standardization processing on the temperature time series data to obtain a normalized temperature data sequence, and inputting the temperature-transmission feature mapping relationship as prior knowledge into the normalized temperature data sequence to obtain a high-dimensional feature space of temperature change; Performing radial basis function transformation and eigendecomposition on the temperature data in the high-dimensional feature space to obtain a kernel function probability density matrix, and inputting the kernel function probability density matrix into a maximum likelihood estimation function to perform hyperparameter optimization to obtain an optimal parameter set of the Gaussian process kernel function; Based on the optimal parameter set, Gaussian process modeling is performed on the temperature change trajectory to obtain a multivariate Gaussian distribution function that describes the dynamic characteristics of temperature, and the multivariate Gaussian distribution function is substituted into the temperature prediction model to calculate the mean and variance to obtain the temperature change statistics; A coupling analysis is performed on the temperature change statistics and the temperature-transmission characteristic mapping relationship to obtain a statistical feature of the temperature dynamic characteristic.

5. The high voltage connector fault prediction method based on thermal management according to claim 4, characterized in that: The step of inputting the temperature dynamic characteristic statistical features into the specified time delay thermal management neural network to predict the signal transmission delay and obtain the delay prediction characteristic parameters includes: Reconstructing the statistical characteristics of the temperature dynamic characteristics to obtain a multi-dimensional input data matrix including temperature change rate, temperature distribution characteristics, signal integrity indicators and initial delay values; Inputting the multidimensional input data matrix into a thermal feature extraction layer of a thermal management neural network with a predetermined time delay, wherein the thermal feature extraction layer includes a temperature dynamic encoding unit, a heat conduction feature analysis unit, and a temperature fluctuation processing unit to obtain a thermal management feature vector; Performing a time series feature learning layer processing on the thermal management feature vector, wherein the time series feature learning layer includes a long short-term memory unit, a time attention mechanism unit, and a time series dependency modeling unit to obtain a time series correlation feature matrix; Inputting the time series correlation feature matrix into a delay prediction backbone network layer, wherein the delay prediction backbone network layer includes a multi-head self-attention mechanism unit, a residual connection unit, and a prescribed time constraint unit, to obtain an initial delay prediction result; Performing temperature compensation layer processing on the initial delay prediction result, the temperature compensation layer comprising a temperature disturbance analysis unit, a nonlinear compensation unit and a dynamic correction unit, to obtain a compensated delay prediction value; Inputting the compensated delay prediction value into a multi-task learning layer, wherein the multi-task learning layer includes a delay prediction branch, a temperature prediction branch, and a stability evaluation branch, to obtain a multi-dimensional prediction feature set; The multi-dimensional prediction feature set is subjected to feature fusion layer processing, wherein the feature fusion layer comprises a cross-modal feature fusion unit, an adaptive weight allocation unit and a global information aggregation unit, so as to obtain delay prediction feature parameters.

6. The high voltage connector fault prediction method based on thermal management according to claim 5, characterized in that: The step of calculating the weighted error value in the transmission feature space according to the delay prediction characteristic parameter and dynamically adjusting the signal transmission delay to obtain the compensated signal transmission delay parameter includes: Performing transmission feature space mapping on the delay prediction feature parameters, and constructing a weighted error calculation model based on delay error, temperature influence and signal integrity features; Based on the weighted error calculation model, error analysis is performed on the transmission feature space, and the weighted coefficients of each feature dimension are calculated by the least square method to obtain the weighted error value in the transmission feature space; Performing three-layer Bayesian inference on the weighted error value and the delay prediction characteristic parameter to obtain a compensation strategy parameter set; Based on the compensation strategy parameter set, Gibbs sampling and variational inference calculations are performed, and probability distribution optimization is performed on the delay compensation amount to obtain the compensated signal transmission delay parameters.

7. The high voltage connector fault prediction method based on thermal management according to claim 6, characterized in that: The three-layer Bayesian inference is performed on the weighted error value and the delay prediction characteristic parameter to obtain a compensation strategy parameter set, including: Inputting the weighted error value and the delay prediction feature parameter into the first layer Bayesian inference device to calculate the distribution parameters to obtain the delay error probability distribution, and performing variational inference operation on the delay error probability distribution to obtain the initial distribution parameters of the delay compensation; Inputting the initial distribution parameters of the delay compensation and the temperature influence characteristics into the second-layer Bayesian inference machine to perform joint distribution calculation to obtain a temperature-delay probability function, and constructing a transition probability matrix for the temperature-delay probability function to obtain a state transfer matrix; Inputting the state transfer matrix and the signal integrity index into the third-layer Bayesian inference device to construct a probability field to obtain a Markov random field structure, and performing conditional probability sampling on the Markov random field structure to obtain a state transfer probability sequence; A maximum expectation calculation is performed based on the state transition probability sequence to obtain a compensation strategy parameter set.

8. A high voltage connector fault prediction system based on thermal management, characterized in that: Used to execute the high-voltage connector fault prediction method based on thermal management according to any one of claims 1 to 7, the high-voltage connector fault prediction system based on thermal management comprises: An acquisition module is used to acquire temperature distribution data and electrical signal transmission characteristic data of the high-voltage connector, and to establish a thermal-electric coupling network model including temperature nodes and signal transmission nodes; A separation optimization module, used to separate and optimize the transmission characteristic function according to the thermal-electric coupling network model to obtain a temperature-transmission characteristic mapping relationship; A regression calculation module, used to perform Gaussian process regression calculation on the temperature time series data based on the temperature-transmission characteristic mapping relationship to obtain statistical characteristics of temperature dynamic characteristics; A delay prediction module, used for inputting the statistical characteristics of the temperature dynamic characteristics into a predetermined time delay thermal management neural network to predict the signal transmission delay and obtain delay prediction characteristic parameters; The dynamic adjustment module is used to calculate the weighted error value in the transmission feature space according to the delay prediction characteristic parameter, and dynamically adjust the signal transmission delay to obtain the compensated signal transmission delay parameter.

Images