SiC power device packaging reliability detection method and system

Abstract

The invention relates to the technical field of semiconductor device detection, and provides a SiC power device packaging reliability detection method and system, and the detection method comprises the following steps: constructing a multi-physics field coupling finite element model, and operating the finite element model to obtain initial simulation data; performing parameter sensitivity analysis on the initial simulation data to determine a key stress parameter combination of the typical failure mode; collecting working parameters of the SiC power device, performing real-time calibration and denoising to obtain measured data, and generating a temperature gradient map of a key area; analyzing the initial simulation data and the measured data through a deep reinforcement learning algorithm, and carrying out iterative optimization on the key stress parameter combination and the key region parameters; and setting a failure criterion threshold value, performing trend prediction, comparing a trend prediction result with the failure criterion threshold value, and executing a multi-level response strategy according to a comparison result. According to the invention, accurate detection of the packaging reliability of the SiC power device in a multi-physics coupling scene can be realized.

Description

Technical Field

[0001] This application relates to the field of semiconductor device testing technology, and in particular to a method and system for testing the reliability of SiC power device packaging. Background Technology

[0002] Silicon carbide (SiC) power devices are semiconductor power devices widely used in electric vehicles, smart grids, and aerospace due to their excellent high-temperature resistance, high-frequency operation, and high power density. The packaging of SiC power devices is fundamental to ensuring their safety and enabling them to fully realize their superior performance; therefore, testing the packaging reliability of SiC power devices is crucial. Currently, there are various methods for testing the packaging reliability of SiC power devices, such as infrared thermography-based thermal resistance analysis or inferring failure trends through electrical performance degradation curves. While these existing testing methods can detect the packaging performance of SiC power devices to some extent, all testing processes rely on the measurement of a single physical parameter, such as simple temperature monitoring or electrical performance testing. These methods cannot comprehensively and accurately reflect the packaging reliability of SiC power devices under complex application scenarios. Summary of the Invention

[0003] This application aims to address the technical problem that existing methods for testing the packaging reliability of SiC power devices rely on measuring only a single physical parameter, making it difficult to comprehensively and accurately reflect the packaging reliability of SiC power devices in complex application scenarios. It proposes a method and system for testing the packaging reliability of SiC power devices to achieve accurate testing of SiC power device packaging reliability in multi-physics coupling scenarios.

[0004] To achieve the above objectives, this application adopts the following technical solution: A method for testing the reliability of SiC power device packaging includes the following steps: Constructing a multiphysics coupled finite element model for SiC power devices; Initial simulation data is obtained by running a multiphysics coupled finite element model, and parameter sensitivity analysis is performed on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The operating parameters of the SiC power device under operating conditions are collected, and the operating parameters are calibrated in real time and denoised to obtain measured data. During the process of collecting the operating parameters of the SiC power device under operating conditions, the key areas of the SiC power device are located by dynamic region tracking technology, and the temperature gradient map of the key areas of the SiC power device is generated based on the measured data. The simulation data and the measured data are compared and analyzed by deep reinforcement learning algorithm. Based on the comparison and analysis results, the key stress parameter combination and the key area parameters of the multiphysics coupled finite element model are iteratively optimized until the error between the initial simulation data and the measured data is within a set range, thereby obtaining the optimized key stress parameter combination and the key area parameters of the multiphysics coupled finite element model. The final simulation data is obtained by performing simulations based on the optimized combination of key stress parameters and the key region parameters of the multiphysics coupled finite element model. The failure criterion threshold is then set based on the final simulation data. The temperature gradient map of the key area is input into the trend prediction model to predict the trend and obtain the prediction data. The prediction data is compared with the failure criterion threshold, and a multi-level response strategy is executed based on the comparison result.

[0005] Furthermore, the process of constructing a multiphysics coupled finite element model of SiC power devices includes: constructing or importing a geometric model of SiC power devices, setting SiC power device parameters, setting boundary conditions, loads and typical failure modes of the model, and performing mesh generation and solution settings to obtain a multiphysics coupled finite element model.

[0006] Furthermore, setting the parameters of the SiC power device includes setting the thermal conductivity, coefficient of thermal expansion, elastic modulus, and Poisson's ratio of its packaging material at different temperatures; the thermal conductivity of the packaging material at different temperatures is obtained by laser scintillation method.

[0007] Furthermore, the typical failure modes include solder layer cracking, thermal resistance degradation, and bond wire detachment. The failure of solder layer cracking is determined by obtaining the strain concentration area in the thermo-mechanical stress simulation. The failure of thermal resistance degradation is determined by obtaining the simulated thermal resistance parameters. The failure of bond wire detachment is determined by obtaining the stress analysis of the bonding area in the electro-thermal coupling simulation.

[0008] Furthermore, the process of performing parameter sensitivity analysis on the initial simulation data to determine the key stress parameter combinations for typical failure modes is as follows: designing a stress parameter sampling matrix using the Morris global sensitivity analysis method; performing finite element simulation on each stress parameter combination and outputting failure indices; calculating the normalized sensitivity coefficient of each stress parameter to failure time, and determining the key stress parameter combinations for typical failure modes.

[0009] Furthermore, the real-time calibration process involves using an environmental radiation compensation algorithm to calibrate the temperature parameter in the operating parameters in real time.

[0010] Furthermore, the process of using the environmental radiation compensation algorithm to calibrate the temperature parameter in the working parameters in real time is as follows: during the acquisition of the temperature parameters of the SiC power device, the radiation intensity at different wavelengths is measured, and the measured temperature is calculated and calibrated according to the blackbody radiation law.

[0011] Furthermore, the denoising process involves: aligning the working parameters with the simulation data in time, and then using a wavelet threshold denoising algorithm to eliminate environmental noise in the working parameters.

[0012] Furthermore, the failure criterion thresholds include a warning threshold and a danger threshold. The multi-level response strategy employs a fuzzy logic control method, and the membership function used in the fuzzy logic control process is: ; In the formula, The membership degree represents the severity of the failure, with a range of values. x represents the predicted data; a represents the warning threshold; and b represents the danger threshold.

[0013] A testing system based on the same inventive concept as the above-mentioned SiC power device packaging reliability testing method includes: The model building module is used to build multiphysics coupled finite element models of SiC power devices; The initial simulation and parameter analysis module is used to run a multiphysics coupled finite element model to obtain initial simulation data, and to perform parameter sensitivity analysis on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The measured data acquisition module includes a working parameter acquisition unit, a data calibration and denoising unit, a dynamic region tracking unit, and a temperature gradient map generation unit; The operating parameter acquisition unit is used to acquire the operating parameters of the SiC power device in operation. The data calibration and denoising unit is used to calibrate and denoise the working parameters in real time to obtain measured data. The dynamic region tracking unit is used to locate the key regions of the SiC power device using dynamic region tracking technology; The temperature gradient map generation unit is used to generate a temperature gradient map of the key region of the SiC power device based on the measured data. The comparative analysis and iterative optimization module is used to compare and analyze the simulation data and the measured data using a deep reinforcement learning algorithm. Based on the comparative analysis results, iterative optimization is performed on the key stress parameter combination and the key region parameters of the multiphysics coupled finite element model until the error between the initial simulation data and the measured data is within a set range, thereby obtaining the optimized key stress parameter combination and the key region parameters of the multiphysics coupled finite element model. The final simulation and failure criterion setting module is used to simulate and obtain the final simulation data based on the optimized combination of key stress parameters and the key region parameters of the multiphysics coupled finite element model, and to set the failure criterion threshold based on the final simulation data. The trend prediction and response execution module includes a trend prediction unit and a comparative response unit; The trend prediction unit is used to input the temperature gradient map of the key area into the trend prediction model to make trend predictions and obtain prediction data. The comparison response unit is used to compare the predicted data with the failure criterion threshold and execute a multi-level response strategy based on the comparison result.

[0014] The beneficial effects of this application are: This application first constructs a multiphysics coupled finite element model for simulation testing of SiC power devices. This model fully considers the packaging reliability of SiC power devices under various complex application scenarios, enabling a more accurate assessment of the thermo-mechanical stress distribution of SiC power devices during actual operation. This provides strong support for optimized device design and reliability improvement. Simultaneously, this application iteratively optimizes and updates the simulation data using measured data, effectively improving the accuracy of the final test results. Furthermore, the measured data is calibrated using an environmental radiation compensation algorithm, effectively avoiding the interference of ambient temperature on the measured data and ensuring its reliability, thereby further improving the accuracy of the final test results.

[0015] This application effectively ensures the consistency and comparability of the two sets of data by aligning the simulation data with the measured data over time. Then, the wavelet threshold denoising algorithm is used to eliminate environmental noise in the data, thereby effectively improving the signal-to-noise ratio and improving the accuracy of subsequent data analysis. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart illustrating the SiC power device packaging reliability testing method provided in this application embodiment; Figure 2 This is a schematic structural block diagram of a SiC power device package reliability testing system provided in an embodiment of this application. Detailed Implementation

[0018] This application provides a method and system for detecting the package reliability of SiC power devices. It uses COMSOL Multiphysics software to construct a multiphysics coupled finite element model to simulate and test the SiC power device. The temperature parameter in the operating parameters is calibrated using an environmental radiation compensation algorithm. Then, wavelet thresholding is used to denoise the operating parameters to obtain measured data. Next, a deep reinforcement learning algorithm is used to analyze and process the measured data and initial simulation data in real time, iteratively optimizing each parameter. Based on the optimized parameters, final simulation data is obtained, and a failure criterion threshold is set using the final simulation data. A prediction model is then used to predict the future value trend of the measured data to obtain predicted data. Finally, the predicted data is compared with the failure criterion threshold, and a multi-level response strategy is executed based on the comparison result, thereby achieving failure prediction of SiC power devices. In use, it can fully consider the package reliability of SiC power devices under various complex application scenarios, achieving accurate detection of SiC power device package reliability in multiphysics coupled scenarios. This solves the technical problem that existing methods for testing the packaging reliability of SiC power devices rely on a single physical parameter for measurement, making it difficult to comprehensively and accurately reflect the packaging reliability of SiC power devices in complex application scenarios.

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.

[0020] like Figure 1 As shown, this embodiment provides a method for testing the reliability of SiC power device packages, applied to a SiC power device package reliability testing system. The SiC power device can be a SiC MOSFET / diode, etc. The testing method specifically includes the following steps: The geometric model of SiC power device is constructed or imported using COMSOL Multiphysics software. The parameters of SiC power device are set, the boundary conditions, loads and typical failure modes of the model are set, and the mesh is generated and the solution is set to obtain a multiphysics coupled finite element model.

[0021] In the above steps, setting the SiC power device parameters may include setting the thermal conductivity, coefficient of thermal expansion, elastic modulus, and Poisson's ratio of the packaging material at different temperatures. In practical applications, those skilled in the art can set the corresponding SiC power device parameters according to actual needs. The thermal conductivity of the packaging material at different temperatures is obtained by laser scintillation method.

[0022] In some embodiments, the parameters of the SiC power device are specifically set as follows:

[0023] In the above steps, the boundary conditions can be thermal boundaries, electrical boundaries, or mechanical boundaries; in practical applications, those skilled in the art can set the boundary conditions according to actual needs.

[0024] In some embodiments, the boundary conditions and loads can be set as follows: Thermal boundary: Forced air cooling or liquid cooling is adopted, the convective heat transfer coefficient is set to 50~200 W / (m²·K), and the ambient temperature is set to 25~85℃; Electrical boundary: For SiC MOSFETs, the drain-source voltage Vds is set to 0~1200V, and the gate-source voltage Vgs is set to -5~20V; Mechanical boundary: The displacement constraint of the fixed edge of the substrate is 0, and the power terminal is subjected to periodic vibration load (frequency 10~200Hz, acceleration 2~5g). Loading step: Perform thermal cycling with a temperature range of -40℃ to 150℃ and a heating / cooling rate of 10℃ / min. Perform electromigration testing using DC or pulsed current with a duty cycle of 50% to 100%.

[0025] The typical failure modes include solder layer cracking, thermal resistance degradation, and bond wire detachment. Solder layer cracking failure is determined by obtaining the strain concentration zone from thermo-mechanical stress simulation; thermal resistance degradation failure is determined by obtaining simulated thermal resistance parameters; and bond wire detachment failure is determined by obtaining stress analysis of the bond region from electro-thermal coupling simulation.

[0026] Initial simulation data is obtained by running a multiphysics coupled finite element model. Parameter sensitivity analysis is then performed on the initial simulation data to determine the key stress parameter combinations for typical failure modes.

[0027] The simulation analysis obtained by running the multiphysics coupled finite element model to obtain the initial simulation data is achieved by simultaneously solving the following set of equations: (1); (2); (3); In equation (1), For stress tensor; Let T be the strain tensor; T be the temperature field; in equation (2), Temperature-dependent thermal conductivity; For gradient temperatures; For the Joule heat source term; For the thermoelastic heat source term; in equation (3), J is the current density; ν is the electrical conductivity; E is the electric field strength; v is the carrier drift velocity; B is the magnetic flux density.

[0028] equation It describes the deformation of the encapsulation material under stress, where the relationship between the stress tensor and the strain tensor reflects the elastic correlation of the material, providing basic data for subsequent thermal and electrophysical field analysis of structural deformation.

[0029] equation It describes the heat conduction process in materials. Temperature-dependent thermal conductivity represents the change of thermal conductivity with temperature, while gradient temperature represents the spatial rate of temperature change. Temperature-dependent thermal conductivity and gradient temperature can accurately simulate heat transfer in different temperature regions, while Joule heat source term and thermoelastic heat source term introduce the effect of electric and mechanical fields on the thermal field.

[0030] equation The current density is obtained by combining factors such as electric field, carrier drift velocity and magnetic induction intensity. The Joule heat source term can be calculated from the current density and electric field intensity, which reflects the heat generated when the current passes through the material.

[0031] When a SiC power device is operating, the current density can be calculated using equation (3) as the current flows through the device. Then, the Joule heat source term Q can be calculated using the current density and the electric field strength. joule The obtained Joule heat source term Q joule The temperature gradient is fed back into equation (2) and drives the change in the temperature gradient. The change in the temperature gradient leads to a change in the temperature field T. The change in the temperature field drives a change in the stress tensor and strain tensor in equation (1), which in turn causes thermal expansion and contraction of the material, generating thermal stress. Thus, the thermoelastic heat source term Q can be calculated. piezoFor most encapsulation materials, under adiabatic or high-speed deformation conditions, their temperature rises when rapidly compressed and falls when rapidly stretched. This thermal effect caused by volume change is called the thermoelastic effect. The thermoelastic heat source term Q is obtained. piezo This is then fed back into equation (2). Since conductivity is a function of temperature, in the above process, when the temperature field T changes, it will also affect the conductivity in equation (3), thus causing the current density in equation (3) to change again. It can be seen that by combining the above three equations, the cyclic coupling process of "electricity generates heat → heat generates force → force generates heat again" is fully covered. The synchronous iterative calculation of multiple physics fields is realized instead of the isolated analysis of a single physics field, which can better reflect the situation of multiple fields interacting in reality, improve the accuracy of model prediction, and provide an important reference for the design and optimization of SiC power devices.

[0032] The initial simulation data obtained by performing multiphysics coupled finite element simulation using the above equations serves as the direct input for subsequent parameter sensitivity analysis. This initial simulation data includes the spatial distribution parameters of the entire SiC power device package structure directly output from the coupled simulation, as well as calculated parameters used to quantify failure risk. These two types of parameters are specifically: 1. The main physical quantity distribution parameters of the entire SiC power device package structure in space, directly output from coupled simulation, include: Temperature field (T): It is obtained by solving the heat conduction equation (2) above and is used to identify overheated regions and calculate temperature gradients; Stress field (σ) and strain field (ε): These are obtained by solving the above mechanical equilibrium equation (1) and are used to locate the mechanical stress / strain concentration area caused by thermal expansion mismatch. Current density field (J): It is obtained by solving the above current field equation (3) and is used to identify the current crowding effect and the high Joule heating region; 2. The parameters obtained for quantifying failure risk are based on the aforementioned physical field distribution parameters and are determined / calculated according to the criteria for typical failure modes. These include solder layer cracks, thermal resistance degradation, and bond wire detachment, etc. These parameters are the direct target response quantities for parameter sensitivity analysis. The specific determination / calculation process for each parameter used to quantify failure risk is as follows: For solder layer cracks: extract the maximum equivalent strain of the solder layer from the strain field (ε) to determine the strain concentration area; To address thermal resistance degradation: Based on the temperature field (T) data, the difference between the chip junction temperature and the case temperature is calculated, and combined with the input power, the simulated thermal resistance value, i.e., the simulated thermal resistance parameter, is obtained. For bond line detachment: extract the maximum principal stress or shear stress from the stress field (σ) in the bond region for stress analysis of the bond region;

[0033] After obtaining the initial simulation data, a failure baseline state under a certain set of fixed input parameters is obtained. Subsequently, parameter sensitivity analysis can be performed on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The specific process is as follows: a stress parameter sampling matrix is ​​designed using the Morris global sensitivity analysis method. The stress parameters include physical variables affecting the failure time of SiC power devices, including: temperature-related parameters such as ambient temperature, junction temperature, and case temperature; current-related parameters such as device operating current and current density; voltage-related parameters such as drain-source voltage and gate-source voltage; thermal cycle count; and mechanical stresses such as interlayer stress caused by thermal expansion coefficient mismatch. Those skilled in the art can design the stress parameter sampling matrix according to actual needs. In some embodiments, it can be designed as follows: temperature 80℃ / 120℃ / 150℃, current 10A / 20A / 30A, and thermal cycle count 100 / 500 / 1000 cycles. After the design is completed, failure time data corresponding to each set of stress parameters is obtained by performing a coupled simulation once for each stress parameter in the stress parameter sampling matrix. Then, by calculating the normalized sensitivity coefficient of each changed stress parameter to the failure time, the key stress parameter combination for typical failure modes is determined. The formula for calculating the sensitivity coefficient is as follows: ; In the formula, Sensitivity coefficient; Y represents the change in failure time; Y represents the baseline failure time. This represents the change in the i-th stress parameter; This is the reference value for the i-th stress parameter; First, a reference value for the i-th stress parameter needs to be set based on the actual application scenario or the datasheet of the SiC power device. and stress parameter variation For example, the temperature reference value can be set to 25℃, and the current reference value can be set to 10A. The change in stress parameters is a percentage or absolute value of the reference value, for example, the temperature change can be set to ±50℃, and the current change can be set to ±5A. When calculating the sensitivity coefficient, the reference failure time is obtained through finite element simulation, then the individual stress parameters are varied, and the change in failure time is re-simulated to obtain the change. The corresponding value is then substituted into the formula for calculation to obtain the normalized sensitivity coefficient. When the stress parameter is determined to be critical, Morris global sensitivity analysis can efficiently identify the stress parameters that have the greatest impact on failure time, providing crucial guidance for optimized design and improved reliability. This helps shorten the development cycle and reduce costs of SiC power devices. By identifying the combination of critical stress parameters, targeted monitoring and protection of these areas can be strengthened, reducing failure risk and improving device lifespan and reliability.

[0034] The Morris global sensitivity analysis method described above is a sensitivity analysis method based on "one factor at a time" changes. It assesses the importance of each parameter by calculating the normalized sensitivity coefficient of each stress parameter to the failure time. In this method, each stress parameter varies within a certain range while other parameters remain constant. The degree of influence of that parameter on the failure time is determined by observing the change in failure time. The larger the sensitivity coefficient, the more significant the effect of that parameter on the failure time, thus identifying it as a critical stress parameter.

[0035] In some embodiments, the key stress parameter combination for the typical failure modes of the SiC power device needs to cover the thermo-electric-mechanical multiphysics coupling effect, specifically including at least the following parameters and typical combinations: Single parameters: temperature, junction temperature, case temperature, current density, device operating current, area, voltage, drain-source voltage Vds, gate-source voltage Vgs, number of thermal cycles, number of temperature fluctuation cycles, mechanical stress, and interlayer stress caused by thermal expansion coefficient mismatch.

[0036] Combined parameters: Temperature-current interaction: High temperature + high current accelerates electromigration failure; Temperature-thermal cycle interaction: Temperature fluctuations + high temperature lead to solder layer fatigue; Interaction between electric current and mechanical stress: Joule heating at high current density intensifies thermal stress.

[0037] The operating parameters of the SiC power device under operating conditions are collected and calibrated in real time. Then, the real-time calibrated operating parameters are time-aligned with the simulation data. The wavelet threshold denoising algorithm is used to eliminate environmental noise in the operating parameters to obtain the measured data. During the collection of SiC power device operating parameters, the key areas of the SiC power device are located by dynamic region tracking technology, and the temperature gradient spectrum of the key areas of the SiC power device is generated based on the measured data.

[0038] In the above steps, the process of acquiring the operating parameters of the SiC power device includes: arranging temperature sensors, current sensors, and voltage sensors at key locations of the SiC power device to acquire the operating temperature, current, and voltage of the SiC power device in real time, thereby obtaining the operating parameters. The key locations are those areas with significant thermo-mechanical coupling and the highest failure risk, identified in previous multiphysics coupling simulations and parameter sensitivity analyses. These include the chip-substrate solder layer experiencing shear stress due to CTE mismatch; the power terminal bonding area experiencing Joule heating due to high current density; the chip surface center forming hotspots at power densities >500 W / cm²; the package edge corners with concentrated mechanical stress; and the gate drive location experiencing voltage spikes due to high-frequency switching.

[0039] Because the electrical parameters of SiC power devices, such as operating current and voltage, are generally less affected by external factors during operation, while temperature parameters are significantly affected by the environment, real-time calibration of operating parameters typically involves real-time calibration of the temperature parameter. The real-time calibration method used is based on an environmental radiation compensation algorithm. Specifically, during the acquisition of SiC power device temperature parameters, an integrated dual-color infrared thermometer is used to measure the radiation intensity at different wavelengths. Then, the measured temperature is calculated and calibrated according to the blackbody radiation law, thereby achieving real-time calibration of the SiC power device's measured temperature to eliminate the influence of ambient temperature. The real-time calibration of the measured temperature is achieved using the following formula: ; In the formula, This is the corrected actual temperature; It is the second radiation constant; , This is the center wavelength of the dual-wavelength filter; , The emissivity of the packaging material corresponding to the center wavelength; , denoted as blackbody radiation spectral density; T is temperature.

[0040] In the above correction formula, the corrected true temperature is calculated using the center wavelength of the dual-wavelength filter, the emissivity of the packaging material at the corresponding wavelength, the blackbody radiation spectral density, and the second radiation constant. The dual-color infrared thermometry method simultaneously measures the radiation intensity at two wavelengths and utilizes the radiation characteristics at these two wavelengths to eliminate the influence of ambient radiation on the temperature measurement results, thereby improving the accuracy of the temperature measurement and ensuring the reliability of the temperature data. This algorithm is applicable to temperature measurement in various complex environments, providing strong support for the reliable operation of SiC power devices.

[0041] In the above steps, the process of locating key regions of SiC power devices using dynamic region tracking technology and generating temperature gradient maps of key regions based on measured data is as follows: An infrared thermometer scans several regions per second (e.g., 100 regions), and combined with image recognition to locate temperature changes, generating temperature gradient maps of key regions with a temperature difference ≥10℃ / mm. The key regions are those areas with significant thermo-mechanical coupling and the highest failure risk, identified in previous multiphysics coupling simulations and parameter sensitivity analyses. These include the chip-substrate solder layer experiencing shear stress due to CTE mismatch, the power terminal bonding area experiencing Joule heating due to high current density, the chip surface center forming hotspots with power density >500 W / cm², package edge corners with concentrated mechanical stress, and the gate drive region experiencing voltage spikes due to high-frequency switching. These key regions are actually related to and progressively related to the aforementioned key locations. A key region describes a range, while a key location refers to a specific point within that region. The temperature gradient map is constructed using the following two-dimensional heat flux density formula: ; In the formula, Let be the heat flux density vector at coordinates (x, y); Temperature-dependent thermal conductivity; , , i and j represent the components of the temperature gradient in the x and y directions, respectively; i and j are the unit vectors of the coordinate axes. The two-dimensional heat flux density formula can intuitively display the temperature distribution and heat flow direction within the SiC power device package structure, providing important basis for analyzing heat conduction paths and hotspot locations, and helping to optimize the thermal design of SiC power devices. By constructing a temperature gradient map, potential thermal problem areas can be identified in a timely manner, and corresponding heat dissipation measures can be taken to improve the thermal stability and reliability of the device.

[0042] The initial simulation data and the measured data are compared and analyzed by a deep reinforcement learning algorithm. Based on the comparison and analysis results, the key stress parameter combination and the key area parameters of the multiphysics coupled finite element model are iteratively optimized until the error between the initial simulation data and the measured data is within a set range, thereby obtaining the optimized key stress parameter combination and the key area parameters of the multiphysics coupled finite element model.

[0043] The deep reinforcement learning algorithm employs an improved DDPG algorithm, whose action value function is updated using the following Bellman equation: ; ; In the formula, Used to measure the mean square error between the current Q value and the target Q value; For network parameters; Let y be the Q-value of action a chosen by the network in state s; y is the target Q-value, determined by the next state. The maximum Q value is calculated; For the samples in the experience playback buffer D Calculate the average; r is the immediate reward, the feedback obtained from the environment after performing action a; As a discount factor, it weighs the importance of future rewards; The next state; The Q-value of the target network; The action output by the target policy network; If the space is a discrete action space, take the following... For the largest action, if it is a continuous action space, the action is directly given by the policy network.

[0044] In the improved DDPG algorithm, the action value function is updated using the Bellman equation to measure the mean square error between the current Q-value and the target Q-value, and the learning process is optimized through an empirical replay buffer. The algorithm dynamically adjusts the combination of key stress parameters and the parameters of key regions through continuous trial and error and optimization, thereby improving the system's adaptability and detection efficiency. The improved DDPG algorithm, through continuous trial and error and optimization, achieves dynamic adjustment of the combination of key stress parameters and the parameters of key regions, improving the system's adaptability and enabling it to better adapt to detection requirements under different working conditions. The application of the Bellman equation allows the algorithm to consider the importance of future rewards, thus making more long-term decisions and improving the overall performance and stability of the system.

[0045] When iteratively optimizing temperature parameters, a deviation comparison analysis is required. The specific process is as follows: the temperature parameters in the measured data are used as the heat flow boundary conditions of the multiphysics coupled finite element model (e.g., the temperature of the temperature measurement point is assigned to the corresponding mesh node to replace the convective heat transfer calculation in the simulation); the deviation between the measured temperature and the simulated temperature (e.g., root mean square error RMSE) is compared. If the deviation exceeds 5%, the thermal conductivity or convective heat transfer coefficient in the multiphysics coupled finite element model is adjusted until the deviation is less than 5% before iterative optimization of the parameters is performed.

[0046] The final simulation data is obtained by performing simulations based on the optimized combination of key stress parameters and the key region parameters of the multiphysics coupled finite element model. Failure criterion thresholds are then set based on this final simulation data; these thresholds include a warning threshold and a danger threshold. The temperature gradient map of the key region is input into a trend prediction model to predict the trend of the measured data and obtain predicted data. The predicted data is compared with the failure criterion thresholds, and a multi-level response strategy is executed based on the comparison results. The multi-level response strategy includes a danger response strategy, a warning response strategy, and a normal response strategy.

[0047] The specific process of comparing the predicted data with the failure criterion threshold and executing a multi-level response strategy based on the comparison result is as follows: The predicted data is compared with the danger threshold. When the predicted data is greater than or equal to the danger threshold, the corresponding danger response strategy in the multi-level response strategy is triggered. When the predicted data is less than the danger threshold, the predicted data is then compared with the warning threshold. At this time, when the predicted data is greater than or equal to the warning threshold, the corresponding warning response strategy in the multi-level response strategy is triggered. When the predicted data is less than the warning threshold, the corresponding normal response strategy in the multi-level response strategy is triggered, and the system operates normally.

[0048] The trend prediction is achieved using sliding window time series analysis, and the state of SiC power devices is predicted through the following autoregressive moving average model: ; In the above formula, These are the detection parameters for the time series at time t; These are the coefficients of the autoregression; These are the historical observations of the time series at time tm; The coefficients of the moving average term; This is the white noise error term; q represents the detection parameters of the white noise error term at time tn; p and q are the model orders, determined by minimizing the Akaike Information Criterion (AIC).

[0049] Sliding window time series analysis is a method for predicting future values ​​of time series using an autoregressive moving average (ARMA) model. This method captures local features of the time series through a sliding window mechanism and uses autoregressive coefficients and moving average coefficients to build a predictive model. White noise error terms reflect the portion of variation that the model cannot explain; adjusting the model order (p, q) can optimize the model's predictive performance. This analysis method can capture the changing trends of SiC power device states in real time, providing strong support for timely early warning and intervention, and helping to prevent the occurrence and development of faults. The application of the autoregressive moving average model improves the accuracy of state prediction, helps to detect potential faults early and take measures to avoid their occurrence, and improves the reliability and safety of the system.

[0050] The multi-level response strategy employs a fuzzy logic control method, and the membership function used in the control process is as follows: ; In the formula, The membership degree represents the severity of the failure, with a range of values. x represents the predicted data; a represents the warning threshold; and b represents the danger threshold.

[0051] Membership functions are used to define a fuzzy set of failure severity levels. The severity of SiC power device failures is determined by comparing predicted data with warning and danger thresholds. Membership functions typically employ trapezoidal or triangular distributions to reflect the impact of predicted data on failure severity across different intervals. The application of membership functions enables the system to more accurately assess failure severity and make reasonable decisions, thus contributing to the optimization of system performance and reliability. The fuzzy logic control method used in this embodiment can take corresponding response measures according to different levels of failure severity, achieving refined and intelligent fault handling and improving the system's response speed and accuracy.

[0052] Based on the same inventive concept as the above-mentioned SiC power device package reliability testing method, this application also provides a SiC power device package reliability testing system, such as... Figure 2 As shown, it includes: The model building module is used to build multiphysics coupled finite element models of SiC power devices; The initial simulation and parameter analysis module, connected to the model building module, is used to run a multiphysics coupled finite element model to obtain initial simulation data, and to perform parameter sensitivity analysis on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The measured data acquisition module includes a working parameter acquisition unit, a data calibration and denoising unit, a dynamic region tracking unit, and a temperature gradient map generation unit; The operating parameter acquisition unit is used to acquire the operating parameters of the SiC power device in operation. The data calibration and denoising unit is connected to the working parameter acquisition unit and is used to perform real-time calibration and denoising of the working parameters to obtain measured data. The dynamic region tracking unit is connected to the working parameter acquisition unit and is used to locate the key regions of the SiC power device through dynamic region tracking technology. The temperature gradient map generation unit is connected to the data calibration and denoising unit and the dynamic region tracking unit, respectively, and is used to generate a temperature gradient map of the key region of the SiC power device based on the measured data. The comparative analysis and iterative optimization module is connected to the initial simulation and parameter analysis module and the measured data acquisition module, respectively. It is used to perform comparative analysis on the simulation data and the measured data through a deep reinforcement learning algorithm. Based on the comparative analysis results, iterative optimization is performed on the key stress parameter combination and the key region parameters of the multiphysics coupled finite element model until the error between the initial simulation data and the measured data is within a set range, thereby obtaining the optimized key stress parameter combination and the key region parameters of the multiphysics coupled finite element model. The final simulation and failure criterion setting module, together with the comparative analysis and iterative optimization module, is used to perform simulations based on the optimized combination of key stress parameters and the key region parameters of the multi-physics coupled finite element model to obtain the final simulation data, and to set the failure criterion threshold based on the final simulation data. The trend prediction and response execution module includes a trend prediction unit and a comparative response unit; The trend prediction unit is connected to the temperature gradient map generation unit and is used to input the temperature gradient map of the key area into the trend prediction model to perform trend prediction and obtain prediction data. The comparison response unit is connected to the final simulation and failure criterion setting module and the trend prediction unit, respectively, and is used to compare the prediction data with the failure criterion threshold, and execute a multi-level response strategy based on the comparison result.

[0053] In the embodiments provided above in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functions, and operations that the systems and methods according to embodiments of this application may implement. In this regard, each block in a flowchart or block diagram may represent a module, segment, or part of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0054] In addition, the functional modules in the embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0055] If a function is implemented as a software module 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 this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0056] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0057] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for testing the reliability of SiC power device packaging, characterized in that, Includes the following steps: Constructing a multiphysics coupled finite element model for SiC power devices; Initial simulation data is obtained by running a multiphysics coupled finite element model, and parameter sensitivity analysis is performed on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The operating parameters of the SiC power device under operating conditions are collected, and the operating parameters are calibrated in real time and denoised to obtain measured data. During the process of collecting the operating parameters of the SiC power device under operating conditions, the key areas of the SiC power device are located by dynamic region tracking technology, and the temperature gradient map of the key areas of the SiC power device is generated based on the measured data. The simulation data and the measured data are compared and analyzed by deep reinforcement learning algorithm. Based on the comparison and analysis results, the key stress parameter combination and the key area parameters of the multiphysics coupled finite element model are iteratively optimized until the error between the initial simulation data and the measured data is within a set range, thereby obtaining the optimized key stress parameter combination and the key area parameters of the multiphysics coupled finite element model. The final simulation data is obtained by performing simulations based on the optimized combination of key stress parameters and the key region parameters of the multiphysics coupled finite element model. The failure criterion threshold is then set based on the final simulation data. The temperature gradient map of the key area is input into the trend prediction model to predict the trend and obtain the prediction data. The prediction data is compared with the failure criterion threshold, and a multi-level response strategy is executed based on the comparison result.

2. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The process of constructing a multiphysics coupled finite element model of SiC power devices includes: constructing or importing the geometric model of SiC power devices, setting the parameters of SiC power devices, setting the boundary conditions, loads and typical failure modes of the model, and performing mesh generation and solution settings to obtain a multiphysics coupled finite element model.

3. The SiC power device packaging reliability testing method as described in claim 2, characterized in that: The parameters for setting SiC power devices include setting the thermal conductivity, coefficient of thermal expansion, elastic modulus, and Poisson's ratio of the encapsulation material at different temperatures; the thermal conductivity of the encapsulation material at different temperatures is obtained by laser flare method.

4. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The typical failure modes include solder layer cracking, thermal resistance degradation, and bond wire detachment. The failure of solder layer cracking is determined by obtaining the strain concentration area in the thermo-mechanical stress simulation. The failure of thermal resistance degradation is determined by obtaining the simulated thermal resistance parameters. The failure of bond wire detachment is determined by obtaining the stress analysis of the bonding area in the electro-thermal coupling simulation.

5. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The process of performing parameter sensitivity analysis on the initial simulation data to determine the key stress parameter combinations for typical failure modes is as follows: a stress parameter sampling matrix is ​​designed using the Morris global sensitivity analysis method. Perform finite element simulations for each set of stress parameters and output failure indices; Calculate the normalized sensitivity coefficients of each stress parameter to failure time, and determine the key stress parameter combinations for typical failure modes.

6. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The real-time calibration process involves using an environmental radiation compensation algorithm to calibrate the temperature parameter in the operating parameters in real time.

7. The SiC power device packaging reliability testing method as described in claim 6, characterized in that: The process of using the environmental radiation compensation algorithm to calibrate the temperature parameter in the working parameters in real time is as follows: during the acquisition of the temperature parameters of the SiC power device, the radiation intensity at different wavelengths is measured, and then the measured temperature is calculated and calibrated according to the blackbody radiation law.

8. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The denoising process is as follows: the working parameters and the simulation data are time-aligned, and then the wavelet threshold denoising algorithm is used to eliminate environmental noise in the working parameters.

9. The SiC power device packaging reliability testing method as described in claim 1, characterized in that: The failure criterion thresholds include a warning threshold and a danger threshold. The multi-level response strategy employs a fuzzy logic control method, and the membership function used in the fuzzy logic control process is: ； In the formula, The membership degree represents the severity of the failure, with a range of values. x represents the predicted data; a represents the warning threshold; and b represents the danger threshold.

10. A reliability testing system for SiC power device packaging, characterized in that, A method for performing the SiC power device package reliability testing method as described in any one of claims 1-9, comprising: The model building module is used to build multiphysics coupled finite element models of SiC power devices; The initial simulation and parameter analysis module is used to run a multiphysics coupled finite element model to obtain initial simulation data, and to perform parameter sensitivity analysis on the initial simulation data to determine the key stress parameter combinations for typical failure modes. The measured data acquisition module is used to collect the operating parameters of the SiC power device under the operating state, and obtain the measured data after real-time calibration and noise reduction of the operating parameters; the key regions of the SiC power device are located by dynamic region tracking technology and a temperature gradient map of the key regions is generated. The comparative analysis and iterative optimization module is used to compare and analyze the simulation data and the measured data using a deep reinforcement learning algorithm. Based on the comparative analysis results, iterative optimization is performed on the key stress parameter combination and the key region parameters of the multiphysics coupled finite element model to obtain the optimized key stress parameter combination and the key region parameters of the multiphysics coupled finite element model. The final simulation and failure criterion setting module is used to simulate and obtain the final simulation data based on the optimized combination of key stress parameters and the key region parameters of the multiphysics coupled finite element model, and to set the failure criterion threshold based on the final simulation data. The trend prediction and response execution module is used to input the temperature gradient map of the key area into the trend prediction model to predict the trend and obtain the prediction data, compare the prediction data with the failure criterion threshold, and execute a multi-level response strategy based on the comparison result.

Images