A novel high-temperature-resistant silicon carbide device packaging structure optimization method and system

Abstract

The application relates to the technical field of silicon carbide device packaging structure optimization, and provides a novel high-temperature-resistant silicon carbide device packaging structure optimization method and system, which comprises the following steps: obtaining sensitive area information according to a physical field sensitivity threshold value; performing calculation domain grid encryption on the sensitive area, keeping the calculation domain grid sparse on the non-sensitive area, and constructing a discretization packaging structure model; obtaining batch difference and microscopic structure unevenness of packaging materials of the packaging structure; generating a preset number of material parameter combinations representing internal variability of the materials according to the batch difference and the microscopic structure unevenness; performing association operation on the material parameter combinations and the discretization packaging structure model to obtain core optimization parameter information, which is used as input of a packaging structure optimization process; and optimizing the packaging structure based on the core optimization parameter information. The application has the effect of improving the stability of the silicon carbide device under extreme working conditions.

Description

Technical Field

[0001] This invention relates to the technical field of silicon carbide device packaging structure optimization, specifically to a novel method and system for optimizing the packaging structure of high-temperature resistant silicon carbide devices. Background Technology

[0002] In the field of power electronics, silicon carbide power devices are gradually replacing traditional silicon-based devices due to their superior high-temperature resistance, high-frequency operation, and high power density. As power electronics technology advances towards higher integration and greater power density, the market is placing increasingly stringent demands on silicon carbide modules. Particularly in new high-voltage direct current (HVDC) transmission systems, silicon carbide devices are required to operate continuously at extremely high switching frequencies. This means that the internal heat load of the device is no longer a simple, uniform distribution, but rather exhibits rapidly changing, localized hot spots. These hot spots are not fixed; their location and intensity dynamically adjust with the device's operating state, such as changes in instantaneous switching current, voltage, or load conditions. This dynamic hot spot phenomenon is fundamentally different from the uniform temperature rise under traditional steady-state heat loads.

[0003] However, traditional gradient descent optimization algorithms encounter significant limitations in applications when faced with such dynamic and nonlinear local hotspot distributions. Due to their inherent local search characteristics and dependence on continuous functions, these algorithms struggle to effectively capture stress concentration regions in encapsulated structures under different transient temperature gradients. They often only find the "optimal solution" under specific steady-state conditions, failing to adequately assess potential failure risks during transient processes. For example, an encapsulated structure that performs well in steady state may experience a rapid formation and dissipation of local hotspots during a high-frequency switching transient, causing a critical interface to withstand a huge temperature gradient in a very short time, leading to instantaneous thermal stress concentration. Traditional optimization algorithms are not sensitive to such transient stress peaks and cannot use them as primary optimization targets.

[0004] Ultimately, in actual operation, although devices designed using traditional optimization methods perform reasonably well in terms of static thermal resistance, accelerated fatigue occurs in specific areas within the package due to the combined effects of thermal cycling caused by high-frequency switching and electromagnetic vibration. This accelerated fatigue leads to intermittent electrical failures in the devices within a timeframe far shorter than the designed expected lifespan, severely impacting the long-term stable operation and maintenance costs of the entire power electronic system.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This application discloses a novel method and system for optimizing the packaging structure of high-temperature silicon carbide devices. It aims to solve the problems in the prior art where traditional optimization algorithms are unable to effectively capture the stress concentration areas of the packaging structure under different transient temperature gradients, and the problem of insufficient assessment of potential failure risks that may occur during the transient process. At the same time, it solves the problem of accelerated fatigue in specific areas inside the package under the combined action of thermal cycling and electromagnetic force vibration caused by high-frequency switching.

[0007] The technical solution of this application is as follows: In a first aspect, this application discloses a novel method for optimizing the packaging structure of high-temperature resistant silicon carbide devices, including: For silicon carbide devices, based on a preset physical field sensitivity threshold, regions in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress are identified, and information on sensitive regions is obtained. Based on information about sensitive areas, the computational domain mesh is densified in sensitive areas and kept sparse in non-sensitive areas, thus constructing a discretized encapsulation structure model. To obtain batch variations and microstructural inhomogeneities of the packaging materials used in the packaging structure; Based on batch differences and microstructural inhomogeneity, a preset number of material parameter combinations representing the inherent variability of the material are generated; By associating the material parameter combination with the discretized packaging structure model, core optimization parameter information is obtained and used as input for the packaging structure optimization process. Based on the core optimization parameter information, the packaging structure is optimized.

[0008] This technical solution can effectively identify sensitive areas in the packaging structure and generate core optimization parameters by combining the inherent variability of the material, thereby achieving fine optimization of the silicon carbide device packaging structure and solving the problem that traditional optimization methods are insensitive to transient stress peaks.

[0009] Furthermore, this application proposes that, based on sensitive area information, the steps of refining the computational domain grid in sensitive areas and maintaining a sparse computational domain grid in non-sensitive areas include: To obtain information on the actual microscopic geometric non-uniformity characteristics in the packaging material; Establish a rule base that associates micro-defect types with potential failure mechanisms; Based on the microscopic geometric non-uniformity characteristic information and the association rule library, the transient working condition combination that maximizes the intensity of the local key physical field around the microscopic defect is searched in the preset working condition parameter space, and an active working condition list is generated. Select the transient condition related to the current package structure to be optimized from the list of active conditions as the pre-simulation input; Before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity characteristics, all identified micro-defect regions are pre-marked in the geometric model of the encapsulated structure and assigned preset mesh refinement weights. Based on the pre-labeled geometric model of the encapsulation structure, the preset mesh refinement weights, and the pre-simulation input, a pre-simulation is performed to identify the sensitive areas for pre-simulation. By combining the pre-simulated sensitive areas, the pre-marked micro-defect areas, and the selected transient conditions, the mesh refinement sensitivity threshold of the corresponding areas is adjusted. Mesh refinement is performed on regions where the mesh refinement sensitivity threshold is adjusted or on pre-marked micro-defect regions, while the computational domain mesh remains sparse in non-sensitive regions.

[0010] This technical solution enables more accurate identification and processing of microscopic defect areas in the packaging structure. By dynamically adjusting the mesh refinement sensitivity threshold, it ensures fine mesh densification in critical areas, thereby improving simulation accuracy and optimization results.

[0011] Based on the above, this application further proposes a step that, before pre-simulating and generating an initial coarse mesh, pre-marks all identified micro-defect regions in the geometric model of the encapsulated structure according to the micro-geometric non-uniformity characteristic information and assigns preset mesh refinement weights, including: During the use of silicon carbide devices, the geometric feature information of newly formed micro-defects in the corresponding packaging material is periodically acquired, and the micro-defect database is updated. Based on the current service time, the characteristics of the identified micro-defects, and the historical operating data of silicon carbide devices, the morphological evolution trend of existing micro-defects and the location of new micro-defects are analyzed. Based on the morphological evolution trend and the location of occurrence, the boundaries and mesh refinement weights of the marked micro-defect regions in the geometric model of the encapsulation structure are adjusted, and newly emerging micro-defect regions are marked and weighted. Before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity characteristics, the adjusted boundaries of the marked micro-defect regions, the adjusted mesh refinement weights, and the marking and weight assignment of the newly formed micro-defect regions, all identified micro-defect regions are pre-marked in the geometric model and assigned preset mesh refinement weights.

[0012] This technical solution enables dynamic monitoring and updating of micro-defects, and combines service data to analyze defect evolution trends, thereby more accurately marking and weighting defect areas before pre-simulation, further improving the adaptability and accuracy of the optimization method.

[0013] Furthermore, this application proposes a process for analyzing the morphological evolution trend of existing microdefects and the location of newly formed microdefects based on the current service time, the characteristics of identified microdefects, and historical operating data of silicon carbide devices. This process includes: During the operation of silicon carbide devices, current, voltage, and temperature are monitored in real time to obtain transient electromagnetic forces, transient thermal loads, and local electric field strength. Based on transient electromagnetic force, transient thermal load, and local electric field intensity, calculate the instantaneous gradient and instantaneous intensity of each physical field in the microscopic defect region; Based on instantaneous gradient and instantaneous intensity, combined with a preset threshold for the synergistic effect of physical fields, the region where there is a nonlinear synergistic effect between transient electromagnetic force-induced micro-vibration, transient thermal load, and local electric field intensity is identified, and the relevant analysis region is obtained; Based on the relevant analysis region, and combined with the existing geometric characteristics of micro-defects and material degradation mechanisms, the analysis parameters corresponding to the evolution rate and direction of the existing micro-defects are adjusted. Based on the adjusted analysis parameters, the morphological evolution trend of existing micro-defects and the location of newly emerging micro-defects were analyzed.

[0014] This technical solution enables a deeper understanding of the evolution mechanism of microscopic defects through real-time monitoring and multi-physics synergistic analysis, thereby more accurately predicting the morphological evolution trend of defects and the location of newly formed defects, providing a more reliable basis for packaging structure optimization.

[0015] In some preferred embodiments, this application also proposes a step of associating material parameter combinations with a discretized packaging structure model to obtain core optimization parameter information, which is then used as input to the packaging structure optimization process. The mesh cells of the discretized packaging structure model are partitioned according to the material layer and interface type, and sensitive and non-sensitive areas are identified. The material parameter combinations are converted to obtain an effective parameter set that matches each material layer. The effective parameter set is projected according to material layer and interface type and bound to the corresponding mesh partition to form model parameter instances; Apply the boundary conditions of the target application to the model parameter instances and solve them to obtain the corresponding key indicators; Key indicators are normalized and weighted to generate core optimization parameter information, which is then used as input for the encapsulation structure optimization process.

[0016] This technical solution can systematically associate material parameters with discrete packaging structure models, and generate more representative core optimization parameters through multi-level parameter processing and aggregation of key indicators, thereby improving the efficiency and accuracy of the optimization process.

[0017] As a technical improvement, this application also proposes a step of adjusting the analytical parameters corresponding to the evolution rate and direction of existing micro-defects based on the relevant analytical region and in combination with the existing geometric characteristics of micro-defects and material degradation mechanisms. For micro-defect regions, the propagation speed and attenuation factor of the elastic wave generated in the encapsulation material are calculated based on transient electromagnetic force, transient thermal load and local electric field intensity. Based on the propagation speed and attenuation factor of elastic waves, the energy concentration effect of elastic waves in the defect edge region is quantified. Calculate the local temperature gradient and electric field intensity in the micro-defect region based on the transient thermal load and local electric field intensity; Based on the local temperature gradient and electric field intensity of the micro-defect region, combined with the plastic flow threshold of the encapsulation material, the rate and direction of plastic deformation accumulation are quantified. Based on the local electric field strength and transient thermal load, combined with the chemical bond dissociation energy and activation energy of the encapsulation material, the rate and location of chemical bond breakage are quantified. Based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breaking, a multi-damage mode coupling factor is constructed. By associating the multi-damage mode coupling factor with the existing geometric characteristics of microdefects and material degradation mechanisms, the analytical parameters corresponding to the evolution rate and direction of existing microdefects are corrected.

[0018] This technical solution enables a more comprehensive consideration of the effects of elastic waves, plastic deformation, and chemical bond breaking on defect evolution by constructing a multi-damage mode coupling factor. This allows for more precise correction of the analytical parameters for defect evolution rate and direction, thereby improving the accuracy of prediction.

[0019] To enhance functionality, this application also proposes a method for constructing a multi-damage mode coupling factor based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breakage. The method includes the following steps: For the three damage modes of elastic wave propagation, plastic deformation accumulation, and chemical bond breakage, their respective time windows and spatial ranges are set; Within the time window of each damage mode, the quantification results of each damage mode are integrated over time to obtain the cumulative damage amount of each damage mode within the corresponding time window. Within the spatial range of each damage mode, the quantification results of each damage mode are spatially averaged to obtain the average damage intensity of each damage mode within the corresponding spatial range. A multi-damage mode coupling factor is constructed based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor.

[0020] This technical solution enables more precise quantification and coupling of different damage modes by setting time windows and spatial ranges and introducing asynchronous correction factors, thereby constructing a more accurate multi-damage mode coupling factor.

[0021] As a further improvement, this application also proposes a step for constructing a multi-damage mode coupling factor based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor, including: During the operation of silicon carbide devices, identify the occurrence of extreme transient overload or short circuit conditions; When extreme transient overload or short-circuit conditions are identified, the high-frequency data acquisition mode is immediately activated to obtain the instantaneous quantization results of each damage mode with a preset higher sampling frequency and a preset shorter time window. Based on the instantaneous quantization results obtained under the high-frequency data acquisition mode, the cumulative damage amount and average damage intensity of each damage mode are recalculated. Based on the recalculated cumulative damage and average damage intensity, as well as the current working condition type, a matching asynchronous correction factor is selected from the preset extreme working condition correction factor library. Based on the selected asynchronous correction factor, a multi-damage mode coupling factor is constructed.

[0022] This technical solution enables high-frequency data acquisition and dynamic correction under extreme conditions, thereby more accurately constructing multi-damage mode coupling factors in special circumstances and improving the robustness of optimization methods.

[0023] To improve the solution, this application also proposes a step of selecting a matching asynchronous correction factor from a pre-set extreme condition correction factor library based on the recalculated cumulative damage amount and average damage intensity, as well as the current working condition type. Extract the asynchronous correction factor corresponding to the current operating condition type from the preset extreme operating condition correction factor library; A consistency check is performed on the extracted asynchronous correction factors to identify and eliminate redundant or conflicting factors in the extreme condition correction factor library. Based on the results of the consistency check, the remaining asynchronous correction factors in the extreme working condition correction factor library are prioritized and selected first, with the asynchronous correction factor having the highest matching degree with the recalculated cumulative damage amount and average damage intensity. Based on the priority ranking results, the asynchrony correction factor with the highest matching degree is selected.

[0024] This technical solution enables the selection of the most suitable asynchronous correction factor from the extreme operating condition correction factor library through consistency checks and priority ranking, thereby further improving the accuracy and reliability of the optimization method.

[0025] Secondly, this application also discloses a novel high-temperature resistant silicon carbide device packaging structure optimization system, used to perform novel high-temperature resistant silicon carbide device packaging structure optimization, including: The sensitive area identification module is used to identify areas in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress, based on a preset physical field sensitivity threshold, and to obtain sensitive area information. The encapsulation model construction module is used to densify the computational domain mesh in sensitive areas based on sensitive area information, and keep the computational domain mesh sparse in non-sensitive areas, thereby constructing a discretized encapsulation structure model. The difference parameter acquisition module is used to acquire batch differences and microstructure non-uniformity of the packaging material in the packaging structure. The material parameter generation module is used to generate a preset number of material parameter combinations that represent the inherent variability of the material based on batch differences and microstructure inhomogeneity. The optimization parameter generation module is used to associate the combination of material parameters with the discretized packaging structure model to obtain core optimization parameter information, which serves as the input for the packaging structure optimization process. The encapsulation structure optimization module is used to optimize the encapsulation structure based on core optimization parameter information.

[0026] This technical solution provides an integrated system for the automated optimization of silicon carbide device packaging structures. Through modular design, it improves the efficiency and operability of the optimization process.

[0027] Beneficial Effects: This application discloses a novel method for optimizing the packaging structure of high-temperature resistant silicon carbide devices. By identifying regions within the packaging structure sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress based on a preset physical field sensitivity threshold, it effectively captures transient stress concentration regions that are difficult to identify using traditional gradient descent optimization algorithms. Based on this, the computational domain mesh is refined using sensitive region information, and material parameter combinations representing the inherent variability of the material are generated by combining batch differences and microstructural inhomogeneities of the packaging material. This allows for a more comprehensive consideration of the packaging structure's response under complex operating conditions. The material parameter combinations are correlated with a discretized packaging structure model to obtain core optimization parameter information, which serves as input to the packaging structure optimization process, enabling optimization of the packaging structure based on this core optimization parameter information. This method overcomes the limitations of existing traditional optimization algorithms, such as insensitivity to transient stress peaks and insufficient assessment of potential failure risks. It effectively solves the problem of accelerated fatigue in specific regions inside the package under the combined effects of thermal cycling and electromagnetic vibration caused by high-frequency switching, significantly improving the long-term stable operation and reliability of silicon carbide devices under extreme conditions. Attached Figure Description

[0028] Figure 1 This is a flowchart of a novel high-temperature resistant silicon carbide device packaging structure optimization method according to one embodiment of the present invention; Figure 2 This is a flowchart of a novel high-temperature resistant silicon carbide device packaging structure optimization method according to another embodiment of the present invention; Figure 3 This is a system block diagram of a novel high-temperature resistant silicon carbide device packaging structure optimization system according to another embodiment of the present invention; Explanation of reference numerals in the attached figures: 1. Novel high-temperature silicon carbide device packaging structure optimization system; 11. Sensitive area identification module; 12. Packaging model construction module; 13. Difference parameter acquisition module; 14. Material parameter generation module; 15. Optimization parameter generation module; 16. Packaging structure optimization module. Detailed Implementation

[0029] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0030] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0031] Traditional optimization methods for silicon carbide power device packaging structures have significant limitations when facing dynamic, nonlinear local hot spots and thermo-mechanical-electrical composite stress environments caused by high-frequency switching. Traditional gradient descent optimization algorithms struggle to effectively capture stress concentration regions in the packaging structure under different transient temperature gradients and fail to adequately assess potential failure risks during transient processes. This leads to accelerated fatigue of the devices during actual operation, severely impacting the long-term stable operation of the system.

[0032] In response, this application proposes a novel method for optimizing the packaging structure of high-temperature resistant silicon carbide devices, combining... Figure 1 As shown, it includes: S1, For silicon carbide devices, based on a preset physical field sensitivity threshold, identify regions in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field strength, and mechanical stress, and obtain sensitive region information. S2, based on sensitive area information, densifies the computational domain grid in sensitive areas and keeps the computational domain grid sparse in non-sensitive areas, thus constructing a discretized encapsulation structure model; S3, obtain the batch variation and microstructure non-uniformity of the packaging material of the packaging structure; S4, based on batch differences and microstructural inhomogeneity, generates a preset number of material parameter combinations representing the inherent variability of the material; S5. The material parameter combination is correlated with the discretized packaging structure model to obtain the core optimization parameter information, which is used as the input for the packaging structure optimization process. S6 optimizes the packaging structure based on core optimization parameter information.

[0033] Silicon carbide (SiC) devices refer to power semiconductor devices manufactured based on SiC materials. They are widely used in power electronics due to their excellent high-temperature resistance, high-frequency operation, and high power density. The packaging structure is a crucial component of SiC devices, protecting the chip and providing electrical connections and heat dissipation channels. The physical field sensitivity threshold is a preset critical value used to determine the sensitivity of different regions within the packaging structure to changes in heat flow, electromagnetic field strength, and mechanical stress. Sensitive region information refers to the geometric location, dimensions, and other data of regions sensitive to these physical field changes, obtained through identification. Computational domain mesh refinement refers to the refinement of the mesh in specific regions during numerical simulation to improve simulation accuracy. Discretized packaging structure models transform continuous packaging structure geometric models into discrete sets of mesh cells for numerical computation. Batch variation refers to the potential performance differences between different production batches of the same packaging material. Microstructural inhomogeneity refers to the degree of inhomogeneity in the internal microstructure of the packaging material (such as grain size and defect distribution). The material parameter combination is a set of parameters generated based on batch variation and microstructural inhomogeneity, used to characterize the inherent variability of the material. The core optimization parameter information is obtained after correlation operations and is key data used to guide the packaging structure optimization process.

[0034] The optimization method of this application first targets silicon carbide devices. Based on preset physical field sensitivity thresholds, it identifies regions in the silicon carbide device's packaging structure that are sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress, thereby obtaining sensitive region information. For example, preliminary physical field analysis of the packaging structure can be performed under typical transient operating conditions using finite element analysis (FEA) or computational fluid dynamics (CFD) simulation. During the analysis, a temperature gradient threshold, an electromagnetic field intensity change rate threshold, and a mechanical strain rate threshold can be set. When a region in the packaging structure exceeds these preset thresholds during the simulation, that region is identified as a sensitive region. Another approach is to perform non-contact monitoring of the silicon carbide device in actual operation using infrared thermal imagers, X-ray diffractometers, or ultrasonic testing equipment. By acquiring real-time data on the temperature distribution, internal stress field, or defect distribution on the device surface and comparing it with preset sensitivity thresholds, regions sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress can be identified.

[0035] Based on information about sensitive regions, the computational domain mesh is refined in sensitive regions and kept sparse in non-sensitive regions, thus constructing a discretized encapsulated structural model. For example, when constructing the finite element model, a smaller mesh size can be set at the geometric boundaries of sensitive regions identified in the previous step, while a larger mesh size is used in non-sensitive regions. This can be achieved by manually specifying the region mesh density in the mesh generation software or by automating mesh generation through script programming. Another approach is to utilize adaptive mesh refinement technology. After the initial simulation is completed, the mesh in sensitive regions is automatically refined based on the physical field gradient or error estimate until the preset accuracy requirements are met.

[0036] Next, the batch variation and microstructural inhomogeneity of the encapsulation materials are obtained. For example, batch variation data can be obtained by testing the material properties of different batches of encapsulation materials, such as tensile strength, thermal conductivity, and dielectric constant. Simultaneously, microscopic analysis techniques such as scanning electron microscopy (SEM) or X-ray computed tomography (X-ray CT) are used to characterize the microstructure of the encapsulation materials, obtaining information on microstructural inhomogeneity such as grain size, porosity, and defect distribution. Another approach is to obtain batch variation information through material quality control reports and production batch data provided by the supplier. For microstructural inhomogeneity, statistical methods can be used to perform multi-point measurements on material samples and calculate their mean and standard deviation to quantify the degree of inhomogeneity.

[0037] Based on batch variation and microstructural inhomogeneity, a predetermined number of material parameter combinations representing the inherent variability of the material are generated. For example, Monte Carlo methods or Latin hypercube sampling can be used to randomly select a predetermined number of material parameter combinations within the statistical distribution range of the material parameters, based on the acquired batch variation and microstructural inhomogeneity data. Another approach is to utilize response surface methodology or a surrogate model to construct a mathematical model based on the relationship between batch variation, microstructural inhomogeneity, and material properties. Then, this model is used to generate a series of material parameter combinations representing the inherent variability of the material.

[0038] By associating material parameter combinations with a discretized packaging structure model, core optimization parameter information is obtained and used as input for the packaging structure optimization process. For example, each generated set of material parameter combinations can be substituted into the discretized packaging structure model for a series of numerical simulations (such as thermo-mechanical-electrical coupling simulations). Each simulation yields a set of key performance indicators, such as maximum temperature, maximum stress, and electric field strength. These key performance indicators are then processed, for example, by calculating their average, maximum, standard deviation, or sensitivity, to obtain the core optimization parameter information. Another approach is to use machine learning algorithms, taking the material parameter combinations and the discretized packaging structure model as input, to train a predictive model. This model can predict the key performance indicators of the packaging structure under different material parameter combinations. Then, the core optimization parameter information is generated through this model.

[0039] Finally, the encapsulation structure is optimized based on the core optimization parameters. For example, a multi-objective optimization algorithm, such as NSGA-II (Non-dominated sorting genetic algorithm II), can be used to optimize the design variables of the encapsulation structure, such as its geometry, material selection, and layout, by using the core optimization parameters as the optimization objective or constraint. Another approach is to use topology optimization methods, using the core optimization parameters as the optimization criterion, to automatically generate the optimal encapsulation structure topology within a given design space.

[0040] Optional, combined Figure 2 As shown, the steps in step S2, which involve refining the computational domain mesh in sensitive areas based on sensitive area information and maintaining a sparse computational domain mesh in non-sensitive areas, include: S21, Obtain information on the actual microscopic geometric non-uniformity characteristics in the packaging material; S22, Establish a rule base for the association between micro-defect types and potential failure mechanisms; S23. Based on the microscopic geometric non-uniformity characteristic information and the association rule library, search for transient working condition combinations that maximize the intensity of local key physical fields around the microscopic defects in the preset working condition parameter space, and generate an active working condition list. S24, Select the transient condition related to the current package structure to be optimized from the list of active conditions as the pre-simulation input; S25, before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity characteristic information, all identified micro-defect regions are pre-marked in the geometric model of the encapsulation structure and assigned preset mesh refinement weights. S26. Based on the pre-marked geometric model of the encapsulation structure, the preset mesh refinement weights, and the pre-simulation input, pre-simulation is performed to identify the sensitive areas of the pre-simulation. S27, combining the pre-simulated sensitive area, the pre-marked micro-defect area and the selected transient condition, adjust the mesh refinement sensitivity threshold of the corresponding area; S28 refines the mesh in areas where the mesh refinement sensitivity threshold is adjusted or in pre-marked micro-defect areas, while keeping the computational domain mesh sparse in non-sensitive areas.

[0041] Specifically, "obtaining information on the actual microscopic geometric non-uniformity characteristics in the encapsulation material" refers to identifying and quantifying various microscopic defects, such as voids, cracks, inclusions, and delamination, existing within the encapsulation material through non-destructive testing techniques (e.g., X-ray computed tomography, ultrasonic testing) or microscopic analysis methods (e.g., scanning electron microscopy, metallographic analysis), and acquiring their geometric characteristic data, such as size, shape, location, and distribution. This information is fundamental to understanding the local behavior of the material and potential failure modes.

[0042] "Establishing a rule base for the association between microscopic defect types and potential failure mechanisms" refers to building a knowledge base that maps different types of microscopic defects (e.g., solder voids, chip adhesive layer cracks, wire bond fatigue) to their potential failure mechanisms (e.g., thermal fatigue, electromigration, mechanical stress concentration, dielectric breakdown). This rule base can be established and updated based on experimental data, physical failure models, finite element analysis results, or expert experience, with the aim of providing a theoretical basis for subsequent defect analysis and mesh refinement.

[0043] "Searching for transient operating condition combinations that maximize the local critical physical field strength around micro-defects within a preset operating condition parameter space, and generating an active operating condition list" refers to systematically exploring various transient operating conditions (e.g., rapid switching, short circuit, overload, temperature shock) under typical or extreme operating conditions of silicon carbide devices through numerical simulation or experimental testing to identify those operating condition combinations that can cause the local temperature gradient, electric field strength, or mechanical stress around micro-defects to reach their maximum values. These operating condition combinations are considered "active operating conditions" most likely to trigger defect propagation or device failure and are compiled into a list.

[0044] "Selecting transient conditions related to the current package structure to be optimized from the list of active conditions as pre-simulation inputs" means selecting the most representative or most stringent transient conditions from the list of active conditions generated above, based on the design characteristics of the current package structure, the expected application scenario, and the optimization objectives. These selected transient conditions will be used as boundary conditions and excitations for subsequent pre-simulation analyses to simulate the response of the package structure under critical operating conditions.

[0045] "Before pre-simulation and initial coarse mesh generation, based on the microscopic geometric non-uniformity characteristics, all identified microscopic defect regions are pre-marked in the geometric model of the package structure, and assigned a preset mesh refinement weight." This means that before any numerical simulation, the geometric location and morphology of the detected microscopic defects are accurately mapped onto the digital geometric model of the package structure. Simultaneously, a preset mesh refinement weight is assigned to these marked defect regions based on the type, size, location of the defects, and their potential impact on device performance and reliability. This weight guides the subsequent mesh generation process, ensuring that these critical regions achieve sufficient refinement during the initial mesh generation.

[0046] "Pre-simulation based on a pre-labeled geometric model of the encapsulation structure, preset mesh refinement weights, and pre-simulation input to identify pre-simulation sensitive areas" refers to using a geometric model with pre-labeled defects and refinement weights, combined with selected transient conditions as input, to run a preliminary numerical simulation, typically with a coarse mesh. Through this pre-simulation, it is possible to initially identify, under these transient conditions, regions within the encapsulation structure where, in addition to known microscopic defect regions, other physical fields (such as thermal stress and electromagnetic stress) are concentrated or have large gradients. These regions are referred to as pre-simulation sensitive areas.

[0047] "Adjusting the mesh refinement sensitivity threshold for corresponding regions by combining pre-simulation sensitive regions, pre-marked micro-defect regions, and selected transient conditions" refers to dynamically adjusting the mesh refinement sensitivity threshold by comprehensively considering the influence of macroscopic sensitive regions identified in the pre-simulation, pre-marked micro-defect regions, and the currently selected transient conditions on these regions. This means that for regions exhibiting high stress, high temperature gradients, or high electric field strength in the pre-simulation, as well as all pre-marked micro-defect regions, their mesh refinement thresholds will be reduced to ensure that these regions are finely divided in the final mesh generation.

[0048] "Refine the mesh in areas with adjusted mesh refinement sensitivity thresholds or pre-marked micro-defect areas, while maintaining a sparse mesh in non-sensitive areas" means that, based on the adjusted mesh refinement sensitivity thresholds, during the final mesh generation stage, all identified sensitive areas (including pre-simulation sensitive areas and pre-marked micro-defect areas) are divided into high-density meshes to accurately capture the details of the local physical field. Simultaneously, for non-sensitive areas that are less critical to the analysis, a relatively sparse mesh is used to effectively control computational resource consumption while ensuring computational accuracy.

[0049] Optionally, before pre-simulating and generating the initial coarse mesh, the steps of pre-marking all identified micro-defect regions in the geometric model of the encapsulation structure based on the micro-geometric non-uniformity characteristics and assigning preset mesh refinement weights include: During the use of silicon carbide devices, the geometric feature information of newly formed micro-defects in the corresponding packaging material is periodically acquired, and the micro-defect database is updated. Based on the current service time, the characteristics of the identified micro-defects, and the historical operating data of silicon carbide devices, the morphological evolution trend of existing micro-defects and the location of new micro-defects are analyzed. Based on the morphological evolution trend and the location of occurrence, the boundaries and mesh refinement weights of the marked micro-defect regions in the geometric model of the encapsulation structure are adjusted, and newly emerging micro-defect regions are marked and weighted. Before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity characteristics, the adjusted boundaries of the marked micro-defect regions, the adjusted mesh refinement weights, and the marking and weight assignment of the newly formed micro-defect regions, all identified micro-defect regions are pre-marked in the geometric model and assigned preset mesh refinement weights.

[0050] Specifically, during the use of silicon carbide devices, periodically acquiring the geometric feature information of newly formed micro-defects in the corresponding packaging material refers to periodically inspecting the packaging material using non-destructive testing techniques, such as ultrasonic scanning, X-ray tomography, or infrared thermography, to capture newly emerging micro-defects, such as cracks, voids, or delamination. The geometric feature information of these newly formed micro-defects, including their size, shape, and location, is recorded and used to update the micro-defect database. This database aims to maintain a dynamic, real-time set of defect information.

[0051] Based on the current service time, the characteristics of identified micro-defects, and historical operating data of silicon carbide devices, this study analyzes the morphological evolution trends of existing micro-defects and the location of newly emerging micro-defects. This can be understood as using historical data and the current state to predict defect behavior. The current service time provides the timescale for defect evolution, the characteristics of identified micro-defects (such as initial size, location, and type) form the basis of the analysis, while historical operating data of silicon carbide devices (such as temperature cycling, electrical stress, and mechanical vibration) provides the driving factors for defect evolution. By establishing physical models or machine learning models, the influence of these factors on defect propagation, shape changes, and the emergence of new defects can be analyzed.

[0052] In practical applications, based on the morphological evolution trend and location of occurrence, the boundaries and mesh refinement weights of the marked micro-defect regions in the geometric model of the encapsulation structure are adjusted. Newly emerging micro-defect regions are also marked and weighted. Specifically, this means dynamically updating the regions representing defects in the geometric model based on the analysis results. If a marked defect is expected to expand, its boundary will be enlarged accordingly; if it is expected to shrink or heal, its boundary may shrink. Simultaneously, the mesh refinement weights are adjusted according to the potential severity and evolution rate of the defect; for example, rapidly expanding defect regions are assigned higher weights. For newly emerging defects, new marked regions are added to the geometric model, and initial mesh refinement weights are assigned based on their characteristics.

[0053] Therefore, before pre-simulating and generating the initial coarse mesh, all identified micro-defect regions are pre-marked in the geometric model based on micro-geometric non-uniformity characteristics, adjusted boundaries of marked micro-defect regions, adjusted mesh refinement weights, and the marking and weight assignment of newly formed micro-defect regions. Pre-defined mesh refinement weights are then assigned to these regions. This step ensures that the geometric model used in the pre-simulation accurately reflects the most realistic defect distribution and potential sensitive areas of the silicon carbide device under its current service condition, providing a precise basis for subsequent mesh refinement.

[0054] As a specific implementation method, a concrete example is given below. Suppose that after 1000 hours of operation, periodic inspections of a silicon carbide device reveal new microcracks within its encapsulation material, and an existing void defect shows a slight tendency to expand under thermal cycling stress. At this point, the system first acquires the geometric characteristics of these newly formed microcracks and adds them to the micro-defect database. Simultaneously, based on the 1000 hours of operational history data (including temperature, current, voltage, etc.) and combined with existing material degradation models, the system analyzes the morphological evolution trend of the existing void defect, for example, its size increasing by 5 micrometers and expanding in a specific direction. Based on these analysis results, the boundary of the existing void defect in the geometric model of the encapsulation structure is adjusted accordingly, and its mesh refinement weight may also be increased due to its increased expansion rate and potential hazard. For newly discovered microcracks, a marked region is added to the geometric model, and an initial mesh refinement weight is assigned based on its size and location. Ultimately, during the pre-simulation before the next packaging structure optimization, the geometric model used will include this dynamically updated defect information, enabling mesh refinement to more accurately focus on these real, dynamically changing sensitive areas, ensuring the accuracy and effectiveness of the optimization process.

[0055] Optionally, based on the current service time, the characteristics of identified micro-defects, and the operating history data of silicon carbide devices, the steps to analyze the morphological evolution trend of existing micro-defects and the location of newly formed micro-defects include: During the operation of silicon carbide devices, current, voltage, and temperature are monitored in real time to obtain transient electromagnetic forces, transient thermal loads, and local electric field strength. Based on transient electromagnetic force, transient thermal load, and local electric field intensity, calculate the instantaneous gradient and instantaneous intensity of each physical field in the microscopic defect region; Based on instantaneous gradient and instantaneous intensity, combined with a preset threshold for the synergistic effect of physical fields, the region where there is a nonlinear synergistic effect between transient electromagnetic force-induced micro-vibration, transient thermal load, and local electric field intensity is identified, and the relevant analysis region is obtained; Based on the relevant analysis region, and combined with the existing geometric characteristics of micro-defects and material degradation mechanisms, the analysis parameters corresponding to the evolution rate and direction of the existing micro-defects are adjusted. Based on the adjusted analysis parameters, the morphological evolution trend of existing micro-defects and the location of newly emerging micro-defects were analyzed.

[0056] Specifically, during the operation of silicon carbide devices, real-time monitoring of current, voltage, and temperature, and acquisition of transient electromagnetic forces, transient thermal loads, and local electric field strengths, refers to the continuous acquisition of key electrical and thermal parameters of silicon carbide devices under actual operating conditions through integrated sensors, external testing equipment, or high-precision simulation models. Among these, current, voltage, and temperature are fundamental physical quantities that directly reflect the device's operating state, and they are the fundamental driving factors for the generation of transient electromagnetic forces, transient thermal loads, and local electric field strengths. Transient electromagnetic forces refer to the electromagnetic effects generated by changes in the internal current of the device, which may lead to minute mechanical vibrations within the material; transient thermal loads refer to the heat accumulation and dissipation generated during instantaneous switching or load changes in the device; and local electric field strength specifically refers to the electric field distribution intensity in the region near microscopic defects. These physical quantities are considered to be the physical stimuli that directly affect the evolution of microscopic defects and the initiation of new defects.

[0057] Furthermore, based on transient electromagnetic force, transient thermal load, and local electric field strength, the instantaneous gradient and instantaneous intensity of each physical field in the micro-defect region are calculated. The instantaneous gradient reflects the rate of change of a physical quantity in space; for example, an instantaneous temperature gradient may lead to local thermal stress concentration. The instantaneous intensity represents the absolute magnitude of a physical quantity at a specific moment. These parameters are crucial for accurately assessing the local load and energy density experienced by the micro-defect region.

[0058] Based on this, and using instantaneous gradients and instantaneous intensities, combined with a preset threshold for synergistic effects of physical fields, regions exhibiting nonlinear synergistic effects among transient electromagnetic force-induced micro-vibrations, transient thermal loads, and local electric field strengths are identified, yielding the relevant analysis regions. Nonlinear synergistic effects refer to the complex interaction of multiple physical fields, rather than simple linear superposition, potentially leading to destructive effects far exceeding those of a single physical field. For example, high temperatures can significantly reduce the mechanical strength of materials, making them more susceptible to fatigue cracks under transient electromagnetic force-induced micro-vibrations; high electric field strength can accelerate electrochemical corrosion, which is further accelerated by transient thermal loads. The preset threshold for synergistic effects of physical fields can be set based on experimental data, material properties, failure mechanisms, and historical operating experience, with the aim of accurately identifying these high-risk regions.

[0059] Therefore, based on the relevant analytical regions and considering the existing geometric characteristics of micro-defects and material degradation mechanisms, the analytical parameters corresponding to the evolution rate and direction of existing micro-defects are adjusted. Once relevant analytical regions with nonlinear synergistic effects are identified, it is necessary to modify the parameters used to analyze defect evolution based on the specific physical field coupling characteristics of these regions, combined with the existing geometric characteristics of micro-defects (e.g., size, shape, location, orientation, etc.) and material degradation mechanisms (e.g., fatigue, creep, electromigration, thermal cracking, oxidation, etc.). For example, in regions with significant synergistic effects, the defect evolution rate parameter may be increased, while the evolution direction parameter may be modified to more accurately reflect the propagation path driven by combined stress or energy.

[0060] Finally, based on the adjusted analysis parameters, the morphological evolution trend of existing microdefects and the location of newly formed microdefects are analyzed. By using these modified analysis parameters, more advanced physics-coupled simulation models or machine learning models can be used to more accurately predict how existing microdefects will develop (e.g., crack propagation speed and direction, void growth pattern), and in which high-risk areas new microdefects are most likely to emerge.

[0061] Optionally, the steps of associating material parameter combinations with a discretized packaging structure model to obtain core optimization parameter information, which serves as input for the packaging structure optimization process, include: The mesh cells of the discretized packaging structure model are partitioned according to the material layer and interface type, and sensitive and non-sensitive areas are identified. The material parameter combinations are converted to obtain an effective parameter set that matches each material layer. The effective parameter set is projected according to material layer and interface type and bound to the corresponding mesh partition to form model parameter instances; Apply the boundary conditions of the target application to the model parameter instances and solve them to obtain the corresponding key indicators; Key indicators are normalized and weighted to generate core optimization parameter information, which is then used as input for the encapsulation structure optimization process.

[0062] Specifically, partitioning the mesh elements of a discretized packaging structure model according to material layers and interface types refers to dividing the entire discretized model into multiple independent sub-regions based on the physical boundaries of different materials within the packaging structure and the contact surfaces between them. For example, a silicon carbide device packaging structure may contain multiple material layers such as a chip, solder layer, substrate, wire bonding, and encapsulating adhesive, as well as the interfaces between these layers. Each material layer and interface has its unique physical and chemical properties. Simultaneously, within these partitions, mesh elements are identified based on pre-identified sensitive region information, distinguishing between regions sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress, and non-sensitive regions. The purpose is to provide a refined structural foundation for subsequent material parameter binding and physics field solving.

[0063] The effective conversion of material parameter combinations can be understood as transforming the original material parameters, which may have statistical distribution characteristics (e.g., combinations of material parameters generated by batch differences and microstructural inhomogeneities), into macroscopically equivalent parameters applicable to a specific material layer or interface type. Specifically, this may involve homogenizing the material parameters, considering anisotropic effects, or calculating based on the material's microstructural model to obtain an effective parameter set that matches the actual physical behavior of each material layer. The aim is to ensure that the material parameters accurately reflect their role in the macroscopic model, while taking into account the inherent variability of the material.

[0064] In practical applications, the effective parameter set is projected and bound to the corresponding mesh partitions according to the material layer and interface type, forming model parameter instances. Specifically, this means that the material parameters, after effective transformation, are precisely mapped and assigned to the corresponding mesh cell partitions in the discretized package structure model according to their material layer or interface type. For example, the effective parameter set of the solder layer will be bound to the mesh partition corresponding to the solder layer, while the interface parameters between the chip and the solder will be bound to the corresponding interface mesh. Thus, each discretized package structure model will be combined with a specific set of finely processed material parameters to form a complete model parameter instance that can be used for simulation. The purpose is to construct a simulation model that can reflect the inherent variability of materials and their specific distribution in the package structure.

[0065] Furthermore, applying boundary conditions to the model parameter instances and solving them to obtain the corresponding key performance indicators (KPIs) involves placing the constructed model parameter instances under boundary conditions simulating the actual working environment for numerical simulation. For example, specific boundary conditions such as temperature cycling, current load, or mechanical vibration can be applied, and then solved using finite element analysis (FEA) or other numerical methods to obtain key performance indicators of the packaging structure under these conditions, such as maximum temperature, maximum stress, electric field distribution, thermal resistance, and fatigue life prediction. The purpose is to evaluate the performance of the packaging structure in practical applications after considering material variability.

[0066] Therefore, normalizing and weighting key indicators to generate core optimization parameters, which then serve as input for the packaging structure optimization process, involves normalizing multiple key indicators obtained from simulations using a unified scale to eliminate dimensional differences. Subsequently, based on the importance of these indicators in the packaging structure optimization objective, different weights are assigned for weighted aggregation, resulting in a comprehensive value—the core optimization parameters. For example, if the optimization objective is to reduce thermal resistance and improve mechanical reliability, thermal resistance and maximum stress might be assigned higher weights. The aim is to transform complex simulation results into a concise, quantifiable input to guide subsequent packaging structure optimization.

[0067] Optionally, the steps of adjusting the analytical parameters corresponding to the evolution rate and direction of existing micro-defects based on the relevant analysis region and in conjunction with the existing geometric characteristics of micro-defects and material degradation mechanisms include: For micro-defect regions, the propagation speed and attenuation factor of the elastic wave generated in the encapsulation material are calculated based on transient electromagnetic force, transient thermal load and local electric field intensity. Among them, transient electromagnetic force, transient thermal load, and local electric field intensity refer to key physical field parameters that are monitored and acquired in real time during the operation of silicon carbide devices. Elastic wave propagation velocity and attenuation factor refer to the propagation characteristics of mechanical waves excited by these physical field changes (especially micro-vibrations induced by transient electromagnetic forces) within the packaging material. These parameters can be calculated using the material's acoustic properties, density, elastic modulus, and damping coefficient, with the aim of assessing the impact of mechanical vibration on micro-defects.

[0068] Based on the propagation speed and attenuation factor of elastic waves, the energy concentration effect of elastic waves in the defect edge region is quantified. Specifically, the energy concentration effect refers to the phenomenon where, when an elastic wave encounters a microscopic defect (such as a crack or void), the stress or strain energy density at the defect edge region increases significantly due to wave scattering and reflection. This effect can be quantified using finite element analysis or boundary element method combined with elastic wave theory, with the aim of identifying potentially high-risk areas for mechanical damage accumulation.

[0069] Calculate the local temperature gradient and electric field intensity in the micro-defect region based on the transient thermal load and local electric field intensity; The local temperature gradient refers to the rate of change of temperature within or near a micro-defect region due to transient thermal load, while the local electric field strength refers to the electric field strength within that region caused by local charge distribution or dielectric constant inhomogeneity. These parameters can be solved using the heat conduction equation and the electromagnetic field equation, with the aim of evaluating thermoelectric damage mechanisms such as thermal stress, thermal fatigue, and electro-induced breakdown.

[0070] Based on the local temperature gradient and electric field intensity of the micro-defect region, combined with the plastic flow threshold of the encapsulation material, the rate and direction of plastic deformation accumulation are quantified. Specifically, plastic deformation accumulation refers to the irreversible plastic deformation that occurs in a material within a micro-defect region under cyclic thermal or electrical stress. The plastic flow threshold is the critical stress or strain at which a material begins to undergo macroscopic plastic deformation. By converting local temperature gradients and electric field strengths into local stress / strain and comparing them with the material's plastic flow threshold, the accumulation rate and direction of plastic deformation can be quantified, with the aim of predicting fatigue damage and creep behavior in materials.

[0071] Based on the local electric field strength and transient thermal load, combined with the chemical bond dissociation energy and activation energy of the encapsulation material, the rate and location of chemical bond breakage are quantified. Among them, the chemical bond dissociation energy refers to the energy required to break chemical bonds, and the activation energy refers to the minimum energy required for a chemical reaction (such as material degradation). Under high temperature and strong electric field, the molecular chains of encapsulation materials may break, leading to material performance degradation. By correlating local electric field strength and transient thermal load with the chemical bond dissociation energy and activation energy of the material, the rate and location of chemical bond breakage can be quantified, with the aim of assessing the risk of chemical aging and dielectric breakdown of the material.

[0072] Based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breaking, a multi-damage mode coupling factor is constructed. Specifically, the multi-damage mode coupling factor is a comprehensive parameter used to characterize the interactions and synergistic effects between different damage mechanisms (mechanical damage, plastic damage, and chemical damage). This factor is obtained by weighting, normalizing, or nonlinearly combining the above quantitative results, and its purpose is to provide a unified index to assess the overall damage state of micro-defects.

[0073] By associating the multi-damage mode coupling factor with the existing geometric characteristics of microdefects and material degradation mechanisms, the analytical parameters corresponding to the evolution rate and direction of existing microdefects are corrected.

[0074] The correlation operation involves using the calculated multi-damage mode coupling factor as input, combined with the actual geometry of micro-defects (such as crack length, width, and tip radius) and known material degradation mechanisms (such as fatigue crack propagation models, creep damage models, and dielectric aging models), to correct the original analytical parameters for the evolution rate and direction of micro-defects. The corrected parameters can more accurately reflect the true evolution behavior of micro-defects under multi-physics coupling, aiming to improve the prediction accuracy of device lifetime and reliability.

[0075] Optionally, the steps for constructing a multi-damage mode coupling factor based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breaking include: For the three damage modes of elastic wave propagation, plastic deformation accumulation, and chemical bond breakage, their respective time windows and spatial ranges are set; Within the time window of each damage mode, the quantification results of each damage mode are integrated over time to obtain the cumulative damage amount of each damage mode within the corresponding time window. Within the spatial range of each damage mode, the quantification results of each damage mode are spatially averaged to obtain the average damage intensity of each damage mode within the corresponding spatial range. A multi-damage mode coupling factor is constructed based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor.

[0076] Specifically, the three damage modes mentioned above—elastic wave propagation, plastic deformation accumulation, and chemical bond breakage—represent different failure mechanisms that silicon carbide device packaging structures may experience during service. Elastic wave propagation mainly focuses on the energy concentration effect caused by transient shocks or vibrations; plastic deformation accumulation emphasizes the irreversible deformation of materials under long-term stress; and chemical bond breakage involves the destruction of the material's microstructure under extreme conditions such as high temperature and strong electric fields. To accurately assess the combined impact of these damage modes, they need to be precisely quantified.

[0077] Specifically, setting individual time windows and spatial ranges for each damage mode aims to capture the temporal and spatial specificities of different damage mechanisms. For example, elastic wave propagation may occur within an extremely short time, and its energy concentration effect may be limited to the edge region of microscopic defects; while plastic deformation accumulation is a relatively slow process that may affect a wider area. By setting appropriate time windows and spatial ranges, higher accuracy and representativeness of the quantification results for each damage mode can be ensured.

[0078] By integrating the quantified results of each damage mode over time within its respective time window, the cumulative damage amount for that mode within the corresponding time window can be obtained. Time integration reflects the cumulative effect of damage over time, which is crucial for assessing material fatigue and aging under long-term service. For example, the cumulative damage amount from plastic deformation can be obtained by integrating the plastic strain rate over the time window.

[0079] Simultaneously, spatial averaging of the quantification results for each damage mode within its spatial range yields the average damage intensity of each mode within that corresponding spatial range. Spatial averaging helps smooth local fluctuations, providing regional damage assessments and thus enabling a better understanding of the damage distribution characteristics within the encapsulated structure. For example, the energy concentration effect of elastic waves at the defect edge can be quantified by averaging the local energy density over a specific spatial range.

[0080] Finally, a multi-damage mode coupling factor is constructed based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor. The asynchronous correction factor is one of the key innovations of this application, aiming to address the problem of asynchronous occurrence or evolution between different damage modes. For example, elastic wave damage may occur rapidly under instantaneous overload, while chemical bond breaking may gradually manifest under prolonged high-temperature conditions. By introducing an asynchronous correction factor, the interaction and superposition effects of these damage modes under actual working conditions can be more accurately reflected, thereby constructing a more physically meaningful multi-damage mode coupling factor.

[0081] Optionally, the step of constructing a multi-damage mode coupling factor based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor, includes: During the operation of silicon carbide devices, identify the occurrence of extreme transient overload or short circuit conditions; When extreme transient overload or short-circuit conditions are identified, the high-frequency data acquisition mode is immediately activated to obtain the instantaneous quantization results of each damage mode with a preset higher sampling frequency and a preset shorter time window. Based on the instantaneous quantization results obtained under the high-frequency data acquisition mode, the cumulative damage amount and average damage intensity of each damage mode are recalculated. Based on the recalculated cumulative damage and average intensity, as well as the current working condition type, a matching asynchronous correction factor is selected from the preset extreme working condition correction factor library. Based on the selected asynchronous correction factor, a multi-damage mode coupling factor is constructed.

[0082] Specifically, "identifying extreme transient overload or short-circuit conditions" refers to real-time monitoring of key operating parameters of silicon carbide devices, such as current, voltage, and temperature, and comparing them with preset thresholds to determine whether transient overload or short-circuit events exceeding normal ranges exist. For example, when the current or voltage exceeds the safe operating limit within a very short time, or when a short-circuit protection mechanism is triggered, it can be identified as an extreme transient condition. Its purpose is to promptly capture special operating states that may lead to rapid damage accumulation.

[0083] The phrase "immediately activate high-frequency data acquisition mode to acquire instantaneous quantization results of each damage mode at a preset higher sampling frequency and a preset shorter time window" can be understood as follows: once extreme conditions are identified, the system will automatically switch to a more intensive data sampling mode. For example, the normal sampling frequency may be 1kHz, but under extreme conditions, the sampling frequency can be increased to 10kHz or higher, while the data acquisition time window is shortened from several seconds to milliseconds to ensure that the drastic changes in the physical field within a very short time can be captured, thereby obtaining more refined instantaneous damage quantification data. The aim is to acquire sufficiently high-resolution data to accurately reflect the dynamic evolution of damage modes under extreme conditions.

[0084] In practical applications, "recalculating the cumulative damage and average damage intensity of each damage mode based on the instantaneous quantization results acquired under high-frequency data acquisition mode" specifically involves re-evaluating the quantization results of damage modes such as the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breakage using instantaneous data acquired at high frequencies. Due to the improved data resolution, the recalculated cumulative damage and average damage intensity will more accurately reflect the actual damage level under extreme conditions. Its purpose is to provide more reliable input for subsequent selection of correction factors.

[0085] Furthermore, "selecting a matching asynchronous correction factor from a pre-set extreme condition correction factor library based on the recalculated cumulative damage and average intensity, and the current operating condition type" means that the system searches a pre-established extreme condition correction factor library based on the recalculated damage and intensity, combined with the specific type of the currently identified extreme condition (e.g., transient overload or short circuit). This factor library stores asynchronous correction factors pre-calibrated or experimentally verified for different extreme conditions and damage states. A matching algorithm selects the correction factor that best matches the current situation. The purpose is to ensure that the selected correction factor accurately reflects the nonlinear coupling relationship between multiple damage modes under extreme conditions.

[0086] Therefore, "constructing a multi-damage mode coupling factor based on the selected asynchronous correction factor" refers to substituting the dynamically selected asynchronous correction factor, which is highly matched to the current extreme operating conditions, into the calculation model of the multi-damage mode coupling factor, thereby obtaining a more accurate and reliable coupling factor under extreme operating conditions. Its purpose is to improve the applicability and accuracy of the multi-damage mode coupling factor in complex and dynamic environments.

[0087] Optionally, the step of selecting a matching asynchronous correction factor from a pre-defined extreme condition correction factor library based on the recalculated cumulative damage and average damage intensity, and the current operating condition type, includes: Extract the asynchronous correction factor corresponding to the current operating condition type from the preset extreme operating condition correction factor library; A consistency check is performed on the extracted asynchronous correction factors to identify and eliminate redundant or conflicting factors in the extreme condition correction factor library. Based on the results of the consistency check, the remaining asynchronous correction factors in the extreme working condition correction factor library are prioritized and selected first, with the asynchronous correction factor having the highest matching degree with the recalculated cumulative damage amount and average damage intensity. Based on the priority ranking results, the asynchrony correction factor with the highest matching degree is selected.

[0088] Specifically, when selecting a matching asynchronous correction factor, it is first necessary to extract the asynchronous correction factor corresponding to the current operating condition type from a pre-defined extreme operating condition correction factor library. This factor library can be pre-established and contains asynchronous correction factors calibrated for different extreme operating condition types (e.g., different degrees of overload, short circuits of different durations, etc.). The extraction operation aims to initially screen candidate factors related to the current operating state.

[0089] Furthermore, a consistency check is performed on the extracted asynchronous correction factors. The purpose of this check is to identify and eliminate potentially redundant or conflicting factors in the extreme condition correction factor library. Redundant factors are those that functionally or effectively overlap, while conflicting factors may produce mutually canceling or unreasonable results under specific conditions. The consistency check can be performed using pre-defined logical rules, expert systems, or machine learning models to ensure the independence and validity of candidate factors.

[0090] Based on this, and according to the consistency check results, the remaining asynchronous correction factors in the extreme condition correction factor library are prioritized. The priority ranking is based on the degree of matching between these factors and the recalculated cumulative damage and average damage intensity. The degree of matching can be calculated using a pre-defined similarity measurement algorithm, such as methods based on statistical correlation, pattern recognition, or fuzzy logic. The factor with the highest degree of matching will be assigned the highest priority, indicating that it best reflects the damage coupling characteristics under the current extreme conditions.

[0091] Ultimately, based on the priority ranking, the asynchrony correction factor with the highest matching degree was selected. This selection ensures that the correction factor used when constructing the multi-damage mode coupling factor best represents the actual damage state of the current device, thereby improving the accuracy and reliability of the coupling factor.

[0092] This application also discloses a novel high-temperature resistant silicon carbide device packaging structure optimization system, used to perform packaging structure optimization for novel high-temperature resistant silicon carbide devices, combined with... Figure 3 The novel high-temperature resistant silicon carbide device packaging structure optimization system 1 includes: Sensitive area identification module 11 is used to identify areas in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field intensity and mechanical stress, based on a preset physical field sensitivity threshold, and to obtain sensitive area information. The encapsulation model construction module 12 is used to encrypt the computational domain grid of sensitive areas based on sensitive area information, and keep the computational domain grid sparse for non-sensitive areas, thereby constructing a discretized encapsulation structure model. The difference parameter acquisition module 13 is used to acquire the batch difference and microstructure non-uniformity of the packaging material of the packaging structure. The material parameter generation module 14 is used to generate a preset number of material parameter combinations representing the inherent variability of the material based on batch differences and microstructure inhomogeneity. The optimization parameter generation module 15 is used to perform correlation operations between the material parameter combination and the discretized packaging structure model to obtain core optimization parameter information, which is used as input for the packaging structure optimization process. The packaging structure optimization module 16 is used to optimize the packaging structure based on the core optimization parameter information.

[0093] Specifically, the sensitive area identification module can be a software program that integrates physical field analysis algorithms, such as finite element analysis or computational fluid dynamics solvers, to perform preliminary simulations of the packaged structure and automatically identify sensitive areas based on preset physical field sensitivity thresholds. Alternatively, this module can be a hardware acceleration unit, such as a programmable gate array (FPGA) or application-specific integrated circuit (ASIC), to process sensor data in real time and quickly identify sensitive areas of operating silicon carbide devices.

[0094] The encapsulation model building module can be a submodule integrated into computer-aided design (CAD) or computer-aided engineering (CAE) software. It receives sensitive area information and automatically adjusts the mesh generation strategy based on this information, refining the mesh in sensitive areas while maintaining a sparse mesh in non-sensitive areas, ultimately generating a discretized encapsulated structure model. In some implementations, this module can also interface with existing mesh generation tools via a scripting language to achieve automated mesh generation.

[0095] The difference parameter acquisition module can be a data acquisition and management system used to obtain batch difference data and microstructure inhomogeneity information of packaging materials from material testing equipment (such as tensile testing machines and thermal conductivity meters) or supplier databases. For example, this module can be configured with a data interface to import performance reports of different batches of materials and process microstructure images to extract inhomogeneity features.

[0096] The material parameter generation module can be a software component based on statistical or machine learning algorithms. It generates a preset number of material parameter combinations representing the inherent variability of the material, based on data provided by the difference parameter acquisition module. For example, this module can implement Monte Carlo simulation or Latin hypercube sampling algorithms to generate diverse parameter combinations within the statistical distribution range of the material parameters.

[0097] The optimization parameter generation module can be a simulation management and data post-processing unit, used to associate material parameter combinations with discretized packaging structure models, automatically execute a series of numerical simulations, and extract key performance indicators from the simulation results. For example, this module can perform statistical analysis, normalization, or weighted aggregation on simulation data such as maximum temperature and maximum stress to generate core optimization parameter information.

[0098] The packaging structure optimization module can be a software platform integrating multiple optimization algorithms, such as genetic algorithms, particle swarm optimization algorithms, or topology optimization algorithms. It receives core optimization parameter information and, based on this, it iteratively optimizes design variables such as the packaging structure's geometry, material selection, or internal layout. For example, this module can automatically adjust design parameters according to optimization objectives and constraints until the optimal packaging structure that meets the requirements is found.

[0099] The novel high-temperature silicon carbide device packaging structure optimization system proposed in this application represents a significant advancement compared to existing technologies. Traditional packaging structure optimization often relies on manual experience or simple iterative trial and error, lacking a systematic consideration of the inherent variability of materials and the coupling effects of complex physical fields. This is particularly problematic when facing the dynamic and transient conditions of high-frequency switching devices, where optimization results are often unsatisfactory. The system in this application, through the introduction of a modular design, integrates key steps such as sensitive area identification, refined modeling, material variability consideration, and multi-parameter correlation analysis into a unified framework. This systematic approach ensures the comprehensiveness, accuracy, and robustness of the optimization process, effectively addressing the shortcomings of traditional methods under dynamic and nonlinear conditions. This system can significantly improve the long-term stable operation and reliability of silicon carbide devices, reduce the risk of early failure, and provide strong technical support for the development of the power electronics field.

[0100] The above are merely embodiments of this application and are not intended to limit the scope of protection of 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 scope of protection of this application.

Claims

1. A novel method for optimizing the packaging structure of high-temperature resistant silicon carbide devices, characterized in that, include: For silicon carbide devices, based on a preset physical field sensitivity threshold, regions in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress are identified, and information on sensitive regions is obtained. Based on the information of the sensitive areas, the computational domain mesh is densified in the sensitive areas, while the computational domain mesh is kept sparse in the non-sensitive areas, thus constructing a discretized encapsulation structure model. To obtain batch variations and microstructural inhomogeneities of the packaging materials used in the packaging structure; Based on the batch differences and microstructure inhomogeneity, a preset number of material parameter combinations representing the inherent variability of the material are generated; The material parameter combination is associated with the discrete packaging structure model to obtain core optimization parameter information, which is used as input for the packaging structure optimization process. Optimize the packaging structure based on core optimization parameter information; The step of associating the material parameter combination with the discretized packaging structure model to obtain core optimization parameter information, which is then used as input for the packaging structure optimization process, includes: The discrete packaging structure model is partitioned into grid cells according to material layers and interface types, and sensitive and non-sensitive areas are identified. The material parameter combinations are converted to obtain an effective parameter set that matches each material layer; The effective parameter set is projected according to material layer and interface type and bound to the corresponding mesh partition to form model parameter instances; Apply the boundary conditions of the target application to the model parameter instance and solve for them to obtain the corresponding key indicators; The key indicators are normalized and weighted to generate core optimization parameter information, which is then used as input for the encapsulation structure optimization process.

2. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 1, characterized in that, The steps of densifying the computational domain grid in sensitive areas and maintaining a sparse computational domain grid in non-sensitive areas based on the sensitive area information include: To obtain information on the actual microscopic geometric non-uniformity characteristics in the packaging material; Establish a rule base that associates micro-defect types with potential failure mechanisms; Based on the microscopic geometric non-uniformity feature information and the association rule library, a transient combination of working conditions that maximizes the intensity of local key physical fields around the microscopic defects is searched in the preset working condition parameter space to generate an active working condition list. Select the transient condition related to the current packaging structure to be optimized from the list of active conditions as the pre-simulation input; Before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity feature information, all identified micro-defect regions are pre-marked in the geometric model of the encapsulation structure and assigned preset mesh refinement weights. Based on the pre-marked geometric model of the encapsulation structure, the preset mesh refinement weights, and the pre-simulation input, a pre-simulation is performed to identify the sensitive areas for pre-simulation. By combining the pre-simulated sensitive area, the pre-marked micro-defect area, and the selected transient condition, the mesh refinement sensitivity threshold of the corresponding area is adjusted. Mesh refinement is performed on regions where the mesh refinement sensitivity threshold is adjusted or on pre-marked micro-defect regions, while the computational domain mesh remains sparse in non-sensitive regions.

3. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 2, characterized in that, The step of pre-marking all identified micro-defect regions in the geometric model of the encapsulation structure and assigning preset mesh refinement weights based on the micro-geometric non-uniformity feature information before pre-simulation and generation of the initial coarse mesh includes: During the use of silicon carbide devices, the geometric feature information of newly formed micro-defects in the corresponding packaging material is periodically acquired, and the micro-defect database is updated. Based on the current service time, the characteristics of the identified micro-defects, and the historical operating data of silicon carbide devices, the morphological evolution trend of existing micro-defects and the location of new micro-defects are analyzed. Based on the morphological evolution trend and the location of occurrence, adjust the boundary and mesh refinement weight of the marked micro-defect regions in the geometric model of the encapsulation structure, and mark and assign weights to newly formed micro-defect regions. Before pre-simulation and generation of the initial coarse mesh, based on the micro-geometric non-uniformity feature information, the adjusted boundaries of the marked micro-defect regions, the adjusted mesh refinement weights, and the marking and weight assignment of the newly formed micro-defect regions, all identified micro-defect regions are pre-marked in the geometric model and assigned preset mesh refinement weights.

4. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 3, characterized in that, The steps for analyzing the morphological evolution trend of existing microdefects and the location of newly formed microdefects based on the current service time, the characteristics of identified microdefects, and the operating history data of silicon carbide devices include: During the operation of silicon carbide devices, current, voltage, and temperature are monitored in real time to obtain transient electromagnetic forces, transient thermal loads, and local electric field strength. Based on the transient electromagnetic force, the transient thermal load, and the local electric field strength, calculate the instantaneous gradient and instantaneous intensity of each physical field in the microscopic defect region; Based on instantaneous gradient and instantaneous intensity, combined with a preset threshold for the synergistic effect of physical fields, the region where there is a nonlinear synergistic effect between transient electromagnetic force-induced micro-vibration, transient thermal load, and local electric field intensity is identified, and the relevant analysis region is obtained; Based on the relevant analysis region, and combined with the existing geometric characteristics of micro-defects and material degradation mechanisms, the analysis parameters corresponding to the evolution rate and direction of the existing micro-defects are adjusted. Based on the adjusted analysis parameters, the morphological evolution trend of existing micro-defects and the location of newly emerging micro-defects were analyzed.

5. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 4, characterized in that, The step of adjusting the analytical parameters corresponding to the evolution rate and direction of existing micro-defects based on the relevant analysis region, combined with the existing geometric characteristics of micro-defects and material degradation mechanisms, includes: For the micro-defect region, the propagation speed and attenuation factor of the elastic wave generated in the encapsulation material are calculated based on the transient electromagnetic force, the transient thermal load, and the local electric field strength. Based on the propagation speed and attenuation factor of elastic waves, the energy concentration effect of elastic waves in the defect edge region is quantified. Based on the transient thermal load and the local electric field intensity, calculate the local temperature gradient and electric field intensity of the micro-defect region; Based on the local temperature gradient and electric field intensity of the micro-defect region, combined with the plastic flow threshold of the encapsulation material, the rate and direction of plastic deformation accumulation are quantified. Based on the local electric field strength and the transient thermal load, combined with the chemical bond dissociation energy and activation energy of the encapsulation material, the rate and location of chemical bond breakage are quantified. Based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breaking, a multi-damage mode coupling factor is constructed. The multi-damage mode coupling factor is correlated with the existing geometric features of microdefects and material degradation mechanisms to correct the analysis parameters corresponding to the evolution rate and direction of existing microdefects.

6. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 5, characterized in that, The steps for constructing a multi-damage mode coupling factor based on the energy concentration effect of elastic wave propagation, the rate and direction of plastic deformation accumulation, and the rate and location of chemical bond breaking include: For the three damage modes of elastic wave propagation, plastic deformation accumulation, and chemical bond breakage, their respective time windows and spatial ranges are set; Within the time window of each damage mode, the quantification results of each damage mode are integrated over time to obtain the cumulative damage amount of each damage mode within the corresponding time window. Within the spatial range of each damage mode, the quantification results of each damage mode are spatially averaged to obtain the average damage intensity of each damage mode within the corresponding spatial range. A multi-damage mode coupling factor is constructed based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor.

7. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 6, characterized in that, The step of constructing a multi-damage mode coupling factor based on the cumulative damage amount and average damage intensity of each damage mode, as well as a preset asynchronous correction factor, includes: During the operation of silicon carbide devices, identify the occurrence of extreme transient overload or short circuit conditions; When extreme transient overload or short-circuit conditions are identified, the high-frequency data acquisition mode is immediately activated to obtain the instantaneous quantization results of each damage mode with a preset higher sampling frequency and a preset shorter time window. Based on the instantaneous quantization results obtained under the high-frequency data acquisition mode, the cumulative damage amount and average damage intensity of each damage mode are recalculated. Based on the recalculated cumulative damage and average damage intensity, as well as the current working condition type, a matching asynchronous correction factor is selected from the preset extreme working condition correction factor library. Based on the selected asynchronous correction factor, a multi-damage mode coupling factor is constructed.

8. The method for optimizing the packaging structure of a novel high-temperature resistant silicon carbide device according to claim 7, characterized in that, The step of selecting a matching asynchronous correction factor from a preset extreme condition correction factor library based on the recalculated cumulative damage amount and average damage intensity, as well as the current operating condition type, includes: Extract the asynchronous correction factor corresponding to the current operating condition type from the preset extreme operating condition correction factor library; A consistency check is performed on the extracted asynchronous correction factors to identify and eliminate redundant or conflicting factors in the extreme condition correction factor library. Based on the results of the consistency check, the remaining asynchronous correction factors in the extreme working condition correction factor library are prioritized and selected first, with the asynchronous correction factor having the highest matching degree with the recalculated cumulative damage amount and average damage intensity. Based on the priority ranking results, the asynchrony correction factor with the highest matching degree is selected.

9. A novel high-temperature resistant silicon carbide device packaging structure optimization system, used to perform packaging structure optimization for novel high-temperature resistant silicon carbide devices, characterized in that, include: The sensitive area identification module is used to identify areas in the packaging structure of silicon carbide devices that are sensitive to changes in heat flow, electromagnetic field intensity, and mechanical stress, based on a preset physical field sensitivity threshold, and to obtain sensitive area information. The encapsulation model construction module is used to encrypt the computational domain grid of the sensitive area based on the sensitive area information, and keep the computational domain grid sparse for the non-sensitive area, thereby constructing a discretized encapsulation structure model. The difference parameter acquisition module is used to acquire batch differences and microstructure non-uniformity of the packaging material in the packaging structure. The material parameter generation module is used to generate a preset number of material parameter combinations representing the inherent variability of the material based on the batch differences and microstructure inhomogeneity. The optimization parameter generation module is used to associate the material parameter combination with the discrete packaging structure model to obtain core optimization parameter information, which is used as input for the packaging structure optimization process. The packaging structure optimization module is used to optimize the packaging structure based on core optimization parameter information; The step of associating the material parameter combination with the discretized packaging structure model to obtain core optimization parameter information, which is then used as input for the packaging structure optimization process, includes: The discrete packaging structure model is partitioned into grid cells according to material layers and interface types, and sensitive and non-sensitive areas are identified. The material parameter combinations are converted to obtain an effective parameter set that matches each material layer; The effective parameter set is projected according to material layer and interface type and bound to the corresponding mesh partition to form model parameter instances; Apply the boundary conditions of the target application to the model parameter instance and solve for them to obtain the corresponding key indicators; The key indicators are normalized and weighted to generate core optimization parameter information, which is then used as input for the encapsulation structure optimization process.

Images