A method, apparatus, and computer device for life assessment of press-fit IGBTs.
By combining an electro-thermal-mechanical multiphysics coupling model and deep learning algorithms, a life assessment method for press-fit IGBTs is constructed, which solves the problem of inaccurate simulation results in existing technologies, achieves accurate and rapid life assessment, and ensures the reliable operation of power electronic equipment.
Patent Information
- Application Number
- CN202311514113.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-14
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-11-14
AI Technical Summary
Existing life assessment methods for press-fit IGBTs suffer from inaccurate simulation results, especially when considering the complexity of electromagnetic interactions. Conventional models struggle to comprehensively account for all factors, leading to life assessment biases.
By combining an electro-thermal-mechanical multiphysics coupling model with a deep learning algorithm, a multiphysics coupling simulation model of a press-fit IGBT is constructed. The controlled variable method is used for simulation, and combined with power cycling tests, a life assessment model is trained using a data-driven deep learning algorithm to achieve the fusion of simulation data and measured data.
It achieves accuracy and speed in life assessment of press-fit IGBTs, providing reliable life predictions for the reliable operation of power electronic equipment and guiding inspection and maintenance.
Smart Images

Figure CN117371294B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of semiconductor technology, and in particular to a method, apparatus, and computer device for assessing the lifetime of a press-fit IGBT. Background Technology
[0002] With the booming development of emerging industries such as new energy power generation, rail transit, smart grids, and electric vehicles, power semiconductor devices have been applied more widely. The reliability of power semiconductor devices is crucial for the reliable and stable operation of power electronic equipment; however, the failure rate of semiconductor devices is as high as 31%, posing a significant threat to the reliable operation of power systems.
[0003] To ensure the stable operation of power electronic systems, a great deal of research has been conducted both domestically and internationally on the reliability and lifespan of IGBT devices.
[0004] Current lifetime assessment methods fall into two categories: one based on multiphysics coupling model data, and the other based on measured data. The former, based on multiphysics coupling modeling, can obtain data that is difficult to obtain through actual measurements, such as junction temperature, stress distribution, and thermal resistance of the entire device. However, electromagnetic forces are relatively small under normal operating conditions, and the electromagnetic interactions inside IGBTs are too complex. Therefore, most studies choose to ignore the influence of magnetic fields and adopt electro-thermal-mechanical coupling models. This will, to some extent, lead to inaccurate physical simulation results of high-power press-fit IGBT modules, and thus cause deviations in lifetime assessment.
[0005] Analytical models do not consider the specific physical failure process of press-fit IGBTs and can only be used under specific conditions; physical models, while considering certain physical characteristics, often ignore other characteristics, making it difficult to comprehensively consider all factors. Furthermore, existing life assessment models for press-fit IGBTs either rely on measured power cycling data or are based on simulation data, lacking a life assessment method that combines more accurate experimental data with more abundant simulation data. Summary of the Invention
[0006] Based on this, it is necessary to provide a method, device, and computer equipment for assessing the lifespan of press-fit IGBTs by using electro-thermal-mechanical multi-physics coupling model data and deep learning algorithms to train a deep learning simulation model, which can realize rapid calculation of simulation data.
[0007] In a first aspect, this application provides a life assessment method for press-fit IGBTs, the life assessment method comprising:
[0008] A multiphysics coupling simulation model of a press-fit IGBT was constructed, and the control variable method was used to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0009] A test platform was constructed to conduct power cycling tests on the simulated press-fit IGBT.
[0010] The multiphysics coupling simulation model is trained using a data-driven deep learning algorithm, and a press-fit IGBT lifetime assessment model is constructed based on the trained multiphysics coupling simulation model.
[0011] The lifespan of the target press-fit IGBT is assessed using the aforementioned press-fit IGBT lifespan assessment model.
[0012] In one embodiment, the construction of a multiphysics coupling simulation model for the press-fit IGBT, and the simulation of the press-fit IGBT under different conditions using the controlled variable method based on the multiphysics coupling simulation model, includes:
[0013] A three-dimensional model of a press-fit IGBT entity is constructed in a COMSOL component, and the three-dimensional model is then simplified.
[0014] The influence of various conditions on spring-loaded compression IGBTs was analyzed using the established multiphysics coupling model and the controlled variable method.
[0015] In one embodiment, the simplification process of the three-dimensional model includes:
[0016] Set the parameters in the three-dimensional model, set the electric field boundary conditions, set the temperature field boundary conditions, and set the stress field boundary conditions.
[0017] In one embodiment, the step of using the constructed multiphysics coupling model and the controlled variable method to analyze the influence of various conditions on the spring-loaded compression type IGBT includes:
[0018] A reference group is constructed using copper and aluminum to maintain the elastic-plastic properties for simulation, and a preset simulation current is set for the reference group;
[0019] Construct the heat transfer coefficients corresponding to the reference group;
[0020] Adjust the stress distribution of the clamp size of the reference group.
[0021] In one embodiment, the construction of the test platform for conducting power cycling tests on the simulated press-fit IGBT includes:
[0022] The base is made of a material that is compatible with different models of press-fit IGBT modules and is stable, reliable, and not prone to slipping or instability.
[0023] The chip is periodically switched on and off to simulate operating conditions;
[0024] Welded water cooling heads are selected to provide continuous and stable heat dissipation for IGBTs, ensuring the orderly cooling process.
[0025] A power cycling test was conducted to simulate the aging process.
[0026] In one embodiment, training the multiphysics coupling simulation model using a data-driven deep learning algorithm, and constructing a press-fit IGBT lifetime assessment model based on the trained multiphysics coupling simulation model, includes:
[0027] Obtain junction temperature and strain fluctuation data from multiphysics simulations, as well as the basic parameters corresponding to the fluctuation data;
[0028] The simulation data is divided into experimental group dataset and test group dataset. The experimental group dataset is further divided into training set and validation set according to a specific ratio. The fluctuation data is preprocessed to obtain feature vector values and corresponding feature values.
[0029] Construct a network framework that includes an input layer, hidden layers, and an output layer;
[0030] Using basic parameters as input information and corresponding feature vector values as output information, the deep learning network model is trained using training and validation sets to form a deep learning multiphysics simulation model trained on finite element simulation data.
[0031] The processed test dataset was then fed into a data-driven deep learning multiphysics simulation model to verify the accuracy and effectiveness of the network model.
[0032] Secondly, this application also provides a life assessment device for press-fit IGBTs, the device comprising:
[0033] The model building module is used to build a multiphysics coupling simulation model of the press-fit IGBT. Based on the multiphysics coupling simulation model, the control variable method is used to simulate the press-fit IGBT under different conditions.
[0034] The cyclic testing module is used to construct a test platform to conduct power cyclic tests on the simulated press-fit IGBT.
[0035] The model training module is used to train the multiphysics coupling simulation model based on a data-driven deep learning algorithm, and to construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0036] The life assessment module is used to assess the life of the target press-fit IGBT using the press-fit IGBT life assessment model.
[0037] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0038] A multiphysics coupling simulation model of a press-fit IGBT was constructed, and the control variable method was used to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0039] A test platform was constructed to conduct power cycling tests on the simulated press-fit IGBT.
[0040] The multiphysics coupling simulation model is trained using a data-driven deep learning algorithm, and a press-fit IGBT lifetime assessment model is constructed based on the trained multiphysics coupling simulation model.
[0041] The lifespan of the target press-fit IGBT is assessed using the aforementioned press-fit IGBT lifespan assessment model.
[0042] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0043] A multiphysics coupling simulation model of a press-fit IGBT was constructed, and the control variable method was used to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0044] A test platform was constructed to conduct power cycling tests on the simulated press-fit IGBT.
[0045] The multiphysics coupling simulation model is trained using a data-driven deep learning algorithm, and a press-fit IGBT lifetime assessment model is constructed based on the trained multiphysics coupling simulation model.
[0046] The lifespan of the target press-fit IGBT is assessed using the aforementioned press-fit IGBT lifespan assessment model.
[0047] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:
[0048] A multiphysics coupling simulation model of a press-fit IGBT was constructed, and the control variable method was used to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0049] A test platform was constructed to conduct power cycling tests on the simulated press-fit IGBT.
[0050] The multiphysics coupling simulation model is trained using a data-driven deep learning algorithm, and a press-fit IGBT lifetime assessment model is constructed based on the trained multiphysics coupling simulation model.
[0051] The lifespan of the target press-fit IGBT is assessed using the aforementioned press-fit IGBT lifespan assessment model.
[0052] By using the electro-thermal-mechanical multiphysics coupled model data built with finite element simulation software and training with deep learning algorithms to obtain a deep learning simulation model, rapid calculation of simulation data is achieved. The life prediction results of the simulation data are fused with the life prediction results of the field measured data to form a multi-source data fusion driven life assessment method for press-fit IGBTs. The results are then integrated into a simple-to-use and fast-computing life assessment software using computer tools, thereby providing guidance for the inspection and maintenance work of engineering technicians and ensuring the reliable operation of power electronic equipment. Attached Figure Description
[0053] Figure 1 This is an application environment diagram of a life assessment method for press-fit IGBTs in one embodiment;
[0054] Figure 2 This is a flowchart illustrating a life assessment method for a press-fit IGBT in one embodiment.
[0055] Figure 3 This is a flowchart illustrating the structure of a deep learning algorithm in one embodiment;
[0056] Figure 4 This is a flowchart illustrating the nonlinear fatigue cumulative damage life calculation process in another embodiment;
[0057] Figure 5 This is a structural block diagram of a life assessment device for a press-fit IGBT in one embodiment;
[0058] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0060] The life assessment method for press-fit IGBTs provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.
[0061] In one embodiment, such as Figure 2 As shown, a life assessment method for press-fit IGBTs is provided, which can be applied to... Figure 1 Taking the server in the example, the following steps are included:
[0062] Step 202: Build a multiphysics coupling simulation model of the press-fit IGBT, and use the control variable method to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0063] A three-dimensional model was constructed in the finite element analysis software COMSOL, referencing the selected 5SNA 3000K452300 spring-loaded IGBT. To reduce computation time and workload while maintaining accuracy, the device model was simplified as follows:
[0064] ① Take a single sub-module for modeling, which includes 8 IGBT chips and 4 FRD chips, with a voltage level of 4500V;
[0065] ② Ignore other components that have little impact on the pressure distribution on the chip surface inside the device, such as: plastic frame, drive connection board, etc.
[0066] ③ Ignore fine details such as chamfers and rounded corners inside the module;
[0067] ④ Ignore the power loss of other structures and contact surfaces outside the active region of the chip.
[0068] Step 204: Construct a test platform and conduct power cycling tests on the simulated press-fit IGBT.
[0069] During the construction of the experimental platform, the base was made of a material that could adapt to different models of press-fit IGBT modules and was stable, reliable, and not prone to slippage or instability. The elasticity of the pressure plate material should be moderate to ensure that pressure is applied evenly without deformation during the pressing process, thus ensuring uniform pressing between the IGBT module and the heat sink. The experimental circuit was built with the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply to simulate the operating conditions by periodically switching the chip on and off. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip, and thermally conductive adhesive was used to fix the K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0070] Step 206: Train the multiphysics coupling simulation model using a data-driven deep learning algorithm, and construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0071] This involves training a deep learning multiphysics simulation model using simulation data to drive deep learning algorithm training. The simulation data is divided into experimental and test datasets. The experimental dataset is further divided into training and validation sets according to a specific ratio. Fluctuation data is preprocessed to obtain feature vector values and their corresponding feature values. A network framework comprising an input layer, hidden layers, and an output layer is constructed. Basic parameters are used as input information, and the corresponding feature vector values are used as output information. The deep learning network model is trained using the training and validation sets, resulting in a deep learning multiphysics simulation model trained on finite element simulation data.
[0072] Step 208: Use the aforementioned press-fit IGBT life assessment model to assess the life of the target press-fit IGBT.
[0073] A deep learning algorithm is trained using simulation data to obtain a multiphysics simulation model, enabling rapid calculation of three-dimensional multi-field information. A nonlinear lifetime model considering simulation results is fused with lifetime results considering measured data to obtain a lifetime assessment model driven by multi-source data fusion. The model and its converted program code are compiled into MATLAB. The LiveLink for MATLAB module is used to implement the deep learning multiphysics simulation model and secondary development of MATLAB. A simple, intuitive, and fast-computing pressure-fit IGBT lifetime assessment software is designed and constructed using a GUI, achieving integrated fatigue life calculation for equipment. Users only need to input basic parameters such as IGBT device parameters, converter valve design parameters, and operating conditions to quickly obtain the expected life of the device, providing a reference for field equipment maintenance and ensuring the safe operation of the power system.
[0074] In one embodiment, step S202 includes:
[0075] A three-dimensional model of a press-fit IGBT entity is constructed in a COMSOL component, and the three-dimensional model is then simplified.
[0076] The influence of various conditions on spring-loaded compression IGBTs was analyzed using the established multiphysics coupling model and the controlled variable method.
[0077] In implementation, a three-dimensional model was constructed in the COMSOL assembly with reference to the selected 5SNA3000K452300 spring-loaded compression type IGBT entity.
[0078] Set the material parameters. The sub-modules, from top to bottom, are: emitter (Cu), spring (high-strength alloy steel), aluminum (Al), molybdenum (Mo), silicon (Si), chip (Solder), and collector (Mo). The material parameters required for the simulation include: ① coefficient of thermal expansion, ② constant-pressure heat capacity, ③ density, ④ thermal conductivity, ⑤ Young's modulus, ⑥ Poisson's ratio, ⑦ electrical conductivity, and ⑧ relative permittivity. COMSOL's built-in library contains all materials; simply import them and select the corresponding domains to complete the material settings.
[0079] Set the electric field boundary conditions. The electric field is set using the "Current (ec)" parameter included in COMSOL, with the emitter set as the "Ground" boundary; the current flowing through the collector is set using the "Current" parameter in the "Termination" boundary; and the contact resistance within the press-fit IGBT is set using the "Electrical Contact" boundary. The main functional expressions included in the electric field are:
[0080]
[0081] J = σE + J e(2);
[0082]
[0083] Where J is the current density, in A / m²; Q j Free charge density, in C / m³ 3 σ represents electrical conductivity, measured in S / m; J e E is the free current density, in A / m²; E is the electric field strength, in V / m; U is the electric potential, in V.
[0084] Set the temperature field boundary conditions. The temperature field is set as "Solid Heat Transfer (ht)". The "Heat Flux" boundary's "Heat Transfer Coefficient" and "External Temperature" conditions are used to simulate convective heat transfer between the collector and emitter in the external cooling system and the IGBT. The remaining external boundaries are set as "Thermal Insulation". The "Thermal Contact" boundary is used to set the contact thermal resistance within the press-fit IGBT. Finally, the chip is set as the "Heat Source", and the "Heat Dissipation Rate" is adjusted to simulate the chip's power loss under different currents. The main function expressions included in the temperature field are:
[0085]
[0086]
[0087] Where ρ is density, in kg / m³ 3 C ρ The constant-pressure heat capacity is expressed in J / (kg·K); u is the velocity, expressed in m / s; T is the temperature, expressed in K; and q represents the heat flux density, expressed in W / m³. 2 Q represents a possible volume source term, in J; Q ted The value represents the possible interface source term, in J; k is the thermal conductivity, in W / (m·K).
[0088] Set stress field boundary conditions. The stress field is set using "Solid Mechanics," and the "Total Force Ftot" in "Boundary Loads" is used to set the collector, and the "Fixed Constraint" is used to set the emitter, simulating the pressure applied to the press-fit IGBT by the fixture. The "Contact" boundary is used to simulate the non-welded condition of the press-fit IGBT. In "Linear Elastic Materials," a "Shaping" condition is added, and an "Initial Yield Stress" is set to simulate copper and aluminum as elasto-plastic materials, ensuring simulation accuracy by closely resembling real-world conditions. The main functional expressions included in the stress field are:
[0089]
[0090]
[0091] Where F is the deformation gradient tensor; S is the Cauchy stress tensor in Pa; T is the temperature in K; V is the velocity of the object in m / s; I is the unit tensor; and U is the displacement in m.
[0092] The finite element analysis software COMSOL has a powerful automatic mesh generation function. In it, select "Physical Field Control Mesh", check "Solid Mechanics", "Solid Heat Transfer", "Current" and "Thermal Expansion" in "Contribution Items", and select "Normal" for "Element Size" to obtain a finite element model mesh that meets the simulation requirements.
[0093] In one embodiment, the step of using the constructed multiphysics coupling model and the controlled variable method to analyze the influence of various conditions on the spring-loaded compression type IGBT includes:
[0094] A reference group is constructed using copper and aluminum to maintain the elastic-plastic properties for simulation, and a preset simulation current is set for the reference group;
[0095] Construct the heat transfer coefficients corresponding to the reference group;
[0096] Adjust the stress distribution of the clamp size of the reference group.
[0097] In practice, under default conditions, the pressure applied to the collector side surface is 1.2 * 12 = 14400 N, and the pressure distributed to each chip is 1.2 kN.
[0098] Heat is dissipated through both the collector and emitter terminals, and the heat transfer coefficient is set to 3000 W / (m2·K), with an ambient temperature of 293.15 K.
[0099] The internal components of the device are configured with corresponding contact pairs, and the coefficient of friction is set to 0.5; the active area of the chip is set as a heat source with a heat dissipation rate of 150W.
[0100] The specific simulations include:
[0101] (1) In the reference group, copper and aluminum were simulated as elasto-plastic materials; copper was simulated as a linear elastic material; aluminum was simulated as a linear elastic material; and copper-aluminum composite was simulated as a linear elastic material. The simulation results were compared and analyzed to study the influence of the material properties of the two materials on stress.
[0102] (2) Considering that the press-fit IGBT module may withstand different current levels during actual operation, and the maximum current it withstands in a short time will not exceed twice the rated current, the simulation currents are set to 80A, 250A and 400A respectively, and the effects of different current levels on stress and temperature are simulated and analyzed.
[0103] (3) In order to further quantify the impact of different power losses caused by different current magnitudes, the loss of IGBT was changed by changing the heat dissipation rate in solid heat transfer. The heat dissipation rates were set to 100W, 125W, 150W, 175W and 200W respectively, and the effects of different power loss magnitudes on stress and temperature were simulated and analyzed.
[0104] (4) Different cooling devices and their working conditions can cause great changes in heat dissipation conditions. The heat dissipation conditions of IGBT can be changed by adjusting the heat transfer coefficient value of the heat flux in solid heat transfer. The heat transfer coefficient (unit W / (m2*K)) includes 3000, 5000, 6000, 7000, and 9000. The effects of different heat dissipation conditions on stress and temperature are simulated and analyzed.
[0105] (5) During the operation of the press-fit IGBT, the main source of convective heat dissipation is at the collector and emitter cooling devices. At this time, the ambient temperature is actually the temperature of the cooling medium provided by the cooling devices. The ambient temperature of the IGBT is changed by adjusting the external temperature value in the solid heat transfer. The external ambient temperature (unit °C) includes 0, 10, 20, 30, 40, and 50. The effects of different ambient temperatures on stress and temperature are simulated and analyzed.
[0106] (6) The magnitude of the clamping pressure not only affects the stress distribution, but also changes the contact heat and resistance, thus affecting the temperature distribution. The external pressure of the IGBT is changed by adjusting the boundary load values in solid mechanics. The boundary loads (in N) include 5760, 10080, 14400, 18720, and 22800. The effects of different boundary loads on stress and temperature are simulated and analyzed.
[0107] (7) Surface roughness parameters can affect stress distribution, contact heat, and electrical resistance. The analysis includes: average surface roughness height (in μm) with four values: 0.1, 0.3, 0.5, and 0.7; average surface roughness slope (in μm) with four values: 0.02, 0.03, 0.05, and 0.07; and microhardness (in GPa) with four values: 1, 3, 5, and 7. Simulation analysis is used to analyze the effects of different surface roughness parameters on stress and temperature.
[0108] In one implementation, step S204 includes:
[0109] The base is made of a material that is compatible with different models of press-fit IGBT modules and is stable, reliable, and not prone to slipping or instability.
[0110] The chip is periodically switched on and off to simulate operating conditions;
[0111] Welded water cooling heads are selected to provide continuous and stable heat dissipation for IGBTs, ensuring the orderly cooling process.
[0112] A power cycling test was conducted to simulate the aging process.
[0113] During the implementation, the experimental platform was constructed using a base made of a stable and reliable material that could accommodate different models of press-fit IGBT modules and was not prone to slippage or instability. The pressure plate material had moderate elasticity, ensuring that pressure was applied evenly without deformation during the pressing process, thus guaranteeing uniform pressing between the IGBT module and the heat sink. The experimental circuit was built using the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply. The chip was periodically switched on and off to simulate operating conditions. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip. Thermally conductive adhesive was used to fix a K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0114] Power cycling tests were conducted to simulate the aging process, with the tests corresponding to the simulations. The main variables studied included chip current, water cooling system flow rate (corresponding to changes in the heat dissipation coefficient in the simulation), ambient temperature, and the pressure applied by the fixture. The collected data primarily included device junction temperature, on-state voltage drop, gate voltage, and collector current. These data were compared with the simulation data for mutual verification.
[0115] In one embodiment, step S206 includes:
[0116] Obtain junction temperature and strain fluctuation data from multiphysics simulations, as well as the basic parameters corresponding to the fluctuation data;
[0117] The simulation data is divided into experimental group dataset and test group dataset. The experimental group dataset is further divided into training set and validation set according to a specific ratio. The fluctuation data is preprocessed to obtain feature vector values and corresponding feature values.
[0118] Construct a network framework that includes an input layer, hidden layers, and an output layer;
[0119] Using basic parameters as input information and corresponding feature vector values as output information, the deep learning network model is trained using training and validation sets to form a deep learning multiphysics simulation model trained on finite element simulation data.
[0120] The processed test dataset was then fed into a data-driven deep learning multiphysics simulation model to verify the accuracy and effectiveness of the network model.
[0121] The deep learning multiphysics simulation model here can quickly obtain the eigenvector values of junction temperature and strain of the device based on the input basic parameters. By inverse processing these values, the simulation fluctuation data of junction temperature and strain can be obtained, realizing the rapid calculation of three-dimensional multi-field information by deep learning to replace finite element simulation.
[0122] In implementation, a deep learning multiphysics simulation model is obtained by training deep learning algorithms using simulation data. The model is then trained to possess associative memory and reasoning capabilities through network training, specifically including the following key steps:
[0123] (1) Obtain junction temperature and strain fluctuation data from multiphysics simulation, as well as the basic parameters corresponding to the fluctuation data, including IGBT device parameters, converter valve design parameters, and operating conditions. More specifically, the basic parameters include, but are not limited to: surface roughness parameters, friction coefficient, gap thermal conductivity, chip current, chip heat dissipation rate, boundary load, external temperature, and heat sink heat transfer coefficient;
[0124] (2) Divide the simulation data into experimental group dataset and test group dataset, and then divide the experimental group dataset into training set and validation set according to a specific ratio. Preprocess the fluctuation data to obtain feature vector values and corresponding feature values.
[0125] (3) Construct a network framework that includes an input layer, a hidden layer and an output layer. Considering the limited input and output parameters, choose one input layer, one hidden layer and one output layer. The number of nodes in the input layer is the same as the number of basic parameters of the compression IGBT that serve as input information. The output layer outputs the feature vector values of junction temperature and strain with two nodes respectively. A deep learning network model is initially built.
[0126] (4) Using basic parameters as input information and corresponding feature vector values as output information, the deep learning network model is trained using training and validation sets. The training loss function and validation loss function of the network are used as validation indicators. First, the number of candidate hidden layer nodes in the model is determined by empirical formulas, and then the parameters are tuned by grid search. The mean square error is selected as the loss function. Based on the simple cross-validation method, the optimal number of hidden layer nodes and the optimal number of training iterations are determined by the trend of the graph of training loss and validation loss, thus forming a deep learning multiphysics simulation model trained with finite element simulation data.
[0127] (5) Substitute the processed test group dataset into the data-driven deep learning multiphysics simulation model to verify the accuracy and effectiveness of its network model.
[0128] The deep learning multiphysics simulation model here can quickly obtain the eigenvector values of junction temperature and strain of a device based on the input basic parameters. Inverse processing of these eigenvectors yields the simulated fluctuation data of junction temperature and strain, enabling rapid computation of three-dimensional multi-field information using deep learning to replace finite element simulation. The deep learning algorithm structure is as follows: Figure 3 As shown, the parameters of the IGBT device, the design parameters of the converter valve, and the parameters of the operating conditions are input into the deep learning multiphysics simulation model for calculation and processing, thereby obtaining the simulated fluctuation data of junction temperature and strain.
[0129] A multi-source data fusion-driven life assessment model for press-fit IGBTs was constructed. For the simulation data, after processing with the rainflow counting method, a nonlinear fatigue cumulative damage life model was adopted as shown in equations (8) and (9):
[0130] m2=α1·ΔT j +α2·T jm +α3·Δε+α4·σ m +A (8);
[0131]
[0132] In the formula, m2 is the degradation index; ΔT j Junction temperature fluctuation, unit: K; T jm The highest junction temperature is expressed in Kelvin (K); Δε is the strain range; σ m It is the average stress, in N; N is the cycle period. f The number of failures is calculated using the Coffin-Manson model; D represents the cumulative fatigue damage variable; α1~α4 and A are the model fitting coefficients. The calculation process for nonlinear cumulative fatigue damage life is as follows: Figure 4 As shown, the fluctuating data such as junction temperature and strain are sequentially processed through a rainflow statistical algorithm, a rainflow statistical spectrum, a lifetime calculation model, and a damage accumulation model to obtain the expected lifetime.
[0133] For the measured data obtained from the experimental platform, the sudden and sharp increase in saturation voltage drop was used as the criterion for ending the test. The lifetime prediction model of the press-fit IGBT chip was adopted as shown in equation (10):
[0134]
[0135] Where c represents the material parameter, a and b represent the model parameters, both of which are obtained by fitting power cycle test data, and P represents the pressure per unit area in MPa.
[0136] Simulated lifetime models, which consider the nonlinear accumulation of damage, theoretically offer higher accuracy. However, being based on simulations, they cannot comprehensively and thoroughly account for the actual operation of IGBTs. Measured data, on the other hand, is more accurate and realistic, and the lifetime models constructed from it more closely reflect the operating conditions. However, these are linear models built under high loads and fail to consider the nonlinear accumulation of damage, leading to discrepancies between lifetime calculations and actual conditions. These errors are particularly pronounced under low-load conditions with relatively low damage values. Therefore, data fusion of the two lifetime assessment results yields a weighted lifetime prediction model based on multi-source data, as follows:
[0137] L = (1 - D) 2 L1+(2D-D 2 L2 (11);
[0138] Where L represents the lifetime result from data fusion, and L1 and L2 represent the simulated nonlinear lifetime result and the measured linear lifetime result, respectively, in years. A weighting function constructed using damage values ensures the accuracy of low-load predictions by primarily using the nonlinear fatigue cumulative damage lifetime model when damage values are low, while using a more closely aligned phenomenological lifetime prediction model when damage values are high. This results in a multi-source data fusion-driven method for assessing the lifetime of press-fit IGBTs, comprehensively considering both simulated and measured data.
[0139] The life assessment model for press-fit IGBTs driven by rainflow counting and multi-source data fusion is converted into program code in the form of mathematical modeling. Optimization is carried out to ensure that the program runs correctly and smoothly. The finished product is then programmed into MATLAB to obtain an automatic life calculation platform for press-fit IGBTs, which can automatically and quickly complete the life assessment after receiving simulation data.
[0140] By utilizing the LiveLink for MATLAB module to perform secondary development on deep learning simulation models and MATLAB, users can automatically control and run deep learning models through MATLAB, achieving interconnection and interoperability of data and results between the two, realizing full coupling between the software, and thus obtaining a multi-functional integrated software that encompasses the entire lifespan calculation process.
[0141] A GUI design was implemented, placing the program's execution code within a black box. Aside from aesthetic design, the user interface only provides a clear and intuitive interface for inputting basic simulation parameters and displaying results and images. This allows users, without understanding the simulation or lifespan assessment principles, to input basic parameters into the GUI software. The integrated system then automatically drives the deep learning simulation model to correct parameters and quickly calculate temperature and junction temperature fluctuations. The results are displayed in text / image format on the software interface and combined with simultaneously retrieved real-time device operating data in the background to perform lifespan calculations. Due to MATLAB's powerful computing capabilities, provided the program code is accurate and concise, the integrated software should be able to obtain various lifespan results almost simultaneously and display them in the software's output interface, exhibiting extremely fast computation speed and strong field applicability. This results in a convenient and fast lifespan assessment software for press-fit IGBTs integrated into a GUI interface.
[0142] Based on the same inventive concept, this application also provides a life assessment device for press-fit IGBTs for implementing the life assessment method of press-fit IGBTs described above.
[0143] In one embodiment, such as Figure 5 As shown, a life assessment device 50 for press-fit IGBTs is provided, comprising:
[0144] The model building module 52 is used to build a multi-physics coupling simulation model of the press-fit IGBT, and to simulate the press-fit IGBT under different conditions using the control variable method based on the multi-physics coupling simulation model.
[0145] A three-dimensional model was constructed in the finite element analysis software COMSOL, referencing the selected 5SNA 3000K452300 spring-loaded IGBT. To reduce computation time and workload while maintaining accuracy, the device model was simplified as follows:
[0146] ① Take a single sub-module for modeling, which includes 8 IGBT chips and 4 FRD chips, with a voltage level of 4500V;
[0147] ② Ignore other components that have little impact on the pressure distribution on the chip surface inside the device, such as: plastic frame, drive connection board, etc.
[0148] ③ Ignore fine details such as chamfers and rounded corners inside the module;
[0149] ④ Ignore the power loss of other structures and contact surfaces outside the active region of the chip.
[0150] Cyclic test module 54 is used to construct a test platform to perform power cyclic tests on the simulated press-fit IGBT.
[0151] During the construction of the experimental platform, the base was made of a material that could adapt to different models of press-fit IGBT modules and was stable, reliable, and not prone to slippage or instability. The elasticity of the pressure plate material should be moderate to ensure that pressure is applied evenly without deformation during the pressing process, thus ensuring uniform pressing between the IGBT module and the heat sink. The experimental circuit was built with the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply to simulate the operating conditions by periodically switching the chip on and off. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip, and thermally conductive adhesive was used to fix the K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0152] The model training module 56 is used to train the multiphysics coupling simulation model based on a data-driven deep learning algorithm, and to construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0153] This involves training a deep learning multiphysics simulation model using simulation data to drive deep learning algorithm training. The simulation data is divided into experimental and test datasets. The experimental dataset is further divided into training and validation sets according to a specific ratio. Fluctuation data is preprocessed to obtain feature vector values and their corresponding feature values. A network framework comprising an input layer, hidden layers, and an output layer is constructed. Basic parameters are used as input information, and the corresponding feature vector values are used as output information. The deep learning network model is trained using the training and validation sets, resulting in a deep learning multiphysics simulation model trained on finite element simulation data.
[0154] The life assessment module 58 is used to assess the life of the target press-fit IGBT using the press-fit IGBT life assessment model.
[0155] A deep learning algorithm is trained using simulation data to obtain a multiphysics simulation model, enabling rapid calculation of three-dimensional multi-field information. A nonlinear lifetime model considering simulation results is fused with lifetime results considering measured data to obtain a lifetime assessment model driven by multi-source data fusion. The model and its converted program code are compiled into MATLAB. The LiveLink for MATLAB module is used to implement the deep learning multiphysics simulation model and secondary development of MATLAB. A simple, intuitive, and fast-computing pressure-fit IGBT lifetime assessment software is designed and constructed using a GUI, achieving integrated fatigue life calculation for equipment. Users only need to input basic parameters such as IGBT device parameters, converter valve design parameters, and operating conditions to quickly obtain the expected life of the device, providing a reference for field equipment maintenance and ensuring the safe operation of the power system.
[0156] Each module in the aforementioned life assessment device for press-fit IGBTs can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.
[0157] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores life assessment data for press-fit IGBTs. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a life assessment method for press-fit IGBTs.
[0158] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0159] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0160] Step 202: Build a multiphysics coupling simulation model of the press-fit IGBT, and use the control variable method to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0161] Specifically, a three-dimensional model was constructed in the finite element analysis software COMSOL component, referring to the selected 5SNA 3000K452300 spring-loaded IGBT entity.
[0162] Step 204: Construct a test platform and conduct power cycling tests on the simulated press-fit IGBT.
[0163] During the construction of the experimental platform, the base was made of a material that could adapt to different models of press-fit IGBT modules and was stable, reliable, and not prone to slippage or instability. The elasticity of the pressure plate material should be moderate to ensure that pressure is applied evenly without deformation during the pressing process, thus ensuring uniform pressing between the IGBT module and the heat sink. The experimental circuit was built with the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply to simulate the operating conditions by periodically switching the chip on and off. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip, and thermally conductive adhesive was used to fix the K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0164] Step 206: Train the multiphysics coupling simulation model using a data-driven deep learning algorithm, and construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0165] This involves training a deep learning multiphysics simulation model using simulation data to drive deep learning algorithm training. The simulation data is divided into experimental and test datasets. The experimental dataset is further divided into training and validation sets according to a specific ratio. Fluctuation data is preprocessed to obtain feature vector values and their corresponding feature values. A network framework comprising an input layer, hidden layers, and an output layer is constructed. Basic parameters are used as input information, and the corresponding feature vector values are used as output information. The deep learning network model is trained using the training and validation sets, resulting in a deep learning multiphysics simulation model trained on finite element simulation data.
[0166] Step 208: Use the aforementioned press-fit IGBT life assessment model to assess the life of the target press-fit IGBT.
[0167] A deep learning algorithm is trained using simulation data to obtain a multiphysics simulation model, enabling rapid calculation of three-dimensional multi-field information. A nonlinear lifetime model considering simulation results is fused with lifetime results considering measured data to obtain a lifetime assessment model driven by multi-source data fusion. The model and its converted program code are compiled into MATLAB. The LiveLink for MATLAB module is used to implement the deep learning multiphysics simulation model and secondary development of MATLAB. A simple, intuitive, and fast-computing pressure-fit IGBT lifetime assessment software is designed and constructed using a GUI, achieving integrated fatigue life calculation for equipment. Users only need to input basic parameters such as IGBT device parameters, converter valve design parameters, and operating conditions to quickly obtain the expected life of the device, providing a reference for field equipment maintenance and ensuring the safe operation of the power system.
[0168] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0169] Step 202: Build a multiphysics coupling simulation model of the press-fit IGBT, and use the control variable method to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0170] Specifically, a three-dimensional model was constructed in the finite element analysis software COMSOL component, referring to the selected 5SNA 3000K452300 spring-loaded IGBT entity.
[0171] Step 204: Construct a test platform and conduct power cycling tests on the simulated press-fit IGBT.
[0172] During the construction of the experimental platform, the base was made of a material that could adapt to different models of press-fit IGBT modules and was stable, reliable, and not prone to slippage or instability. The elasticity of the pressure plate material should be moderate to ensure that pressure is applied evenly without deformation during the pressing process, thus ensuring uniform pressing between the IGBT module and the heat sink. The experimental circuit was built with the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply to simulate the operating conditions by periodically switching the chip on and off. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip, and thermally conductive adhesive was used to fix the K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0173] Step 206: Train the multiphysics coupling simulation model using a data-driven deep learning algorithm, and construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0174] This involves training a deep learning multiphysics simulation model using simulation data to drive deep learning algorithm training. The simulation data is divided into experimental and test datasets. The experimental dataset is further divided into training and validation sets according to a specific ratio. Fluctuation data is preprocessed to obtain feature vector values and their corresponding feature values. A network framework comprising an input layer, hidden layers, and an output layer is constructed. Basic parameters are used as input information, and the corresponding feature vector values are used as output information. The deep learning network model is trained using the training and validation sets, resulting in a deep learning multiphysics simulation model trained on finite element simulation data.
[0175] Step 208: Use the aforementioned press-fit IGBT life assessment model to assess the life of the target press-fit IGBT.
[0176] A deep learning algorithm is trained using simulation data to obtain a multiphysics simulation model, enabling rapid calculation of three-dimensional multi-field information. A nonlinear lifetime model considering simulation results is fused with lifetime results considering measured data to obtain a lifetime assessment model driven by multi-source data fusion. The model and its converted program code are compiled into MATLAB. The LiveLink for MATLAB module is used to implement the deep learning multiphysics simulation model and secondary development of MATLAB. A simple, intuitive, and fast-computing pressure-fit IGBT lifetime assessment software is designed and constructed using a GUI, achieving integrated fatigue life calculation for equipment. Users only need to input basic parameters such as IGBT device parameters, converter valve design parameters, and operating conditions to quickly obtain the expected life of the device, providing a reference for field equipment maintenance and ensuring the safe operation of the power system.
[0177] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0178] Step 202: Build a multiphysics coupling simulation model of the press-fit IGBT, and use the control variable method to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model.
[0179] Specifically, a three-dimensional model was constructed in the finite element analysis software COMSOL component, referring to the selected 5SNA 3000K452300 spring-loaded IGBT entity.
[0180] Step 204: Construct a test platform and conduct power cycling tests on the simulated press-fit IGBT.
[0181] During the construction of the experimental platform, the base was made of a material that could adapt to different models of press-fit IGBT modules and was stable, reliable, and not prone to slippage or instability. The elasticity of the pressure plate material should be moderate to ensure that pressure is applied evenly without deformation during the pressing process, thus ensuring uniform pressing between the IGBT module and the heat sink. The experimental circuit was built with the IGBT module under test, the drive circuit, the acquisition circuit, and the programmable DC power supply to simulate the operating conditions by periodically switching the chip on and off. An oscilloscope was used to monitor parameters such as the output current, control signal, and voltage of the IGBT module. A highly reliable welded water-cooling head was selected to provide continuous and stable heat dissipation for the IGBT, ensuring the orderly cooling process. A FLIR SC7000 infrared imager was used to measure and record the surface temperature of the IGBT chip, and thermally conductive adhesive was used to fix the K-type thermocouple to the edges of the upper and lower copper electrodes to measure the case temperature.
[0182] Step 206: Train the multiphysics coupling simulation model using a data-driven deep learning algorithm, and construct a press-fit IGBT life assessment model based on the trained multiphysics coupling simulation model.
[0183] This involves training a deep learning multiphysics simulation model using simulation data to drive deep learning algorithm training. The simulation data is divided into experimental and test datasets. The experimental dataset is further divided into training and validation sets according to a specific ratio. Fluctuation data is preprocessed to obtain feature vector values and their corresponding feature values. A network framework comprising an input layer, hidden layers, and an output layer is constructed. Basic parameters are used as input information, and the corresponding feature vector values are used as output information. The deep learning network model is trained using the training and validation sets, resulting in a deep learning multiphysics simulation model trained on finite element simulation data.
[0184] Step 208: Use the aforementioned press-fit IGBT life assessment model to assess the life of the target press-fit IGBT.
[0185] A deep learning algorithm is trained using simulation data to obtain a multiphysics simulation model, enabling rapid calculation of three-dimensional multi-field information. A nonlinear lifetime model considering simulation results is fused with lifetime results considering measured data to obtain a lifetime assessment model driven by multi-source data fusion. The model and its converted program code are compiled into MATLAB. The LiveLink for MATLAB module is used to implement the deep learning multiphysics simulation model and secondary development of MATLAB. A simple, intuitive, and fast-computing pressure-fit IGBT lifetime assessment software is designed and constructed using a GUI, achieving integrated fatigue life calculation for equipment. Users only need to input basic parameters such as IGBT device parameters, converter valve design parameters, and operating conditions to quickly obtain the expected life of the device, providing a reference for field equipment maintenance and ensuring the safe operation of the power system.
[0186] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0187] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0188] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for evaluating the lifespan of a press-fit IGBT, characterized in that, The life assessment method includes: A multiphysics coupling simulation model of a press-fit IGBT was constructed, and the control variable method was used to simulate the press-fit IGBT under different conditions based on the multiphysics coupling simulation model. A test platform was constructed to conduct power cycling tests on the simulated press-fit IGBT. The multiphysics coupling simulation model is trained using a data-driven deep learning algorithm, and a life assessment model for press-fit IGBTs is constructed based on the trained multiphysics coupling simulation model. The deep learning multiphysics coupling simulation model is configured to output eigenvector values of junction temperature and strain based on the input basic parameters. The lifespan of the target press-fit IGBT is assessed using the aforementioned press-fit IGBT lifespan assessment model. The life assessment model for the press-fit IGBT is: L = (1-D) 2 L1+(2D-D 2 L2; L represents the lifetime result from data fusion, L1 represents the simulated nonlinear lifetime result, and L2 represents the measured linear lifetime result; D represents the fatigue cumulative damage variable. The simulated nonlinear lifetime result is obtained by calculating the simulated fluctuation data of junction temperature and strain using a nonlinear fatigue cumulative damage lifetime model; the simulated fluctuation data of junction temperature and strain is obtained by inverse processing the eigenvector values of junction temperature and strain. The nonlinear fatigue cumulative damage life model is as follows: , ; m2 is the degradation index; ΔT j For junction temperature fluctuations; T jm The highest junction temperature; Δε is the strain range; σ m It is the average stress; N is the cycle period, N f The number of failures is calculated using the Coffin-Manson model; α1~α4 and A are the model fitting coefficients. The measured linear lifetime result was calculated using the measured data obtained from the experimental platform, with a sudden and sharp increase in saturation voltage drop as the criterion for ending the test, and using a lifetime prediction model for the pressure-fit IGBT chip. The lifetime prediction model for the press-fit IGBT chip is as follows: ; c represents the material parameters, a and b represent the model parameters, both of which are obtained by fitting power cycle test data, and P represents the pressure per unit area.
2. The life assessment method for press-fit IGBTs according to claim 1, characterized in that, The multiphysics coupling simulation model for the press-fit IGBT is constructed, and based on this model, the controlled variable method is used to simulate the press-fit IGBT under different conditions, including: A three-dimensional model of a press-fit IGBT entity is constructed in a COMSOL component, and the three-dimensional model is then simplified. The influence of various conditions on spring-loaded compression IGBTs was analyzed using the established multiphysics coupling model and the controlled variable method.
3. The life assessment method for press-fit IGBTs according to claim 2, characterized in that, The simplification process of the three-dimensional model includes: Set the parameters in the three-dimensional model, set the electric field boundary conditions, set the temperature field boundary conditions, and set the stress field boundary conditions.
4. The life assessment method for press-fit IGBTs according to claim 2, characterized in that, The analysis of the influence of various conditions on spring-loaded compression IGBTs using the established multiphysics coupling model and the controlled variable method includes: A reference group is constructed using copper and aluminum to maintain the elastic-plastic properties for simulation, and a preset simulation current is set for the reference group; Construct the heat transfer coefficients corresponding to the reference group; Adjust the stress distribution of the clamp size of the reference group.
5. The life assessment method for press-fit IGBTs according to claim 1, characterized in that, The aforementioned test platform is constructed to conduct power cycling tests on the simulated press-fit IGBT, including: The base is made of a material that is compatible with different models of press-fit IGBT modules and is stable, reliable, and not prone to slipping or instability. The chip is periodically switched on and off to simulate operating conditions; Welded water cooling heads are selected to provide continuous and stable heat dissipation for IGBTs, ensuring the orderly cooling process. A power cycling test was conducted to simulate the aging process.
6. The life assessment method for press-fit IGBTs according to claim 1, characterized in that, The step of training the multiphysics coupling simulation model using a data-driven deep learning algorithm, and then constructing a life assessment model for press-fit IGBTs based on the trained multiphysics coupling simulation model, includes: Obtain junction temperature and strain fluctuation data from multiphysics simulations, as well as the basic parameters corresponding to the fluctuation data; The simulation data is divided into experimental group dataset and test group dataset. The experimental group dataset is further divided into training set and validation set according to the ratio. The fluctuation data is preprocessed to obtain feature vector values and corresponding feature values. Construct a network framework that includes an input layer, hidden layers, and an output layer; Using basic parameters as input information and corresponding feature vector values as output information, the deep learning network model is trained using training and validation sets to form a deep learning multiphysics simulation model trained on finite element simulation data. The processed test dataset was then fed into a data-driven deep learning multiphysics simulation model to verify the accuracy and effectiveness of the network model.
7. A life assessment device for press-fit IGBTs, characterized in that, The device includes: The model building module is used to build a multiphysics coupling simulation model of the press-fit IGBT. Based on the multiphysics coupling simulation model, the control variable method is used to simulate the press-fit IGBT under different conditions. The cyclic testing module is used to construct a test platform to conduct power cyclic tests on the simulated press-fit IGBT. The model training module is used to train the multiphysics coupling simulation model based on a data-driven deep learning algorithm, and to construct a life assessment model for press-fit IGBTs based on the trained multiphysics coupling simulation model; the deep learning multiphysics coupling simulation model is configured to output feature vector values of junction temperature and strain according to the input basic parameters. The life assessment module is used to assess the life of the target press-fit IGBT using the press-fit IGBT life assessment model. The life assessment model for the press-fit IGBT is: L = (1-D) 2 L1+(2D-D 2 L2; L represents the lifetime result from data fusion, L1 represents the simulated nonlinear lifetime result, and L2 represents the measured linear lifetime result; D represents the fatigue cumulative damage variable. The life assessment module is also used to calculate the nonlinear fatigue cumulative damage life model using the simulation fluctuation data of junction temperature and strain to obtain the nonlinear life result of the simulation. The nonlinear fatigue cumulative damage life model is as follows: , ; m2 is the degradation index; ΔT j For junction temperature fluctuations; T jm The highest junction temperature; Δε is the strain range; σ m It is the average stress; N is the cycle period, N f The number of failures is calculated using the Coffin-Manson model; α1~α4 and A are the model fitting coefficients. The model training module is also used to perform inverse processing on the feature vector values of the junction temperature and strain to obtain the simulated fluctuation data of the junction temperature and strain. The lifetime assessment module is also used to calculate the measured linear lifetime results by using the lifetime prediction model of the press-fit IGBT chip, with the sudden and sharp increase in saturation voltage drop as the criterion for ending the test, based on the measured data obtained from the experimental platform. The lifetime prediction model for the press-fit IGBT chip is as follows: ; c represents the material parameters, a and b represent the model parameters, both of which are obtained by fitting power cycle test data, and P represents the pressure per unit area.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
