A method and system for evaluating the health status of a power device

Abstract

The present invention discloses a method and system for evaluating the health status of a power device. The method takes the health status of a power device as the research object, adopts simulation to generate data to solve the problem of sample shortage, and adopts a neural network algorithm to solve the problem of rapid and accurate status evaluation under complex working conditions. By using the trained neural network, a conclusion on the health status of the power device can be quickly given based on the measured data of the device, and the degradation degree and degradation position of the device under actual variable working conditions can also be analyzed. The method solves the problems of difficulty in preparing samples for internal structure degradation of traditional test samples and small data samples in health status evaluation, and provides a quantitative evaluation method for objectively evaluating the health status of power devices.

Description

Technical Field

[0001] The present invention belongs to the technical field of power device health status assessment, and in particular relates to a power device health status assessment method and system. Background Art

[0002] Power devices are widely used in industry, military, aerospace and other fields. They play important roles such as power conversion and are the core components of power electronic power circuits. In important fields such as aerospace, the reliability of power devices has a very large impact on the reliability of the system in which they are located. Therefore, these important fields have higher requirements for the reliability of power devices. The reliability and health status evaluation methods of power devices mainly include reliability prediction method, failure physics method and data-driven method. Among them, the data-driven method can characterize the degree of degradation of the device through its measurable characteristic parameters. However, in the actual process of device evaluation, there are problems such as difficulty in preparing degradation test samples and lack of data. Summary of the invention

[0003] The technology of the present invention solves the problem: Overcoming the shortcomings of the prior art, providing a method and system for evaluating the health status of a power device, generating electric and thermal characteristic parameter data of the power device under different working conditions and different degrees of degradation through a simulation method, extracting the data as a data set for neural network training, and using it for device health status evaluation.

[0004] In order to solve the above technical problems, the present invention discloses a method for evaluating the health status of a power device, comprising:

[0005] Step 1, analyzing the failure location and failure mode of the power device to determine the change trend between the characteristic parameters of the power device and the structural degradation;

[0006] Step 2: Based on the finite element analysis method, multi-physics field coupling simulation of the power device is carried out to obtain the benchmark evaluation data of the power device in a healthy state;

[0007] Step 3: According to the variation trend between the characteristic parameters of the power device and the structural degradation, different structural degradations are injected into the finite element analysis method, and the multi-physics field coupling simulation of the power device is traversed to obtain the benchmark evaluation data of the power device in different degradation states;

[0008] Step 4, using an improved generalized regression neural network algorithm for training based on the obtained benchmark evaluation data of healthy power devices and the benchmark evaluation data of power devices in different degradation states to obtain a health status evaluation model;

[0009] Step 5: Evaluate the health status of the power device according to the health status evaluation model.

[0010] In the above power device health status assessment method, the failure location and failure mode of the power device are analyzed to determine the change trend between the characteristic parameters of the power device and the structural degradation, including:

[0011] Sub-step 11, analyzing the weak packaging structure of the power device, including: solder layer structure and bonding wires;

[0012] Sub-step 12, determining degradation locations and related stresses corresponding to each degradation location based on the analysis results of the weak packaging structure;

[0013] Sub-step 13, analyzing and obtaining the variation trend between the characteristic parameters of the power device and the structural degradation according to the determined degradation positions and the relevant stresses corresponding to the degradation positions.

[0014] In the above power device health status assessment method, the characteristic parameters of the power device include: port voltage drop and device junction-case thermal resistance.

[0015] In the above power device health status assessment method, based on the finite element analysis method, multi-physics field coupling simulation of the power device is carried out to obtain the benchmark assessment data of the healthy power device, including:

[0016] Sub-step 21, constructing a three-dimensional structural model of the power device in a multi-physics field finite element analysis software, and assigning corresponding material properties to the three-dimensional structural model according to the materials of various parts of the power device;

[0017] Sub-step 21, based on the constructed three-dimensional structural model with material properties, conduct electrothermal coupling simulation, and use the results of the electrothermal coupling simulation as benchmark evaluation data for healthy power devices; wherein the physical field includes: electric field and temperature field.

[0018] In the above power device health status assessment method, an electrothermal coupling simulation is carried out based on a constructed three-dimensional structural model with material properties, and the results obtained by the electrothermal coupling simulation are used as benchmark assessment data for healthy power devices, including:

[0019] Sub-step 211, setting a coupling mode: the electric field and the temperature field are coupled via the Joule heating effect and the material temperature characteristics;

[0020] Sub-step 212, setting temperature-related parameters: for material properties that vary with temperature, setting temperature-related parameters in the multi-physics finite element analysis software;

[0021] Sub-step 213, setting boundary conditions: boundary conditions are set according to actual working conditions and application environment;

[0022] Sub-step 214, performing an electrothermal coupling simulation based on the constructed three-dimensional structure model with material properties according to the set coupling mode, temperature-related parameters and boundary conditions, to obtain the temperature distribution and electrical parameter distribution of the power device;

[0023] Sub-step 215, changing the actual working conditions and application environment, setting different boundary conditions, repeating the electrothermal coupling simulation process, and obtaining the temperature distribution and electrical parameter distribution of the power device under different working conditions;

[0024] Sub-step 216, extracting the chip junction temperature, case temperature and port voltage under different operating conditions from the temperature distribution and electrical parameter distribution of the power device under different operating conditions as the benchmark evaluation data of the healthy power device.

[0025] In the above power device health status assessment method, according to the change trend between the characteristic parameters and structural degradation of the power device, different structural degradations are injected into the finite element analysis method, and the multi-physics field coupling simulation of the power device is traversed to obtain the benchmark assessment data of the power devices in different degradation states, including:

[0026] Sub-step 31, analyzing the main degradation modes and degradation positions of the device under different working conditions according to the variation trends between the characteristic parameters of the power device and the structural degradation, and injecting different degrees of structural degradation into the finite element analysis method;

[0027] Sub-step 32, repeating sub-steps 211 to 216, extracting different degrees of structural degradation, chip junction temperature, case temperature and port voltage under different operating conditions as benchmark evaluation data for power devices in different degradation states.

[0028] In the above power device health status assessment method, based on the obtained benchmark assessment data of healthy power devices and the benchmark assessment data of power devices in different degradation states, an improved generalized regression neural network algorithm is used for training to obtain a health status assessment model, including:

[0029] Sub-step 41, constructing a model training set based on the obtained benchmark evaluation data of power devices in healthy states and the benchmark evaluation data of power devices in different degradation states;

[0030] Sub-step 42, based on the constructed model training set, adopting an improved generalized regression neural network algorithm to train the neural network model and determine each weight coefficient of the neural network model;

[0031] Sub-step 43, assigning each weight coefficient of the determined neural network model to the neural network model to obtain a health status assessment model.

[0032] Correspondingly, the present invention also discloses a power device health status assessment system, comprising:

[0033] An analysis module is used to analyze the failure location and failure mode of the power device and determine the change trend between the characteristic parameters of the power device and the structural degradation;

[0034] The first finite element analysis module is used to carry out multi-physics field coupling simulation of power devices based on the finite element analysis method to obtain benchmark evaluation data of power devices in a healthy state;

[0035] The second finite element analysis module is used to inject different structural degradations into the finite element analysis method according to the change trend between the characteristic parameters of the power device and the structural degradation, traverse the multi-physics field coupling simulation of the power device, and obtain the benchmark evaluation data of the power device in different degradation states;

[0036] A model training module is used to train the obtained benchmark evaluation data of power devices in healthy state and the benchmark evaluation data of power devices in different degradation states using an improved generalized regression neural network algorithm to obtain a health state evaluation model;

[0037] The evaluation module is used to evaluate the health status of the power device according to the health status evaluation model.

[0038] The present invention has the following advantages:

[0039] In view of the problems of difficulty in preparing samples due to internal structure degradation of traditional test samples and small data sample size in health status assessment, the present invention proposes a power device health status assessment system based on simulation and neural network, and establishes a domestically developed aerospace power component selection optimization method by combining various types of information on power components such as function and electrical parameters, quality and reliability, radiation resistance, anti-static ability, design capability, manufacturing process, and process control capability. This method provides a complete and quantitative method support for the rapid positioning and improvement of weak links in power components, and for the selection optimization of domestically developed aerospace power components for aerospace models. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flowchart of the steps of a method for evaluating the health status of a power device in an embodiment of the present invention;

[0041] Figure 2 It is a schematic diagram of a structural geometric model of an IGBT power module device in an embodiment of the present invention;

[0042] Figure 3 is a schematic diagram of a welding layer plane with different stratification degrees in an embodiment of the present invention;

[0043] Figure 4is a flow chart of an adaptive particle swarm optimization generalized neural network algorithm in an embodiment of the present invention;

[0044] Figure 5 It is a schematic diagram of the structural degradation degree evaluation result of an IGBT power module device in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments disclosed in the present invention will be further described in detail below with reference to the accompanying drawings.

[0046] like Figure 1 In this embodiment, the power device health status assessment method includes:

[0047] Step 1: Analyze the failure location and failure mode of the power device to determine the change trend between the characteristic parameters of the power device and the structural degradation.

[0048] In this embodiment, the failure location and failure mode of the power device can be analyzed to clarify the variation trend between the characteristic parameters of the power device and the structural degradation:

[0049] Sub-step 11, analyzing the structure of the power device, the structural objects to be analyzed include but are not limited to: bonding wires, chips, solder layers, packaging structures, etc. of the power device; based on the structural analysis results of the power device, the weak packaging structures of the power device (such as weak packaging structures such as solder layer structures and bonding wires) are identified, and the failure modes and failure mechanisms of the weak packaging structures under different working conditions are determined.

[0050] Sub-step 12, based on the analysis results of the weak packaging structure, that is, determining the failure mode and failure mechanism of the weak packaging structure under different working conditions, determining the degradation position, and the relevant stress corresponding to each degradation position, and then obtaining the change trend between the characteristic parameters of the power device (such as port voltage drop, device junction-case thermal resistance, etc.) and the structural degradation.

[0051] Step 2: Based on the finite element analysis method, multi-physics field coupling simulation of power devices is carried out to obtain benchmark evaluation data of power devices in a healthy state.

[0052] In this embodiment, a finite element analysis method is used to carry out multi-physics field coupling simulation, and parameter data such as port voltage and current, shell temperature, and junction temperature under different working conditions are extracted:

[0053] Sub-step 21, in the multi-physics field finite element analysis software, construct a three-dimensional structural model of the power device according to the geometric structure of the power device, and assign corresponding material properties to the three-dimensional structural model according to the materials of various parts of the power device.

[0054] Sub-step 21, based on the constructed three-dimensional structural model with material properties, conduct electrothermal coupling simulation, and use the results of the electrothermal coupling simulation as benchmark evaluation data for healthy power devices; wherein the physical field includes: electric field and temperature field.

[0055] Preferably, the specific process of the electrothermal coupling simulation is as follows:

[0056] Sub-step 211, setting the coupling mode: the electric field and the temperature field are coupled via the Joule heating effect and the material temperature characteristics.

[0057] Sub-step 212, setting temperature-related parameters: For material properties that vary with temperature, setting temperature-related parameters in the multi-physics finite element analysis software.

[0058] Sub-step 213, setting boundary conditions: boundary conditions are set according to actual working conditions and application environment. The actual working conditions refer to parameters and their ranges such as external environment temperature, convection mode, convection parameters, bottom heat dissipation mode, bottom heat dissipation parameters, applied current, control voltage, etc. determined according to the use conditions of the power device.

[0059] Sub-step 214 , according to the set coupling mode, temperature-related parameters and boundary conditions, based on the constructed three-dimensional structure model with material properties, conduct an electrothermal coupling simulation to obtain the temperature distribution and electrical parameter distribution of the power device.

[0060] Sub-step 215, changing the actual working conditions and application environment, setting different boundary conditions, repeating the electrothermal coupling simulation process, and obtaining the temperature distribution and electrical parameter distribution of the power device under different working conditions;

[0061] Sub-step 216, extracting the chip junction temperature, case temperature and port voltage under different operating conditions from the temperature distribution and electrical parameter distribution of the power device under different operating conditions as the benchmark evaluation data of the healthy power device.

[0062] Step 3: According to the changing trend between the characteristic parameters of the power device and the structural degradation, different structural degradations are injected into the finite element analysis method, and the multi-physics field coupling simulation of the power device is traversed to obtain the benchmark evaluation data of the power device in different degradation states.

[0063] In this embodiment, different structural degradations can be injected into the finite element analysis method, and the electrothermal coupling simulation is traversed to obtain parameter data such as port voltage and current, shell temperature, junction temperature, etc. of the power device under different working conditions after degradation:

[0064] Sub-step 31, according to the variation trend between the characteristic parameters of the power device and the structural degradation, analyze the main degradation modes and degradation positions of the device under different working conditions, and inject different degrees of structural degradation into the finite element analysis method.

[0065] Sub-step 32, repeating sub-steps 211 to 216, extracting different degrees of structural degradation, chip junction temperature, case temperature and port voltage under different operating conditions as benchmark evaluation data for power devices in different degradation states.

[0066] Step 4: Based on the obtained benchmark evaluation data of power devices in healthy state and the benchmark evaluation data of power devices in different degradation states, an improved generalized regression neural network algorithm is used for training to obtain a health state evaluation model.

[0067] In this embodiment, based on the data set obtained by simulation, an improved generalized regression neural network algorithm is used to train a health status assessment model:

[0068] Sub-step 41, constructing a model training set based on the obtained benchmark evaluation data of power devices in healthy state and the benchmark evaluation data of power devices in different degradation states. The obtained benchmark evaluation data of power devices in healthy state and the benchmark evaluation data of power devices in different degradation states can be divided into a training set and a test set.

[0069] Sub-step 42, based on the constructed model training set, adopts the improved generalized regression neural network algorithm to train the neural network model and determine the weight coefficients of the neural network model. During the training, the degradation index can be constructed based on the degree of structural degradation, and the generalized regression network can be optimized using particle swarm optimization.

[0070] Sub-step 43, assigning each weight coefficient of the determined neural network model to the neural network model to obtain a health status assessment model.

[0071] Step 5: Evaluate the health status of the power device according to the health status evaluation model.

[0072] In this embodiment, the degradation degree of the power device to be evaluated can be evaluated by the health status assessment model, thereby achieving an objective assessment of the health status of the power device and providing complete and quantitative technical support for the selection and optimization of the power device.

[0073] In summary, the present invention takes the health status of power devices as the research object, uses simulation to generate data to solve the problem of sample scarcity, and uses a neural network algorithm to solve the problem of rapid and accurate status evaluation under complex working conditions. By using the trained neural network, the health status conclusion of the power device can be quickly given for the measured data of the device, and the degradation degree and degradation position of the device under actual variable working conditions can also be analyzed. This solves the problems of the difficulty in preparing samples for internal structure degradation of traditional test samples and the small data sample in health status evaluation, and provides a quantitative evaluation method for objectively evaluating the health status of power devices.

[0074] Based on the above embodiment, a specific example is described below.

[0075] The health status assessment process of a certain type of power device based on finite element simulation and neural network is as follows:

[0076] Step (1), select a certain type of IGBT power module device, conduct failure mode and failure mechanism analysis, and identify weak structures. In the actual use of IGBT power module devices, the repeated alternating temperature cycle stress caused by the periodic switching of the device is the main cause of device aging. The key structures of IGBT power module devices are bonding wires and chip solder layer structures, among which the degradation of the chip solder layer structure is the main degradation mode for analysis.

[0077] Step (2), using COMSOL Multiphysics finite element analysis software to perform finite element modeling and simulation analysis on the IGBT power module device in step (1):

[0078] 21) Measure the dimensions of each part of the actual structure of the IGBT power module device and establish the structural geometric model of the IGBT power module device, such as Figure 2 shown.

[0079] 22) The various parts of the established structural geometric model are assigned corresponding material properties according to the actual materials of the IGBT power module device. Among them, the relationship between the voltage, current and temperature of the silicon chip is measured at different immersion temperatures, and the conductivity characteristic equation of the device is established by function fitting.

[0080] 23) Select physical fields and apply boundary conditions: The physical fields involved in the model analysis include electric fields and temperature fields. The boundary conditions include ambient temperature, substrate temperature, load current, etc. Boundary conditions are applied at corresponding positions according to the actual thermoelectric stresses to which the IGBT power module devices are subjected.

[0081] 24) Divide the grid and conduct simulation analysis. Divide the model into grids with reasonable density and simulate the distribution of parameters such as potential, current, and temperature.

[0082] 25) Input parameter scanning and characteristic parameter extraction. Change the ambient temperature and load current and other conditions to obtain the device port data under different working conditions.

[0083] Step (3), inject structural degradation, traverse the simulation, and obtain the port voltage and current, case temperature, junction temperature and other parameter data of the power device under different working conditions after degradation:

[0084] 31) Inject relevant degradation forms into the key structure analyzed in step (1), and take the degradation of the solder layer under the chip as an example for analysis. During the degradation process of the solder layer under the chip, the delamination caused by cracks mainly expands gradually from the periphery to the center. The schematic diagram of the remaining soldering part of the solder layer under different delamination degrees is shown in Figure 3 shown.

[0085] 32) The electrical characteristics port data and junction temperature changes of the IGBT power module device under different working conditions and degradation degrees are obtained by simulation, as shown in Table 1:

[0086]

[0087] Table 1

[0088] Among them, T a Indicates the ambient temperature, I c represents the collector current, Φ Rthja Indicates the degree of thermal resistance degradation of the device, which is calculated by dividing the change in thermal resistance by the initial thermal resistance value. The thermal resistance value is calculated by dividing the difference between the device chip junction temperature and the ambient temperature by the power consumption value. The health status level labels are 1, 2, and 3, representing low, medium, and high degradation levels, respectively.

[0089] Step (4), health status assessment. The adaptive particle swarm algorithm is used to find the optimal GRNN expansion factor, and the adaptive particle swarm optimization generalized neural network is established to avoid the inefficiency and inaccuracy of manually adjusting the expansion factor. The health status of the IGBT power module device is evaluated through a comprehensive analysis of the IGBT power module device operating information and external characteristic parameters. The established adaptive particle swarm optimization generalized neural network algorithm process is as follows: Figure 4 The GRNN algorithm with optimized expansion factor is used to predict the degradation degree of IGBT power module devices. The test set used is the remaining 45 test set sample data. The prediction value results are as follows: Figure 5 shown.

[0090] On the basis of the above embodiments, the present invention also discloses a power device health status assessment system, including: an analysis module, which is used to analyze the failure location and failure mode of the power device, and determine the change trend between the characteristic parameters of the power device and the structural degradation. A first finite element analysis module, which is used to carry out multi-physics field coupling simulation of the power device based on the finite element analysis method, and obtain the benchmark assessment data of the power device in a healthy state. A second finite element analysis module, which is used to inject different structural degradations into the finite element analysis method according to the change trend between the characteristic parameters of the power device and the structural degradation, traverse the multi-physics field coupling simulation of the power device, and obtain the benchmark assessment data of the power device in different degradation states. A model training module, which is used to train with an improved generalized regression neural network algorithm according to the obtained benchmark assessment data of the power device in a healthy state and the benchmark assessment data of the power device in different degradation states, and obtain a health status assessment model. An assessment module, which is used to assess the health status of the power device according to the health status assessment model.

[0091] As for the system embodiment, since it corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment part.

[0092] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

[0093] The contents not described in detail in the specification of the present invention belong to the common knowledge of professionals in the field.

Claims

1. A method for evaluating the health status of a power device, characterized in that: include: Step 1, analyzing the failure location and failure mode of the power device to determine the change trend between the characteristic parameters of the power device and the structural degradation; Step 2: Based on the finite element analysis method, multi-physics field coupling simulation of the power device is carried out to obtain the benchmark evaluation data of the power device in a healthy state; Step 3: According to the variation trend between the characteristic parameters of the power device and the structural degradation, different structural degradations are injected into the finite element analysis method, and the multi-physics field coupling simulation of the power device is traversed to obtain the benchmark evaluation data of the power device in different degradation states; Step 4, using an improved generalized regression neural network algorithm for training based on the obtained benchmark evaluation data of healthy power devices and the benchmark evaluation data of power devices in different degradation states to obtain a health status evaluation model; Step 5, evaluating the health status of the power device according to the health status evaluation model; Based on the finite element analysis method, multi-physics field coupling simulation of power devices is carried out to obtain benchmark evaluation data of healthy power devices, including: Sub-step 21, constructing a three-dimensional structural model of the power device in a multi-physics field finite element analysis software, and assigning corresponding material properties to the three-dimensional structural model according to the materials of various parts of the power device; Sub-step 21, based on the constructed three-dimensional structural model with material properties, conduct electrothermal coupling simulation, and use the results of the electrothermal coupling simulation as benchmark evaluation data for healthy power devices; wherein the physical field includes: electric field and temperature field; Based on the constructed three-dimensional structural model with material properties, electrothermal coupling simulation is carried out, and the results of electrothermal coupling simulation are used as benchmark evaluation data for healthy power devices, including: Sub-step 211, setting a coupling mode: the electric field and the temperature field are coupled via the Joule heating effect and the material temperature characteristics; Sub-step 212, setting temperature-related parameters: for material properties that vary with temperature, setting temperature-related parameters in the multi-physics finite element analysis software; Sub-step 213, setting boundary conditions: boundary conditions are set according to actual working conditions and application environment; Sub-step 214, performing an electrothermal coupling simulation based on the constructed three-dimensional structure model with material properties according to the set coupling mode, temperature-related parameters and boundary conditions, to obtain the temperature distribution and electrical parameter distribution of the power device; Sub-step 215, changing the actual working conditions and application environment, setting different boundary conditions, repeating the electrothermal coupling simulation process, and obtaining the temperature distribution and electrical parameter distribution of the power device under different working conditions; Sub-step 216, extracting the chip junction temperature, case temperature and port voltage under different operating conditions from the temperature distribution and electrical parameter distribution of the power device under different operating conditions as the benchmark evaluation data of the healthy power device; According to the changing trend between the characteristic parameters of power devices and structural degradation, different structural degradations are injected into the finite element analysis method, and the multi-physics field coupling simulation of power devices is traversed to obtain the benchmark evaluation data of power devices in different degradation states, including: Sub-step 31, analyzing the main degradation modes and degradation positions of the device under different working conditions according to the variation trends between the characteristic parameters of the power device and the structural degradation, and injecting different degrees of structural degradation into the finite element analysis method; Sub-step 32, repeating sub-steps 211 to 216, extracting different degrees of structural degradation, chip junction temperature, case temperature and port voltage under different operating conditions as benchmark evaluation data for power devices in different degradation states.

2. The power device health status assessment method according to claim 1, characterized in that: Analyze the failure location and failure mode of power devices to determine the change trend between the characteristic parameters of power devices and structural degradation, including: Sub-step 11, analyzing the weak packaging structure of the power device, including: solder layer structure and bonding wires; Sub-step 12, determining degradation locations and related stresses corresponding to each degradation location based on the analysis results of the weak packaging structure; Sub-step 13, analyzing and obtaining the variation trend between the characteristic parameters of the power device and the structural degradation according to the determined degradation positions and the relevant stresses corresponding to the degradation positions.

3. The method for evaluating the health status of a power device according to claim 1 or 2, characterized in that: Characteristic parameters of power devices, including: port voltage drop and device junction-case thermal resistance.

4. The method for evaluating the health status of a power device according to claim 1, characterized in that: Based on the obtained benchmark evaluation data of power devices in healthy state and the benchmark evaluation data of power devices in different degradation states, an improved generalized regression neural network algorithm is used for training to obtain a health state evaluation model, including: Sub-step 41, constructing a model training set based on the obtained benchmark evaluation data of power devices in healthy states and the benchmark evaluation data of power devices in different degradation states; Sub-step 42, based on the constructed model training set, adopting an improved generalized regression neural network algorithm to train the neural network model and determine each weight coefficient of the neural network model; Sub-step 43, assigning each weight coefficient of the determined neural network model to the neural network model to obtain a health status assessment model.

Images