Dynamic method and system for thermal fatigue assessment of integrated circuit packages

Abstract

The present invention provides a dynamic method and system for thermal fatigue assessment of integrated circuit packaging, which relates to the field of integrated circuit technology, including using a genetic algorithm to determine the arrangement position of a temperature sensor, collecting a temperature data sequence of the temperature sensor array and its corresponding time information; establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship and a load boundary, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model; based on the finite element model, using a nonlinear iterative algorithm to periodically solve to obtain a solder joint stress-strain field data sequence and its corresponding time information, establishing a modified Coffin-Manson model that considers the temperature-stress-time coupling effect, calculating the fatigue damage increment of the solder joint and accumulating it in real time, and evaluating the thermal fatigue state of the integrated circuit packaging structure according to the rate of change of the accumulated fatigue damage value.

Description

Technical Field

[0001] The present invention relates to integrated circuit technology, and in particular to a dynamic method and system for evaluating thermal fatigue of integrated circuit packaging. Background Art

[0002] Commonly used thermal fatigue assessment methods mainly rely on finite element analysis simulations. These methods are usually based on simplified temperature cycle conditions. However, the temperature changes in the actual working environment are often non-uniform and complex, and simplified simulations are difficult to accurately reflect the actual temperature field distribution, resulting in a large deviation between the assessment results and the actual situation. In addition, the traditional Coffin-Manson model does not consider the temperature-stress-time coupling effect, and its prediction accuracy is also limited. These defects lead to the following deficiencies in the existing technology in evaluating the thermal fatigue life of integrated circuit packages:

[0003] Insufficient accuracy of temperature field simulation: Existing methods usually use simplified temperature cycle conditions, which cannot accurately reflect the complex temperature field distribution of the chip under the actual working state, resulting in errors in subsequent stress-strain calculations and fatigue life predictions.

[0004] The temperature-stress-time coupling effect is not considered: The traditional Coffin-Manson model only considers the influence of strain range and number of cycles on fatigue life, ignoring the coupling effect of factors such as temperature and time. Especially in high temperature environment, this coupling effect will significantly affect the fatigue performance of the material, resulting in inaccurate life prediction results.

[0005] Lack of real-time performance: Traditional finite element simulation methods are computationally intensive and time-consuming, making it difficult to conduct real-time thermal fatigue status assessment, and unable to promptly detect potential risks and take corresponding measures. Summary of the invention

[0006] The embodiments of the present invention provide a dynamic method and system for evaluating thermal fatigue of integrated circuit packaging, which can solve the problems in the prior art.

[0007] According to a first aspect of the embodiments of the present invention,

[0008] Provides a dynamic approach to IC package thermal fatigue assessment, including:

[0009] An infrared thermal imager is used to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, a hierarchical clustering algorithm is used to identify hotspot areas of the thermal image sequence to obtain hotspot positions and temperature gradient data, a hotspot feature evolution model is constructed based on the hotspot positions and temperature gradient data, a temperature field reconstruction sensitivity distribution is calculated according to the hotspot feature evolution model, a genetic algorithm is used to determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution, a temperature sensor array is set at the layout position, and a temperature data sequence of the temperature sensor array and its corresponding time information are collected;

[0010] Performing filtering preprocessing on the temperature data sequence to obtain corrected temperature data, and reconstructing the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, converting the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model;

[0011] Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information; based on the solder point stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, and the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence, and the fatigue damage increment is accumulated in real time to obtain the cumulative fatigue damage value, and the thermal fatigue state of the integrated circuit packaging structure is evaluated according to the change rate of the cumulative fatigue damage value.

[0012] In an optional embodiment,

[0013] The steps of constructing a hotspot feature evolution model based on the hotspot position and temperature gradient data, calculating a temperature field reconstruction sensitivity distribution according to the hotspot feature evolution model, and determining the arrangement position of the temperature sensor using a genetic algorithm based on the temperature field reconstruction sensitivity distribution include:

[0014] The hotspot position and temperature gradient data are constructed as a data set containing time series information; based on the data set containing time series information, a state equation and an observation equation are established by using Kalman filtering combined with Gaussian process regression, wherein the state equation includes a state vector, a state transfer matrix based on kernel function optimization, and process noise, and the observation equation includes an observation matrix and observation noise based on kernel function prediction, and the state equation and the observation equation form a hotspot feature evolution model;

[0015] Based on the prediction data of the hotspot feature evolution model, the radial basis function is used to interpolate and reconstruct the temperature field, and a temperature field reconstruction error function is established, wherein the temperature field reconstruction error function represents the difference between the actual temperature field and the reconstructed temperature field;

[0016] Introducing adjoint variables into the temperature field reconstruction error function to construct a Lagrangian functional, deriving an adjoint equation based on the Lagrangian functional, discretizing the adjoint equation to obtain a stiffness matrix and a load vector, and solving a linear equation system consisting of the stiffness matrix and the load vector to obtain an adjoint variable;

[0017] Calculating the variation of the temperature field reconstruction error to the temperature of any position point based on the accompanying variable to obtain the sensitivity distribution of the position point, wherein the sensitivity distribution indicates the influence of the temperature change of the position point on the temperature field reconstruction accuracy;

[0018] The sensitivity distribution is used as the temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal layout position of the temperature sensor.

[0019] In an optional embodiment,

[0020] The sensitivity distribution is used as a temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal arrangement position of the temperature sensor, and the steps include:

[0021] Normalizing the sensitivity distribution to obtain a temperature-sensitive weight coefficient, and constructing an optimization objective function based on the temperature-sensitive weight coefficient, wherein the optimization objective function includes a sensitivity target term and a penalty term, wherein the sensitivity target term is the product sum of the temperature-sensitive weight coefficient and the sensitivity of the corresponding position point, and the penalty term is the square sum of the difference between the adjacent sensor spacing and the minimum allowable spacing;

[0022] An improved genetic algorithm with real number coding is used to solve the optimization objective function, and the chromosomes of the improved genetic algorithm are encoded with sensor position coordinates; the improved genetic algorithm includes an arithmetic crossover operator and a non-uniform mutation operator, the arithmetic crossover operator generates offspring individuals through linear combinations of parent individuals, and the non-uniform mutation operator dynamically adjusts the mutation amplitude according to the evolutionary algebra;

[0023] An elite retention strategy and adaptive control parameters are introduced into the improved genetic algorithm. The elite retention strategy directly retains the individuals with the highest fitness to the next generation. The adaptive control parameters dynamically adjust the crossover probability and the mutation probability according to the individual fitness, thereby obtaining the optimal arrangement position of the temperature sensors that meets the spacing constraints.

[0024] In an optional embodiment,

[0025] The steps of establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship and a load boundary, converting the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model include:

[0026] A multilayer composite structure geometric model including a silicon chip layer, a solder layer and a substrate layer is constructed, wherein displacement continuity is achieved among the silicon chip layer, the solder layer and the substrate layer through interface constraints;

[0027] A temperature-dependent elastic-plastic constitutive model is established based on the geometric model of the multi-layer composite structure, wherein the elastic-plastic constitutive model includes the temperature-dependent elastic deformation of each layer of material, the thermal strain caused by thermal expansion, and the plastic deformation based on the temperature-dependent flow criterion, wherein the plastic deformation is described by a temperature-dependent yield function;

[0028] The complete temperature field distribution is mapped into a temperature load in the continuous time domain through an interpolation function, and the multilayer composite structure geometric model and the elastic-plastic constitutive model are combined to impose a fixed constraint on the outer boundary of the substrate and a symmetric constraint on the symmetric surface to construct a thermal-mechanical coupling analysis model;

[0029] The mesh size distribution is determined based on the temperature gradient criterion and the stress concentration criterion. The temperature gradient criterion determines the mesh size of the temperature sensitive area according to the temperature gradient, and the stress concentration criterion determines the mesh size of the stress sensitive area according to the stress gradient. The thermal-mechanical coupling analysis model is meshed using a hexahedron and pentahedron mixed mesh. A structured hexahedron mesh is used in a regular area, and a pentahedron transition unit is used in a transition area. The mesh quality is evaluated by mesh distortion, aspect ratio and orthogonality. Local mesh optimization is performed on the area that does not meet the quality requirements to obtain a finite element model.

[0030] In an optional embodiment,

[0031] Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve to obtain a solder joint stress-strain field data sequence and its corresponding time information; based on the solder joint stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder joint stress-strain field data sequence, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the step of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value includes:

[0032] Based on the finite element model, an improved Newton-Raphson nonlinear iterative algorithm is used for solving, and an adaptive time step control strategy is introduced to adjust the iteration step, wherein the adaptive time step control strategy is determined according to the step adjustment coefficient, the previous time step, and the ratio of the current step iteration number to the expected iteration number, and the step adjustment coefficient has a value range of 0.5 to 2. When the current step iteration number is greater than the expected iteration number, the value is less than 1 to reduce the step length, and when the current step iteration number is less than the expected iteration number, the value is greater than 1 to increase the step length, and the stress and strain field data sequence of the weld point and its corresponding time information are obtained by periodic solution;

[0033] Based on the stress-strain field data sequence of the solder joint and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, wherein the modified Coffin-Manson model includes an inelastic strain term, an elastic strain term, and a temperature correction function, a stress correction function, and a time correction function;

[0034] The modified Coffin-Manson model is used in combination with a linear fatigue damage accumulation criterion to calculate the fatigue damage increment corresponding to the stress-strain field data sequence of the weld point, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the rate of change of the cumulative fatigue damage value relative to time is calculated to obtain a damage accumulation rate;

[0035] The thermal fatigue state of the integrated circuit packaging structure is evaluated according to the damage accumulation rate, and the thermal fatigue process of the solder joint is divided into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance.

[0036] In an optional embodiment,

[0037] The steps of modifying the Coffin-Manson model including an inelastic strain term, an elastic strain term, a temperature correction function, a stress correction function and a time correction function include:

[0038] The inelastic strain term is determined by the fatigue ductility coefficient and the number of failure cycles, and the elastic strain term is determined by the fatigue strength coefficient, the elastic modulus and the number of failure cycles;

[0039] Establish a temperature correction function, which is based on the Arrhenius relationship and uses the product of the reciprocal difference between the working temperature and the reference temperature and the activation energy, divided by the Boltzmann constant to obtain an exponential value; the working temperature is the time-weighted average value of the temperature field data in a complete load cycle, and the reference temperature is the room temperature value;

[0040] Establishing a stress correction function, first obtaining an average stress, the average stress is calculated by time-integrated averaging of stress-strain field data within a complete load cycle, then dividing the average stress by the material fatigue strength to obtain a stress ratio, and subtracting the stress ratio from 1 to perform a power calculation of a material correlation coefficient, wherein the material correlation coefficient and the material fatigue strength are determined by a standard fatigue test;

[0041] A time correction function is established, wherein the time correction function is determined by calculating the ratio of the high temperature residence time to the standard time and taking the time effect exponential power, wherein the high temperature residence time is the cumulative time that the temperature exceeds the critical temperature threshold, the critical temperature threshold is the starting temperature of the material creep, and the standard time is the loading cycle in the standard fatigue test.

[0042] In an optional embodiment,

[0043] The steps of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the damage accumulation rate and dividing the solder joint thermal fatigue process into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance include:

[0044] Establishing a solder joint thermal fatigue accumulation characteristic parameter database, wherein the solder joint thermal fatigue accumulation characteristic parameter database stores a constant value of a damage accumulation rate under stable working conditions, a reference interval of a damage increment ratio between adjacent cycles, a microscopic grain boundary structure evolution map, and a critical value of a strain energy density release rate;

[0045] Collecting real-time working state parameters of the solder joints of the integrated circuit packaging structure, the real-time working state parameters include damage accumulation rate, damage increment data of multiple consecutive cycles, microstructure image and strain energy density data; performing multi-level analysis on the thermal fatigue state of the solder joints, including: calculating the degree of constancy of the damage accumulation rate, obtaining the damage increment ratio of multiple consecutive cycles and its change trend, analyzing the grain boundary sliding and microcrack extension length in the microstructure image, calculating the strain energy density release rate and comparing it with the critical value;

[0046] Determine the thermal fatigue state of the solder joint based on the multi-level analysis results of the thermal fatigue state of the solder joint, including:

[0047] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate does not exceed the preset fluctuation threshold, the damage increment ratio of multiple consecutive cycles is within the reference interval, the area of ​​the grain boundary sliding region in the microstructure image is greater than a first preset percentage of the total observation area, and the strain energy density release rate is less than a first preset ratio of the critical value, it is determined to be a stable period with a constant damage accumulation rate;

[0048] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a first preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than the first preset ratio, the extension length of the microcracks in the microstructure image reaches a first preset length, and the strain energy density release rate is greater than a first preset ratio of the critical value and less than a second preset ratio of the critical value, it is determined to be an accelerated period of microcrack extension;

[0049] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a second preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than a second preset ratio, the length of the microcracks in the microstructure image exceeds a second preset length, and the strain energy density release rate is greater than a second preset ratio of a critical value, it is determined to be a failure period requiring immediate maintenance;

[0050] The evaluation results are output based on the thermal fatigue status of the solder joint. When the accelerated period characteristics of microcrack extension are detected, a warning signal is issued and the monitoring frequency is increased. When the failure period requiring immediate maintenance is reached, a maintenance instruction is triggered.

[0051] According to a second aspect of the embodiments of the present invention,

[0052] Provides dynamic systems for integrated circuit package thermal fatigue assessment, including:

[0053] The first unit is used to use an infrared thermal imager to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, use a hierarchical clustering algorithm to identify hot spot areas of the thermal image sequence to obtain hot spot positions and temperature gradient data, build a hot spot feature evolution model based on the hot spot positions and temperature gradient data, calculate the temperature field reconstruction sensitivity distribution according to the hot spot feature evolution model, determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution by using a genetic algorithm, set a temperature sensor array at the layout position, and collect the temperature data sequence of the temperature sensor array and its corresponding time information;

[0054] The second unit is used to filter and preprocess the temperature data sequence to obtain corrected temperature data, and reconstruct the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establish a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, convert the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and mesh the thermal-mechanical coupling analysis model to obtain a finite element model;

[0055] The third unit is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information based on the finite element model using a nonlinear iterative algorithm; establish a modified Coffin-Manson model that considers the temperature-stress-time coupling effect based on the solder point stress-strain field data sequence and its corresponding time information, calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence using the modified Coffin-Manson model, accumulate the fatigue damage increment in real time to obtain the cumulative fatigue damage value, and evaluate the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value.

[0056] According to a third aspect of the embodiments of the present invention,

[0057] An electronic device is provided, comprising:

[0058] processor;

[0059] a memory for storing processor-executable instructions;

[0060] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0061] A fourth aspect of the embodiments of the present invention is:

[0062] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0063] The present invention accurately locates the hot spot area and temperature gradient through infrared thermal imaging, hierarchical clustering algorithm and hot spot feature evolution model, and optimizes the temperature sensor arrangement based on the temperature field reconstruction sensitivity distribution, so as to more accurately capture the temperature field change. In addition, the radial basis function interpolation algorithm is used to reconstruct the complete temperature field, avoiding the error of the traditional method in simplifying the temperature field, and the thermal-mechanical coupling model is solved based on the nonlinear iterative algorithm, which more accurately reflects the stress-strain state of the solder joint. Finally, through the modified Coffin-Manson model and fatigue damage increment accumulation, a more accurate thermal fatigue life prediction is achieved.

[0064] The present invention analyzes the thermal image sequence, temperature data sequence and solder point stress-strain field data sequence, and can capture the changes in temperature, stress and strain in real time, and accumulate and evaluate fatigue damage in real time, thereby realizing dynamic monitoring and evaluation of the thermal fatigue state of the integrated circuit packaging structure, rather than just providing a static life prediction result. This makes it possible to take timely measures based on the real-time evaluation results to avoid thermal fatigue failure.

[0065] The present invention optimizes the arrangement of temperature sensors by reconstructing the sensitivity distribution of the temperature field and the genetic algorithm, reduces the number of sensors required, and reduces the testing cost. At the same time, the temperature information can be obtained non-contactly through infrared thermal imaging technology, which simplifies the testing process and improves the testing efficiency. In addition, the method can perform real-time monitoring during the operation of the product, avoids additional testing links, and further reduces the testing cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic flow chart of a dynamic method for evaluating thermal fatigue of integrated circuit packaging according to an embodiment of the present invention;

[0067] Figure 2 It is a schematic diagram of the structure of a dynamic system for thermal fatigue assessment of integrated circuit packaging according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0070] Figure 1 FIG. 1 is a flow chart of a dynamic method for evaluating thermal fatigue of integrated circuit packaging according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0071] S1. Use an infrared thermal imager to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, use a hierarchical clustering algorithm to identify hot spot areas in the thermal image sequence to obtain hot spot positions and temperature gradient data, build a hot spot feature evolution model based on the hot spot position and temperature gradient data, calculate the temperature field reconstruction sensitivity distribution according to the hot spot feature evolution model, use a genetic algorithm to determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution, set a temperature sensor array at the layout position, and collect the temperature data sequence of the temperature sensor array and its corresponding time information;

[0072] S2. Filter and preprocess the temperature data sequence to obtain corrected temperature data, and reconstruct the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establish a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, convert the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and mesh the thermal-mechanical coupling analysis model to obtain a finite element model;

[0073] S3. Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information; based on the solder point stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, and the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence, and the fatigue damage increment is accumulated in real time to obtain the cumulative fatigue damage value, and the thermal fatigue state of the integrated circuit packaging structure is evaluated according to the rate of change of the cumulative fatigue damage value.

[0074] Exemplarily, first, an infrared thermal imager is used to periodically scan the target integrated circuit package to obtain a series of thermal images. Image processing techniques, such as hierarchical clustering algorithms, are used to identify hotspot areas in the thermal image sequence and extract hotspot location and temperature gradient data. These data will be used to construct a hotspot feature evolution model that describes the temperature variation of the hotspot area over time and space.

[0075] Based on the hotspot feature evolution model, the temperature field reconstruction sensitivity distribution is calculated. The temperature field reconstruction sensitivity indicates the degree of influence of temperature sensors at different positions on the temperature field reconstruction accuracy. Using genetic algorithms, the optimal layout of temperature sensors is determined according to the temperature field reconstruction sensitivity distribution to maximize the temperature field reconstruction accuracy.

[0076] The temperature sensor array is arranged at the determined optimal position, and the temperature data sequence and its corresponding time information are collected. For example, the sensor array collects temperature data every 1 second for 10 minutes.

[0077] The collected temperature data sequence is filtered and preprocessed to remove noise and outliers to obtain the corrected temperature data. The radial basis function interpolation algorithm is used to interpolate the corrected temperature data and reconstruct the complete temperature field distribution. For example, the thin plate spline function is used as the radial basis function to interpolate 600 temperature data points collected within 10 minutes to obtain the complete temperature field distribution at each time point.

[0078] Establish a thermomechanical coupling analysis model that includes a geometric model, material constitutive relations, and load boundaries. Convert the reconstructed complete temperature field distribution and its corresponding time information into the temperature load of the thermomechanical coupling analysis model. For example, import the temperature field data at each time point into the finite element software as the temperature boundary condition of the model.

[0079] The thermomechanical coupling analysis model is meshed to obtain a finite element model. The finite element model is periodically solved using a nonlinear iterative algorithm to obtain a stress-strain field data sequence of the weld and its corresponding time information. For example, the Newton-Raphson iterative algorithm is used to solve the nonlinear equations at each time step to obtain the stress-strain distribution of the weld.

[0080] Based on the obtained solder joint stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature, stress and time coupling effects is established. The fatigue damage increment corresponding to the solder joint stress-strain field data sequence is calculated using the modified Coffin-Manson model. For example, the stress-strain data of each time step is substituted into the modified Coffin-Manson model to calculate the fatigue damage increment of the time step.

[0081] The calculated fatigue damage increments are accumulated in real time to obtain the cumulative fatigue damage value. For example, the fatigue damage increments of each time step are accumulated to obtain the cumulative fatigue damage value at the current moment. The thermal fatigue state of the integrated circuit package structure is evaluated based on the rate of change of the cumulative fatigue damage value. For example, based on the growth rate of the cumulative fatigue damage value, it is determined whether the package is in a dangerous state. Assuming that during a 10-minute test, the cumulative fatigue damage value increases from 0 to 0.1, the thermal fatigue state of the package can be evaluated and its remaining life can be predicted.

[0082] The present invention can obtain temperature field information more accurately by combining infrared thermal imaging and temperature sensor array, thereby improving the accuracy of thermal fatigue assessment; it can collect temperature data in real time, calculate and accumulate fatigue damage increments, thereby realizing real-time assessment of the thermal fatigue state of integrated circuit packages; by analyzing the rate of change of the cumulative fatigue damage value, it can predict the remaining life of the integrated circuit package, providing guidance for preventive maintenance.

[0083] In an optional embodiment,

[0084] The steps of constructing a hotspot feature evolution model based on the hotspot position and temperature gradient data, calculating a temperature field reconstruction sensitivity distribution according to the hotspot feature evolution model, and determining the arrangement position of the temperature sensor using a genetic algorithm based on the temperature field reconstruction sensitivity distribution include:

[0085] The hotspot position and temperature gradient data are constructed as a data set containing time series information; based on the data set containing time series information, a state equation and an observation equation are established by using Kalman filtering combined with Gaussian process regression, wherein the state equation includes a state vector, a state transfer matrix based on kernel function optimization, and process noise, and the observation equation includes an observation matrix and observation noise based on kernel function prediction, and the state equation and the observation equation form a hotspot feature evolution model;

[0086] Based on the prediction data of the hotspot feature evolution model, the radial basis function is used to interpolate and reconstruct the temperature field, and a temperature field reconstruction error function is established, wherein the temperature field reconstruction error function represents the difference between the actual temperature field and the reconstructed temperature field;

[0087] Introducing adjoint variables into the temperature field reconstruction error function to construct a Lagrangian functional, deriving an adjoint equation based on the Lagrangian functional, discretizing the adjoint equation to obtain a stiffness matrix and a load vector, and solving a linear equation system consisting of the stiffness matrix and the load vector to obtain an adjoint variable;

[0088] Calculating the variation of the temperature field reconstruction error to the temperature of any position point based on the accompanying variable to obtain the sensitivity distribution of the position point, wherein the sensitivity distribution indicates the influence of the temperature change of the position point on the temperature field reconstruction accuracy;

[0089] The sensitivity distribution is used as the temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal layout position of the temperature sensor.

[0090] For example, in order to more accurately reconstruct the temperature field and optimize the sensor layout, a sensor layout method based on the hotspot feature evolution model and the temperature field reconstruction sensitivity distribution is proposed. This method uses the hotspot location and temperature gradient data to construct a hotspot feature evolution model, and calculates the temperature field reconstruction sensitivity distribution based on the model, and finally uses a genetic algorithm to determine the optimal layout position of the temperature sensor.

[0091] First, the hotspot location and temperature gradient data are collected and organized into a data set in chronological order. For example, the hotspot location and temperature gradient are recorded every 1 second for 1 hour, resulting in 3600 data points, each of which contains the hotspot location coordinates and temperature gradient values.

[0092] Then, the hotspot feature evolution model is established using the sorted data set. Kalman filtering combined with Gaussian process regression method is used. The state equation of Kalman filtering describes the evolution of the system state over time, including the state vector (such as hotspot position and temperature), the state transfer matrix and the process noise. The state transfer matrix is ​​obtained by kernel function optimization, such as using the radial basis kernel function. The process noise represents the random disturbance in the system state evolution process. The observation equation describes the relationship between the observation value and the system state, including the observation matrix and observation noise. The observation matrix is ​​predicted by the kernel function, such as using the linear kernel function. The observation noise represents the random error in the measurement process. The state equation and the observation equation together constitute the hotspot feature evolution model.

[0093] Next, based on the predicted data of the hotspot feature evolution model, the temperature field is interpolated and reconstructed. For example, radial basis functions are used for interpolation. Then, a temperature field reconstruction error function is established. This function represents the difference between the actual temperature field and the reconstructed temperature field.

[0094] In order to calculate the temperature field reconstruction sensitivity distribution, the adjoint variable is introduced and the Lagrangian functional is constructed. Based on the functional, the adjoint equation is derived, and the adjoint equation is discretized to obtain the stiffness matrix and the load vector. The linear equation system consisting of the stiffness matrix and the load vector is solved to obtain the adjoint variable. For example, the adjoint equation is discretized using the finite element method.

[0095] Using the calculated adjoint variables, the variation of the temperature field reconstruction error to the temperature at any position is calculated to obtain the sensitivity distribution of the position. The sensitivity distribution indicates the degree of influence of the temperature change at the position on the accuracy of temperature field reconstruction. For example, if the sensitivity of a certain position is high, the temperature change at that position will have a greater impact on the accuracy of temperature field reconstruction.

[0096] Finally, the sensitivity distribution is used as the temperature sensitivity weight coefficient to construct an optimization objective function that considers the sensor spacing constraint. For example, the objective function can be set to minimize the number of sensors while ensuring that the sensitivity of all positions reaches a certain threshold. The improved genetic algorithm with real number coding is used to solve the optimization objective function. The improved genetic algorithm uses arithmetic crossover operators and non-uniform mutation operators. For example, two parent individuals perform a crossover operation to generate new offspring individuals. The mutation operation can change certain genes of the individual and increase the diversity of the population. Through the iterative calculation of the genetic algorithm, the optimal layout position of the temperature sensor is finally obtained.

[0097] The present invention calculates the temperature field reconstruction sensitivity distribution based on the hotspot feature evolution model, and can more accurately capture the changing trend of the temperature field, thereby improving the temperature field reconstruction accuracy; by using the sensitivity distribution as the weight coefficient of the optimization objective function, the positions that have the greatest impact on the temperature field reconstruction accuracy can be found, and sensors are arranged at these positions, thereby optimizing the sensor layout and improving the sensor utilization efficiency; by optimizing the sensor layout, the required number of sensors can be reduced, thereby reducing the sensor cost.

[0098] In an optional embodiment,

[0099] The sensitivity distribution is used as a temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal arrangement position of the temperature sensor, and the steps include:

[0100] Normalizing the sensitivity distribution to obtain a temperature-sensitive weight coefficient, and constructing an optimization objective function based on the temperature-sensitive weight coefficient, wherein the optimization objective function includes a sensitivity target term and a penalty term, wherein the sensitivity target term is the product sum of the temperature-sensitive weight coefficient and the sensitivity of the corresponding position point, and the penalty term is the square sum of the difference between the adjacent sensor spacing and the minimum allowable spacing;

[0101] An improved genetic algorithm with real number coding is used to solve the optimization objective function, and the chromosomes of the improved genetic algorithm are encoded with sensor position coordinates; the improved genetic algorithm includes an arithmetic crossover operator and a non-uniform mutation operator, the arithmetic crossover operator generates offspring individuals through linear combinations of parent individuals, and the non-uniform mutation operator dynamically adjusts the mutation amplitude according to the evolutionary algebra;

[0102] An elite retention strategy and adaptive control parameters are introduced into the improved genetic algorithm. The elite retention strategy directly retains the individuals with the highest fitness to the next generation. The adaptive control parameters dynamically adjust the crossover probability and the mutation probability according to the individual fitness, thereby obtaining the optimal arrangement position of the temperature sensors that meets the spacing constraints.

[0103] Exemplarily, first, obtain the temperature sensitivity distribution data of the area to be tested. For example, obtain the temperature sensitivity distribution of a rectangular area through finite element simulation or experimental measurement, and discretize the area into 100 grid points, each grid point corresponding to a sensitivity value. Assume that the obtained sensitivity value is between 0.5 and 1.5.

[0104] Next, the acquired sensitivity distribution data is normalized to obtain the temperature sensitivity weight coefficient. The normalization method can be the maximum and minimum value normalization, that is, all sensitivity values ​​are linearly transformed to between 0 and 1. For example, the point with a sensitivity value of 0.5 is converted to 0, the point with a sensitivity value of 1.5 is converted to 1, and the sensitivity values ​​of other points are linearly transformed in proportion.

[0105] Based on the normalized temperature sensitivity weight coefficient, an optimization objective function is constructed. The optimization objective function consists of two parts: the sensitivity target term and the penalty term. The goal of the sensitivity target term is to maximize the weighted sum of the sensitivity of the sensor location, that is, the sum of the products of the temperature sensitivity weight coefficient corresponding to each sensor location and the sensitivity value of the location. The goal of the penalty term is to minimize the sum of the squares of the difference between the distance between adjacent sensors and the minimum allowed distance. Assume that the minimum allowed distance is set to 0.1 meters. The introduction of the penalty term is to ensure that the sensors are not too close to each other to avoid mutual interference.

[0106] The improved genetic algorithm with real number coding is used to solve the optimization objective function and obtain the optimal layout of the temperature sensor. The chromosomes of the genetic algorithm are encoded with the sensor position coordinates.

[0107] The improved genetic algorithm uses arithmetic crossover operator and non-uniform mutation operator. The arithmetic crossover operator generates two offspring individuals through the linear combination of two parent individuals.

[0108] The elite retention strategy and adaptive control parameters are introduced into the improved genetic algorithm. The elite retention strategy directly copies the individuals with the highest fitness to the next generation to ensure the convergence of the algorithm. The adaptive control parameters dynamically adjust the crossover probability and mutation probability according to the individual fitness. For example, the crossover probability and mutation probability of individuals with high fitness are lower, and the crossover probability and mutation probability of individuals with low fitness are higher.

[0109] Through the above steps, the optimal arrangement position of the temperature sensor that meets the spacing constraint is finally obtained.

[0110] The present invention arranges sensors at positions with higher sensitivity by considering the temperature sensitivity distribution, thereby effectively improving the accuracy of temperature measurement; by optimizing the objective function, an optimal arrangement scheme of sensors that meets the spacing constraint can be obtained, thereby avoiding mutual interference between sensors; an improved genetic algorithm is adopted, and an elite retention strategy and adaptive control parameters are introduced, thereby improving the convergence speed and optimization ability of the algorithm.

[0111] In an optional embodiment,

[0112] The steps of establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship and a load boundary, converting the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model include:

[0113] A multilayer composite structure geometric model including a silicon chip layer, a solder layer and a substrate layer is constructed, wherein displacement continuity is achieved among the silicon chip layer, the solder layer and the substrate layer through interface constraints;

[0114] A temperature-dependent elastic-plastic constitutive model is established based on the geometric model of the multi-layer composite structure, wherein the elastic-plastic constitutive model includes the temperature-dependent elastic deformation of each layer of material, the thermal strain caused by thermal expansion, and the plastic deformation based on the temperature-dependent flow criterion, wherein the plastic deformation is described by a temperature-dependent yield function;

[0115] The complete temperature field distribution is mapped into a temperature load in the continuous time domain through an interpolation function, and the multilayer composite structure geometric model and the elastic-plastic constitutive model are combined to impose a fixed constraint on the outer boundary of the substrate and a symmetric constraint on the symmetric surface to construct a thermal-mechanical coupling analysis model;

[0116] The mesh size distribution is determined based on the temperature gradient criterion and the stress concentration criterion. The temperature gradient criterion determines the mesh size of the temperature sensitive area according to the temperature gradient, and the stress concentration criterion determines the mesh size of the stress sensitive area according to the stress gradient. The thermal-mechanical coupling analysis model is meshed using a hexahedron and pentahedron mixed mesh. A structured hexahedron mesh is used in a regular area, and a pentahedron transition unit is used in a transition area. The mesh quality is evaluated by mesh distortion, aspect ratio and orthogonality. Local mesh optimization is performed on the area that does not meet the quality requirements to obtain a finite element model.

[0117] Exemplarily, first, a geometric model of a multi-layer composite structure is constructed. This model consists of a silicon chip layer, a solder layer, and a substrate layer. In order to ensure the accuracy of the model, it is necessary to accurately measure and define the geometric dimensions of each layer, such as length, width, and thickness. For example, the size of the silicon chip layer can be set to 10mmx10mmx0.5mm, the size of the solder layer can be set to 12mmx12mmx0.1mm, and the size of the substrate layer can be set to 20mmx20mmx2mm. The three layers of materials are connected by interface constraints to ensure displacement continuity and simulate the real physical connection state.

[0118] Next, a temperature-dependent elastic-plastic constitutive model is established. This model needs to consider the temperature dependence of the material, including elastic deformation, thermal strain caused by thermal expansion, and plastic deformation based on the temperature-dependent flow criterion. For the silicon chip layer, it can be assumed that its elastic modulus is 150 GPa at room temperature, the Poisson's ratio is 0.25, and the thermal expansion coefficient is For the solder layer, it can be assumed that its elastic modulus is 30 GPa at room temperature, Poisson’s ratio is 0.3, and thermal expansion coefficient is The Anand model is used to describe its temperature-dependent plastic deformation behavior. For the substrate layer, its elastic modulus can be assumed to be 100 GPa at room temperature, Poisson's ratio is 0.3, and thermal expansion coefficient is In the plastic deformation part, a temperature-dependent yield function is used to describe the change in yield strength of the material at different temperatures.

[0119] Then, the complete temperature field distribution is converted into the temperature load of the thermal-mechanical coupling analysis model. Assume that the complete temperature field distribution data of the chip in the working state is obtained through thermal analysis, including the temperature values ​​at different positions and different times. Use interpolation functions, such as linear interpolation or spline interpolation, to convert discrete temperature data into temperature loads in the continuous time domain. For example, if the temperature at the center of the chip is 25°C at 0s, 100°C at 1s, and 125°C at 2s, linear interpolation can be used to obtain the temperature value of the point at any time between 0s and 2s. Apply these temperature loads to the previously established multi-layer composite structure geometry model. At the same time, a fixed constraint is applied to the outer boundary of the substrate to limit its displacement in all directions; a symmetry constraint is applied to the symmetry plane to simplify the model and improve the calculation efficiency. At this point, the complete thermal-mechanical coupling analysis model is built.

[0120] Finally, the thermal-mechanical coupling analysis model is meshed to generate a finite element model. The meshing needs to take into account the temperature gradient and stress concentration areas. For example, near the interface between the chip and the solder, due to the large temperature and stress gradients, a finer mesh is required. Assume that according to the temperature gradient criterion, the mesh size is set to 0.01mm near the interface between the chip and the solder; and the mesh size is set to 0.1mm in the area far from the interface. A mixed hexahedral and pentahedral mesh is used for meshing, with structured hexahedral meshes used in regular areas and pentahedral transition units used in transition areas. After the meshing is completed, the mesh quality needs to be evaluated, such as checking the mesh distortion, aspect ratio, and orthogonality. For areas that do not meet the quality requirements, such as units with a distortion greater than 0.5 or an aspect ratio greater than 10, local mesh optimization is required, such as by adjusting the node position or re-meshing, until the mesh quality meets the requirements.

[0121] The present invention can more accurately predict the deformation and stress distribution of multilayer composite structures under thermal-mechanical coupling by considering the temperature dependence and plastic deformation of the material; the hybrid meshing technology can effectively reduce the amount of calculation and improve the simulation efficiency while ensuring the calculation accuracy; by evaluating and optimizing the mesh quality, the reliability of the finite element model can be improved, ensuring the accuracy and credibility of the simulation results.

[0122] In an optional embodiment,

[0123] Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve to obtain a solder joint stress-strain field data sequence and its corresponding time information; based on the solder joint stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder joint stress-strain field data sequence, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the step of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value includes:

[0124] Based on the finite element model, an improved Newton-Raphson nonlinear iterative algorithm is used for solving, and an adaptive time step control strategy is introduced to adjust the iteration step, wherein the adaptive time step control strategy is determined according to the step adjustment coefficient, the previous time step, and the ratio of the current step iteration number to the expected iteration number, and the step adjustment coefficient has a value range of 0.5 to 2. When the current step iteration number is greater than the expected iteration number, the value is less than 1 to reduce the step length, and when the current step iteration number is less than the expected iteration number, the value is greater than 1 to increase the step length, and the stress and strain field data sequence of the weld point and its corresponding time information are obtained by periodic solution;

[0125] Based on the stress-strain field data sequence of the solder joint and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, wherein the modified Coffin-Manson model includes an inelastic strain term, an elastic strain term, and a temperature correction function, a stress correction function, and a time correction function;

[0126] The modified Coffin-Manson model is used in combination with a linear fatigue damage accumulation criterion to calculate the fatigue damage increment corresponding to the stress-strain field data sequence of the weld point, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the rate of change of the cumulative fatigue damage value relative to time is calculated to obtain a damage accumulation rate;

[0127] The thermal fatigue state of the integrated circuit packaging structure is evaluated according to the damage accumulation rate, and the thermal fatigue process of the solder joint is divided into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance.

[0128] For example, first, a finite element model of the integrated circuit packaging structure is established. The model should include the solder joints and surrounding materials, such as chips, substrates, packaging materials, etc. Material properties, such as Young's modulus, Poisson's ratio, thermal expansion coefficient, etc., need to be accurately defined according to the actual materials used. Geometric structures, such as the shape, size, and position of the solder joints, also need to be accurately modeled.

[0129] Next, the improved Newton-Raphson nonlinear iterative algorithm is used to solve the finite element model. Set the initial temperature conditions and boundary conditions, such as temperature cycle curves, fixed constraints, etc. In the iterative solution process, an adaptive time step control strategy is introduced. This strategy dynamically adjusts the time step based on the step adjustment coefficient, the previous time step, and the ratio of the current step iteration number to the expected number of iterations. For example, the expected number of iterations is set to 10. The step adjustment coefficient ranges from 0.5 to 2. If the current step iteration number is 12 times, which is greater than the expected number of iterations of 10 times, the step adjustment coefficient is less than 1, such as 0.8, to reduce the time step of the next step. If the current step iteration number is 8 times, which is less than the expected number of iterations of 10 times, the step adjustment coefficient is greater than 1, such as 1.2, to increase the time step of the next step. Assuming that the time step of the previous step is 0.1 seconds, the current step is 0.1 seconds * 0.8 = 0.08 seconds or 0.1 seconds * 1.2 = 0.12 seconds. By periodically solving, for example, solving once every certain period of time, the stress-strain field data sequence of the solder joint at different time points and its corresponding time information are obtained. For example, the corresponding solder joint stress-strain data are obtained at time points such as 0 seconds, 0.1 seconds, 0.18 seconds, and 0.3 seconds.

[0130] Then, based on the obtained solder joint stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established. The model includes inelastic strain terms, elastic strain terms, and temperature correction functions, stress correction functions, and time correction functions. For example, the temperature correction function can be in the form of an exponential function, the stress correction function can be in the form of a power function, and the time correction function can be in the form of a linear function. The parameters in these functions need to be determined based on experimental data or empirical formulas.

[0131] Next, the fatigue damage increment of the solder joint at each time step is calculated using the established modified Coffin-Manson model combined with the linear fatigue damage accumulation criterion. For example, at 0.1 seconds, based on the stress-strain data and temperature at that time, the fatigue damage increment of that time step is calculated to be 0.0001 by the modified Coffin-Manson model. Then, the fatigue damage increment of each time step is accumulated in real time to obtain the cumulative fatigue damage value. For example, at 0.18 seconds, the cumulative fatigue damage value is 0.0001+0.00015=0.00025.

[0132] After that, the rate of change of the cumulative fatigue damage value relative to time is calculated to obtain the damage accumulation rate. For example, in the time period from 0.18 seconds to 0.3 seconds, the cumulative fatigue damage value increases from 0.00025 to 0.0004, and the time increases by 0.12 seconds. The damage accumulation rate in this time period is (0.0004-0.00025) / 0.12=0.00125 / second.

[0133] Finally, the thermal fatigue state of the integrated circuit packaging structure is evaluated based on the damage accumulation rate. The thermal fatigue process of the solder joint is divided into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period that requires immediate maintenance. For example, when the damage accumulation rate is lower than a certain threshold, it is considered to be a stable period; when the damage accumulation rate exceeds the threshold but is lower than another higher threshold, it is considered to be an accelerated period; when the damage accumulation rate exceeds the highest threshold, it is considered to be a failure period.

[0134] The present invention considers the temperature-stress-time coupling effect, and the modified Coffin-Manson model can more accurately reflect the fatigue damage evolution law of the solder joint under the actual working environment, thereby improving the thermal fatigue life prediction accuracy; by accumulating fatigue damage increments in real time and calculating the damage accumulation rate, the thermal fatigue state of the integrated circuit packaging structure can be evaluated in real time, and potential failure risks can be discovered in time; according to the thermal fatigue state evaluation result, the solder joint thermal fatigue process can be divided into different stages, and corresponding maintenance strategies can be formulated for different stages, thereby effectively extending the service life of the integrated circuit packaging structure and avoiding major economic losses.

[0135] In an optional embodiment,

[0136] The steps of modifying the Coffin-Manson model including an inelastic strain term, an elastic strain term, a temperature correction function, a stress correction function and a time correction function include:

[0137] The inelastic strain term is determined by the fatigue ductility coefficient and the number of failure cycles, and the elastic strain term is determined by the fatigue strength coefficient, the elastic modulus and the number of failure cycles;

[0138] Establish a temperature correction function, which is based on the Arrhenius relationship and uses the product of the reciprocal difference between the working temperature and the reference temperature and the activation energy, divided by the Boltzmann constant to obtain an exponential value; the working temperature is the time-weighted average value of the temperature field data in a complete load cycle, and the reference temperature is the room temperature value;

[0139] Establishing a stress correction function, first obtaining an average stress, the average stress is calculated by time-integrated averaging of stress-strain field data within a complete load cycle, then dividing the average stress by the material fatigue strength to obtain a stress ratio, and subtracting the stress ratio from 1 to perform a power calculation of a material correlation coefficient, wherein the material correlation coefficient and the material fatigue strength are determined by a standard fatigue test;

[0140] A time correction function is established, wherein the time correction function is determined by calculating the ratio of the high temperature residence time to the standard time and taking the time effect exponential power, wherein the high temperature residence time is the cumulative time that the temperature exceeds the critical temperature threshold, the critical temperature threshold is the starting temperature of the material creep, and the standard time is the loading cycle in the standard fatigue test.

[0141] For example, the Coffin-Manson model is often used to predict the low-cycle fatigue life of materials. In order to improve the prediction accuracy, especially in high-temperature environments, the influence of factors such as temperature, stress state and time needs to be considered. This method modifies the Coffin-Manson model by introducing inelastic strain terms, elastic strain terms, and temperature correction functions, stress correction functions, and time correction functions to more accurately predict the fatigue life of materials.

[0142] First, determine the fatigue ductility coefficient and fatigue strength coefficient of the material. These two coefficients can be obtained through standard fatigue tests. For example, a series of strain-controlled fatigue tests are performed on a specific material, and the number of failure cycles at different strain amplitudes is recorded. Then, the two coefficients are obtained by data fitting. Assume that the fatigue ductility coefficient of the material is measured to be 0.4 and the fatigue strength coefficient is 900MPa.

[0143] Next, calculate the inelastic strain term and the elastic strain term. The inelastic strain term is equal to the fatigue ductility coefficient multiplied by the absolute value of the negative logarithm of the number of failure cycles. The elastic strain term is equal to the fatigue strength coefficient divided by the elastic modulus, multiplied by the absolute value of the negative logarithm of the number of failure cycles. Assuming the number of failure cycles is 1000 times and the elastic modulus is 200GPa, the inelastic strain amplitude is 0.12 and the elastic strain amplitude is 0.0045.

[0144] Then, a temperature correction function is established. This function is based on the Arrhenius relationship. The product of the reciprocal difference between the working temperature and the reference temperature and the activation energy is divided by the Boltzmann constant and then the exponential value is obtained. The working temperature is the time-weighted average of the temperature field data within a complete load cycle. The reference temperature is the room temperature value, such as 25 degrees Celsius. The activation energy and the Boltzmann constant are material-related constants. Assuming that the temperature field data within a complete load cycle is: 500 degrees Celsius for 10 seconds, 600 degrees Celsius for 20 seconds, and 550 degrees Celsius for 30 seconds, the working temperature is (500*10+600*20+550*30) / (10+20+30)=550 degrees Celsius. Assuming that the activation energy is 100kJ / mol and the Boltzmann constant is 8.314J / (mol·K), the temperature correction factor is about 2.5.

[0145] Next, establish a stress correction function. First, obtain the average stress. The average stress is calculated by time-integrating and averaging the stress-strain field data within a complete load cycle. Then divide the average stress by the material fatigue strength to get the stress ratio, and after subtracting the stress ratio from 1, perform a power calculation of the material correlation coefficient. The material correlation coefficient and the material fatigue strength are determined by standard fatigue tests. Assuming that the stress cycles between 800MPa and -200MPa in a complete load cycle, the average stress is 300MPa. Assuming that the material fatigue strength is 900MPa and the material correlation coefficient is 0.8, the stress correction factor is approximately 0.88.

[0146] Then, a time correction function is established. The time correction function is determined by calculating the ratio of the high temperature dwell time to the standard time and taking the power of the time effect exponent. The high temperature dwell time is the cumulative time that the temperature exceeds the critical temperature threshold. The critical temperature threshold is the starting temperature of the material creep. The standard time is the loading cycle in the standard fatigue test. Assuming that the critical temperature threshold is 400 degrees Celsius, in a complete load cycle, the time that the temperature exceeds 400 degrees Celsius is 60 seconds, the standard time is 1 second, and the time effect exponent is 0.5, the time correction factor is about 7.7.

[0147] Finally, the inelastic strain term and the elastic strain term are added together and multiplied by the temperature correction function, the stress correction function, and the time correction function to obtain the total strain amplitude of the modified Coffin-Manson model. In this example, the total strain amplitude is approximately 0.1245*2.5*0.88*7.7=2.1.

[0148] By considering the influence of factors such as temperature, stress state and time, the present invention can more accurately predict the fatigue life of materials under complex load conditions, especially in high temperature environments; it not only takes into account the cumulative effect of fatigue damage, but also takes into account the influence of factors such as temperature, stress state and time on fatigue damage, and can more comprehensively evaluate the fatigue damage condition of materials.

[0149] In an optional embodiment,

[0150] The steps of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the damage accumulation rate and dividing the solder joint thermal fatigue process into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance include:

[0151] Establishing a solder joint thermal fatigue accumulation characteristic parameter database, wherein the solder joint thermal fatigue accumulation characteristic parameter database stores a constant value of a damage accumulation rate under stable working conditions, a reference interval of a damage increment ratio between adjacent cycles, a microscopic grain boundary structure evolution map, and a critical value of a strain energy density release rate;

[0152] Collecting real-time working state parameters of the solder joints of the integrated circuit packaging structure, the real-time working state parameters include damage accumulation rate, damage increment data of multiple consecutive cycles, microstructure image and strain energy density data; performing multi-level analysis on the thermal fatigue state of the solder joints, including: calculating the degree of constancy of the damage accumulation rate, obtaining the damage increment ratio of multiple consecutive cycles and its change trend, analyzing the grain boundary sliding and microcrack extension length in the microstructure image, calculating the strain energy density release rate and comparing it with the critical value;

[0153] Determine the thermal fatigue state of the solder joint based on the multi-level analysis results of the thermal fatigue state of the solder joint, including:

[0154] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate does not exceed the preset fluctuation threshold, the damage increment ratio of multiple consecutive cycles is within the reference interval, the area of ​​the grain boundary sliding region in the microstructure image is greater than a first preset percentage of the total observation area, and the strain energy density release rate is less than a first preset ratio of the critical value, it is determined to be a stable period with a constant damage accumulation rate;

[0155] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a first preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than the first preset ratio, the extension length of the microcracks in the microstructure image reaches a first preset length, and the strain energy density release rate is greater than a first preset ratio of the critical value and less than a second preset ratio of the critical value, it is determined to be an accelerated period of microcrack extension;

[0156] When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a second preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than a second preset ratio, the length of the microcracks in the microstructure image exceeds a second preset length, and the strain energy density release rate is greater than a second preset ratio of a critical value, it is determined to be a failure period requiring immediate maintenance;

[0157] The evaluation results are output based on the thermal fatigue status of the solder joint. When the accelerated period characteristics of microcrack extension are detected, a warning signal is issued and the monitoring frequency is increased. When the failure period requiring immediate maintenance is reached, a maintenance instruction is triggered.

[0158] Exemplarily, first, a database of thermal fatigue accumulation characteristic parameters of solder joints is established. This step requires a large number of experiments in a strictly controlled laboratory environment to obtain thermal fatigue data of different types of solder joints under stable working conditions. For example, for a certain type of ball grid array package (BGA) solder joint, the operating temperature cycle range is set to -40°C to 125°C, and the cycle period is 1 hour. By performing thousands of cycle tests on multiple samples, the damage of the solder joints after each cycle, such as crack length, resistance change, etc., is recorded. Using these data, a constant value of the damage accumulation rate can be calculated. For example, the constant value of the damage accumulation rate of this type of BGA solder joint under stable working conditions is an increase of 0.001 mm crack length for each cycle. At the same time, the damage increment between adjacent cycles is recorded, and the damage increment ratio is calculated. For example, the damage increment ratio of the 100th cycle and the 99th cycle is 1.05. By statistically analyzing the damage increment ratio of a large number of samples, a reference interval can be determined, for example, 0.95 to 1.05. In addition, a high-resolution microscope is used to observe the evolution of the microstructure of the solder joint and draw a microscopic grain boundary structure evolution map. For example, phenomena such as grain boundary sliding and void formation can be observed. Finally, the strain energy density of the solder joint after each cycle is calculated by finite element analysis and other methods, and the critical value of the strain energy density release rate is determined, for example, 0.5 joules / cubic millimeter.

[0159] Next, collect the real-time working state parameters of the solder joints of the integrated circuit packaging structure in the actual working environment. This requires the development of an online monitoring system that can monitor the temperature, strain and other parameters of the solder joints in real time and convert them into data such as damage accumulation rate, damage increment, microstructure image and strain energy density. For example, the system can collect data every hour and transmit the data to a central server for storage and analysis.

[0160] Then, the collected data is analyzed at multiple levels. First, the degree of constancy of the damage accumulation rate is calculated. For example, the current damage accumulation rate is compared with the constant value stored in the database and its deviation is calculated. Secondly, the damage increment ratio and its change trend for multiple consecutive cycles are obtained. For example, the damage increment ratio for the past 10 cycles is calculated and its change curve is plotted. Thirdly, the grain boundary sliding and microcrack extension length in the microstructure image are analyzed. For example, the area of ​​the grain boundary sliding region and the length of the microcrack are identified using image recognition technology. Finally, the strain energy density release rate is calculated and compared with the critical value.

[0161] Based on the multi-level analysis results, the thermal fatigue state of the solder joint is determined. For example, assuming that the preset fluctuation threshold is 5%, the reference interval is 0.95 to 1.05, the first preset percentage is 10%, the first preset ratio of the critical value is 50%, the first preset growth threshold is 10%, the first preset ratio is 1.1, the first preset length is 0.01 mm, the second preset growth threshold is 20%, the second preset ratio is 1.2, the second preset length is 0.02 mm, and the second preset ratio of the critical value is 80%. If the current damage accumulation rate is within 5% of the constant value, the damage increment ratio of 10 consecutive cycles is between 0.95 and 1.05, the grain boundary slip area is greater than 10% of the total observation area, and the strain energy density release rate is less than 50% of the critical value, it is determined to be a stable period. If the damage accumulation rate is more than 10% of the constant value, the damage increment ratio of 10 consecutive cycles is greater than 1.1, the microcrack extension length reaches 0.01 mm, and the strain energy density release rate is greater than 50% of the critical value and less than 80%, it is determined to be an accelerated period. If the deviation of the damage accumulation rate from the constant value exceeds 20%, the damage increment ratio for 10 consecutive cycles is greater than 1.2, the microcrack length exceeds 0.02 mm, and the strain energy density release rate is greater than 80% of the critical value, the failure period is determined.

[0162] Finally, the evaluation results are output according to the thermal fatigue state of the solder joint. When the accelerated period characteristics are detected, an early warning signal is issued and the monitoring frequency is increased, for example, the monitoring frequency is increased from 1 hour to 0.5 hours. When entering the failure period, a maintenance instruction is triggered, for example, an alarm message is sent to the maintenance personnel, prompting the need to replace the solder joint immediately.

[0163] The present invention can detect potential failure risks early and take corresponding preventive measures by real-time monitoring and evaluation of the thermal fatigue state of solder joints, thereby improving the reliability of the integrated circuit packaging structure; by issuing early warning signals during the acceleration period, unnecessary maintenance can be avoided, thereby reducing maintenance costs. At the same time, by triggering maintenance instructions during the failure period, greater losses caused by solder joint failure can be avoided.

[0164] Figure 2 FIG. 1 is a schematic diagram of a dynamic system for evaluating thermal fatigue of integrated circuit packaging according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0165] The first unit is used to use an infrared thermal imager to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, use a hierarchical clustering algorithm to identify hot spot areas of the thermal image sequence to obtain hot spot positions and temperature gradient data, build a hot spot feature evolution model based on the hot spot positions and temperature gradient data, calculate the temperature field reconstruction sensitivity distribution according to the hot spot feature evolution model, determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution by using a genetic algorithm, set a temperature sensor array at the layout position, and collect the temperature data sequence of the temperature sensor array and its corresponding time information;

[0166] The second unit is used to filter and preprocess the temperature data sequence to obtain corrected temperature data, and reconstruct the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establish a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, convert the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and mesh the thermal-mechanical coupling analysis model to obtain a finite element model;

[0167] The third unit is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information based on the finite element model using a nonlinear iterative algorithm; establish a modified Coffin-Manson model that considers the temperature-stress-time coupling effect based on the solder point stress-strain field data sequence and its corresponding time information, calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence using the modified Coffin-Manson model, accumulate the fatigue damage increment in real time to obtain the cumulative fatigue damage value, and evaluate the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value.

[0168] According to a third aspect of the embodiments of the present invention,

[0169] An electronic device is provided, comprising:

[0170] processor;

[0171] a memory for storing processor-executable instructions;

[0172] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0173] A fourth aspect of the embodiments of the present invention is:

[0174] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0175] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic method for thermal fatigue assessment of integrated circuit packages, characterized in that: include: An infrared thermal imager is used to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, a hierarchical clustering algorithm is used to identify hotspot areas of the thermal image sequence to obtain hotspot positions and temperature gradient data, a hotspot feature evolution model is constructed based on the hotspot positions and temperature gradient data, a temperature field reconstruction sensitivity distribution is calculated according to the hotspot feature evolution model, a genetic algorithm is used to determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution, a temperature sensor array is set at the layout position, and a temperature data sequence of the temperature sensor array and its corresponding time information are collected; Performing filtering preprocessing on the temperature data sequence to obtain corrected temperature data, and reconstructing the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, converting the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model; Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information; based on the solder point stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, and the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence, and the fatigue damage increment is accumulated in real time to obtain the cumulative fatigue damage value, and the thermal fatigue state of the integrated circuit packaging structure is evaluated according to the change rate of the cumulative fatigue damage value.

2. The method according to claim 1, characterized in that The steps of constructing a hotspot feature evolution model based on the hotspot position and temperature gradient data, calculating a temperature field reconstruction sensitivity distribution according to the hotspot feature evolution model, and determining the arrangement position of the temperature sensor using a genetic algorithm based on the temperature field reconstruction sensitivity distribution include: The hotspot position and temperature gradient data are constructed as a data set containing time series information; based on the data set containing time series information, a state equation and an observation equation are established by using Kalman filtering combined with Gaussian process regression, wherein the state equation includes a state vector, a state transfer matrix based on kernel function optimization, and process noise, and the observation equation includes an observation matrix and observation noise based on kernel function prediction, and the state equation and the observation equation form a hotspot feature evolution model; Based on the prediction data of the hotspot feature evolution model, the radial basis function is used to interpolate and reconstruct the temperature field, and a temperature field reconstruction error function is established, wherein the temperature field reconstruction error function represents the difference between the actual temperature field and the reconstructed temperature field; Introducing adjoint variables into the temperature field reconstruction error function to construct a Lagrangian functional, deriving an adjoint equation based on the Lagrangian functional, discretizing the adjoint equation to obtain a stiffness matrix and a load vector, and solving a linear equation system consisting of the stiffness matrix and the load vector to obtain an adjoint variable; Calculating the variation of the temperature field reconstruction error to the temperature of any position point based on the accompanying variable to obtain the sensitivity distribution of the position point, wherein the sensitivity distribution indicates the influence of the temperature change of the position point on the temperature field reconstruction accuracy; The sensitivity distribution is used as the temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal layout position of the temperature sensor.

3. The method according to claim 2, characterized in that The sensitivity distribution is used as a temperature sensitivity weight coefficient, an optimization objective function considering the sensor spacing constraint is constructed, and an improved genetic algorithm with real number coding is used to solve the optimization objective function, wherein the improved genetic algorithm uses an arithmetic crossover operator and a non-uniform mutation operator to obtain the optimal arrangement position of the temperature sensor, and the steps include: Normalizing the sensitivity distribution to obtain a temperature-sensitive weight coefficient, and constructing an optimization objective function based on the temperature-sensitive weight coefficient, wherein the optimization objective function includes a sensitivity target term and a penalty term, wherein the sensitivity target term is the product sum of the temperature-sensitive weight coefficient and the sensitivity of the corresponding position point, and the penalty term is the square sum of the difference between the adjacent sensor spacing and the minimum allowable spacing; An improved genetic algorithm with real number coding is used to solve the optimization objective function, and the chromosomes of the improved genetic algorithm are encoded with sensor position coordinates; the improved genetic algorithm includes an arithmetic crossover operator and a non-uniform mutation operator, the arithmetic crossover operator generates offspring individuals through linear combinations of parent individuals, and the non-uniform mutation operator dynamically adjusts the mutation amplitude according to the evolutionary algebra; An elite retention strategy and adaptive control parameters are introduced into the improved genetic algorithm. The elite retention strategy directly retains the individuals with the highest fitness to the next generation. The adaptive control parameters dynamically adjust the crossover probability and the mutation probability according to the individual fitness, thereby obtaining the optimal arrangement position of the temperature sensors that meets the spacing constraints.

4. The method according to claim 1, characterized in that: The steps of establishing a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship and a load boundary, converting the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and meshing the thermal-mechanical coupling analysis model to obtain a finite element model include: A multilayer composite structure geometric model including a silicon chip layer, a solder layer and a substrate layer is constructed, wherein displacement continuity is achieved among the silicon chip layer, the solder layer and the substrate layer through interface constraints; A temperature-dependent elastic-plastic constitutive model is established based on the geometric model of the multi-layer composite structure, wherein the elastic-plastic constitutive model includes the temperature-dependent elastic deformation of each layer of material, the thermal strain caused by thermal expansion, and the plastic deformation based on the temperature-dependent flow criterion, wherein the plastic deformation is described by a temperature-dependent yield function; The complete temperature field distribution is mapped into a temperature load in the continuous time domain through an interpolation function, and the multilayer composite structure geometric model and the elastic-plastic constitutive model are combined to impose a fixed constraint on the outer boundary of the substrate and a symmetric constraint on the symmetric surface to construct a thermal-mechanical coupling analysis model; The mesh size distribution is determined based on the temperature gradient criterion and the stress concentration criterion. The temperature gradient criterion determines the mesh size of the temperature sensitive area according to the temperature gradient, and the stress concentration criterion determines the mesh size of the stress sensitive area according to the stress gradient. The thermal-mechanical coupling analysis model is meshed using a hexahedron and pentahedron mixed mesh. A structured hexahedron mesh is used in a regular area, and a pentahedron transition unit is used in a transition area. The mesh quality is evaluated by mesh distortion, aspect ratio and orthogonality. Local mesh optimization is performed on the area that does not meet the quality requirements to obtain a finite element model.

5. The method according to claim 1, characterized in that Based on the finite element model, a nonlinear iterative algorithm is used to periodically solve to obtain a solder joint stress-strain field data sequence and its corresponding time information; based on the solder joint stress-strain field data sequence and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, the modified Coffin-Manson model is used to calculate the fatigue damage increment corresponding to the solder joint stress-strain field data sequence, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the step of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value includes: Based on the finite element model, an improved Newton-Raphson nonlinear iterative algorithm is used for solving, and an adaptive time step control strategy is introduced to adjust the iteration step, wherein the adaptive time step control strategy is determined according to the step adjustment coefficient, the previous time step, and the ratio of the current step iteration number to the expected iteration number, and the step adjustment coefficient has a value range of 0.5 to 2. When the current step iteration number is greater than the expected iteration number, the value is less than 1 to reduce the step length, and when the current step iteration number is less than the expected iteration number, the value is greater than 1 to increase the step length, and the stress and strain field data sequence of the weld point and its corresponding time information are obtained by periodic solution; Based on the stress-strain field data sequence of the solder joint and its corresponding time information, a modified Coffin-Manson model considering the temperature-stress-time coupling effect is established, wherein the modified Coffin-Manson model includes an inelastic strain term, an elastic strain term, and a temperature correction function, a stress correction function, and a time correction function; The modified Coffin-Manson model is used in combination with a linear fatigue damage accumulation criterion to calculate the fatigue damage increment corresponding to the stress-strain field data sequence of the weld point, the fatigue damage increment is accumulated in real time to obtain a cumulative fatigue damage value, and the rate of change of the cumulative fatigue damage value relative to time is calculated to obtain a damage accumulation rate; The thermal fatigue state of the integrated circuit packaging structure is evaluated according to the damage accumulation rate, and the thermal fatigue process of the solder joint is divided into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance.

6. The method according to claim 5, characterized in that The steps of modifying the Coffin-Manson model including an inelastic strain term, an elastic strain term, a temperature correction function, a stress correction function and a time correction function include: The inelastic strain term is determined by the fatigue ductility coefficient and the number of failure cycles, and the elastic strain term is determined by the fatigue strength coefficient, the elastic modulus and the number of failure cycles; Establish a temperature correction function, which is based on the Arrhenius relationship and uses the product of the reciprocal difference between the working temperature and the reference temperature and the activation energy, divided by the Boltzmann constant to obtain an exponential value; the working temperature is the time-weighted average value of the temperature field data in a complete load cycle, and the reference temperature is the room temperature value; Establishing a stress correction function, first obtaining an average stress, the average stress is calculated by time-integrated averaging of stress-strain field data within a complete load cycle, then dividing the average stress by the material fatigue strength to obtain a stress ratio, and subtracting the stress ratio from 1 to perform a power calculation of a material correlation coefficient, wherein the material correlation coefficient and the material fatigue strength are determined by a standard fatigue test; A time correction function is established, wherein the time correction function is determined by calculating the ratio of the high temperature residence time to the standard time and taking the time effect exponential power, wherein the high temperature residence time is the cumulative time that the temperature exceeds the critical temperature threshold, the critical temperature threshold is the starting temperature of the material creep, and the standard time is the loading cycle in the standard fatigue test.

7. The method according to claim 5, characterized in that The steps of evaluating the thermal fatigue state of the integrated circuit packaging structure according to the damage accumulation rate and dividing the solder joint thermal fatigue process into a stable period with a constant damage accumulation rate, an accelerated period of microcrack expansion, and a failure period requiring immediate maintenance include: Establishing a solder joint thermal fatigue accumulation characteristic parameter database, wherein the solder joint thermal fatigue accumulation characteristic parameter database stores a constant value of a damage accumulation rate under stable working conditions, a reference interval of a damage increment ratio between adjacent cycles, a microscopic grain boundary structure evolution map, and a critical value of a strain energy density release rate; Collecting real-time working state parameters of the solder joints of the integrated circuit packaging structure, the real-time working state parameters include damage accumulation rate, damage increment data of multiple consecutive cycles, microstructure image and strain energy density data; performing multi-level analysis on the thermal fatigue state of the solder joints, including: calculating the degree of constancy of the damage accumulation rate, obtaining the damage increment ratio of multiple consecutive cycles and its change trend, analyzing the grain boundary sliding and microcrack extension length in the microstructure image, calculating the strain energy density release rate and comparing it with the critical value; Determine the thermal fatigue state of the solder joint based on the multi-level analysis results of the thermal fatigue state of the solder joint, including: When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate does not exceed the preset fluctuation threshold, the damage increment ratio of multiple consecutive cycles is within the reference interval, the area of ​​the grain boundary sliding region in the microstructure image is greater than a first preset percentage of the total observation area, and the strain energy density release rate is less than a first preset ratio of the critical value, it is determined to be a stable period with a constant damage accumulation rate; When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a first preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than the first preset ratio, the extension length of the microcracks in the microstructure image reaches a first preset length, and the strain energy density release rate is greater than a first preset ratio of the critical value and less than a second preset ratio of the critical value, it is determined to be an accelerated period of microcrack extension; When the deviation of the damage accumulation rate from the constant value of the damage accumulation rate exceeds a second preset growth threshold, the damage increment ratio of a plurality of consecutive cycles is greater than a second preset ratio, the length of the microcracks in the microstructure image exceeds a second preset length, and the strain energy density release rate is greater than a second preset ratio of a critical value, it is determined to be a failure period requiring immediate maintenance; The evaluation results are output based on the thermal fatigue status of the solder joint. When the accelerated period characteristics of microcrack extension are detected, a warning signal is issued and the monitoring frequency is increased. When the failure period requiring immediate maintenance is reached, a maintenance instruction is triggered.

8. A dynamic system for thermal fatigue assessment of integrated circuit packages, used to implement the method according to any one of claims 1 to 7, characterized in that: include: The first unit is used to use an infrared thermal imager to periodically scan the integrated circuit packaging structure to obtain a thermal image sequence, use a hierarchical clustering algorithm to identify hot spot areas of the thermal image sequence to obtain hot spot positions and temperature gradient data, build a hot spot feature evolution model based on the hot spot positions and temperature gradient data, calculate the temperature field reconstruction sensitivity distribution according to the hot spot feature evolution model, determine the layout position of the temperature sensor based on the temperature field reconstruction sensitivity distribution by using a genetic algorithm, set a temperature sensor array at the layout position, and collect the temperature data sequence of the temperature sensor array and its corresponding time information; The second unit is used to filter and preprocess the temperature data sequence to obtain corrected temperature data, and reconstruct the corrected temperature data using a radial basis function interpolation algorithm to obtain a complete temperature field distribution; establish a thermal-mechanical coupling analysis model including a geometric model, a material constitutive relationship, and a load boundary, convert the complete temperature field distribution and its corresponding time information into a temperature load of the thermal-mechanical coupling analysis model, and mesh the thermal-mechanical coupling analysis model to obtain a finite element model; The third unit is used to periodically solve the solder point stress-strain field data sequence and its corresponding time information based on the finite element model using a nonlinear iterative algorithm; establish a modified Coffin-Manson model that considers the temperature-stress-time coupling effect based on the solder point stress-strain field data sequence and its corresponding time information, calculate the fatigue damage increment corresponding to the solder point stress-strain field data sequence using the modified Coffin-Manson model, accumulate the fatigue damage increment in real time to obtain the cumulative fatigue damage value, and evaluate the thermal fatigue state of the integrated circuit packaging structure according to the change rate of the cumulative fatigue damage value.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images