A high temperature test method for a power semiconductor device

Abstract

The application discloses a high-temperature test method of a power semiconductor device, and relates to the technical field of quantum thermal control, which comprises the following steps: collecting temperature field and surface conductivity data of the edge region of the power semiconductor device in real time, and synchronously applying electric stress of a specified working voltage to the power semiconductor device; reconstructing an internal thermal displacement field based on the temperature field, generating a calibrated thermal displacement field, and calculating the topological index of the edge region in combination with the surface conductivity data; generating a thermal failure risk distribution map according to the spatial position of the peak value of the thermodynamic entropy production rate, and locking the high-entropy variable power dense region; dynamically adjusting the electric stress application direction in association with the high-entropy variable power dense region, and outputting a thermal failure diagnosis report after completing the full-cycle high-temperature test. Through the collaborative mapping of the quantum-calibrated thermal displacement field and the tensor field of the classical temperature gradient, the application realizes the real-time positioning of the peak value of the thermodynamic entropy production under high-temperature working conditions at the submicron level.

Description

Technical Field

[0001] The present invention relates to the field of quantum thermal control technology, in particular to a high-temperature testing method for power semiconductor devices. Background Art

[0002] Thermal failure diagnosis of power semiconductor devices has evolved from traditional thermal imaging to multi-field coupling analysis. Existing methods are mainly based on using infrared thermal imagers to collect surface temperature field distribution and combining conductivity measurements to construct thermal-electric coupling models. In recent years, the application of quantum thermodynamics theory has gradually emerged. The lattice vibration spectrum is inverted through the phonon dispersion relation to predict the thermal displacement field, and the thermal transport edge state is described by topological invariants. Synchronous technologies such as X-ray diffraction lattice measurement and molecular dynamics simulation can verify local heat flow characteristics. Thermal failure risk assessment relies on the calculation of the spatial distribution of entropy production and the characterization of material microstructure. Electric field safety regulation is achieved based on pulse decoherence time calibration.

[0003] Current technology is unable to quantify the quantum-classical thermal behavior correlation in high-temperature areas in real time when the device is energized. Traditional temperature field reconstruction lacks quantum non-local coupling correction, resulting in insufficient sensitivity of the topological Chern number and difficulty in capturing the critical point of thermodynamic phase transition at the sub-microsecond scale. The existing entropy production rate model does not integrate the dynamic characteristics of heat flow vortices and the entropy change gradient tensor, and the mislocation rate of thermal failure risk is high, which cannot meet the needs of accurate diagnosis of junction temperature of third-generation semiconductors. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a high-temperature testing method for power semiconductor devices to solve the problems of topological invariant drift and thermodynamic entropy generation peak positioning misalignment caused by quantum thermal fluctuation effects under high-temperature conditions of power semiconductor devices.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a high-temperature testing method for a power semiconductor device, comprising: collecting temperature field and surface conductivity data of an edge region of the power semiconductor device in real time; and simultaneously applying an electrical stress of a specified operating voltage to the power semiconductor device;

[0008] The internal thermal displacement field is reconstructed based on the temperature field to generate a calibrated thermal displacement field, and the topological Chern number of the edge region is calculated in combination with the surface conductivity data.

[0009] When the absolute deviation of the topological Chern number of the edge area and the preset benchmark value exceeds the preset deviation threshold, the entropy flow monitoring protocol is activated, otherwise the static parameter test continues;

[0010] According to the activated entropy flow monitoring protocol, the local heat flow balance equation is constructed in combination with the calibrated thermal displacement field, the heat flow curl distribution and entropy gradient field are calculated in real time, and the spatial position of the thermodynamic entropy production peak is located;

[0011] Generate a thermal failure risk distribution map based on the spatial position of the thermodynamic entropy production peak, and lock in the high entropy change power intensive area;

[0012] The direction of electrical stress application is dynamically adjusted in the associated high-entropy change power-intensive area, and a thermal failure diagnosis report is output after completing the full-cycle high-temperature test.

[0013] As a preferred solution of the high temperature testing method of the power semiconductor device of the present invention, wherein: the synchronous application of electrical stress of a specified working voltage to the power semiconductor device is carried out in the following specific steps:

[0014] The temperature field of the edge area of ​​the power semiconductor device is collected in real time through the temperature acquisition equipment, and the surface conductivity data is collected synchronously through the surface conductivity sensor;

[0015] After the acquisition is completed, the control unit applies an electrical stress of a specified working voltage to the electrodes of the power semiconductor device, and enters the energized working state.

[0016] As a preferred solution of the high-temperature testing method for power semiconductor devices of the present invention, wherein: the edge region topological Chern number is calculated by combining surface conductivity data, and the specific steps are as follows:

[0017] Based on the real-time collected temperature field, the internal thermal displacement field is reconstructed through the pre-trained quantum thermal fluctuation model to obtain the thermal displacement wave function;

[0018] The thermal displacement wave function is integrated with the real-time surface conductivity data to generate a calibrated thermal displacement field.

[0019] The calibrated thermal displacement field is combined with real-time surface conductivity data to perform dynamic Berry curvature integral operations and calculate the topological Chern number of the edge region.

[0020] As a preferred solution of the high temperature testing method for power semiconductor devices of the present invention, wherein: the entropy flow monitoring protocol is activated, otherwise the static parameter test is continued, the specific steps are as follows:

[0021] Obtain the topological Chen number of the current edge area, compare it with the absolute deviation of the preset benchmark value, and generate a comparison result;

[0022] When the comparison result shows that the topological Chern number exceeds the preset deviation threshold, it is determined that the power semiconductor device has entered a non-equilibrium state, and the entropy flow monitoring protocol is activated to perform a thermodynamic state scan on the edge area of ​​the power semiconductor device;

[0023] When the comparison result shows that the topological Chern number does not exceed the preset deviation threshold, the static parameter test is continued.

[0024] As a preferred solution of the high-temperature testing method for power semiconductor devices described in the present invention, the continued execution of static parameter testing refers to a standardized test of continuously monitoring and recording temperature field, surface conductivity data and topological Chern number parameters under a stable working state of the device.

[0025] As a preferred solution of the high-temperature testing method for power semiconductor devices of the present invention, the specific steps of locating the spatial position of the peak value of thermodynamic entropy production rate are as follows:

[0026] According to the real-time collected temperature field, the temperature field gradient is obtained by using the Sobel operator gradient algorithm;

[0027] According to the activated entropy flow monitoring protocol, the calibrated thermal displacement field is tensor-coupled with the temperature field gradient to construct the local heat flow balance equation and generate the heat flux density field;

[0028] The heat flux density field is calculated using the curl operator to calculate the heat flux vortex distribution, and the entropy gradient field is calculated by combining the temperature field gradient;

[0029] The heat flux vortex distribution and entropy gradient field are input into the pre-trained field coupling analysis model to output the spatial distribution of entropy production rate;

[0030] Multi-scale convolution is performed on the spatial distribution of entropy production rate to locate the spatial position of the peak of thermodynamic entropy production rate.

[0031] As a preferred solution of the high-temperature testing method for power semiconductor devices of the present invention, wherein: the high entropy change power intensive area is locked, and the specific steps are as follows:

[0032] Based on the spatial position of the thermodynamic entropy production rate peak, the neighborhood range of each thermodynamic entropy production rate peak is dynamically delineated;

[0033] In each defined neighborhood, the heat flow curl, entropy change flux and gradient characteristics are extracted simultaneously, and the risk transfer weight between neighborhoods is obtained by combining the heat transfer path curvature and the transient entropy change rate.

[0034] Taking the risk transfer weight as the connection relationship, fractional-order risk diffusion is performed on the topological network constrained by fractal dimension to generate the risk field;

[0035] Perform time-domain integration of the risk field to generate a thermal failure risk distribution map, and verify the thermal failure risk distribution map using a confidence evaluation function;

[0036] When the confidence level of the thermal failure risk distribution map exceeds a preset confidence threshold, the high entropy change power intensive area is locked.

[0037] As a preferred solution of the high-temperature testing method for power semiconductor devices of the present invention, wherein: after completing the full-cycle high-temperature test, a thermal failure diagnosis report is output, and the specific steps are as follows:

[0038] Associate the spatial layout of high-entropy change power-intensive areas, lay out quantum dot arrays and apply microsecond pulse electric fields, and calibrate the safe application area of ​​the electric field based on the quantum decoherence time;

[0039] According to the safety area calibration results, the electrical stress application direction is dynamically adjusted through the swarm decision algorithm to generate the optimal electrical stress application direction;

[0040] When executing the optimal electrical stress application direction, the four-dimensional failure signal coordinates are captured through the biomimetic sensing network;

[0041] The spatial overlap of the calibrated electric field safety application area and the four-dimensional failure signal coordinates is compared. When the spatial overlap exceeds the preset overlap threshold, a thermal failure diagnosis report is output; otherwise, the electric field safety application area is recalibrated.

[0042] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the high-temperature testing method for power semiconductor devices as described in the first aspect of the present invention is implemented.

[0043] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the high-temperature testing method for power semiconductor devices as described in the first aspect of the present invention is implemented.

[0044] The beneficial effects of the present invention are: through the coordinated mapping of the tensor field of the quantized calibration thermal displacement field and the classical temperature gradient, submicron-level real-time positioning of the thermodynamic entropy production peak under high-temperature conditions is achieved, the quantum-classical heat flow coupling mechanism is used to accurately separate the lattice vibration quantum state and the macroscopic heat diffusion effect, the heat flow curl distribution analysis capability is improved by an order of magnitude, the configuration entropy gradient and the transient temperature driving force are synchronously integrated, the spatial resolution of the entropy change gradient field reaches the microscopic scale, the ellipsoid neighborhood is dynamically delineated to avoid the edge diffusion error of traditional thermal field analysis, and the positioning robustness of the core area of ​​thermal failure under extreme temperature conditions is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 Flowchart of a high-temperature testing method for power semiconductor devices.

[0047] Figure 2 Flowchart for calculating the topological Chern number of the edge region.

[0048] Figure 3 Flowchart for locating the spatial position of the peak of thermodynamic entropy production.

[0049] Figure 4 Flowchart for locking high entropy power-intensive areas. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0053] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides a high-temperature testing method for a power semiconductor device, comprising the following steps:

[0054] S1. Real-time collection of temperature field and surface conductivity data of the edge area of ​​the power semiconductor device, and simultaneous application of electrical stress of a specified working voltage to the power semiconductor device.

[0055] Furthermore, the temperature field of the edge area of ​​the power semiconductor device is collected in real time through the temperature acquisition equipment, and the surface conductivity data is collected synchronously through the surface conductivity sensor;

[0056] Specifically, an infrared thermal imager is used to scan the edge area of ​​the power semiconductor device, and the temperature value is recorded pixel by pixel to form a temperature field. Simultaneously, a four-point probe array is used to contact the surface of the power semiconductor device. A constant current is applied between adjacent probes and the voltage drop is measured. The surface conductivity data is converted according to Ohm's law. The two devices are triggered for collection through a nanosecond synchronized clock, and the temperature field and surface conductivity data share the same spatial coordinate system and timestamp.

[0057] After the collection is completed, the control unit applies an electrical stress of a specified operating voltage to the electrodes of the power semiconductor device, and enters a charged operating state.

[0058] Specifically, after the collection is completed, the control unit connects the source electrode and the drain electrode of the power semiconductor device, applies a trigger pulse to the gate electrode, and the direct current power supply delivers a specified operating voltage to the electrodes of the power semiconductor device. When the monitored current value reaches a rated operating current threshold, the power semiconductor device enters a charged operating state.

[0059] It should be noted that the specified operating voltage refers to the maximum allowed continuous operating voltage marked in the rated parameter table of the power semiconductor device, which usually corresponds to the drain-source breakdown voltage or reverse repetitive peak voltage value in the device data manual.

[0060] The rated operating current threshold is set based on the thermal limit of the power semiconductor device package and the current carrying capacity of the semiconductor material, for example, the typical rated operating current threshold of a 1200V SiC MOSFET is 100A.

[0061] S2, based on the temperature field, reconstruct the internal thermal displacement field to generate a calibrated thermal displacement field, and calculate the edge region topological index in combination with the surface conductivity data.

[0062] Furthermore, based on the real-time collected temperature field, the internal thermal displacement field is reconstructed through a pre-trained quantum thermal fluctuation model to obtain a thermal displacement wave function.

[0063] It should be noted that a plurality of power semiconductor device samples are obtained, and an infrared thermal imager is used in a constant temperature vacuum chamber to measure the temperature field of the samples under different thermal stress conditions, and an X-ray diffractometer is used to record the lattice vibration spectrum simultaneously. The atomic motion trajectory of the sample is calculated by molecular dynamics simulation, and the phonon dispersion relationship parameters are extracted; the temperature field distribution data is used as input features, the phonon effective mass and the electron-phonon coupling density are used as labels, the three-dimensional finite element grid divides the space region, and the non-local coupling coefficient in the quantum thermal fluctuation model is optimized by the gradient descent method. The loss function is to minimize the root mean square error between the experimental temperature field distribution and the reconstructed temperature field distribution, and the quantum thermal fluctuation model with the ability to map the temperature field to the quantized thermal displacement wave function is obtained.

[0064] Specifically, the internal thermal displacement field is reconstructed through a pre-trained quantum thermal fluctuation model to obtain a thermal displacement wave function, and the expression is:

[0065] ;

[0066] In the formula, represents the position vector of the observation point The thermal displacement wave function at time , is expressed as represents the position vector of the observation point, Indicates time, represents Planck's constant, represents the Laplace operator, Represents the position vector at the observation point collected in real time and time The temperature field at represents the phonon effective mass, represents the Boltzmann constant, represents the electron-phonon coupling density, represents the nonlocal coupling coefficient, represents the normalization factor, Represents the spatial region Perform integration, represents a spatial region, represents the position vector of the source point, represents the exponential function, Represents the position vector of the observation point and the position vector of the source point The square of the Euclidean distance, represents the squared difference, represents the quantum decoherence length, represents the square of the standard deviation of the quantum decoherence length, The position vector representing the source point In time The temperature fluctuations at Indicates temperature fluctuations, The volume element representing the position vector of the source point.

[0067] It should be noted that the nonlocal coupling coefficient comes from the quantum correction of the electron-phonon scattering cross section, and the example value is 0.35; the normalization factor is derived from the Gaussian kernel function after normalization of the full-space integral, and the example value is 0.564.

[0068] Specifically, the thermal displacement wave function is subjected to a canonical connection integration operation with the real-time surface conductivity data to generate a calibrated thermal displacement field;

[0069] ;

[0070] Where, represents the calibrated thermal displacement field, represents the imaginary inner product integral, represents the thermal displacement wave function and time The partial derivative of represents the thermal Green's function operator, represents the tensor direct product, represents the square of the variance of the real-time surface conductivity data fluctuation over time, represents the variance, represents real-time surface conductivity data, represents the determinant operation, represents the surface metric tensor, represents the square root of the metric determinant, Indicates the integration of the preset measurement path. Indicates the preset measurement path, represents the differential operator, represents the projection vector, represents the gradient operator.

[0071] Specifically, the calibrated thermal displacement field is combined with real-time surface conductivity data to perform dynamic Berry curvature integral operations and calculate the topological Chern number of the edge region.

[0072] ;

[0073] Where, represents the edge region topological Chern number, represents the normalization coefficient, Indicates along the target area Closed integration is performed on the boundary of Indicates the target area, represents the boundary operator, represents the Fermi level, represents the curl operator, Indicates Berry contact, represents the dynamic Berry curvature, Represents the vertical path differential element.

[0074] It should be noted that the normalization coefficient comes from the topological quantization condition, and the example value is 0.15915.

[0075] S3. When the absolute deviation of the topological Chern number of the edge area and the preset reference value exceeds the preset deviation threshold, the entropy flow monitoring protocol is activated, otherwise the static parameter test is continued.

[0076] Furthermore, the topological Chen number of the current edge area is obtained, and the absolute deviation is compared with the preset reference value to generate a comparison result;

[0077] It should be noted that the absolute deviation of the preset reference value refers to the absolute value of the algebraic difference between the topological Chern number of the current edge region obtained in the comparison stage and the preset reference value.

[0078] Specifically, the current edge region topological entropy is obtained, and a preset reference value is taken as the average value of five measured topological entropies in the initial health state of the power semiconductor device. In the comparison stage, the absolute value of the algebraic difference between the current edge region topological entropy and the preset reference value is obtained, and the absolute deviation value in the form of a positive real number is output. For example, if the absolute deviation value reaches 3%, a thermal failure warning sign is triggered, and if it is lower than 0.5%, it is considered to be in a normal state.

[0079] When the comparison result shows that the topological entropy exceeds the preset deviation threshold, it is determined that the power semiconductor device enters a non-equilibrium state, and the entropy flow monitoring protocol is activated to perform thermodynamic state scanning on the edge region of the power semiconductor device.

[0080] It should be noted that the preset deviation threshold is set based on the upper limit of the statistical fluctuation of the topological entropy of the power semiconductor device in the standard aging test. The example value is 3%.

[0081] Specifically, when the absolute deviation value of the edge region topological entropy reaches the preset deviation threshold, it is determined that the power semiconductor device enters a non-equilibrium state, and the entropy flow monitoring protocol is immediately activated. The control unit drives the thermal imager and the four-point probe array to perform thermodynamic state scanning on the edge region of the power semiconductor device. The scanning process combines real-time temperature field and surface conductivity data.

[0082] When the comparison result shows that the topological entropy does not exceed the preset deviation threshold, the static parameter test is continued.

[0083] Specifically, when the comparison result shows that the absolute deviation value of the edge region topological entropy does not exceed the preset deviation threshold, the control unit keeps the power semiconductor device in an on-state, and continues to perform the static parameter test process. It should be noted that continuing to perform the static parameter test means continuously monitoring and recording the temperature field, surface conductivity data and topological entropy parameters in the standard test under the stable working state of the device; the static parameter test process includes collecting temperature field and surface conductivity data at fixed time intervals, recalculating the current edge region topological entropy and continuously comparing, maintaining the specified working voltage of the power semiconductor device, updating the thermal displacement wave function by the quantum thermal fluctuation model, and cyclically executing the standard connection integral operation and dynamic Berry curvature integral operation, and the topological entropy deviation result is refreshed and recorded to the test log in real time.

[0084] S4, according to the activated entropy flow monitoring protocol, a local heat flow balance equation is constructed combining the calibrated thermal displacement field, the heat flow curl distribution and the entropy change gradient field are calculated in real time, and the peak space position of the thermodynamic entropy production rate is located.

[0085] Further, the temperature field gradient is obtained by the Sobel operator gradient algorithm according to the real-time collected temperature field.

[0086] Specifically, the temperature field collected in real time is input into the Sobel operator gradient algorithm, and a fixed three-dimensional convolution kernel is used to process the temperature field. The convolution kernel contains weight matrices in three dimensions: horizontal, vertical, and depth. Each weight matrix is ​​configured with specific coefficients in a centrally symmetrical manner. Convolution operations are performed on the temperature field along the x-axis, y-axis, and z-axis respectively. The convolution operation outputs the gradient component of the temperature field in the x-direction, the y-direction, and the z-direction. The three directional gradient components are synthesized to form a temperature field gradient in vector form. The temperature field gradient serves as the basic input data for the calculation of the heat diffusion term and the entropy change gradient field in the local heat flow balance equation.

[0087] According to the activated entropy flow monitoring protocol, the calibrated thermal displacement field is tensor-coupled with the temperature field gradient to construct the local heat flow balance equation and generate the heat flux density field;

[0088] It should be noted that the tensor coupling operation is realized through the outer product operation in the Cartesian coordinate system. The vector components of the calibrated thermal displacement field and the vector components of the temperature field gradient are combined according to the Einstein summation rule. The combination result is substituted into the energy storage term on the left and the source term on the right of the local heat flow balance equation. The material density, constant-pressure specific heat capacity, and partial derivatives of the temperature field and time are used as known quantities, and the thermal conductivity tensor, piezoelectric coupling coefficient, material stiffness, calibrated thermal displacement field divergence and electric field strength are used as parameter variables. The local heat flow balance equation is solved to obtain the heat flux density field, which is expressed as a function of the spatial position vector.

[0089] Specifically, the calibrated thermal displacement field is tensor-coupled with the temperature field gradient to construct the local heat flow balance equation, which is expressed as:

[0090] ;

[0091] Where, represents the material density, represents the specific heat capacity at constant pressure, represents the partial derivative of the temperature field with time, represents the divergence operator, represents the thermal conductivity tensor, represents the temperature field gradient, represents the piezoelectric coupling coefficient, represents the material stiffness, represents the calibrated thermal displacement field divergence, Indicates the electric field strength.

[0092] It should be noted that the piezoelectric coupling coefficient is derived from the linear response relationship between mechanical stress and electric polarization intensity in the piezoelectric material, and the example value is 5.5.

[0093] The heat flux density field is calculated using the curl operator to calculate the heat flux vortex distribution, and the entropy gradient field is calculated by combining the temperature field gradient;

[0094] Specifically, the heat flux vortex distribution is calculated using the curl operator on the heat flux density field, and the expression is:

[0095] ;

[0096] Where, represents the heat flux vortex distribution, represents the transverse heat flux density field, represents the unit vector in the x direction of the Cartesian coordinate system, represents the unit vector in the y direction of the Cartesian coordinate system, represents the unit vector in the z direction of the Cartesian coordinate system, Representation Space Partial derivative operator in the coordinate direction, Representation Space Partial derivative operator in the coordinate direction, Representation Space Partial derivative operator in the coordinate direction, Represents the heat flux field in space The components of the coordinate direction, Represents the heat flux field in space The components of the coordinate direction, Represents the heat flux field in space The components of the coordinate direction, represents the heat flux field.

[0097] Specifically, the entropy gradient field is calculated by combining the temperature field gradient, and the expression is:

[0098] ;

[0099] Where, represents the entropy gradient field, Indicates the real-time temperature. represents the reference temperature gradient, represents the reference temperature, represents the coefficient of thermal expansion, represents the natural logarithm function, Indicates the ratio of real-time temperature to reference temperature. represents the entropy density, represents the entropy density gradient.

[0100] It should be noted that the thermal expansion coefficient comes from the change in atomic spacing with temperature caused by the anharmonicity of lattice vibration. The example value is .

[0101] The heat flux vortex distribution and entropy gradient field are input into the pre-trained field coupling analysis model to output the spatial distribution of entropy production rate;

[0102] It should be noted that power semiconductor device samples with known thermal failure states were prepared, and a thermal imager was used to record the temperature field of the samples under thermal stress, and surface conductivity data was collected simultaneously; the temperature field distribution was processed by the quantum thermal fluctuation model to reconstruct the thermal displacement wave function, and the canonical connection integral operation was performed in combination with the surface conductivity data to obtain a calibrated thermal displacement field. The Sobel operator gradient algorithm was used to obtain the temperature field gradient, and the heat flux density field was generated according to the local heat flux balance equation. The heat flux curl distribution and the entropy change gradient field were calculated, and molecular dynamics simulation software was used to output the true entropy production rate spatial distribution in the verification area. The heat flux vortex distribution and the entropy change gradient field were used as input features, and the true entropy production rate spatial distribution was used as the target label. The three-dimensional convolutional neural network architecture processed the feature mapping relationship, and the network weight parameters were optimized using the root mean square error loss function to obtain a field coupling analysis model capable of receiving the heat flux vortex distribution and the entropy change gradient field to output the entropy production rate spatial distribution.

[0103] Specifically, the field coupling analysis model receives the heat flux vortex distribution and the entropy change gradient field as input features, and performs entropy production rate kernel function calculation internally in the field coupling analysis model. The entropy production rate kernel function integrates the energy dissipation component of the heat flux vortex distribution and the irreversible process contribution of the entropy change gradient field. The entropy production rate kernel function calculation is based on the Clausius integral relationship, taking the square of the curl modulus of the heat flux vortex distribution as the numerator, the square of the absolute temperature of the temperature field as the denominator, and multiplying it by the thermal conductivity tensor. At the same time, the dot product term of the entropy change gradient field and the heat flux density field is superimposed as an additional dissipation source. The field coupling analysis model outputs the entropy production rate scalar value at each spatial position to form the spatial distribution of entropy production rate.

[0104] Multi-scale convolution is performed on the spatial distribution of entropy production rate to locate the spatial position of the peak of thermodynamic entropy production rate.

[0105] Specifically, the spatial distribution of entropy production rate is input into the multi-scale convolution algorithm. The multi-scale convolution algorithm uses convolution kernels of different sizes to perform convolution operations on the spatial distribution matrix of entropy production rate layer by layer. The first layer uses a small-size convolution kernel to capture detail features, the second layer uses a medium-size convolution kernel to extract regional features, and the third layer uses a large-size convolution kernel to identify global features. The convolution operation outputs a multi-resolution entropy production rate feature map. The multi-resolution entropy production rate feature map is superimposed and fused to generate an enhanced entropy production rate distribution. The global maximum value of the enhanced entropy production rate distribution is obtained, and the three-dimensional coordinate value of the position point corresponding to the maximum value is found through spatial coordinate search. The output three-dimensional coordinate value is marked as the spatial position of the thermodynamic entropy production rate peak.

[0106] S5. Generate a thermal failure risk distribution map based on the spatial position of the thermodynamic entropy production peak and lock in the high entropy change power intensive area.

[0107] Furthermore, based on the spatial position of the thermodynamic entropy production rate peak, the neighborhood range of each thermodynamic entropy production rate peak is dynamically delineated;

[0108] Specifically, the coordinates of the neighborhood origin are defined based on the spatial position of the thermodynamic entropy production peak. The main axis direction and semi-axis length of the neighborhood ellipsoid are determined based on the eigenvalue of the second-order derivative curvature matrix of the current thermodynamic entropy production spatial distribution. With the thermodynamic entropy production peak point as the center, the ellipsoidal neighborhood boundary is extended along the three main axis directions. The main axis extension distance is proportional to the square root of the inverse of the second-order derivative eigenvalue. The proportional factor is three times the spatial resolution value of the current temperature field distribution. The ellipsoid neighborhood range is dynamically adjusted in real time as the spatial distribution of the thermodynamic entropy production changes. For example, the main axis lengths of the ellipsoid neighborhood generated by the GaN device at the 478K hotspot peak position are 15μm, 9μm, and 5μm, respectively.

[0109] In each defined neighborhood, the heat flow curl, entropy change flux and gradient characteristics are extracted simultaneously, and the risk transfer weight between neighborhoods is obtained by combining the heat transfer path curvature and the transient entropy change rate.

[0110] Specifically, within the scope of each designated neighborhood ellipsoid, the distribution form of the heat flux curl field vector and the spatial characteristics of the entropy change flux are directly related to the entropy production diffusion direction, the heat transfer path curvature is manifested as the geometric curvature visual form of the heat flux density field streamline, the transient entropy change rate is reflected as the relative change trend of the spatial distribution of the entropy production rate at continuous moments, and the risk transfer weight between neighborhoods is formed by the coupling effect of three types of physical field characteristics: the spatial synergy intensity of the heat flux curl field and the entropy change flux, the inherent curvature of the heat transfer path, and the time evolution trend of the entropy production rate. The spatial distribution characteristics of thermal conductivity are used as a physical reference benchmark, and the weight results are directly generated through the spatial topological correlation characteristics of the field characteristics.

[0111] Taking the risk transfer weight as the connection relationship, fractional-order risk diffusion is performed on the topological network constrained by fractal dimension to generate the risk field;

[0112] Specifically, the risk transfer weight between neighbors is used as the connection strength parameter of the topological network, and it is executed on a topological network structure constrained by fractal dimension. The topological network nodes correspond to the center points of each neighborhood ellipsoid, and the network edges reflect the spatial adjacent relationship of the neighborhood. The fractional-order risk diffusion uses the Riemann-Liouville fractional-order derivative to define the risk transfer flow. The initial risk value is set to the thermodynamic entropy production rate value of each neighborhood center. The risk flow is transmitted along the network connection direction, and the transmission intensity is proportional to the product of the risk transfer weight and the fractional-order derivative operator. The diffusion process maintains the conservation of the total risk between the peak values ​​of the thermodynamic entropy production rate, and generates a continuously distributed risk field in the entire domain of power semiconductor devices.

[0113] Perform time-domain integration of the risk field to generate a thermal failure risk distribution map, and verify the thermal failure risk distribution map using a confidence evaluation function;

[0114] Specifically, the risk field is performed Riemann integration in the time dimension, and the risk field values of continuous time slices are accumulated to generate a thermal failure risk distribution map. The thermal failure risk distribution map in the historical high-confidence thermal failure sample library is called as a reference template. The Jaccard similarity index of the real-time thermal failure risk distribution map and the reference template is obtained. If the similarity index is lower than a preset similarity threshold, the risk field generation process is triggered again.

[0115] It should be noted that the preset similarity threshold is set based on the spatial coincidence degree statistical distribution of the historical high-confidence thermal failure sample library and the real-time map. The example value is 0.92.

[0116] When the confidence of the thermal failure risk distribution map exceeds a preset confidence threshold, the high-entropy variable power intensive region is locked.

[0117] It should be noted that the preset confidence threshold is set based on the Jaccard similarity index statistical characteristics of the historical thermal failure samples of the power semiconductor device. The example value is 0.92.

[0118] Specifically, when the output value of the confidence evaluation function of the thermal failure risk distribution map exceeds the preset confidence threshold, the control unit immediately identifies a peak coordinate point set in the thermal failure risk distribution map. The peak coordinate point set is simultaneously mapped to the real-time surface conductivity data and the thermal flow density field spatial distribution data. The high-entropy variable power intensive region is matched and positioned in the spatial position. The three-dimensional coordinate values corresponding to the peak points of the thermal failure risk distribution map are directly extracted. The three-dimensional coordinate values are used as the center position coordinates of the high-entropy variable power intensive region. The spatial geometric range boundary coordinates of the high-entropy variable power intensive region are locked.

[0119] S6, the dynamic adjustment of the electric stress application direction is associated with the high-entropy variable power intensive region. After the full-cycle high-temperature test is completed, a thermal failure diagnosis report is output.

[0120] Furthermore, the spatial layout of the high-entropy variable power intensive region is associated. A quantum dot array is laid out and a microsecond pulse electric field is applied. The electric field safe application region is determined according to the quantum decoherence time;

[0121] Specifically, after the spatial layout coordinates of the high-entropy variable power intensive region are locked, the quantum dot array deposition path is planned according to the spatial layout coordinates. The microprobe platform directly deposits colloidal quantum dots to form an array at the coordinate position. The pulse power supply applies a square wave pulse electric field with a duration of microseconds to the quantum dot array. The quantum dot fluorescence lifetime decay curve is recorded synchronously. The quantum decoherence time is obtained by measuring the fluorescence lifetime decay time constant. The safe boundary is determined by comparing the difference between the quantum decoherence time and the pulse electric field time width. When the pulse electric field time width is less than 55% of the quantum decoherence time, the region is marked as the electric field safe application region.

[0122] According to the safety area calibration results, the electrical stress application direction is dynamically adjusted through the swarm decision algorithm to generate the optimal electrical stress application direction;

[0123] It should be noted that the safe area calibration result refers to the safe electric field application area calibrated according to the quantum decoherence time.

[0124] Specifically, based on the safety area calibration results, multiple sets of electric stress application direction solution sets are initialized through the swarm decision algorithm. Each solution contains a combination of pulse voltage amplitude and electric field direction angle parameters. The swarm individuals update the current direction parameters based on the historical optimal solution, and at the same time receive guidance from the successful trajectory pheromones of neighboring individuals. Each round of iteration measures three evaluation indicators under each electric stress direction: quantum dot fluorescence uniformity index, thermal drift compensation amount, and thermodynamic entropy production rate decay rate. The weighted sum of the evaluation indicators is used as the direction fitness function value, and the weight coefficient is automatically adjusted with reference to the boundary constraint ratio of the electric field safety application area. The global optimal solution is screened through a roulette wheel selection mechanism. For example, when the fitness function change of the optimal direction solution is lower than the 0.5% threshold after three consecutive iterations, the optimal electric stress application direction is locked.

[0125] When executing the optimal electrical stress application direction, the four-dimensional failure signal coordinates are captured through the biomimetic sensing network;

[0126] It should be noted that power semiconductor devices with known thermal failure morphology are obtained as training samples, quantum dot arrays are arranged on the sample surface and a pulsed electric field with a thermal failure sensitive frequency is applied, a controllable ion concentration gradient distribution is generated through a microfluidic channel, and a mechanical stress field is applied simultaneously using nanoindentation. The laser interferometer accurately calibrates the three-dimensional spatial coordinates of the failure point, and the neural pulse encoder collects the quantum dot fluorescence phase response, the ion concentration gradient change rate and the stress-induced displacement to form a timestamp pulse sequence. The timestamp pulse sequence is input into the pulse neural network layer of the bionic sensing network, and the real space coordinates are used as supervision signals. The network weights are updated using the pulse time-dependent plasticity rule, and the error function is defined as the Minkowski distance between the predicted four-dimensional coordinates and the laser interferometer coordinates. The bionic sensing network is capable of converting stimulus responses into four-dimensional failure signal coordinates.

[0127] Specifically, when executing the optimal electric stress application direction, the microfluidic neuromorphology is deployed on the surface of the power semiconductor device to collect the ion flux density distribution. Each node couples the fluorescence oscillation phase information of the quantum dot array and the displacement response signal of the piezoresistive strain sensing network. The bionic sensing network neuron-like pulse encoder encodes the ion concentration gradient, fluorescence phase offset and strain displacement into a timestamp sequence. The timestamp sequence is input into the pulse nerve to perform spatiotemporal correlation matching, and the spatial coordinate fusion information is obtained through the synaptic weight matrix. The output signal contains three-dimensional spatial coordinate points and corresponding timestamps to form four-dimensional failure signal coordinates.

[0128] The spatial overlap of the calibrated electric field safety application area and the four-dimensional failure signal coordinates is compared. When the spatial overlap exceeds the preset overlap threshold, a thermal failure diagnosis report is output; otherwise, the electric field safety application area is recalibrated.

[0129] It should be noted that the preset coincidence threshold is set based on the statistical characteristics of the spatial matching between the safe application area of ​​the electric field and the four-dimensional failure signal in the historical thermal failure samples of power semiconductor devices, and the example value is 0.92.

[0130] Specifically, the spatial coordinate range of the calibrated electric field safety application area and the three-dimensional spatial position component of the four-dimensional failure signal coordinates are used to perform spatial overlap calculation. The spatial overlap is defined as the fraction of the intersection volume of the two coordinate sets to the total area. When the spatial overlap value exceeds the preset overlap threshold, the diagnostic control unit immediately triggers the thermal failure diagnosis report generation process. The thermal failure diagnosis report content includes the position coordinates of the high entropy change power intensive area, the quantum decoherence time experimental value, the thermodynamic entropy production rate peak data and the four-dimensional failure signal timestamp sequence. When the spatial overlap value is lower than the preset overlap threshold, the quantum dot array fluorescence lifetime measurement link is returned to recalibrate the electric field safety application area, and the swarm decision algorithm is re-executed to update the optimal electric stress application direction.

[0131] This embodiment also provides a computer device suitable for the high-temperature testing method of power semiconductor devices, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the high-temperature testing method of power semiconductor devices proposed in the above embodiment.

[0132] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0133] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the processor implements the high-temperature testing method for power semiconductor devices proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0134] In summary, the present invention realizes submicron-level real-time positioning of the thermodynamic entropy production peak under high-temperature conditions through: quantized calibration of the thermal displacement field and the tensor field collaborative mapping of the classical temperature gradient, and uses the quantum-classical heat flow coupling mechanism to accurately separate the lattice vibration quantum state and the macroscopic thermal diffusion effect, and the heat flow curl distribution analysis capability is improved by an order of magnitude, and the configuration entropy gradient and the transient temperature driving force are synchronously integrated. The spatial resolution of the entropy change gradient field reaches the microscopic scale, and the ellipsoid neighborhood is dynamically delineated to avoid the edge diffusion error of traditional thermal field analysis, thereby enhancing the positioning robustness of the core area of ​​thermal failure under extreme temperature conditions.

[0135] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A high-temperature testing method for power semiconductor devices, characterized in that: include, Real-time collection of temperature field and surface conductivity data at the edge of power semiconductor devices, and simultaneous application of electrical stress of a specified working voltage to the power semiconductor devices; The internal thermal displacement field is reconstructed based on the temperature field to generate a calibrated thermal displacement field, and the topological Chern number of the edge region is calculated in combination with the surface conductivity data. When the absolute deviation of the topological Chern number of the edge area and the preset benchmark value exceeds the preset deviation threshold, the entropy flow monitoring protocol is activated, otherwise the static parameter test continues; According to the activated entropy flow monitoring protocol, the local heat flow balance equation is constructed in combination with the calibrated thermal displacement field, the heat flow curl distribution and entropy gradient field are calculated in real time, and the spatial position of the thermodynamic entropy production peak is located; Generate a thermal failure risk distribution map based on the spatial position of the thermodynamic entropy production peak, and lock in the high entropy change power intensive area; The direction of electrical stress application is dynamically adjusted in the associated high-entropy change power-intensive area, and a thermal failure diagnosis report is output after completing the full-cycle high-temperature test.

2. The high-temperature testing method for a power semiconductor device according to claim 1, wherein: The specific steps of synchronously applying the electrical stress of the specified working voltage to the power semiconductor device are as follows: The temperature field of the edge area of ​​the power semiconductor device is collected in real time through the temperature acquisition equipment, and the surface conductivity data is collected synchronously through the surface conductivity sensor; After the acquisition is completed, the control unit applies an electrical stress of a specified working voltage to the electrodes of the power semiconductor device, and enters the energized working state.

3. The high-temperature testing method for a power semiconductor device according to claim 2, wherein: The specific steps for calculating the edge region topological Chern number by combining surface conductivity data are as follows: Based on the real-time collected temperature field, the internal thermal displacement field is reconstructed through the pre-trained quantum thermal fluctuation model to obtain the thermal displacement wave function; The thermal displacement wave function is integrated with the real-time surface conductivity data to generate a calibrated thermal displacement field. The calibrated thermal displacement field is combined with real-time surface conductivity data to perform dynamic Berry curvature integral operations and calculate the topological Chern number of the edge region.

4. The high-temperature testing method for a power semiconductor device according to claim 3, wherein: The entropy flow monitoring protocol is activated, otherwise the static parameter test is continued. The specific steps are as follows: Obtain the topological Chen number of the current edge area, compare it with the absolute deviation of the preset benchmark value, and generate a comparison result; When the comparison result shows that the topological Chern number exceeds the preset deviation threshold, it is determined that the power semiconductor device has entered a non-equilibrium state, and the entropy flow monitoring protocol is activated to perform a thermodynamic state scan on the edge area of ​​the power semiconductor device; When the comparison result shows that the topological Chern number does not exceed the preset deviation threshold, the static parameter test is continued.

5. The high-temperature testing method for a power semiconductor device according to claim 4, wherein: The continued execution of the static parameter test refers to a standardized test of continuously monitoring and recording the temperature field, surface conductivity data and topological Chern number parameters under a stable working state of the device.

6. The high-temperature testing method for a power semiconductor device according to claim 1, wherein: The specific steps of locating the spatial position of the thermodynamic entropy production peak are as follows: According to the real-time collected temperature field, the temperature field gradient is obtained by using the Sobel operator gradient algorithm; According to the activated entropy flow monitoring protocol, the calibrated thermal displacement field is tensor-coupled with the temperature field gradient to construct the local heat flow balance equation and generate the heat flux density field; The heat flux density field is calculated using the curl operator to calculate the heat flux vortex distribution, and the entropy gradient field is calculated by combining the temperature field gradient; The heat flux vortex distribution and entropy gradient field are input into the pre-trained field coupling analysis model to output the spatial distribution of entropy production rate; Multi-scale convolution is performed on the spatial distribution of entropy production rate to locate the spatial position of the peak of thermodynamic entropy production rate.

7. The high-temperature testing method for a power semiconductor device according to claim 6, wherein: The specific steps of locking the high entropy change power intensive area are as follows: Based on the spatial position of the thermodynamic entropy production rate peak, the neighborhood range of each thermodynamic entropy production rate peak is dynamically delineated; Within each defined neighborhood, the heat flow curl, entropy change flux, and gradient characteristics are extracted simultaneously, and the risk transfer weight between neighborhoods is obtained by combining the heat transfer path curvature and the transient entropy change rate. Taking the risk transfer weight as the connection relationship, fractional-order risk diffusion is performed on the topological network constrained by fractal dimension to generate the risk field; Perform time-domain integration of the risk field to generate a thermal failure risk distribution map, and verify the thermal failure risk distribution map using a confidence evaluation function; When the confidence level of the thermal failure risk distribution map exceeds a preset confidence threshold, the high entropy change power intensive area is locked.

8. The high-temperature testing method for a power semiconductor device according to claim 7, wherein: After completing the full cycle high temperature test, the thermal failure diagnosis report is output. The specific steps are as follows: Associate the spatial layout of high-entropy change power-intensive areas, lay out quantum dot arrays and apply microsecond pulse electric fields, and calibrate the safe application area of ​​the electric field based on the quantum decoherence time; According to the safety area calibration results, the electrical stress application direction is dynamically adjusted through the swarm decision algorithm to generate the optimal electrical stress application direction; When executing the optimal electrical stress application direction, the four-dimensional failure signal coordinates are captured through the biomimetic sensing network; The spatial overlap of the calibrated electric field safety application area and the four-dimensional failure signal coordinates is compared. When the spatial overlap exceeds the preset overlap threshold, a thermal failure diagnosis report is output; otherwise, the electric field safety application area is recalibrated.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the high-temperature testing method for a power semiconductor device according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the high-temperature testing method for a power semiconductor device according to any one of claims 1 to 8 are implemented.

Images