A method and system for predicting the crack propagation tendency of a heat sink using fractal geometry

Abstract

This invention belongs to the technical field of material physical property analysis, and relates to a method and system for predicting the crack propagation trend of heat sinks using fractal geometry. The method includes the following steps: generating a two-dimensional original matrix of the infrared temperature field; inverting the stress source deformation kernel matrix using a preset empirical Green's space derivative function to generate a six-component moment tensor; generating a principal stress azimuth vector; generating an anisotropic rectangular metric grid; generating a multifractal singular extremum clustering image of the thermal field; generating the steepest fractal gradient vector; and coupling the normalized principal stress azimuth vector and the steepest fractal gradient vector into a preset multi-field parameter feature weighted superposition model to output the current crack propagation amount and crack propagation direction vector. This invention solves the problem that existing nondestructive testing technologies suffer from uncertainty in predicting crack propagation paths and rates due to the failure to effectively integrate microscopic mechanical stress field and macroscopic heat dissipation field information.

Description

Technical Field

[0001] This invention belongs to the technical field of material physical property analysis, and relates to a method and system for predicting the crack propagation trend of heat sinks using fractal geometry. Background Technology

[0002] In engineering structures, especially critical components such as radiators subjected to complex thermo-mechanical coupling loads, the initiation and propagation of microcracks within the material are one of the main causes of structural failure. To ensure the safe operation of equipment, accurate monitoring and prediction of the propagation trend of these microcracks has become a persistent challenge in the field of structural health monitoring. The core issue is that crack propagation behavior is a complex process driven by multiple physical mechanisms, including local stress fields, material microstructure, and thermodynamic dissipation. A single physical observation method is insufficient to fully capture its dynamic evolution characteristics.

[0003] Currently, the industry's commonly adopted solution is to monitor the structural condition using various non-destructive testing (NDT) techniques. Among these, acoustic emission testing (AE) captures the transient elastic waves released during crack propagation, enabling real-time detection and location of crack events, and allowing for the inversion of the source mechanism based on waveform parameters. Another commonly used method is infrared thermal imaging, which monitors the temperature field distribution on the structural surface to identify thermal anomalies caused by localized energy dissipation or impeded heat flow due to crack propagation. These technologies provide certain technical means for crack detection and have been applied within their respective applicable ranges.

[0004] However, traditional methods also have certain limitations in application. While acoustic emission detection technology alone can determine the time and spatial location of crack events, it suffers from uncertainties in retrieving the crack's geometry and precise stress state, offering limited support for predicting subsequent propagation paths. On the other hand, relying solely on infrared thermal imaging to identify cracks is susceptible to interference from the macroscopic thermal background generated by equipment operation and environmental convection. The weak thermal signals generated by microcracks are often submerged in noise, leading to room for improvement in detection sensitivity and reliability. More importantly, these two technologies are typically used independently in traditional applications, failing to establish an intrinsic physical connection between microscopic mechanical events and macroscopic thermal field responses, thus limiting the ability to comprehensively predict crack propagation—a multi-physics coupled behavior.

[0005] Based on the above problems, the present invention aims to solve the problem that existing nondestructive testing technologies have uncertainties in predicting crack propagation paths and rates because they fail to effectively integrate information from microscopic mechanical stress fields and macroscopic heat dissipation fields. Summary of the Invention

[0006] In a first aspect, the present invention provides a method for predicting the crack propagation trend of a heat sink using fractal geometry, comprising the following steps: S1. Obtain the transient acoustic emission voltage time series and continuous infrared temperature field video stream of the heat sink. Extract the single frame image from the continuous infrared temperature field video stream based on the first arrival time of the transient acoustic emission voltage time series and generate the two-dimensional original matrix of the infrared temperature field. S2. Analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combine the preset empirical Green space derivative function to invert the stress wave source deformation kernel matrix and generate a six-component moment tensor. S3. Decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform plane projection, and generate the principal stress azimuth vector. S4. Using the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector, scale and rotate the preset isotropic square measurement cell mesh template to generate an anisotropic rectangular measurement mesh. S5. Align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, calculate the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derive the Lipschitz exponent field to generate a multifractal singular extremum clustering image of the thermal field. S6. Extract the outer closed contour line of the thermal field multifractal singular extremum clustering image, calculate the spatial derivative of the outer closed contour line based on the Lipschitz exponential field, and select the maximum modulus vector to generate the steepest fractal gradient vector. S7. The normalized principal stress azimuth vector and the steepest fractal gradient vector are coupled into the preset multi-field parameter feature weighted superposition model to output the current crack propagation amount and crack extension direction vector.

[0007] A further aspect of the present invention generates a two-dimensional original matrix of the infrared temperature field, comprising the following steps: The preset piezoelectric sensor array is activated to capture burst wave trains and output transient acoustic emission voltage time series; The preset infrared focal plane thermal imager is activated to capture surface thermal radiation images and generate a continuous infrared temperature field video stream. The transient acoustic emission voltage time series is analyzed using a preset long and short time window arrival determination algorithm to lock the first arrival time; Extract the single-frame image from the continuous infrared temperature field video stream whose timestamp matches the first arrival time, and convert it into a two-dimensional original matrix of the infrared temperature field.

[0008] A further aspect of the present invention generates a six-component moment tensor, comprising the following steps: The effective shock wave signal in the transient acoustic emission voltage time series is separated, the absolute time difference of arrival is read, and the three-dimensional absolute physical coordinates are defined by combining the preset sound velocity medium model. The absolute poloidal parameter and absolute amplitude of the first wave of transient acoustic emission voltage time series are tracked to assemble a stress wave source deformation core matrix that reflects the lattice tearing mode. The empirical Green space derivative function is loaded into the stress wave source deformation kernel matrix to perform wave source mechanism inversion iterative calculation, and the six-component moment tensor is generated by solving.

[0009] A further aspect of the present invention generates the principal stress azimuth vector, comprising the following steps: Transform the six-component moment tensor into a positive definite symmetric frame matrix, and separate three sets of independent orthogonal real eigenvalues ​​and their associated non-zero physical eigenvectors in space; Compare three sets of independent orthogonal real eigenvalues, select the largest and smallest micro eigenvalues, calculate the quotient of the smallest micro eigenvalue divided by the largest micro eigenvalue, and generate the ratio of deviatoric stress eigenvalues. Extract the spatial non-zero physical eigenvector corresponding to the maximum microscopic eigenvalue, and project the spatial non-zero physical eigenvector along the main fracture normal interface onto the preset reference two-dimensional plane field coordinate system to generate the principal stress azimuth vector.

[0010] A further aspect of the present invention generates an anisotropic rectangular measure mesh, comprising the following steps: Obtain an isotropic square measure cell mesh template, inject the deviatoric stress eigenvalue ratio as a geometric scaling boundary condition, and change the ratio of the geometric minor axis to the geometric major axis of the isotropic square measure cell mesh template to the deviatoric stress eigenvalue ratio. The principal stress azimuth vector is introduced as a two-dimensional rotation deflection kernel to construct a coordinate adaptive affine transformation matrix, which guides the mesh after the scale change to perform rigid body rotation, so that the mesh major axis boundary is parallel to the principal stress azimuth vector. An anisotropic rectangular measurement grid is generated using the rotated boundary outline coordinates.

[0011] A further aspect of the present invention generates a multifractal singular extremum clustering image of a thermal field, comprising the following steps: The anisotropic rectangular measurement grid is translated into the two-dimensional original matrix of the infrared temperature field. The effective grayscale analysis range is forcibly defined with the absolute physical coordinates in three-dimensional space as the geometric masking center, and the noise floor masking is truncated. Within the effective grayscale analysis range, a set of neighboring pixels is extracted as a fractal isolated computational domain. The nonlinear partition integral function corresponding to the steady-state temperature gradient of local lattice grid points within the fractal isolated computational domain is calculated. The scaling slope of the nonlinear partition integral function on the double logarithmic metric coordinate system is obtained, the Lipschitz exponential field is derived, the high-frequency fixed-point heat flux abrupt change group is quantized and filtered, and a multifractal singular extremum clustering image of the thermal field is generated.

[0012] A further aspect of this invention involves selecting the vector with the largest modulus to generate the steepest fractal gradient vector, comprising the following steps: The preset connected component calibration operator is used to traverse the multifractal singular extreme value clustering image of the thermal field, identify the extreme variation core clusters and extract the outer closed contour lines; Discrete state observation nodes are retained at the edge of the closed contour line. Second-order difference calculations are performed in the Lipschitz exponential field with sub-pixel grid precision. The Lipschitz spatial derivative coordinate determination matrix is ​​then demixed and separated. The orientation of the peak value of the modulus limit in the Lipschitz spatial derivative coordinate determination matrix and its absolute scalar value are determined, and the steepest fractal gradient vector is generated by geometrically oriented combination.

[0013] A further aspect of the present invention outputs the current crack propagation amount and crack propagation direction vector, including the following steps: Perform vector normalization and scaling calculations on the principal stress azimuth vectors to obtain the unit vector of the mechanical stress direction; The unit vector of mechanical stress direction is transformed into the left term of the operator, and the steepest fractal gradient vector is substituted as the right term of the operator. This is then loaded into the multi-field parametric feature weighted superposition model to perform thermal dissipation isomorphic interaction operation to generate a composite extended vector. The dot product magnitude scalar of the composite propagation vector is calculated and assigned to the current crack propagation amount. The multidimensional deflection direction of the composite propagation vector is used as the crack propagation direction vector.

[0014] A further aspect of the present invention involves locking the first arrival time, including the following steps: At each sampling point of the transient acoustic emission voltage time series, the average signal energy within a preset short time window and the average signal energy within a preset long time window are calculated. The average signal energy within the short time window is divided by the average signal energy within the long time window to generate the short-to-long time window energy ratio. By comparing the energy ratio of the short and long time windows with the preset arrival determination threshold, the timestamp corresponding to the sampling point where the energy ratio of the short and long time windows first exceeds the arrival determination threshold is locked as the first arrival time.

[0015] Secondly, the present invention provides a radiator crack propagation trend prediction system using fractal geometry, comprising the following modules: The multi-source physical data synchronous acquisition module is used to acquire the transient acoustic emission voltage time series and the continuous infrared temperature field video stream of the heat sink. Based on the first arrival time of the transient acoustic emission voltage time series, a single frame image in the continuous infrared temperature field video stream is extracted to generate a two-dimensional original matrix of the infrared temperature field. The acoustic emission source mechanism inversion module is used to analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combined with the preset empirical Green space derivative function, the stress wave source deformation kernel matrix is ​​inverted to generate a six-component moment tensor. The mechanical eigenfactor extraction module is used to decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform planar projection, and generate the principal stress azimuth vector. The anisotropic computational mesh generation module uses the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector to scale and rotate the preset isotropic square metric cell mesh template to generate anisotropic rectangular metric mesh. The thermal field multifractal analysis module is used to align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, it calculates the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derives the Lipschitz exponent field to generate a thermal field multifractal singular extremum clustering image. The steepest fractal gradient vector extraction module is used to extract the outer closed contour line of the thermal field multifractal singular extreme value clustering image. It calculates the spatial derivative of the outer closed contour line based on the Lipschitz exponential field and selects the vector with the largest modulus to generate the steepest fractal gradient vector. The crack propagation trend prediction module couples the normalized principal stress azimuth vector with the steepest fractal gradient vector into a preset multi-field parameter feature weighted superposition model, and outputs the current crack propagation amount and crack extension direction vector.

[0016] In summary, the present invention has the following beneficial technical effects: 1. This invention acquires acoustic emission and infrared radiation field data simultaneously, and inverts the six-component moment tensor of the acoustic emission source. The ratio of the principal stress azimuth vector to the decomposition of this tensor is used to construct an anisotropic computational grid. This mechanism utilizes the underlying material fracture mechanics information extracted from the acoustic emission signal to directly constrain and guide the analytical geometry of subsequent thermal field data, ensuring that the direction and scale of the thermal field analysis are no longer isotropic, but rather match the direction and intensity ratio of the physically real stress field.

[0017] 2. This invention utilizes an anisotropic rectangular metric grid to perform multi-scale distortion mapping on the original infrared temperature field matrix, enabling targeted selection and calculation of thermal image pixels based on mechanical constraints determined by acoustic emission events. This metric grid, whose morphology is isomorphic to the physical characteristics of cracks, can amplify weak thermal gradient signals consistent with crack direction and morphology during multifractal analysis, while suppressing isotropic background thermal noise unrelated to cracks. This achieves the separation and extraction of submerged thermal field multifractal singular extremum clustering images under strong thermal noise backgrounds, improving the sensitivity for identifying microcrack thermal features.

[0018] 3. This invention extracts the steepest fractal gradient vector from the multifractal singular extremum clustering image of the thermal field and performs tensor inner product mapping calculation with the principal stress azimuth vector, thereby fusing two physical quantities representing the macroscopic mechanical driving direction and the microscopic thermodynamic dissipation most unstable direction, respectively. The calculation model generates a composite propagation vector by weighted summing of the two vectors using a material-dependent coupling coefficient. Its modulus and direction are defined as the crack propagation rate and path, respectively. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0020] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0021] Figure 2 Structural schematic diagrams of embodiments of this application are disclosed. Detailed Implementation

[0022] The following is in conjunction with the appendix Figures 1-2 A preferred description of the present invention is provided below.

[0023] See attached document Figure 1 This invention proposes a method for predicting the crack propagation trend of heat sinks using fractal geometry, comprising the following steps: S1. Obtain the transient acoustic emission voltage time series and continuous infrared temperature field video stream of the heat sink. Extract the single frame image from the continuous infrared temperature field video stream based on the first arrival time of the transient acoustic emission voltage time series and generate the two-dimensional original matrix of the infrared temperature field. S2. Analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combine the preset empirical Green space derivative function to invert the stress wave source deformation kernel matrix and generate a six-component moment tensor. S3. Decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform plane projection, and generate the principal stress azimuth vector. S4. Using the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector, scale and rotate the preset isotropic square measurement cell mesh template to generate an anisotropic rectangular measurement mesh. S5. Align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, calculate the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derive the Lipschitz exponent field to generate a multifractal singular extremum clustering image of the thermal field. S6. Extract the outer closed contour line of the thermal field multifractal singular extremum clustering image, calculate the spatial derivative of the outer closed contour line based on the Lipschitz exponential field, and select the maximum modulus vector to generate the steepest fractal gradient vector. S7. The normalized principal stress azimuth vector and the steepest fractal gradient vector are coupled into the preset multi-field parameter feature weighted superposition model to output the current crack propagation amount and crack extension direction vector.

[0024] In one embodiment of the present invention, step S1 involves simultaneously acquiring the acoustic parametric flow of the substrate of the heat sink and the infrared radiation field of the heat dissipation surface to construct a closed loop of original multi-source physical data for coupling of underlying geometric features, including the following steps: A piezoelectric sensor array, constantly attached to the substrate of the target heat sink, is activated to capture burst wave trains, and a transient acoustic emission voltage time series containing high-frequency initial motion signals of micro-fractures is acquired and output. An infrared focal plane thermal imager is continuously activated to capture surface thermal radiation images of the heat sink's convection working surface at a predetermined integration frame rate, and mapped to generate a continuous infrared temperature field video stream that is tightly aligned with the time dimension. The first arrival time of the transient acoustic emission voltage time series is analyzed, and the corresponding single-frame digital image matching that arrival time is synchronously and uniquely extracted from the continuous infrared temperature field video stream, and directly converted into a two-dimensional original matrix of the infrared temperature field containing parameters of non-uniform distortion of the local thermal barrier.

[0025] Specifically, an integrated data acquisition and control system synchronously acquires the acoustic parametric flow of the heat sink substrate and the infrared radiation field of the heat dissipation surface, and coordinates the sensing devices deployed on the target heat sink to construct a data closed loop. The main component performing this step is an embedded processing unit, such as a hardware controller developed based on a field-programmable gate array (FPGA). This controller manages the timestamp alignment of its subordinate data acquisition devices through a synchronous clock protocol. The embedded processing unit activates a piezoelectric sensor array pre-coupled to the non-heat-transferring area of ​​the target heat sink substrate with a specific geometric layout. These sensors, such as resonant sensors with center frequencies in the range of 150kHz to 300kHz, are suitable for capturing high-frequency stress waves generated by microscopic fractures in metallic materials. The processing unit instructs a high-speed data acquisition card to acquire the analog voltage signal output by the piezoelectric sensor array at a sampling rate of no less than 1MS / s. The converted digital signal stream is temporarily stored in an internal high-speed cache, forming a digital waveform record that evolves over time, i.e., a transient acoustic emission voltage time series.

[0026] The processing unit continuously drives a long-wave infrared focal plane imager facing the main convective heat transfer surface of the heat sink to perform continuous imaging. The imager periodically captures the thermal radiation energy distribution on the heat sink surface according to a pre-configured integration frame rate, such as 100 frames per second, and converts it into a series of digital thermal images arranged in chronological order, aggregating them to form a continuous infrared temperature field video stream. The processing unit continuously performs real-time data stream analysis on the transient acoustic emission voltage time series. It uses a short-window energy to long-window energy ratio criterion algorithm to determine the arrival time of the first valid burst wave in the transient acoustic emission voltage time series. This algorithm calculates the ratio of the average signal energy in a short time window to the average signal energy in a longer time window at each sampling point. When this ratio first exceeds a preset trigger threshold, the timestamp corresponding to that sampling point is locked as the first arrival time. Once the first arrival time is uniquely determined, the processing unit indexes it in the storage sequence of the continuous infrared temperature field video stream based on this timestamp.

[0027] By comparing the timestamp of each infrared image frame with the first arrival time, the system extracts the single-frame digital image whose timestamp is closest to the first arrival time. The system directly outputs the pixel data array of this single-frame digital image as raw data. Each element value of this data array represents the radiation intensity or equivalent temperature at the corresponding spatial location, thereby generating a two-dimensional raw matrix of the infrared temperature field that encapsulates the parameters of the non-uniform distortion of the local thermal barrier at the moment the event occurred. This matrix will serve as the baseline input for subsequent analysis.

[0028] Determining the first arrival time requires calculating the ratio of short-window energy to long-window energy. Its definition is as follows: First time the condition is met At that time, the corresponding timestamp This is the first arrival time. This is the index of the current sampling point. The transient acoustic emission voltage time series is in the 1st... Voltage amplitude at each sampling point The number of sampling points for the short time window. The number of sampling points for the long time window, and . The preset arrival determination threshold, The sampling rate of the acoustic emission system is based on... Extract the corresponding frame and its frame index number from the continuous infrared temperature field video stream. Calculated using the following formula: in, The frame rate of the infrared thermal imager. This indicates the floor function.

[0029] The transient acoustic emission voltage time series is a discrete digital sequence obtained by analog-to-digital conversion of the continuous voltage signal output by the piezoelectric sensor using an acoustic emission data acquisition card at a specific sampling rate. The continuous infrared temperature field video stream is a series of two-dimensional digital matrices arranged in chronological order, generated by the image acquisition unit of the infrared thermal imager. The original two-dimensional matrix of the infrared temperature field refers to a specific frame in the aforementioned video stream that is synchronized with the acoustic emission event. Its dimension is consistent with the pixel resolution of the thermal imager sensor, and each element in the matrix is ​​an unsigned integer representing the infrared radiation intensity count value received by that pixel.

[0030] Short window length The value is determined based on its ability to cover the rising edge of the effective acoustic emission signal, assuming that the typical rise time of the signal in the application scenario is 10. s, and the sampling rate is 1MS / s, then It can be set to 10. Long window length The value of is chosen based on its ability to stably reflect the average energy level of background noise, and is usually taken as . It is 50 to 100 times that, and can be set to 500 here.

[0031] Portal determination threshold The setting balances sensitivity and anti-interference. Based on the experimental statistical results of a large number of background noise signals and typical microcrack signals, it is assumed that in the low-noise experimental environment of this invention, in order to effectively detect signals and control the false trigger rate below 0.1%, the threshold can be set to 8.

[0032] For example, suppose the sampling rate of the acoustic emission data acquisition system is... The frame rate of the infrared thermal imager is 1 MS / s. The system's short time window length is set to 200Hz. For 10 sampling points, the long time window length For 500 sampling points, the arrival determination threshold The value is 8. During a monitoring session, the processing unit, while analyzing the transient acoustic emission voltage time series data stream, detected that before the 35100th sampling point, the signal mean was 0, and its short-to-long time window energy ratio was... It fluctuates around 1.2. Starting from the 35102nd sampling point, the signal amplitude rapidly increases from 0.01V, as calculated by the processing unit. The value is 9.5.

[0033] because The system determines that the acoustic emission event has occurred for the first time and records this moment as the first arrival time, with its timestamp... Calculated as The system calculates the index number of the infrared image frame to be extracted based on this timestamp. The system extracts the single-frame digital image at index 7 from the continuous infrared temperature field video stream, which is the two-dimensional original matrix of the infrared temperature field, such as an integer array with 640 rows and 480 columns. Most of the element values ​​are between 3000 and 3100, but a count value of 3150 appears in a local area in the center of the matrix. This matrix completely encapsulates the thermal field distribution state of the radiator surface at the moment the acoustic emission event occurs.

[0034] In one embodiment of the present invention, step S2, reconstructing the forced stress elastic shock wavefront carried by the transient acoustic emission voltage time series, and inverting and decoding to obtain the six-component moment tensor that determines the underlying spatial topology mechanism, includes the following steps: The effective shock wave signal in the transient acoustic emission voltage time series is separated by a long-short time window arrival determination algorithm. The absolute arrival time difference of each channel is read and the three-dimensional absolute physical coordinates of the triggering shock source are defined by combining the acoustic velocity medium model. The absolute poloidal parameter of the first wave initial motion of the longitudinal wave of lattice cleavage in the transient acoustic emission voltage time series and the absolute amplitude of the displacement peak of the envelope surface are tracked. The stress source deformation kernel matrix reflecting the tearing and opening mode of the underlying lattice is assembled according to the Green transform inverse method. The empirical Green space derivative function of the heat sink solid material is loaded into the stress source deformation kernel matrix to perform the source mechanism inversion iterative calculation. The solution outputs a six-component moment tensor that uniquely expresses the current transient mechanical stress field distribution characteristics of microcrack rupture.

[0035] Specifically, this step uses the transient acoustic emission voltage time series acquired and stored in step S1 as input to reconstruct the physical source of the elastic shock wave. This process is executed by the processing unit, which first independently runs a long-short time window arrival determination algorithm on the voltage time series data recorded by each sensor channel in the piezoelectric sensor array to determine the first wave arrival time of the effective shock wave signal that is effectively separated from the background noise for each channel. The processing unit reads the first wave arrival times between each sensor channel distributed at different physical locations, calculates the absolute difference between them, and forms the absolute arrival time difference. Combined with the acoustic velocity medium model of the heat sink substrate called from a preset material property library, which stores the P-wave propagation velocity of the material... The processing unit constructs a system of nonlinear equations with the three-dimensional spatial coordinates of the shock source to be determined as unknowns. This system of equations is solved using iterative optimization algorithms, such as the Gauss-Newton method or the simplex method, until the residual of the solution is less than a preset convergence tolerance, thus obtaining the three-dimensional absolute physical coordinates that uniquely identify the microcrack event occurrence point.

[0036] The processing unit shifts its focus to the detailed characteristics of the acoustic emission waveform itself. The system tracks the transient acoustic emission voltage time series, identifying the initial motion characteristics of the fastest-propagating longitudinal wave component generated by lattice cleavage events. The system determines whether the voltage polarity of the first peak or trough is positive or negative, quantifying this as the absolute poloidal parameter describing the initial motion of compression or stretching. The system applies a Hilbert transform to the original voltage signal to calculate its instantaneous envelope, locates and reads the displacement peak from this envelope, and obtains the absolute amplitude of the displacement peak in physical displacement units after sensor sensitivity coefficient calibration. The absolute poloidal parameters of the initial motion of all sensor channels are integrated with the absolute amplitude of the displacement peak into an observation vector. Based on the Green's transform inverse method of solid elasticity theory, the system combines this observation vector with the geometric configuration data of the sensor array to initially construct a mathematical expression describing the static displacement discontinuity at the wave source. This expression formally constitutes an intermediate product, namely the stress wave source deformation kernel matrix.

[0037] To convert the stress wave source deformation kernel matrix into a mechanical quantity with clear physical meaning, the system performs the final wave source mechanism inversion operation. Based on the determined three-dimensional absolute physical coordinates and the coordinates of each sensor, the processing unit loads an empirical Green's space derivative function matching the solid phase material of the heat sink into the computational core. This empirical Green's space derivative function describes the displacement field generated by a unit couple source at the source point relative to the observation point in a specific material. The system constructs a linear inversion equation using this derivative function and the observation vector. By performing the least-squares solution of this equation, iteratively solving for the 3x3 symmetric tensor that best fits the observation data, this tensor is a six-component moment tensor that explicitly expresses the distribution characteristics and intensity of the equivalent body couple released by the current microcrack at the moment of rupture. Its six independent components completely define the mechanical mechanism of this microcrack rupture event.

[0038] During the shock source localization process, the coordinates of the shock source Determined by the following system of equations, It is the first The coordinates of each sensor, It is its time of arrival, with the j-th reference sensor: In the moment tensor inversion step, the first P-wave displacement observations from individual sensors Vectorization of the six-component moment tensor The relationship between them is represented as: in, It includes The vector, It is an empirical Green's space derivative component that depends on the direction cosine between the source and the sensor and the material properties. This is achieved by solving the overdetermined equations. You can get The least squares solution.

[0039] The long / short time window arrival determination algorithm is the same as the algorithm defined in step S1. Here, it is applied to each sensor channel to obtain its own arrival time. The absolute arrival time difference is the arrival time obtained by different channels. The algebraic difference between them: The sound velocity medium model is a data structure that stores the acoustic parameters of different materials at specific temperatures. For a typical aluminum alloy heat sink, the P-wave velocity... The range is typically 6100-6400 The initial absolute polar parameter is a sign parameter, where +1 represents compression (initial motion direction towards the sensor) and -1 represents tension (initial motion direction away from the sensor). The absolute amplitude of the peak displacement is the maximum displacement value after correction by the sensor transfer function. The empirical Green space derivative function is a pre-calculated numerical matrix, the value of which depends on the material properties of the heat sink and the geometric relationship between the source and the sensor. The six-component moment tensor is a second-order symmetric tensor. , representing the equivalent body couple on the crack surface, whose six independent components describe the relative magnitude and orientation of the shear and tensile / compressive components of the crack origin.

[0040] For example, the transient acoustic emission voltage time series is derived from an array of four piezoelectric sensors (numbered S1 to S4). The coordinates of the four sensors are known: S1(0, 50, 0), S2(50, 100, 0), S3(100, 50, 0), and S4(50, 0, 0), in mm. The heat sink is made of 6061 aluminum alloy, and its P-wave velocity... The current operating condition is set to 6320 m / s. After the processing unit runs the long and short time window arrival determination algorithm on the data from the four channels, the corrected arrival times are obtained as follows: s, s, s, s. The system is based on The absolute time difference of arrival is calculated based on the following: , , .

[0041] Substituting these values ​​into the nonlinear equations and solving them, we obtain the absolute physical coordinates in three-dimensional space as (48.5, 52.1, -8.2) mm. Analyzing the waveforms of each channel, we determine the absolute polar parameters of the initial motion of the first wave to be +1, -1, -1, and +1, respectively. The calculated absolute amplitudes of the peak displacements are 2.7, -1.9, -2.2, and 2.4 mm, respectively. (Units: mm) m. These amplitude values ​​constitute the observation vector. The system, based on the obtained source coordinates and known sensor coordinates, loads the empirical Green space derivative function of the aluminum alloy material to construct... matrix By solving the linear inversion problem, the final solution vector is reorganized into a six-component moment tensor. Its value is (unit: ): In one embodiment of the present invention, step S3, extracting the orthogonal mapping mechanical eigenfactors of the six-component moment tensor and extracting the set of oriented physical constraint parameters used to shape the forced distortion mesh, includes the following steps: The algebraic transformation of the six-component moment tensor into a diagonalized positive definite symmetric frame matrix is ​​used to establish the characteristic polynomial of the square matrix for equation root finding. This forces the separation of three sets of independent orthogonal real eigenvalues ​​along with their respective spatial non-zero physical eigenvectors. By comparing the absolute magnitudes of the three sets of independent orthogonal real eigenvalues, the maximum microscopic eigenvalue describing the strongest fracture opening mode in three-dimensional space and the minimum microscopic eigenvalue describing the thickness compression mode are selected. The real quotient of the minimum microscopic eigenvalue divided by the maximum microscopic eigenvalue is used to obtain the ratio of deviatoric stress eigenvalues.

[0042] The spatial non-zero physical characteristic vectors with the maximum microscopic eigenvalues ​​are extracted and projected along the main fracture normal interface onto the reference two-dimensional plane field coordinate system to generate the principal stress azimuth vector characterizing the true normal vector of the instantaneous tear fracture surface of the lattice.

[0043] Specifically, the six-component moment tensor obtained in step S2 is introduced as input data to extract the core directional physical constraint parameters. The processing unit performs an algebraic transformation operation to explicitly express the six-component moment tensor, consisting of six independent scalars, as a 3x3 positive definite symmetric frame matrix. The system constructs a characteristic polynomial equation about the eigenvalues ​​around this matrix and uses numerical algorithms, such as the Jacobi iteration method or QR decomposition, to perform root-finding operations on this cubic equation. The result of this operation is the forced separation of three sets of mutually independent and orthogonal real eigenvalues, and a spatial non-zero physical eigenvector uniquely associated with each set of eigenvalues. These three sets of eigenvalues ​​and eigenvectors completely describe the magnitude and direction of the principal stresses in the local stress field induced by the microcrack event.

[0044] The system compares three sets of independent orthogonal real eigenvalues. The processing unit sorts the eigenvalues ​​according to their algebraic values, selecting the largest eigenvalue as the maximum microscopic eigenvalue representing the strongest fracture propagation mode in three-dimensional space, and the smallest eigenvalue as the minimum microscopic eigenvalue representing the compression or minimum tension mode in the thickness direction. The system performs scalar division, dividing the minimum microscopic eigenvalue by the maximum microscopic eigenvalue. The dimensionless real quotient obtained is the deviatoric stress eigenvalue ratio. After calculating the ratio, the system focuses on the eigenvectors. From the three sets of eigenvectors, the system extracts the spatially non-zero physical eigenvector that coincides with or corresponds to the maximum microscopic eigenvalue. This three-dimensional vector points in physical space towards the normal direction of the microcrack propagation surface.

[0045] To apply it to two-dimensional thermal field analysis, the system orthogonally projects this three-dimensional vector along the principal fracture normal interface onto a preset reference two-dimensional plane field coordinate system, which is usually parallel to the imaging plane of the infrared thermal imager. The projection operation is achieved by retaining the two coordinate components of the three-dimensional vector in the reference two-dimensional plane field coordinate system and discarding the perpendicular component, ultimately generating a two-dimensional vector. This two-dimensional vector is the principal stress azimuth vector representing the projection of the true normal vector of the instantaneous tear fracture surface of the lattice onto the observation plane.

[0046] The process of finding eigenvalues ​​boils down to solving the characteristic equation: , It is the 3x3 matrix form corresponding to the six-component moment tensor. These are the eigenvalues ​​to be determined. It is an identity matrix, and the three roots obtained are: .

[0047] If the eigenvalues ​​are sorted as follows Maximum microscopic eigenvalue Minimal microscopic eigenvalues eigenvalue ratio of deviatoric stress The calculation formula is: If the maximum microscopic eigenvalue The corresponding non-zero physical eigenvectors in space are If the reference two-dimensional plane field coordinate system is the XY plane, then the principal stress azimuth vector obtained after projection is... for: The positive definite symmetric frame matrix is ​​the equivalent mathematical representation of the six-component moment tensor, used for standard linear algebra operations. Three sets of independent orthogonal real eigenvalues ​​are measures of the three principal stress components in the principal stress coordinate system; their values ​​and signs reflect whether the stress state is tensile or compressive. The maximum microscopic eigenvalue corresponds to the maximum principal tensile stress, the primary mechanical quantity driving crack opening. The minimum microscopic eigenvalue corresponds to the minimum principal tensile stress or the maximum principal compressive stress. The deviatoric stress eigenvalue ratio is a key dimensionless parameter, typically ranging from -1 to 1, used to quantify the type of crack initiation; for example, a ratio close to 0 may indicate an opening crack, while other values ​​(such as large positive / negative values) tend to indicate a shear crack. The spatially non-zero physical eigenvector is the unit vector defining the principal stress direction. The principal stress azimuth vector ultimately serves as input to the geometric transformation, determining the rotation angle of the anisotropic analysis mesh in the two-dimensional plane.

[0048] For example, the six-component moment tensor output in step S2 is used. Performing eigenvalue decomposition involves solving the characteristic equation of the matrix to obtain three sets of real eigenvalues, which are then sorted by value to obtain... N·m, N·m, N·m. Based on this, the system determines the maximum microscopic eigenvalue to be 1.401 N·m and the minimum microscopic eigenvalue to be 0.695 N·m. The system calculates the ratio of the minimum microscopic eigenvalue to the maximum microscopic eigenvalue to obtain the deviatoric stress eigenvalue ratio. .

[0049] The system extracts the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue of 1.401 N·m, with normalized coordinates of (0.812, 0.570, -0.129). Assuming the radiator surface is parallel to the XY plane, i.e., the reference two-dimensional planar field coordinate system is the XY plane, this three-dimensional vector is projected, and the z-component is discarded to generate the principal stress azimuth vector, with two-dimensional coordinates of (0.812, 0.570). The two output results are the scalar value 0.496 and the two-dimensional vector (0.812, 0.570).

[0050] In one embodiment of the present invention, step S4, which involves jointly rewriting the computational domain topology using the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vectors to output a highly anisotropic computational grid that is forced to conform to the phonon thermal truncation physical field of the underlying material, includes the following steps: A conventional isotropic square metric cell mesh template used for region bounding calculations is retrieved, and the ratio of the deviatoric stress eigenvalue is injected as a geometric scaling boundary condition. The dimensional ratio of the geometric minor axis to the geometric major axis of this mesh template is forcibly locked and changed to a scalar value of the deviatoric stress eigenvalue ratio. A coordinate adaptive affine transformation matrix based on two-dimensional space is constructed, and the principal stress azimuth vector is introduced as a two-dimensional rotation deflection kernel to guide the mesh with the dimensional ratio to perform rigid body rotation so that its extended major axis boundary is strictly parallel to the normal of the latent crack surface in physical space. The boundary outline coordinates after the topological morphology is reorganized are used to generate an anisotropic rectangular metric mesh for partitioning filtering and interception, thus obtaining an anisotropic rectangular metric mesh suitable for directional filtering of infrared images.

[0051] Specifically, by combining the deviatoric stress eigenvalue ratio and principal stress azimuth vector extracted in step S3, the geometric topology of the computational domain is rewritten to generate a computational mesh that is geometrically highly compatible with the phonon thermal truncation physical field induced by the underlying microcracks. This process is automatically executed by the processing unit, which retrieves the isotropic square metric cell mesh template for the region frame calculation from its internal memory. The processing unit uses the calculated deviatoric stress eigenvalue ratio as a mandatory geometric scaling boundary condition, directly applying it to the basic elements of the mesh template. This operation changes the dimensional ratio between the minor and major axes of each square element from the original 1:1 to a scalar value defined by the deviatoric stress eigenvalue ratio, thus generating rectangular mesh basic elements with a defined aspect ratio.

[0052] The system constructs a coordinate-adaptive affine transformation matrix based on two-dimensional space, using the angular information implicit in the principal stress azimuth vector as the two-dimensional rotation deflection kernel of this matrix. The processing unit applies this transformation matrix to guide the previously scaled rectangular mesh to perform a rigid body rotation. The rotation angle is calculated to ensure that the extended major axis boundaries of the transformed mesh cells are strictly parallel to the direction indicated by the principal stress azimuth vector in physical space, i.e., the normal direction of the latent crack surface. After topological remodeling, the system uses the set of all outer contour coordinate points constituting the mesh cell boundaries to generate the final analysis tool, namely the anisotropic rectangular metric mesh. The structural characteristics of this mesh allow it to function like a key in subsequent temperature field analysis; therefore, it is called the acoustic-thermal isomorphic analytical key used to isolate macroscopic convective temperature variation noise.

[0053] To perform mesh rotation, based on the principal stress azimuth vector Calculate the rotation angle : Then construct a two-dimensional rotation matrix. : Original coordinates of any node in the grid After rotation transformation, new coordinates are obtained. : The isotropic square measure cell mesh template is a digitized reference geometry composed of square elements with equal side lengths, whose aspect ratio is strictly defined as 1.0 before any transformation is applied. The minor and major geometric axes refer to the two orthogonal side lengths of the rectangles constituting the basic mesh elements after scaling by the deviatoric stress eigenvalue ratio. The coordinate adaptive affine transformation matrix is, in a specific embodiment, a pure rotation transformation matrix, whose parameters are uniquely determined entirely by the principal stress azimuth vector. The two-dimensional rotation deflection kernel is the rotation angle calculated using the principal stress azimuth vector. .

[0054] The anisotropic rectangular metric grid is the final output of this step, defining a series of rectangular regions with specific aspect ratios and rotation angles for selective analysis of the infrared temperature field data in step S5. The acoustic-thermal isomorphic analytical key is a functional description of the anisotropic rectangular metric grid, emphasizing its structural isomorphism with the mechanical state revealed by the acoustic emission source and the resulting heat conduction barrier.

[0055] For example, the processing unit receives the deviatoric stress eigenvalue ratio of 0.496 and the principal stress azimuth vector of (0.812, 0.570) output from step S3. The processing unit retrieves an isotropic square measure cell mesh template with a base element of 1x1 unit length, forcibly sets the ratio of the minor axis to the major axis of the elements in the template to 0.496, and deforms the base element into a rectangle with a minor axis of 0.496 units and a major axis of 1 unit. The processing unit calculates the rotation angle based on the principal stress azimuth vector (0.812, 0.570). The radius is approximately 35.08 radians. The system constructs a rotation matrix based on this and applies this rotation transformation to the deformed rectangular element mesh, ensuring that the major axis of each rectangular element is aligned with a 35.08-degree direction. For example, the vertex coordinates of the unrotated deformed rectangle are (1, 0.496). After rotation, the new coordinates... Calculated as ,as well as By transforming the coordinates of all grid nodes in this way, an anisotropic rectangular metric grid consisting of rectangular elements with an aspect ratio of 0.496 and an overall rotation of 35.08 degrees is generated. This grid is the acoustic-thermal isomorphic analytical key, which will be used for subsequent fine-tuning operations.

[0056] In one embodiment of the present invention, step S5 involves intervening in the conventional coarse-grained temperature variation data screening operation with an anisotropic rectangular metric grid, applying a multi-scale distortion mapping to the original two-dimensional matrix of the infrared temperature field, and separating and extracting the submerged thermal field multifractal singular extremum clustering image, including the following steps: An anisotropic rectangular metric grid is driven to translate to the two-dimensional original matrix of the infrared temperature field. The effective grayscale analysis range is forcibly defined with the absolute physical coordinates in three-dimensional space as the geometric masking center, and a noise masking truncation is implemented. The anisotropic rectangular metric grid after center registration slides across the effective grayscale analysis range to extract the neighborhood pixel set as a fractal isolated computational domain. The nonlinear partition integral function is calculated for the steady-state temperature gradient of the local lattice grid points after multi-scale difference. The scale slope of the nonlinear partition integral function on the double logarithmic metric coordinate system is obtained to derive the Lipschitz exponent field. The high-frequency fixed-point heat flux abrupt change group caused by the truncation of extremely fine cracks is quantized and filtered to generate a multifractal singular extremum clustering image of the thermal field characterizing the weak extrema of the fracture surface.

[0057] Specifically, using the anisotropic rectangular metric mesh generated in step S4, multi-scale distortion mapping analysis is performed on the two-dimensional original matrix of the infrared temperature field obtained in step S1 to separate and extract singular thermal field features caused by microcracks that are difficult to distinguish against the background of macroscopic thermal noise. The entire process is scheduled and executed by the processing unit. The processing unit drives the anisotropic rectangular metric mesh to perform a translation operation within the coordinate system of the two-dimensional original matrix of the infrared temperature field, so that its geometric center coincides with the two-dimensional projection point of the absolute physical coordinates in three-dimensional space determined in S2 on the infrared image plane. This coincidence point is the geometric masking center. This action forces the focus of the analysis to be locked in the source region of the acoustic emission event, and the area covered by the mesh is defined as the effective grayscale analysis range. Before performing subsequent calculations, the system performs a noise masking truncation on the pixel values ​​within this range, setting all grayscale values ​​below the preset noise threshold to zero or ignoring them, thereby eliminating interference introduced by the sensor's inherent noise.

[0058] The processing unit, based on the centrally registered anisotropic rectangular metric grid, performs multi-scale sliding truncation within its effective grayscale analysis range, using anisotropic rectangular cells of different sizes as window functions, i.e., fractal isolated computational domains, and traverses this range. Within each fractal isolated computational domain, the system calculates the steady-state temperature gradient of the neighboring pixel set contained within it, typically expressed as the difference between the maximum and minimum grayscale values ​​within that domain. The system then assigns a moment order to these steady-state temperature gradients of the local lattice grid points obtained through multi-scale differentiation. Calculate its corresponding nonlinear partition integral function. Let be the given moment order. Calculate its corresponding nonlinear partition integral function, and systematically obtain the nonlinear partition integral function in a double logarithmic metric coordinate system and its relationship with the scale. The scaling relation.

[0059] By analyzing log( ) and log( Linear regression analysis was performed, and the slope of the resulting line was determined as the Lipschitz exponent. The processing unit calculated the local Lipschitz exponent for each pixel within the effective grayscale analysis range, and these were aggregated into a Lipschitz exponent field image. By applying a screening threshold to this exponent field, the system quantified and filtered out regions with abnormally low exponent values. These regions correspond to high-frequency fixed-point thermal flux abrupt changes caused by extremely fine cracks interrupting the lattice phonon transport path. All pixels filtered by this threshold were retained, forming several clusters, ultimately generating a thermal field multifractal singular extremum cluster image that only highlights the weak extreme points of the fracture.

[0060] For a given moment order Nonlinear partition integral function According to scale Calculated using rectangular measurement units, assuming the first... Each size is The normalized measure within the measure unit is The partition function is calculated as the ratio of the temperature gradient within the specified unit to the sum of the temperature gradients of all units within the entire analysis range. Lipschitz Index scaling exponent of the partition function Related, The following relationship was obtained by fitting it onto logarithmic coordinates: For local computation, the system's goal is to be constant for each pixel. estimate The value reflects the degree of singularity of temperature measurements in the neighborhood centered at that point.

[0061] The geometric masking center is a two-dimensional coordinate point obtained by projecting the absolute physical coordinates of three-dimensional space onto the pixel plane of the two-dimensional original matrix of the infrared temperature field. The effective grayscale analysis range is defined by the boundaries of the pre-located anisotropic rectangular measure grid. Noise floor masking truncation is an image preprocessing operation, and its threshold is pre-calibrated based on the noise statistical characteristics of the infrared thermal imager under stable operating conditions. The neighborhood pixel set refers to all pixels falling within anisotropic rectangular measure cells of a specific size and location. The nonlinear partition integral function is a core mathematical tool in multifractal analysis, used to quantify the statistical moments of the measure distribution at different scales.

[0062] The Lipschitz exponent field is a two-dimensional matrix with the same size as the original infrared image. Each element is a local singularity index for the corresponding pixel. The smaller the exponent value, the stronger the signal singularity at that point and the more drastic the temperature gradient change.

[0063] For example, the processing unit receives the original two-dimensional matrix of the infrared temperature field, the absolute physical coordinates in three-dimensional space (48.5, 52.1, -8.2) mm obtained in step S2, and the anisotropic rectangular measurement grid generated in step S4. Assuming the pixel-to-physical-size correspondence of the infrared camera is 10 pixels / mm and the imaging plane is the XY plane, the geometric masking center obtained after projecting the three-dimensional coordinates is the pixel coordinates (485, 521). The generated anisotropic rectangular measurement grid is translated to the position centered on this pixel, defining an effective grayscale analysis range of, for example, 100x100 pixels. The system sets the noise floor masking cutoff value to 2900. Within this range, the system uses rectangular measurement units of multiple scales, from 2x2 to 32x32 pixels, with an aspect ratio of 0.496 and rotated by 35.08 degrees for analysis.

[0064] Taking (490, 525) as an example, the system calculates the temperature gradient measure within different scale neighborhoods centered on it, and obtains the nonlinear partition integral function value when the moment order q=2. This is achieved by applying the equation to Z(2, ) in the log-log coordinate system. )and Linear fitting is performed on the data points to obtain the slope. And from this, the Lipschitz exponent at that point is calculated. The value is 0.25. The system's preset singularity threshold is 0.4, because 0.25 is less than 0.4, indicating a strong thermal discontinuity at that point. The processing unit performs this calculation for each pixel within the effective grayscale analysis range, ultimately processing all pixels... Pixels with values ​​less than 0.4 are marked as 1, and the rest are marked as 0, thereby generating a thermal field multifractal singular extremum clustering image, which shows sparse pixel chains along a direction of about 35 degrees.

[0065] In one embodiment of the present invention, step S6, stripping the continuous distortion envelope contour line of the multifractal singular extremum clustering image of the thermal field, and tracing back to extract the steepest fractal gradient vector representing the material's thermodynamic dissipation exceeding the limit, includes the following steps: By applying the connected component calibration operator to traverse the main pixel distribution matrix of the thermal field multifractal singular extreme value clustering image, the outer closed contour line of the extreme variant core cluster composed of highly dense Lipschitz exponents is identified and screened.

[0066] All discrete observation nodes at the edge of the closed contour line are retained. The local small neighborhood is forced to continuously advance the second-order difference calculation for the tangent plane with sub-pixel grid precision. The Lipschitz spatial derivative coordinate measurement matrix characterizing the local position intensity of the heat flow anomaly interruption point is demixed and separated. The orientation of the peak value of the maximum modulus limit in the Lipschitz spatial derivative coordinate measurement matrix and its absolute scaling value are determined and captured. The steepest fractal gradient vector pointing to the edge point of the heat blockage is formed by the geometric directional combination of the vectors.

[0067] Specifically, the thermal field multifractal singular extremum clustering image is the direct processing object, aiming to extract key vector information representing the thermal dissipation limit breach. The processing unit applies, for example, a connected component calibration operator based on eight-neighborhood connectivity analysis, to globally traverse the binary pixel matrix of the thermal field multifractal singular extremum clustering image. This operator identifies and groups all spatially connected singular extremum points (regions with a pixel value of 1) into independent connected components. By comparing the total number of pixels contained in each independent connected component, the system automatically identifies and locks the connected component with the largest number of pixels as the extreme mutation core cluster. The system deploys a contour tracking algorithm, such as the Suzuki contour tracking algorithm, on this core cluster to track and capture the outermost boundary pixels of the cluster, extracting a series of discrete pixel coordinates that constitute the outer boundary of the cluster, i.e., the outer closed contour line.

[0068] The system uses each discrete observation node on the closed contour line as a base point. The original Lipschitz exponent field calculated in step S5 is used to obtain a more refined Lipschitz exponent distribution within the node and its immediate small neighborhood using sub-pixel grid precision interpolation methods, such as bilinear interpolation. For each observation node, the system performs second-order difference calculations in its small neighborhood, determining its local spatial gradient by calculating the partial derivatives of the Lipschitz exponent field in the x and y directions. This operation unmixes and separates the spatial derivative vector characterizing the intensity of local changes around the point of interruption of heat flow anomalies. All contour points and their corresponding spatial derivative vectors are stored together to form the Lipschitz spatial derivative coordinate determination matrix.

[0069] The processing unit performs the final judgment and capture operation on the Lipschitz spatial derivative coordinate measurement matrix, traversing each spatial derivative vector recorded in the matrix and calculating its Euclidean norm, i.e., the magnitude of the vector. By comparing all the magnitudes, it identifies and locks the vector whose magnitude reaches the limit maximization peak. The direction of this vector is the orientation, and its magnitude is the absolute scaling value. Through geometrically oriented combination, this vector is finally confirmed as the steepest fractal gradient vector pointing to the edge point of the material's thermal blocking failure.

[0070] When calculating the Lipschitz spatial derivative, for any point on the closed contour line... Its performance in the Lipschitz index field gradient in Approximate calculation using the finite difference method: steepest fractal gradient vector It is all contour points The gradient vector with the largest gradient magnitude: in, The Lipschitz index represents the set of all points along a closed contour line. For dimensionless pure numbers, coordinates This is an absolute scale mapped to the real physical plane, with units of length (mm). Partial derivatives. The unit is the reciprocal of length. gradient vector The units of measurement are all Its modulus unit is also The steepest fractal gradient vector in the final output Inherited from this unit, its physical meaning is the rate of change of the singularity index per unit length.

[0071] The connected component calibration operator refers to an image processing algorithm used to identify clusters of pixels in a binary image. The outer closed contour line is an ordered list of pixel coordinates describing the boundary of the connected region. The discrete state observation node is each pixel in this coordinate list. Subpixel grid precision interpolation is a technique that estimates function values ​​by mathematically interpolating between integer pixel coordinates to achieve higher positioning accuracy than the original image resolution. The Lipschitz spatial derivative coordinate determination matrix is ​​a data structure, typically an N×4 matrix, where N is the number of points on the contour line, and each row records the x and y coordinates of a point and the x and y components of the gradient vector at that point. The steepest fractal gradient vector is a two-dimensional vector whose direction indicates the direction of the most drastic change in the Lipschitz exponent, i.e., the direction of the most thermodynamically unstable dissipation; the magnitude quantifies the degree of this instability.

[0072] For example, the processing unit receives a clustering image of thermal field multifractal singular extrema presented as a sparse pixel chain along an approximately 35-degree direction. A connected component labeling operator identifies this pixel chain as a single connected component containing 38 pixels. A contour tracking algorithm traverses this connected component, generating a closed contour line containing 22 pixel coordinates, for example, points (490, 525), (491, 526), ​​(492, 527), etc. Taking point (491, 526) as an example, the system reads the values ​​around it in the original Lipschitz exponential field, assuming... , , , The system calculates the gradient components at that point based on this: The system repeats this calculation for all 22 contour points, storing the results in the Lipschitz spatial derivative coordinate determination matrix. After traversing the matrix and calculating the magnitude of each gradient vector, the system finds that the gradient vector at pixel (498, 531) has the largest magnitude, which is 0.092 (1 / pixel), and the corresponding vector components are (-0.065, -0.067). The steepest fractal gradient vector output is (-0.065, -0.067), with units per pixel. This vector indicates the direction in which heat dissipation is most likely to break through, and its intensity value is 0.092.

[0073] In one embodiment of the present invention, step S7, synthesizing and fusing the steepest fractal gradient vector and the principal stress azimuth vector to perform tensor Hamiltonian metric inner product mapping calculation, and outputting the physical extension rate and piecewise deflection migration path of fracture propagation in the current working condition across the entire domain, includes the following steps: The macroscopic fracture propagation tendency of the force field carried by the principal stress azimuth vector is calculated by vector normalization to obtain the unit vector of mechanical stress direction. The obtained unit vector of mechanical stress direction is converted into the left term of the operator and substituted into the steepest fractal gradient vector representing the microscopic failure tendency of thermo-mechanical coupling as the right term of the operator. The left and right terms are loaded into the multi-field parameter feature weighted superposition model to perform the thermo-dissipation isomorphic interaction operation. The magnitude of the composite vector generated by the multi-field parameter feature weighted superposition model is calculated as the current crack propagation amount. The multidimensional deflection direction left after the coupling inner product calculation is used as the crack propagation direction vector for the next physical stage of failure in the micro-lattice selection to complete the closed-loop holographic prediction.

[0074] Specifically, the aim is to synthesize and fuse the key vectors extracted from the mechanical and thermal fields in the aforementioned steps, and to output a quantitative prediction of the physical extension rate and broken-line deflection migration path of fracture propagation across the entire domain under the current working condition through tensor Hamiltonian metric inner product mapping. The processing flow is executed by the processing unit. The system introduces the principal stress azimuth vector extracted in step S3, and performs vector normalization scaling calculation on the macroscopic fracture propagation tendency it carries. This calculation is achieved by dividing each component of the vector by the Euclidean norm of the vector to obtain a unit-length vector, i.e., a unit vector of mechanical stress direction. This factor eliminates arbitrary amplitude effects and retains only the pure physical tearing direction information.

[0075] The system uses the acquired unit vector of mechanical stress direction as the left term of the operator and introduces the steepest fractal gradient vector representing the tendency of microscopic damage caused by thermo-coupling obtained in step S6 as the right term of the operator. These two terms are loaded into a preset multi-field parameter feature weighted superposition model. This model performs a specific thermo-dissipation isomorphic interaction operation, the core of which is to construct a composite characterization vector that integrates two physical driving forces. This operation weights and sums the left and right terms according to their respective physical contributions to generate a new prediction vector.

[0076] The system deconstructs and assigns values ​​to the composite vector generated by the model to obtain the final prediction result. The system calculates the Euclidean modulus of the composite vector and assigns this scalar modulus (derived from the dot product) to the final failure process as the current crack propagation amount. The direction of the composite vector itself, i.e., the multidimensional deflection direction remaining after the coupled inner product calculation, is designated as the crack propagation direction vector for the next physical stage of failure in the microcrystalline lattice. Through this closed-loop operation, the system completes the transformation from multi-source physical signals to crack holographic prediction.

[0077] In this embodiment, the function of the multi-field parametric feature weighted superposition model is achieved by constructing a composite extended vector. To achieve this, the calculation formula is as follows: in, It is a normalized unit vector of mechanical stress direction. It is the steepest fractal gradient vector. and This is a preset material-related coupling coefficient. The current crack propagation rate. The magnitude of the composite vector determines the following: The crack propagation direction vector is determined by the vector. The direction is given.

[0078] Unit vector of mechanical stress direction This is a dimensionless unit vector. The steepest fractal gradient vector. The unit is the reciprocal of the length, for example... In order to make The two added terms have the same dimensions, and make the final expansion rate consistent. It has the length unit mm and the coupling coefficient. The unit must be length in mm, and the coupling coefficient... The unit must be the square of the length. . The unit is mm. The unit is After adding the two together The unit is mm, and its mold length The unit is also mm.

[0079] The mechanical stress direction unit vector is a two-dimensional unit vector, representing only the most probable direction of crack propagation tendency dominated by the macroscopic stress field. The multi-field parametric characteristic weighted superposition model is the computational framework used to couple the two physical vector fields. Through weighted summation, mechanical action and thermodynamic dissipation are linearly superimposed to simulate crack propagation behavior driven by both effects. Material coupling coefficient. and It is an empirical constant obtained by calibrating the fracture toughness, thermal conductivity, and thermal expansion coefficient of the radiator material through a large number of tensile fatigue fracture experiments. It quantifies the crack propagation length caused by the contribution of a unit mechanical stress direction and the contribution of a unit heat flow gradient.

[0080] The current crack propagation is a scalar value, predicting the physical distance the crack tip will extend forward in the next typical load cycle or per unit time. The crack propagation direction vector is a two-dimensional vector that provides the three-dimensional spatial direction of this propagation.

[0081] For example, the system receives the principal stress azimuth vector (0.812, 0.570) and the steepest fractal gradient vector (-0.065, -0.067), in units of 1 / pixel. The system normalizes the principal stress azimuth vector, with a magnitude of... The normalized unit vector of mechanical stress direction is After conversion to real-world physical dimensions, the steepest fractal gradient vector... The unit is The system retrieves the material coupling coefficient corresponding to the 6061 aluminum alloy material of the heat sink from the database and sets it to [value missing]. mm, .

[0082] The processing unit substitutes these values ​​into the multi-field parameter feature weighted superposition model to calculate the composite extended vector: This composite propagation vector (-0.0111, -0.02485) is the crack propagation direction vector. The system calculates the magnitude of this vector to determine the current crack propagation amount. mm.

[0083] The final prediction is that the crack will extend by 0.0272 mm in the next stage along the (-0.0111, -0.02485) direction.

[0084] See appendix Figure 2 The present invention also proposes a radiator crack propagation trend prediction system using fractal geometry, comprising the following modules: The multi-source physical data synchronous acquisition module is used to acquire the transient acoustic emission voltage time series and the continuous infrared temperature field video stream of the heat sink. Based on the first arrival time of the transient acoustic emission voltage time series, a single frame image in the continuous infrared temperature field video stream is extracted to generate a two-dimensional original matrix of the infrared temperature field. The acoustic emission source mechanism inversion module is used to analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combined with the preset empirical Green space derivative function, the stress wave source deformation kernel matrix is ​​inverted to generate a six-component moment tensor. The mechanical eigenfactor extraction module is used to decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform planar projection, and generate the principal stress azimuth vector. The anisotropic computational mesh generation module uses the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector to scale and rotate the preset isotropic square metric cell mesh template to generate anisotropic rectangular metric mesh. The thermal field multifractal analysis module is used to align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, it calculates the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derives the Lipschitz exponent field to generate a thermal field multifractal singular extremum clustering image. The steepest fractal gradient vector extraction module is used to extract the outer closed contour line of the thermal field multifractal singular extreme value clustering image. It calculates the spatial derivative of the outer closed contour line based on the Lipschitz exponential field and selects the vector with the largest modulus to generate the steepest fractal gradient vector. The crack propagation trend prediction module couples the normalized principal stress azimuth vector with the steepest fractal gradient vector into a preset multi-field parameter feature weighted superposition model, and outputs the current crack propagation amount and crack extension direction vector.

[0085] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0086] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting the crack propagation trend of a heat sink using fractal geometry, characterized in that, Includes the following steps: S1. Obtain the transient acoustic emission voltage time series and continuous infrared temperature field video stream of the heat sink. Extract the single frame image from the continuous infrared temperature field video stream based on the first arrival time of the transient acoustic emission voltage time series and generate the two-dimensional original matrix of the infrared temperature field. S2. Analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combine the preset empirical Green space derivative function to invert the stress wave source deformation kernel matrix and generate a six-component moment tensor. S3. Decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform plane projection, and generate the principal stress azimuth vector. S4. Using the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector, scale and rotate the preset isotropic square measurement cell mesh template to generate an anisotropic rectangular measurement mesh. S5. Align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, calculate the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derive the Lipschitz exponent field to generate a multifractal singular extremum clustering image of the thermal field. S6. Extract the outer closed contour line of the thermal field multifractal singular extremum clustering image, calculate the spatial derivative of the outer closed contour line based on the Lipschitz exponential field, and select the maximum modulus vector to generate the steepest fractal gradient vector. S7. The normalized principal stress azimuth vector and the steepest fractal gradient vector are coupled into the preset multi-field parameter feature weighted superposition model to output the current crack propagation amount and crack extension direction vector.

2. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Generating the original two-dimensional matrix of the infrared temperature field includes the following steps: The preset piezoelectric sensor array is activated to capture burst wave trains and output transient acoustic emission voltage time series; The preset infrared focal plane thermal imager is activated to capture surface thermal radiation images and generate a continuous infrared temperature field video stream. The transient acoustic emission voltage time series is analyzed using a preset long and short time window arrival determination algorithm to lock the first arrival time; Extract the single-frame image from the continuous infrared temperature field video stream whose timestamp matches the first arrival time, and convert it into a two-dimensional original matrix of the infrared temperature field.

3. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Generating a six-component moment tensor involves the following steps: The effective shock wave signal in the transient acoustic emission voltage time series is separated, the absolute time difference of arrival is read, and the three-dimensional absolute physical coordinates are defined by combining the preset sound velocity medium model. The absolute poloidal parameter and absolute amplitude of the first wave of transient acoustic emission voltage time series are tracked to assemble a stress wave source deformation core matrix that reflects the lattice tearing mode. The empirical Green space derivative function is loaded into the stress wave source deformation kernel matrix to perform wave source mechanism inversion iterative calculation, and the six-component moment tensor is generated by solving.

4. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Generating the principal stress azimuth vector includes the following steps: Transform the six-component moment tensor into a positive definite symmetric frame matrix, and separate three sets of independent orthogonal real eigenvalues ​​and their associated non-zero physical eigenvectors in space; Compare three sets of independent orthogonal real eigenvalues, select the largest and smallest micro eigenvalues, calculate the quotient of the smallest micro eigenvalue divided by the largest micro eigenvalue, and generate the ratio of deviatoric stress eigenvalues. Extract the spatial non-zero physical eigenvector corresponding to the maximum microscopic eigenvalue, and project the spatial non-zero physical eigenvector along the main fracture normal interface onto the preset reference two-dimensional plane field coordinate system to generate the principal stress azimuth vector.

5. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Generating anisotropic rectangular measure mesh includes the following steps: Obtain an isotropic square measure cell mesh template, inject the deviatoric stress eigenvalue ratio as a geometric scaling boundary condition, and change the ratio of the geometric minor axis to the geometric major axis of the isotropic square measure cell mesh template to the deviatoric stress eigenvalue ratio. The principal stress azimuth vector is introduced as a two-dimensional rotation deflection kernel to construct a coordinate adaptive affine transformation matrix, which guides the mesh after the scale change to perform rigid body rotation, so that the mesh major axis boundary is parallel to the principal stress azimuth vector. An anisotropic rectangular measurement grid is generated using the rotated boundary outline coordinates.

6. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Generating a multifractal singular extremum clustering image of the thermal field includes the following steps: The anisotropic rectangular measurement grid is translated into the two-dimensional original matrix of the infrared temperature field. The effective grayscale analysis range is forcibly defined with the absolute physical coordinates in three-dimensional space as the geometric masking center, and the noise floor masking is truncated. Within the effective grayscale analysis range, a set of neighboring pixels is extracted as a fractal isolated computational domain. The nonlinear partition integral function corresponding to the steady-state temperature gradient of local lattice grid points within the fractal isolated computational domain is calculated. The scaling slope of the nonlinear partition integral function on the double logarithmic metric coordinate system is obtained, the Lipschitz exponential field is derived, the high-frequency fixed-point heat flux abrupt change group is quantized and filtered, and a multifractal singular extremum clustering image of the thermal field is generated.

7. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, The process of selecting the vector with the largest modulus to generate the steepest fractal gradient vector includes the following steps: The preset connected component calibration operator is used to traverse the multifractal singular extreme value clustering image of the thermal field, identify the extreme variation core clusters and extract the outer closed contour lines; Discrete state observation nodes are retained at the edge of the closed contour line. Second-order difference calculations are performed in the Lipschitz exponential field with sub-pixel grid precision. The Lipschitz spatial derivative coordinate determination matrix is ​​then demixed and separated. The orientation of the peak value of the modulus limit in the Lipschitz spatial derivative coordinate determination matrix and its absolute scalar value are determined, and the steepest fractal gradient vector is generated by geometrically oriented combination.

8. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 1, characterized in that, Output the current crack propagation amount and crack propagation direction vector, including the following steps: Perform vector normalization and scaling calculations on the principal stress azimuth vectors to obtain the unit vector of the mechanical stress direction; The unit vector of mechanical stress direction is transformed into the left term of the operator, and the steepest fractal gradient vector is substituted as the right term of the operator. This is then loaded into the multi-field parametric feature weighted superposition model to perform thermal dissipation isomorphic interaction operation to generate a composite extended vector. The dot product magnitude scalar of the composite propagation vector is calculated and assigned to the current crack propagation amount. The multidimensional deflection direction of the composite propagation vector is used as the crack propagation direction vector.

9. The method for predicting the crack propagation trend of a heat sink using fractal geometry according to claim 2, characterized in that, Locking in the first Porta time involves the following steps: At each sampling point of the transient acoustic emission voltage time series, the average signal energy within a preset short time window and the average signal energy within a preset long time window are calculated. The average signal energy within the short time window is divided by the average signal energy within the long time window to generate the short-to-long time window energy ratio. By comparing the energy ratio of the short and long time windows with the preset arrival determination threshold, the timestamp corresponding to the sampling point where the energy ratio of the short and long time windows first exceeds the arrival determination threshold is locked as the first arrival time.

10. A radiator crack propagation trend prediction system using fractal geometry, applied to a radiator crack propagation trend prediction method using fractal geometry as described in any one of claims 1-9, characterized in that, Includes the following modules: The multi-source physical data synchronous acquisition module is used to acquire the transient acoustic emission voltage time series and the continuous infrared temperature field video stream of the heat sink. Based on the first arrival time of the transient acoustic emission voltage time series, a single frame image in the continuous infrared temperature field video stream is extracted to generate a two-dimensional original matrix of the infrared temperature field. The acoustic emission source mechanism inversion module is used to analyze the transient acoustic emission voltage time series to determine the absolute physical coordinates in three-dimensional space and construct the stress wave source deformation kernel matrix. Combined with the preset empirical Green space derivative function, the stress wave source deformation kernel matrix is ​​inverted to generate a six-component moment tensor. The mechanical eigenfactor extraction module is used to decompose the six-component moment tensor to calculate the ratio of deviatoric stress eigenvalues, extract the spatial non-zero physical feature vector corresponding to the maximum microscopic eigenvalue, perform planar projection, and generate the principal stress azimuth vector. The anisotropic computational mesh generation module uses the ratio of deviatoric stress eigenvalues ​​and the principal stress azimuth vector to scale and rotate the preset isotropic square metric cell mesh template to generate anisotropic rectangular metric mesh. The thermal field multifractal analysis module is used to align the anisotropic rectangular metric grid to the three-dimensional absolute physical coordinate projection point of the two-dimensional original matrix of the infrared temperature field. At the same time, it calculates the nonlinear partition integral function of the pixels in the anisotropic rectangular metric grid and derives the Lipschitz exponent field to generate a thermal field multifractal singular extremum clustering image. The steepest fractal gradient vector extraction module is used to extract the outer closed contour line of the thermal field multifractal singular extreme value clustering image. It calculates the spatial derivative of the outer closed contour line based on the Lipschitz exponential field and selects the vector with the largest modulus to generate the steepest fractal gradient vector. The crack propagation trend prediction module couples the normalized principal stress azimuth vector with the steepest fractal gradient vector into a preset multi-field parameter feature weighted superposition model, and outputs the current crack propagation amount and crack extension direction vector.

Images