Methods and systems for creep fatigue assessment of deteriorated base materials

By acquiring multiple sets of test data for symbolic regression encoding and microscopic scanning detection, and combining the binarization processing of lightweight assessment and depth assessment, the accurate identification of creep fatigue damage stages and failure mechanism transition points of deteriorated base material was achieved, thus improving the accuracy of the assessment.

CN123072589APending Publication Date: 2026-11-10ZHONGDIAN HUACHUANG ELECTRIC POWER TECH RES +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611302289.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-26
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies are insufficient to accurately identify the creep fatigue damage stages and failure mechanism transition points of deteriorated base materials, and relying solely on macroscopic mechanical parameters or local image features is insufficient to fully reflect the damage evolution process.

Method used

By acquiring multiple sets of test data for symbolic regression encoding, deploying a lightweight detector for microscopic scanning detection, and combining the threshold judgment and depth assessment instructions of the lightweight assessment results, the binarization processing and persistent feature extraction of the microscopic damage area are performed to achieve joint judgment of creep fatigue stage, dominant failure mechanism and transition point.

Benefits of technology

It improves the accuracy of creep fatigue damage stage determination and failure mechanism transition point identification, and solves the problem of inaccurate identification in the prior art.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123072589A_ABST
    Figure CN123072589A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for creep fatigue assessment of deteriorated base materials, relating to the field of fatigue assessment technology. The method includes: acquiring multiple sets of creep-fatigue test data and performing symbolic regression encoding; determining the expression through population iteration and multi-round fitting verification; performing microscopic scanning and detection on the in-service base material; performing variable feature identification and expression calculation to obtain a lightweight assessment result; performing threshold judgment to generate a depth assessment command; performing image backhaul and activation of the depth assessor; jointly judging the damage stage, dominant failure mechanism, and transition point by performing binarization processing and persistent feature extraction of the microscopic damage area to obtain the depth assessment result; and visualizing the lightweight and depth assessment results. This invention solves the technical problem of accurately identifying the creep fatigue damage stage and failure mechanism transition point of deteriorated base materials in the prior art, achieving the technical effect of improving the accuracy of creep fatigue damage stage determination and failure mechanism transition point identification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of fatigue assessment technology, and more specifically to a method and system for creep fatigue assessment of deteriorated base materials. Background Technology

[0002] Degraded base materials are prone to creep-fatigue coupled damage under high temperature, alternating loads, and long-term service conditions. Their microstructure gradually exhibits evolutionary phenomena such as void nucleation, void growth and aggregation, and microcrack connectivity. Since creep damage and fatigue damage may alternately dominate at different service stages, and the dominant failure mechanism shifts with changes in microscopic damage morphology and connectivity, relying solely on macroscopic mechanical parameters or local image features is insufficient to comprehensively reflect the damage evolution process. This leads to inaccurate identification of creep-fatigue damage stages and transition points of failure mechanisms in degraded base materials. Summary of the Invention

[0003] This application provides a method and system for creep fatigue assessment of deteriorated base materials, which addresses the technical problem of difficulty in accurately identifying creep fatigue damage stages and failure mechanism transition points of deteriorated base materials in the prior art.

[0004] In view of the above problems, this application provides a method and system for creep fatigue assessment of deteriorated base materials.

[0005] The first aspect of this application provides a method for creep fatigue assessment of deteriorated base material, the method comprising:

[0006] Multiple sets of test data based on creep-fatigue are acquired, and symbolic regression encoding is performed using basic operators and variable parameters as primitives. An evaluation expression is determined through population iteration and multi-round fitting verification. Based on the evaluation expression, a lightweight detector is deployed at the edge, and a lightweight evaluation result is obtained by performing microscopic scanning detection on the service substrate, performing variable feature recognition and expression calculation. A depth evaluation command is generated by thresholding the lightweight evaluation result. Based on the depth evaluation command, scan image backhaul and depth evaluator activation are performed. A depth evaluation result is obtained by performing binarization processing and persistent feature extraction of the microscopic damage area, and by jointly judging the creep fatigue stage, dominant failure mechanism, and transition point. The lightweight evaluation result and the depth evaluation result are visualized on the terminal interface.

[0007] A second aspect of this application provides a creep fatigue assessment system for deteriorated base materials, the system comprising:

[0008] The data acquisition module acquires multiple sets of test data based on creep-fatigue, performs symbolic regression encoding using basic operators and variable parameters as primitives, and determines the evaluation expression through population iteration and multi-round fitting verification. The recognition and calculation module deploys a lightweight detector at the edge based on the evaluation expression, performs microscopic scanning detection on the service base material, performs variable feature recognition and expression calculation, and obtains a lightweight evaluation result. The threshold judgment module generates a depth evaluation instruction by performing threshold judgment on the lightweight evaluation result. The joint judgment module performs scan image backhaul and depth evaluator activation based on the depth evaluation instruction, performs binarization processing and durability feature extraction of the micro-damage area, and obtains a depth evaluation result by jointly judging the creep fatigue stage, dominant failure mechanism, and transition point. The visualization module visualizes the lightweight evaluation result and the depth evaluation result on the terminal interface.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application acquires multiple sets of test data based on creep-fatigue, performs symbolic regression encoding using basic operators and variable parameters as primitives, and determines the evaluation expression through population iteration and multi-round fitting verification. Based on the evaluation expression, a lightweight detector is deployed at the edge, and a lightweight evaluation result is obtained by performing microscopic scanning detection on the service base material, performing variable feature recognition and expression calculation. A depth evaluation command is generated by thresholding the lightweight evaluation result. Based on the depth evaluation command, the scanned image is transmitted back and the depth evaluator is activated. A depth evaluation result is obtained by performing binarization processing and persistent feature extraction of the microscopic damage area, and jointly judging the creep fatigue stage, dominant failure mechanism, and transition point. The lightweight evaluation result and the depth evaluation result are visualized on the terminal interface. This invention solves the technical problem of accurately identifying the creep fatigue damage stage and failure mechanism transition point of deteriorated base material in the prior art. By triggering depth evaluation through lightweight evaluation and combining binarization processing of the microscopic damage area and persistent feature extraction for joint judgment, the technical effect of improving the accuracy of creep fatigue damage stage determination and failure mechanism transition point identification is achieved. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1A schematic diagram of the creep fatigue assessment method for deteriorated base material provided in the embodiments of this application;

[0013] Figure 2 A schematic diagram of the creep fatigue assessment system for deteriorated base material provided in an embodiment of this application;

[0014] Figure 3 A comparison chart of grayscale values ​​and damage segmentation threshold data for the creep fatigue assessment method for deteriorated base material provided in the embodiments of this application.

[0015] Figure labeling: Data acquisition module 11, recognition and calculation module 12, threshold judgment module 13, joint judgment module 14, visualization module 15. Detailed Implementation

[0016] This application provides a creep fatigue assessment method and system for deteriorated base materials, addressing the technical problem of accurately identifying creep fatigue damage stages and failure mechanism transition points in existing technologies. By triggering in-depth assessment with lightweight assessment and combining binarization of micro-damage areas with extraction of persistent features for joint judgment, the technical effect of improving the accuracy of creep fatigue damage stage determination and failure mechanism transition point identification is achieved.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0019] Example 1, as Figure 1 As shown, this application provides a method for creep fatigue assessment of deteriorated base materials, the method comprising:

[0020] Step S100: Obtain multiple sets of test data based on creep-fatigue, perform symbolic regression encoding using basic operators and variable parameters as primitives, and determine the evaluation expression through population iteration and multi-round fitting verification.

[0021] In this embodiment, multiple sets of test data are first extracted from a pre-stored creep-fatigue test database. Each set of test data is a data sequence formed for the deteriorated base material under different test conditions. The data sequence includes variables such as stress amplitude, holding time, temperature, deterioration parameters, and actual life. By extracting the above multiple sets of test data, a sample basis can be provided for the construction of subsequent evaluation expressions.

[0022] After obtaining multiple sets of test data, the basic operators used for expression construction are determined, and the basic operators and variable parameters are used as primitives for symbolic regression encoding to randomly generate the first expression cluster. Then, the first expression cluster is fitted and verified based on multiple sets of test data, and the first effective fitted expression cluster is selected. Next, a population iteration process based on mutation crossover is performed on the first effective fitted expression cluster to generate the second expression cluster, and multiple rounds of fitting and verification are continued using multiple sets of test data until the expression converges. Finally, the expression with the highest fit to the test data is selected as the evaluation expression for creep fatigue assessment of deteriorated base material.

[0023] Furthermore, in the method provided in the application embodiments, determining the evaluation expression further includes:

[0024] Multiple sets of creep-fatigue test data are retrieved, where each set of test data is a variable data sequence based on stress amplitude, holding time, temperature, degradation parameters and actual life; basic operators are determined, and symbolic regression encoding is performed based on the multiple sets of test data using the basic operators and variable parameters as primitives to obtain the evaluation expression.

[0025] In this embodiment, when determining the evaluation expression, multiple sets of creep-fatigue test data are retrieved. Each set of test data is a sequence of variable data based on stress amplitude, holding time, temperature, degradation parameters, and actual life, forming a data basis for symbolic regression encoding. Subsequently, basic operators are determined, and symbolic regression encoding is performed based on the multiple sets of test data using the basic operators and variable parameters as primitives.

[0026] In the symbolic regression encoding process, a first expression cluster is randomly generated based on basic operators. This first expression cluster is then validated using multiple sets of test data to select the first effective fitting expression cluster. Next, iterative iteration based on mutation crossover is performed on the first effective fitting expression cluster to determine the second expression cluster. This second expression cluster is then validated using multiple sets of test data. This iterative fitting process continues until convergence, and finally, the expression with the highest fit to the test data is selected as the evaluation expression. Thus, through multiple sets of test data retrieval, basic operator determination, symbolic regression encoding, expression cluster validation, and iterative fitting, an evaluation expression for subsequent creep fatigue assessment is obtained.

[0027] Furthermore, the method provided in the application embodiment, which uses basic operators and variable parameters as primitives and performs symbolic regression encoding based on the multiple sets of test data, further includes:

[0028] Based on the basic operators, a first expression cluster is randomly generated; based on the multiple sets of test data, the first expression cluster is verified, and a first effective fitting expression cluster is selected; the first effective fitting expression cluster is subjected to iterative processing based on mutation crossover to determine a second expression cluster; based on the multiple sets of test data, the second expression cluster is verified, and through multiple rounds of iterative fitting processing, until convergence, the expression with the highest fit to the test data is selected as the evaluation expression.

[0029] In this embodiment, a first expression cluster is first randomly generated based on basic operators. This generation is achieved using genetic programming. Data fields corresponding to stress amplitude, holding time, temperature, degradation parameters, and actual lifespan are read, and a set of variable nodes is established. Simultaneously, a set of operator nodes is established, including addition, subtraction, multiplication, division, logarithmic, and exponential operators. Then, operator nodes are randomly selected as parent nodes according to a preset expression depth, and variable nodes are randomly selected as child nodes to form an expression tree structure. When the expression tree reaches a preset number of levels, expansion stops, and the expression tree is converted into a corresponding mathematical expression. Multiple mathematical expressions are generated repeatedly in the same manner to form the first expression cluster.

[0030] Next, the first expression cluster is validated based on multiple sets of test data, and the first effective fitting expression cluster is selected. In this process, the mean squared error (MSE) evaluation method is used for validation. First, the variable parameters in each candidate expression are replaced with the actual values ​​from multiple sets of test data, and the expression calculation is performed group by group to obtain the corresponding calculation result sequence. Then, the calculation result sequence is compared with the corresponding test data sequence to calculate the error value between each data point. Finally, all error values ​​are squared and the average value is calculated to obtain the mean squared error. After calculating the mean squared error of all candidate expressions, they are sorted in ascending order of mean squared error, and the candidate expressions with the smallest mean squared error are retained as the first effective fitting expression cluster.

[0031] Next, an iterative process based on mutation and crossover is performed on the first effective fit expression cluster to determine the second expression cluster. In this process, a genetic operation method is used for iteration. First, according to the fitness ranking, two candidate expressions are selected from the first effective fit expression cluster as parent expressions. Then, the crossover nodes in the two parent expressions are determined, and the expression structures below the crossover nodes are swapped to generate new expressions; for example, the temperature branch in one expression is swapped with the load-bearing time branch in another expression, thus forming a new variable combination relationship. Subsequently, a node is randomly selected in the new expression, and the variable parameter corresponding to that node is replaced with another variable parameter, or the operator corresponding to that node is replaced with another operator, to complete the mutation operation. After repeating the crossover and mutation processing on all candidate expressions, the second expression cluster is obtained.

[0032] Finally, the second expression cluster is validated based on multiple sets of test data. Through multiple rounds of iterative fitting, convergence is achieved, and the expression with the highest fit to the test data is selected as the evaluation expression. Specifically, a cyclic iterative optimization method is used. First, each candidate expression in the second expression cluster is substituted into all test data for calculation, and the corresponding error value is recalculated using the mean squared error evaluation method. Then, candidate expressions with smaller error values ​​are retained, and crossover and mutation operations are performed again to generate a new round of candidate expressions. Subsequently, the newly generated candidate expressions are substituted into all test data for calculation, and the mean squared error is calculated again. This process is continuously repeated to reduce the mean squared error of the candidate expressions round by round. When the change in the minimum mean squared error within a preset number of consecutive rounds is less than a preset threshold, the iteration process is considered to have reached convergence. Finally, the candidate expression with the smallest mean squared error after convergence is extracted as the evaluation expression.

[0033] Step S200: Based on the evaluation expression, deploy a lightweight detector at the edge end, perform microscopic scanning detection on the service base material, perform variable feature recognition and expression calculation, and obtain the lightweight evaluation result.

[0034] In this embodiment, a lightweight detector is deployed at the edge according to the evaluation expression. Specifically, a microscopic scanning test is first performed during the service of the base material to acquire a scanned image; then, the scanned image is transmitted to the edge, where the lightweight detector performs microscopic phase feature identification on the scanned image and converts the identification result into a primitive vector; subsequently, the primitive vector is calculated based on the evaluation expression to obtain a lightweight evaluation result.

[0035] Furthermore, the method provided in the application embodiments, which involves performing microscopic scanning and detection on the service base material, performing variable feature identification and expression calculation to obtain lightweight evaluation results, also includes:

[0036] During the service process of the base material, a microscopic scanning test is performed to obtain a scanned image; the scanned image is transmitted to the edge end, and microscopic phase feature recognition and primitive vector transformation are performed on the scanned image according to the embedded lightweight detector. The lightweight evaluation result is obtained by using the calculation based on the evaluation expression.

[0037] In this embodiment, a microscopic scanning test is performed during the service life of the base material. Specifically, a scanning electron microscope is used to periodically scan and acquire images of the target monitoring area of ​​the base material to reflect the microstructure of the material. The scanned images are then subjected to size standardization and grayscale normalization to ensure that the scanned images obtained at different acquisition times have a consistent data format. The scanned images contain microstructural information such as grain boundary morphology, precipitate distribution, pore structure, and microcrack initiation characteristics, providing input data for subsequent feature identification.

[0038] Next, the scanned image is transmitted to the edge computing device. Based on the embedded lightweight detector, microscopic phase feature recognition and primitive vector transformation are performed on the scanned image. Specifically, after the scanned image is transmitted to the edge computing device, it is input into the lightweight detector, which is a pre-trained lightweight convolutional neural network. First, the convolutional layer is used to extract features from the scanned image, extracting texture features, edge features, and regional morphological features. Then, the feature mapping results are used to identify the pore regions, precipitate regions, and microcrack regions in the scanned image, and determine the corresponding location distribution and morphological information of each region. Then, feature parameters such as the number of pores, average pore size, pore area ratio, precipitate area ratio, and microcrack length are statistically analyzed. After the feature parameters are extracted, they are arranged according to a preset variable order and converted into a uniformly formatted numerical vector to obtain the primitive vector. Each vector component in the primitive vector corresponds to a microscopic phase feature parameter, used to characterize the microscopic organization state of the current scanned region.

[0039] Finally, operations based on the evaluation expression are used. Each variable in the evaluation expression is read, and the correspondence between the variable and each vector component of the primitive vector is established. Then, each feature parameter in the primitive vector is substituted into the corresponding position in the evaluation expression, and the calculation is performed according to the operation relationship in the evaluation expression. After all variable substitutions and expression calculations are completed, the corresponding calculation results are output as the lightweight evaluation results.

[0040] Step S300: Generate a deep evaluation instruction by performing a threshold judgment on the lightweight evaluation result.

[0041] In this embodiment, when performing threshold judgment on the lightweight assessment results, creep-fatigue-based attenuation data is calculated by comparing the first lightweight assessment result and the second lightweight assessment result; wherein, the first lightweight assessment result is the initial assessment result for any damage stage, and the second lightweight assessment result is the real-time assessment result. Then, it is determined whether the attenuation data is greater than the attenuation threshold. If the attenuation data is less than or equal to the attenuation threshold, edge-end lightweight assessment based on the lightweight detector continues; if the attenuation data is greater than the attenuation threshold, a depth assessment instruction based on the damage stage is generated.

[0042] Furthermore, in the method provided in the application embodiments, the process of determining a threshold based on the lightweight evaluation result and generating a deep evaluation instruction further includes:

[0043] By comparing the first lightweight assessment result and the second lightweight assessment result, creep-fatigue-based attenuation data is obtained, wherein the first lightweight assessment result is the initial assessment result based on any damage stage, and the second lightweight assessment result is the real-time assessment result; it is determined whether the attenuation data is greater than the attenuation threshold; if it is less than or equal to, an edge-end lightweight assessment based on the lightweight detector is performed; if it is greater than, a depth assessment instruction based on the damage stage is generated.

[0044] In this embodiment, decay data based on creep-fatigue is first obtained by comparing the first lightweight assessment result and the second lightweight assessment result. The first lightweight assessment result is the initial assessment result based on any damage stage, and the second lightweight assessment result is the real-time assessment result. Specifically, the result calculated by the lightweight detector combined with the assessment expression at the beginning of any damage stage of the base material is used as the first lightweight assessment result, and the result calculated by the lightweight detector combined with the assessment expression at the current sampling time is used as the second lightweight assessment result. Then, the difference between the first and second lightweight assessment results is calculated to obtain the assessment difference, and the decay data is obtained by dividing the assessment difference by the corresponding time interval. Thus, the decay data characterizes the rate of change of the assessment result over time during the creep-fatigue process.

[0045] Next, it is determined whether the attenuation data is greater than the attenuation threshold. The preset attenuation threshold is read, and the attenuation data is compared with the attenuation threshold to determine whether the rate of change of the evaluation result exceeds the preset control range.

[0046] If the value is less than or equal to the value, perform a lightweight assessment at the edge based on a lightweight detector. Maintain the continuous operation of the edge processing flow. The lightweight detector performs microscopic phase feature recognition and primitive vector transformation on the newly acquired scanned image, and calculates and outputs a new lightweight assessment result based on the assessment expression, thereby continuously updating the current service status of the base material.

[0047] If the value is greater than the target value, extract the current damage stage identifier, associate it with the current scan image number, sampling time, and the second lightweight assessment result to generate a depth assessment command, and send the depth assessment command to the control center to trigger the subsequent processing flow for that damage stage.

[0048] Step S400: According to the depth evaluation instruction, perform scan image back transmission and depth evaluator activation. By performing binarization processing of micro-damage areas and extraction of persistent features, and by jointly judging creep fatigue stage, dominant failure mechanism and transition point, the depth evaluation result is obtained.

[0049] In this embodiment, the scan image backhaul and the activation of the depth evaluator are executed synchronously according to the depth evaluation instruction. First, based on the scan image number and damage stage identifier carried in the depth evaluation instruction, the corresponding scan image is retrieved from the edge storage unit and backhauled to the control center. At the same time, the operation of the depth evaluator is triggered. The depth evaluator consists of a first feature extraction layer and a second evaluation layer. After activation, it directly performs continuous processing on the backhauled scan image.

[0050] After the returned scanned image enters the depth evaluator, the first feature extraction layer identifies the micro-organism regions in the image, divides the identified micro-damaged regions and non-damaged regions into pixel-level segments and assigns values ​​of 1 and 0 respectively, thereby generating a binarized image. Based on this binarized image, persistent features are further extracted. Persistent features include the change in the number of damage clusters, the growth rate of the largest damage cluster size, and the change in damage connectivity. Subsequently, the above features are organized into feature vectors and passed to the second evaluation layer.

[0051] After receiving the binarized image and feature vector, the second evaluation layer performs joint analysis and simultaneously completes creep fatigue stage identification, dominant failure mechanism determination, and transition point identification. The creep fatigue stage is used to characterize the stage evolution process of diffuse void nucleation, void growth and aggregation, and microcrack connectivity. The dominant failure mechanism is used to distinguish whether the current state is creep-dominated or fatigue-dominated. The transition point is used to characterize the critical position where the dominant failure mechanism switches. Under the combined effect of the three types of judgment results, the final depth evaluation result is output.

[0052] Furthermore, in the method provided in the application embodiments, the depth estimator further includes:

[0053] The creep-fatigue process of the base material is divided into stages to determine the creep-fatigue stage; a first feature extraction layer is deployed based on image binarization reconstruction and durable feature recognition; a second evaluation layer is deployed based on the stage judgment of the creep-fatigue stage; the first feature extraction layer and the second evaluation layer are cascaded to obtain a depth evaluator, which is then embedded in the control platform.

[0054] Furthermore, the method provided in the application embodiments also includes:

[0055] The creep fatigue stage includes at least three stages: the first stage is characterized by diffuse void nucleation, the second stage is characterized by void growth and aggregation, and the third stage is characterized by microcrack connectivity.

[0056] In this embodiment, the creep-fatigue process of the base material is first divided into stages to determine the creep-fatigue stage. During this process, scanned images of the base material at at least two acquisition times are acquired, and each scanned image is mapped to the same monitoring area. The scanned images are sequentially processed by grayscale conversion, size unification, and grayscale normalization. The average grayscale value and grayscale standard deviation of the pixels within the monitoring area are calculated. The grayscale value obtained by subtracting the grayscale standard deviation from the average grayscale value is used as the damage segmentation threshold. Pixels below the damage segmentation threshold are marked as candidate damaged pixels, and pixels not below the damage segmentation threshold are marked as candidate normal pixels. Figure 3 As shown, Figure 3This is a comparison chart of grayscale values ​​and damage segmentation thresholds. The horizontal axis represents pixel samples in the monitored area, and the vertical axis represents grayscale values. The bar charts represent the grayscale values ​​corresponding to pixel samples in each monitored area. In this example, the average grayscale value is 72, and the damage segmentation threshold is 58. The grayscale values ​​corresponding to regions 1, 2, 3, 5, and 7 are 88, 74, 63, 82, and 70, respectively, all of which are not lower than the damage segmentation threshold and correspond to candidate normal pixels. The grayscale values ​​corresponding to regions 4, 6, and 8 are 58, 49, and 55, respectively. Regions 6 and 8 are lower than the damage segmentation threshold and correspond to candidate damaged pixels. Region 4 is located at the threshold threshold and is processed into either candidate damaged pixels or candidate normal pixels according to a preset judgment rule. The threshold line in the figure represents the damage segmentation threshold. Pixel samples below the threshold line correspond to candidate damaged pixels, and pixel samples not below the threshold line correspond to candidate normal pixels. Opening and closing operations are performed on the candidate damaged pixel regions to remove isolated noise and fill in local fracture boundaries, resulting in micro-damaged areas. Connected component labeling is performed on the micro-damaged areas, and then... Connected region area, aspect ratio, and circularity are used to distinguish between void regions and microcrack regions. The connected region area is obtained by multiplying the number of damaged pixels by the actual area of ​​a single pixel. The aspect ratio is obtained by the ratio of the longer side to the shorter side of the smallest bounding rectangle of the connected region. Circularity is calculated from the connected region area, perimeter, and pi. For adjacent void regions, the shortest distance between their boundaries is calculated. When the shortest distance is less than a preset clustering distance, adjacent void regions are merged into a void clustering region. Based on the identification results of void regions, void clustering regions, and microcrack regions... The damage was assessed in stages. Stage 1 was defined as the period when cavities were scattered, did not form a cluster of cavities, and did not form damage connectivity. Stage 1 was characterized by diffuse cavity nucleation. Stage 2 was defined as the period when the cavity area increased, adjacent cavities merged into a cavity cluster, the area of ​​the largest damage cluster increased, and damage connectivity had not yet been formed. Stage 2 was characterized by the growth and aggregation of cavities. Stage 3 was defined as the period when the microcrack area connected to adjacent damage clusters, and the damage state changed from separation to connectivity. Stage 3 was characterized by microcrack connectivity.

[0057] Next, the first feature extraction layer is deployed based on image binarization reconstruction and persistent feature recognition, and the second evaluation layer is deployed based on the stage judgment of creep fatigue. In this process, scanned images from at least two acquisition times are input into the first feature extraction layer. The first feature extraction layer identifies micro-damage regions in the order of grayscale conversion, grayscale normalization, threshold segmentation, morphological processing, and connected component labeling, assigning a value of 1 to the pixels corresponding to micro-damage regions and a value of 0 to the pixels corresponding to normal regions, thus obtaining a binary image. Connected component segmentation is performed on the regions with a pixel value of 1 in the binary image to obtain the damage cluster distribution results corresponding to each acquisition time. Persistent features are extracted based on the damage cluster distribution results from at least two acquisition times. Persistent features include the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold for damage connectivity. The evolution of the number of damage clusters is obtained by comparing the number of damage clusters at different acquisition times, and the growth rate of the maximum damage cluster size is obtained by subtracting the maximum area of ​​the maximum damage cluster at the previous acquisition time from the area of ​​the maximum damage cluster at the later acquisition time. The damage cluster area is obtained by dividing the time interval between the previous and subsequent acquisition times. The critical threshold for damage connectivity is obtained by judging whether the damage skeleton line in the later acquisition time connects the damage clusters that were separated in the previous acquisition time. The first feature extraction layer organizes the binary image, the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold for damage connectivity into a feature vector. The second evaluation layer reads the binary image and the feature vector. When the damage pixels are distributed in a scattered point-like manner, no connection is formed between damage clusters, and the critical threshold for damage connectivity is not triggered, the diffuse void nucleation is output. When multiple void regions are merged into damage clusters, the growth rate of the maximum damage cluster size reaches the growth judgment condition, and the critical threshold for damage connectivity is not triggered, void growth and aggregation are output. When a continuous damage skeleton line is formed between damage clusters and the critical threshold for damage connectivity is triggered, microcrack connectivity is output.

[0058] Finally, the first feature extraction layer and the second evaluation layer are cascaded to obtain a depth estimator, which is then embedded in the control platform. Specifically, the output of the first feature extraction layer is connected to the input of the second evaluation layer, so that the binary image and feature vector output by the first feature extraction layer enter the second evaluation layer in the order of acquisition time. After the scanned image enters the cascaded structure, the first feature extraction layer first completes image binarization reconstruction and persistent feature extraction to obtain a binary image and feature vector. Then, the second evaluation layer performs stage judgment based on the creep fatigue stage to obtain the corresponding results in diffuse cavity nucleation, cavity growth and aggregation, or microcrack connectivity. The first feature extraction layer, the second evaluation layer, the binary image input field, the feature vector input field, the creep fatigue stage output field, and the stage judgment parameters are encapsulated into a depth estimator, which is then embedded in the control platform.

[0059] Furthermore, in the method provided in the application embodiments, the binarization processing and persistent feature extraction of the micro-damage area further include:

[0060] According to the depth evaluation command, the scanned images at at least two acquisition times are transmitted back to the control center to activate the embedded depth evaluator; based on the first feature extraction layer, the scanned images are subjected to micro-damage region identification and binarization processing to obtain a binary image, wherein the micro-damaged regions are binarized as 1 and the normal regions are binarized as 0; based on the binary image, persistent features are extracted as feature vectors, wherein the persistent features include the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold of damage connectivity.

[0061] In this embodiment, scanned images from at least two acquisition times are first transmitted back to the control center according to the depth assessment instruction, activating the embedded depth evaluator. During this process, the material number, monitoring area number, acquisition time identifier, and scanned image number carried in the depth assessment instruction are read first, and the corresponding scanned images are retrieved from the edge storage unit according to the acquisition time sequence. The at least two acquisition times include a previous acquisition time and a subsequent acquisition time. The scanned image from the previous acquisition time serves as a damage evolution comparison image, and the scanned image from the subsequent acquisition time serves as the current damage state image. After receiving the transmitted scanned images, the control center verifies the material number, monitoring area number, and acquisition time identifier of the scanned images. Scanned images belonging to the same monitoring area are then grouped into an image sequence according to their acquisition time, and this image sequence is input into the embedded depth evaluator, allowing the scanned images to enter the first feature extraction layer according to their acquisition time sequence.

[0062] Next, based on the first feature extraction layer, the scanned image undergoes micro-damage region identification and binarization. The first feature extraction layer first performs grayscale conversion, size unification, and grayscale normalization on the scanned image, ensuring that scanned images from different acquisition times have the same pixel size and grayscale range. Then, the average grayscale value and grayscale standard deviation of the pixels within the monitoring area are calculated. The grayscale value obtained by subtracting the grayscale standard deviation from the average grayscale value is used as the damage segmentation threshold. Pixels with grayscale values ​​below the damage segmentation threshold are marked as candidate damaged pixels, and pixels with grayscale values ​​not lower than the threshold are marked as candidate normal pixels. Opening and closing operations are performed on the candidate damaged pixel regions to remove isolated noise and fill in local fracture boundaries, obtaining the micro-damaged region. Connected component labeling is performed on the micro-damaged region, dividing adjacent or contiguous candidate damaged pixels into the same damaged region, and separating candidate damaged pixels into different damaged regions. Finally, pixels corresponding to the micro-damaged region are assigned a value of 1, and pixels corresponding to the normal region are assigned a value of 0, resulting in a binary image. Among them, the area with a pixel value of 1 corresponds to the void area, void accumulation area or microcrack area, and the area with a pixel value of 0 corresponds to the normal area of ​​the parent material that is not identified as a micro-damage area.

[0063] Finally, persistent features are extracted from the binary images as feature vectors. Specifically, the regions with a pixel value of 1 in the binary images at each acquisition time are first divided into connected components. Adjacent or contiguous damaged pixels are grouped into the same damaged cluster, while separated damaged pixels are grouped into different damaged clusters. The number of damaged clusters at the previous and subsequent acquisition times is counted. The number of damaged clusters at the subsequent acquisition time is compared with the number at the previous acquisition time to obtain the results of increased, decreased, maintained, or merged damage clusters, which is used as the evolution of the number of damaged clusters. Subsequently, the area of ​​each damaged cluster at the previous and subsequent acquisition times is calculated. During the calculation, the number of pixels with a pixel value of 1 within the same damaged cluster is counted, and this number is multiplied by the actual area corresponding to a single pixel to obtain the area of ​​the damaged cluster. Then, the maximum value is selected from all damaged cluster areas at the same acquisition time to obtain the maximum damaged cluster area at that acquisition time. The maximum damaged cluster area at the subsequent scan image is subtracted from the maximum damaged cluster area at the previous scan image to obtain the increment of the maximum damaged cluster area. The increment of the maximum damaged cluster area is then divided by the time interval between the two acquisition times to obtain the growth rate of the maximum damaged cluster size. Subsequently, skeletonization processing was performed on the binary images from the two acquisition times. Skeletonization involves thinning the damage regions with a pixel value of one in the binary image, resulting in a single-pixel-width damage skeleton line located in the middle of the damage region. During skeletonization, pixels on the boundary of the damage region are read one by one, and the distribution of adjacent pixels in the eight-neighbor domain is determined. If deleting the boundary pixel will not break the originally connected damage region, and the boundary pixel is not an endpoint of the damage region, the boundary pixel is marked as a deletable pixel. After one round of scanning, all deletable pixels are removed from the binary image. Then, the boundary pixel reading, eight-neighbor domain judgment, deletable pixel marking, and pixel removal are continued for the remaining damage pixels until the damage region can no longer be thinned, resulting in a damage skeleton line that retains the connectivity of the original damage region. The damage connectivity critical threshold is determined based on the damage skeleton lines from the two acquisition times. When the damage skeleton line in the later acquisition time connects two or more damage clusters that were separate in the earlier acquisition time, it is recorded as reaching the damage connectivity critical threshold. Finally, they are arranged in the order of the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold of damage connectivity to form a feature vector.

[0064] Furthermore, the method provided in the application embodiments also includes:

[0065] The binary image and feature vector are transferred to the second evaluation layer. The creep fatigue stage is used as the first judgment target, the dominant failure mechanism is used as the second judgment target, and the transition point is used as the third judgment target to generate a deep evaluation result. The dominant failure mechanism is creep-dominated or fatigue-dominated, and the third judgment target is whether there is a transition point. The transition point is defined when the dominant failure mechanism changes.

[0066] In this embodiment, the binary image and feature vector are first transferred to the second evaluation layer, with the creep fatigue stage as the first judgment target. During this process, the second evaluation layer first reads the binary image output by the first feature extraction layer, the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold for damage connectivity. It then matches the feature vectors from the previous acquisition time with the feature vectors from the next acquisition time according to the acquisition time sequence. The second evaluation layer reads the connected components of the damage regions with pixel values ​​of one in the binary image and counts the number of damage clusters, the area of ​​the maximum damage cluster, and the connection status between damage clusters. When the damage regions are distributed in a scattered point pattern, there is no connection path between any two damage clusters consisting of consecutive pixels with pixel values ​​of one, and the critical threshold for damage connectivity is not triggered, the creep fatigue stage is judged as the first stage, which is characterized by diffuse void nucleation. When multiple void regions merge into the same damage cluster, the area of ​​the maximum damage cluster increases, and the growth rate of the maximum damage cluster size reaches the growth judgment condition at the next acquisition time relative to the previous acquisition time, the creep fatigue stage is judged as the first stage. The first stage is characterized by diffuse void nucleation. When no continuous connection path is formed between any two damage clusters, the creep fatigue stage is determined to be the second stage, which is characterized by the growth and aggregation of cavities. When there is a continuous damage skeleton line connecting two or more damage clusters in the binary image at the next acquisition time, and the continuous damage skeleton line is formed by the continuous arrangement of adjacent pixels, and the two ends of the continuous damage skeleton line fall into different damage clusters or are connected to the boundaries of different damage clusters respectively, and the length of the continuous damage skeleton line reaches the preset connectivity length threshold, the damage connectivity critical threshold is recorded as triggered, and the creep fatigue stage is determined to be the third stage, which is characterized by the connectivity of microcracks.

[0067] When the dominant failure mechanism is used as the second judgment target, after obtaining the creep fatigue stage, the second evaluation layer continues to read the maximum damage cluster area, number of damage clusters, damage cluster merging status, continuous damage skeleton line length, and damage connectivity critical threshold at the previous and subsequent acquisition times. The second evaluation layer first calculates the cavity expansion amount, which is obtained by subtracting the maximum damage cluster area at the previous acquisition time from the maximum damage cluster area at the subsequent acquisition time; then it calculates the cavity expansion rate, which is obtained by dividing the cavity expansion amount by the time interval between the two acquisition times; subsequently, it calculates the crack connectivity, which is obtained by the length of the continuous damage skeleton line connecting two or more damage clusters at the subsequent acquisition time. The length of the continuous damage skeleton line is obtained by counting the number of skeleton line pixels and multiplying it by the actual length corresponding to a single pixel. When the cavity propagation rate reaches the preset cavity propagation judgment condition, the damage cluster is formed by the merging of scattered cavities, and the crack connectivity is less than the preset connectivity length threshold, or the continuous damage skeleton line does not connect more than two damage clusters, the dominant failure mechanism at that acquisition moment is determined to be creep-dominated. When the crack connectivity reaches the preset connectivity length threshold, the continuous damage skeleton line connects more than two damage clusters, and the damage connectivity critical threshold is triggered, the dominant failure mechanism at that acquisition moment is determined to be fatigue-dominated. When both the cavity propagation rate and the crack connectivity reach the preset connectivity length threshold at the same acquisition moment, the difference between the cavity propagation rate and the preset cavity propagation judgment condition and the difference between the crack connectivity and the preset connectivity length threshold are compared. If the difference corresponding to the cavity propagation rate is large and the damage connectivity critical threshold is not triggered, it is determined to be creep-dominated. If the difference corresponding to the crack connectivity is large and the damage connectivity critical threshold is triggered, it is determined to be fatigue-dominated.

[0068] When using the transition point as the third judgment target, the second evaluation layer reads the dominant failure mechanism corresponding to each acquisition time in the order of acquisition time and compares the dominant failure mechanisms of two adjacent acquisition times. If the dominant failure mechanism of the previous acquisition time is creep-dominated, and the dominant failure mechanism of the next acquisition time is still creep-dominated, it is determined that there is no transition point; if the dominant failure mechanism of the previous acquisition time is fatigue-dominated, and the dominant failure mechanism of the next acquisition time is still fatigue-dominated, it is determined that there is no transition point; if the dominant failure mechanism of the previous acquisition time is creep-dominated, and the dominant failure mechanism of the next acquisition time changes to fatigue-dominated, the next acquisition time is recorded as a transition point; if the dominant failure mechanism of the previous acquisition time is fatigue-dominated, and the dominant failure mechanism of the next acquisition time changes to creep-dominated, the next acquisition time is recorded as a transition point. The transition point is defined by the change of the dominant failure mechanism, and the acquisition time, the dominant failure mechanism before the change, the dominant failure mechanism after the change, the creep-fatigue stage, the binary image, and the feature vector corresponding to the transition point are recorded.

[0069] When generating the final in-depth evaluation result, the second evaluation layer combines the creep fatigue stage corresponding to the first judgment target, the dominant failure mechanism corresponding to the second judgment target, and the transition point judgment result corresponding to the third judgment target. During combination, the creep fatigue stage corresponding to the current acquisition time is first written in, which is either diffuse void nucleation, void growth and aggregation, or microcrack connectivity. Then, the dominant failure mechanism corresponding to the current acquisition time is written in, which is either creep-dominated or fatigue-dominated. Finally, the judgment result indicating whether a transition point exists is written in. If a transition point exists, the acquisition time corresponding to the transition point and the direction of the dominant failure mechanism conversion are recorded. If no transition point exists, it is recorded that the dominant failure mechanism has not converted, thus obtaining the in-depth evaluation result.

[0070] Step S500: On the terminal interface, visualize the lightweight evaluation results and the deep evaluation results.

[0071] In this embodiment, when visualizing the lightweight assessment results on the terminal interface, the lightweight assessment results output from the edge terminal, the corresponding acquisition time, the parent material number, and the monitoring area number are read, and a lightweight assessment result sequence is established according to the acquisition time. The lightweight assessment results at each acquisition time are displayed as a trend curve in chronological order, and the lightweight assessment result at the current acquisition time is displayed as the current assessment value. The attenuation threshold is displayed as a comparison benchmark on the same interface. When the attenuation data corresponding to the lightweight assessment result is less than or equal to the attenuation threshold, it is marked as a lightweight assessment state on the terminal interface. When the attenuation data is greater than the attenuation threshold, it is marked as a deep assessment triggered state on the terminal interface, and the acquisition time and monitoring area corresponding to the deep assessment trigger command are displayed.

[0072] When visualizing the depth assessment results on the terminal interface, the creep fatigue stage, dominant failure mechanism, transition point judgment result, binary image, and feature vector output by the control center are read and displayed in association with the corresponding scan image. The creep fatigue stage is displayed as diffuse void nucleation, void growth and aggregation, or microcrack connectivity; the dominant failure mechanism is displayed as creep-dominated or fatigue-dominated; and the transition point judgment result is displayed as the presence or absence of a transition point. When a transition point exists, the terminal interface displays the acquisition time corresponding to the transition point, the dominant failure mechanism before conversion, and the dominant failure mechanism after conversion, and simultaneously displays the micro-damage area in the binary image, so that the lightweight assessment results and the depth assessment results are presented correspondingly on the same terminal interface.

[0073] In summary, the embodiments of this application have at least the following technical effects:

[0074] This application acquires multiple sets of test data based on creep-fatigue, performs symbolic regression encoding using basic operators and variable parameters as primitives, and determines the evaluation expression through population iteration and multi-round fitting verification. Based on the evaluation expression, a lightweight detector is deployed at the edge, and a lightweight evaluation result is obtained by performing microscopic scanning detection on the service base material, performing variable feature recognition and expression calculation. A depth evaluation command is generated by thresholding the lightweight evaluation result. Based on the depth evaluation command, the scanned image is transmitted back and the depth evaluator is activated. A depth evaluation result is obtained by performing binarization processing and persistent feature extraction of the microscopic damage area, and jointly judging the creep fatigue stage, dominant failure mechanism, and transition point. The lightweight evaluation result and the depth evaluation result are visualized on the terminal interface. This invention solves the technical problem of accurately identifying the creep fatigue damage stage and failure mechanism transition point of deteriorated base material in the prior art. By triggering depth evaluation through lightweight evaluation and combining binarization processing of the microscopic damage area and persistent feature extraction for joint judgment, the technical effect of improving the accuracy of creep fatigue damage stage determination and failure mechanism transition point identification is achieved.

[0075] Example 2, based on the same inventive concept as the creep fatigue assessment method for deteriorated base material in the foregoing examples, such as... Figure 2 As shown, this application provides a creep fatigue assessment system for deteriorated base materials. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0076] The data acquisition module 11 is used to acquire multiple sets of test data based on creep-fatigue, perform symbolic regression encoding using basic operators and variable parameters as primitives, and determine the evaluation expression through population iteration and multi-round fitting verification; the recognition and operation module 12 is used to deploy a lightweight detector at the edge end according to the evaluation expression, perform microscopic scanning detection on the service base material, perform variable feature recognition and expression operation, and obtain a lightweight evaluation result; the threshold judgment module 13 is used to generate a depth evaluation instruction by performing threshold judgment on the lightweight evaluation result; the joint judgment module 14 is used to perform scan image back transmission and depth evaluator activation according to the depth evaluation instruction, and obtain a depth evaluation result by performing binarization processing and persistence feature extraction of the micro-damage area, and jointly judging the creep fatigue stage, the dominant failure mechanism and the transition point; the visualization module 15 is used to visualize the lightweight evaluation result and the depth evaluation result on the terminal interface.

[0077] Furthermore, the system is also used to implement the following functions:

[0078] Multiple sets of creep-fatigue test data are retrieved, where each set of test data is a variable data sequence based on stress amplitude, holding time, temperature, degradation parameters and actual life; basic operators are determined, and symbolic regression encoding is performed based on the multiple sets of test data using the basic operators and variable parameters as primitives to obtain the evaluation expression.

[0079] Furthermore, the system is also used to implement the following functions:

[0080] Based on the basic operators, a first expression cluster is randomly generated; based on the multiple sets of test data, the first expression cluster is verified, and a first effective fitting expression cluster is selected; the first effective fitting expression cluster is subjected to iterative processing based on mutation crossover to determine a second expression cluster; based on the multiple sets of test data, the second expression cluster is verified, and through multiple rounds of iterative fitting processing, until convergence, the expression with the highest fit to the test data is selected as the evaluation expression.

[0081] Furthermore, the system is also used to implement the following functions:

[0082] During the service process of the base material, a microscopic scanning test is performed to obtain a scanned image; the scanned image is transmitted to the edge end, and microscopic phase feature recognition and primitive vector transformation are performed on the scanned image according to the embedded lightweight detector. The lightweight evaluation result is obtained by using the calculation based on the evaluation expression.

[0083] Furthermore, the system is also used to implement the following functions:

[0084] By comparing the first lightweight assessment result and the second lightweight assessment result, creep-fatigue-based attenuation data is obtained, wherein the first lightweight assessment result is the initial assessment result based on any damage stage, and the second lightweight assessment result is the real-time assessment result; it is determined whether the attenuation data is greater than the attenuation threshold; if it is less than or equal to, an edge-end lightweight assessment based on the lightweight detector is performed; if it is greater than, a depth assessment instruction based on the damage stage is generated.

[0085] Furthermore, the system is also used to implement the following functions:

[0086] The creep-fatigue process of the base material is divided into stages to determine the creep-fatigue stage; a first feature extraction layer is deployed based on image binarization reconstruction and durable feature recognition; a second evaluation layer is deployed based on the stage judgment of the creep-fatigue stage; the first feature extraction layer and the second evaluation layer are cascaded to obtain a depth evaluator, which is then embedded in the control platform.

[0087] Furthermore, the system is also used to implement the following functions:

[0088] The creep fatigue stage includes at least three stages: the first stage is characterized by diffuse void nucleation, the second stage is characterized by void growth and aggregation, and the third stage is characterized by microcrack connectivity.

[0089] Furthermore, the system is also used to implement the following functions:

[0090] According to the depth evaluation command, the scanned images at at least two acquisition times are transmitted back to the control center to activate the embedded depth evaluator; based on the first feature extraction layer, the scanned images are subjected to micro-damage region identification and binarization processing to obtain a binary image, wherein the micro-damaged regions are binarized as 1 and the normal regions are binarized as 0; based on the binary image, persistent features are extracted as feature vectors, wherein the persistent features include the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold of damage connectivity.

[0091] Furthermore, the system is also used to implement the following functions:

[0092] The binary image and feature vector are transferred to the second evaluation layer. The creep fatigue stage is used as the first judgment target, the dominant failure mechanism is used as the second judgment target, and the transition point is used as the third judgment target to generate a deep evaluation result. The dominant failure mechanism is creep-dominated or fatigue-dominated, and the third judgment target is whether there is a transition point. The transition point is defined when the dominant failure mechanism changes.

[0093] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for assessing creep fatigue in deteriorated base materials, characterized in that, The method includes: Multiple sets of test data based on creep-fatigue were acquired, and symbolic regression encoding was performed using basic operators and variable parameters as primitives. The evaluation expression was determined through population iteration and multi-round fitting verification. According to the evaluation expression, a lightweight detector is deployed at the edge. By performing microscopic scanning and detection on the service base material, variable feature recognition and expression calculation are performed to obtain the lightweight evaluation result. A deep evaluation instruction is generated by applying a threshold to the lightweight evaluation results. According to the depth assessment instruction, the scanned image is returned and the depth evaluator is activated. By performing binarization processing of the micro-damage area and extraction of persistent features, the depth assessment result is obtained by jointly judging the creep fatigue stage, the dominant failure mechanism and the transition point. The lightweight evaluation results and the in-depth evaluation results are visualized on the terminal interface.

2. The creep fatigue assessment method for deteriorated base material as described in claim 1, characterized in that, Determine the evaluation expression, including: Multiple sets of creep-fatigue test data were retrieved, where each set of test data is a variable data sequence based on stress amplitude, holding time, temperature, degradation parameters and actual life; The basic operators are determined, and symbolic regression encoding is performed based on the basic operators and variable parameters as primitives to obtain the evaluation expression.

3. The creep fatigue assessment method for deteriorated base material as described in claim 2, characterized in that, Using basic operators and variable parameters as primitives, symbolic regression encoding is performed based on the aforementioned multiple sets of test data, including: Based on the basic operators, a first expression cluster is randomly generated; Based on the multiple sets of test data, the first expression cluster is verified, and the first effective fitted expression cluster is selected. Perform iterative mutation-based crossover on the first effective fit expression cluster to determine the second expression cluster; Based on the multiple sets of test data, the second expression cluster is verified. Through multiple rounds of iterative fitting, until convergence, the expression with the highest degree of fit with the test data is selected as the evaluation expression.

4. The creep fatigue assessment method for deteriorated base material as described in claim 1, characterized in that, Microscopic scanning and inspection of the service base material are performed, variable feature identification and expression calculation are carried out, and lightweight evaluation results are obtained, including: During the service life of the base material, microscopic scanning tests are performed to obtain scanned images; The scanned image is transmitted to the edge, and microscopic phase feature recognition and primitive vector transformation are performed on the scanned image based on the embedded lightweight detector. The lightweight evaluation result is obtained by using the calculation based on the evaluation expression.

5. The creep fatigue assessment method for deteriorated base material as described in claim 1, characterized in that, The lightweight evaluation result is used to determine a threshold, and a deep evaluation instruction is generated, including: By comparing the first lightweight assessment result and the second lightweight assessment result, decay data based on creep-fatigue is obtained. The first lightweight assessment result is the initial assessment result based on any damage stage, and the second lightweight assessment result is the real-time assessment result. Determine whether the attenuation data is greater than the attenuation threshold; If it is less than or equal to, perform edge-end lightweight evaluation based on the lightweight detector; If the value is greater than the specified value, generate a depth assessment instruction based on the damage stage.

6. The creep fatigue assessment method for deteriorated base material as described in claim 1, characterized in that, The depth estimator includes: The creep-fatigue process of the base material is divided into stages to determine the creep-fatigue stages; A first feature extraction layer is deployed based on image binarization reconstruction and persistent feature recognition, and a second evaluation layer is deployed based on the stage judgment of the creep fatigue stage. The first feature extraction layer and the second evaluation layer are cascaded to obtain a depth evaluator, which is then embedded in the control platform.

7. The creep fatigue assessment method for deteriorated base material as described in claim 6, characterized in that, The creep fatigue stage includes at least three stages: the first stage is characterized by diffuse void nucleation, the second stage is characterized by void growth and aggregation, and the third stage is characterized by microcrack connectivity.

8. The creep fatigue assessment method for deteriorated base material as described in claim 7, characterized in that, Perform binarization and persistent feature extraction on the micro-damage area, including: According to the depth evaluation instruction, the scanned images at at least two acquisition times are transmitted back to the control center to activate the embedded depth evaluator; Based on the first feature extraction layer, the scanned image is subjected to micro-damage area identification and binarization processing to obtain a binary image, wherein the micro-damage area is binarized as 1 and the normal area is binarized as 0. Based on the binary image, persistent features are extracted as feature vectors, where the persistent features include the evolution of the number of damage clusters, the growth rate of the maximum damage cluster size, and the critical threshold of damage connectivity.

9. The creep fatigue assessment method for deteriorated base material as described in claim 8, characterized in that, The binary image and feature vector are transferred to the second evaluation layer, with creep fatigue stage as the first judgment target, dominant failure mechanism as the second judgment target, and transition point as the third judgment target, to generate a deep evaluation result; The dominant failure mechanism is either creep-dominated or fatigue-dominated, and the third judgment objective is whether a transition point exists, with the transition point defined by the change in the dominant failure mechanism.

10. A creep fatigue assessment system for deteriorated base materials, characterized in that, The system is used to perform the creep fatigue assessment method for deteriorated base material as described in any one of claims 1-9, and the system comprises: The data acquisition module is used to acquire multiple sets of test data based on creep-fatigue, perform symbolic regression encoding using basic operators and variable parameters as primitives, and determine the evaluation expression through population iteration and multi-round fitting verification. The identification and calculation module is used to deploy a lightweight detector at the edge end according to the evaluation expression, perform variable feature identification and expression calculation by performing microscopic scanning detection on the service base material, and obtain a lightweight evaluation result; The threshold determination module is used to determine the threshold of the lightweight evaluation result and generate a deep evaluation instruction. The joint judgment module is used to perform scan image back transmission and depth evaluator activation according to the depth evaluation instruction. By performing binarization processing and persistent feature extraction of the micro-damage area, the depth evaluation result is obtained through joint judgment of creep fatigue stage, dominant failure mechanism and transition point. The visualization module is used to visualize the lightweight evaluation results and the deep evaluation results on the terminal interface.