A Gradient Heat Treatment Method and System for Cutting Tooth Body Based on Stress Cloud Diagram

Abstract

This invention belongs to the field of heat treatment, and particularly relates to a gradient heat treatment method and system for cutting teeth based on stress cloud maps. The method includes: acquiring a three-dimensional model of the initial stress of the cutting tooth; generating simulated stress distribution cloud maps and failure risk probabilities under multiple working conditions through finite element simulation combined with a first relational coupling model and constraint space; outputting a partitioned result sequence based on stress cloud map feature extraction and partitioned performance mapping; calling a process template knowledge graph and generating a partitioned process parameter sequence through a twin-matching model; implementing real-time heat treatment using a temperature-controlled fuzzy control model, which incorporates a temperature-stress change rate mapping function, synchronously monitoring the processing stress cloud map and feeding it back to the coupling model for dynamic simulation of failure risk; and adjusting process parameters in real-time according to the risk probability threshold until a safety threshold is met. This invention achieves closed-loop intelligent heat treatment driven by failure risk, significantly improving the wear resistance and fatigue resistance of the cutting tooth, and reducing the probability of early fracture.

Description

Technical Field

[0001] This invention belongs to the field of heat treatment, and particularly relates to a gradient heat treatment method and system for cutting tooth bodies based on stress cloud diagrams. Background Technology

[0002] In mining operations, cutting tools, as key cutting instruments for coal mining machines and tunneling machines, must continuously withstand friction and impact from coal seams or rocks in harsh environments. Their main failure modes include the detachment of the alloy head due to rapid wear, or bending and breakage due to insufficient bending resistance. Traditional heat treatment for cutting tools often employs uniform heating and quenching processes, which are difficult to adapt to the stress distribution differences in different parts of the cutting tool, resulting in poor performance balance between the head and shank, failing to simultaneously achieve both wear resistance and impact resistance. While existing technologies improve cutting tool performance through improved heat treatment processes, they lack precise control methods based on real-time stress analysis. For example, ordinary cutting tools easily generate sparks when cutting rock and have limited bending resistance; while plasma cladding technology can enhance wear resistance, it still requires optimization of the overall tooth strength. Furthermore, traditional processes do not fully utilize simulation analysis to guide heat treatment parameter settings, making it difficult to achieve gradient optimization of tooth material properties, leading to short service life, frequent replacements, and increased mining costs. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes a gradient heat treatment method and system for cutting teeth based on stress cloud maps. The method includes: acquiring a three-dimensional model of the initial stress of the cutting tooth; generating simulated stress distribution cloud maps and failure risk probabilities under multiple working conditions through finite element simulation combined with a first-relationship coupling model and constraint space; outputting a partitioned result sequence based on stress cloud map feature extraction and partitioned performance mapping; calling a process template knowledge graph and generating a partitioned process parameter sequence through a twin-matching model; implementing real-time heat treatment using a temperature-controlled fuzzy control model, which incorporates a temperature-stress change rate mapping function, synchronously monitoring the processing stress cloud map and feeding it back to the coupling model for dynamic simulation of failure risk; and adjusting process parameters in real-time according to the risk probability threshold until a safety threshold is met. This invention achieves closed-loop intelligent heat treatment driven by failure risk, significantly improving the wear resistance and fatigue resistance of the cutting tooth, and reducing the probability of early fracture.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A gradient heat treatment method for cutting tooth bodies based on stress cloud diagrams includes:

[0006] Obtain the three-dimensional model of the initial stress of the cutting tooth, and obtain the simulated stress distribution cloud map and the failure risk probability of the corresponding scenario by combining the preset first relationship coupling model and constraint space through the finite element simulation algorithm.

[0007] Based on the simulated stress distribution cloud map and the corresponding failure risk probability in each scenario, the tooth cutting partition result sequence in each scenario is obtained by combining the regional cloud map feature extraction model with the preset partition performance mapping.

[0008] Based on the sequence of cutting tooth partitioning results for each scenario and the process template knowledge graph, the partitioning process parameter sequence for each scenario is obtained through a twin matching model.

[0009] Based on the process parameter sequence of each scenario, the cutting teeth are heat-treated in real time through a temperature control fuzzy control model. At the same time, the heat treatment process is monitored, and a real-time processing stress distribution cloud map is obtained. The real-time processing stress distribution cloud map is fed back to the first relational coupling model to simulate the failure risk probability. The heat treatment process is adjusted in real time according to the simulation results until the failure risk probability threshold is met.

[0010] Specifically, the first relationship coupling model and the constraint space are constructed and trained by combining the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction and stress distribution map information of the target impact object on the cutting tooth and the corresponding failure risk probability obtained from simulations under different scenarios with a BP neural network optimized by particle swarm optimization.

[0011] The process template knowledge graph is constructed by combining scene parameters, cutting tooth model, stress cloud map of each heat treatment, heating temperature gradient, holding time, cooling rate and cutting tooth performance index with graph algorithm;

[0012] The sequence of cutting tooth partitioning results for each scenario includes partitioning range information and regional heat treatment difference target performance index information; regional heat treatment difference target performance index information includes cutting tooth hardness, wear resistance, impact toughness and bending strength; partitioning process parameter sequence includes heating temperature gradient, holding time, cooling rate and cutting tooth performance index.

[0013] Specifically, the construction process of the first relational coupling model includes:

[0014] Based on the collected data from different scenarios, including the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction and stress distribution of the cutting tooth, and the corresponding failure risk probability of the target impacted by the target with different hardness, combined with the preprocessing algorithm, we obtain the preprocessed numerical sequence and the preprocessed image sequence.

[0015] The stress distribution map information includes peak stress, average stress, high stress area ratio, and residual stress in the tooth head region; stress concentration factor, alternating stress amplitude, and fatigue safety factor in the tooth shank region; stress gradient, stress uniformity variation coefficient, and shear stress in the transition zone; and equivalent plastic strain, fracture toughness, and stress amplitude at coordinate points in different parts of the overall cutting tooth.

[0016] The failure risk probability is used as the target variable, the cutting tooth performance index as the first principal variable, the stress distribution map information as the second principal variable, the hardness of the target impactor as the background variable, and the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction of the cutting tooth, and influence coefficient of the target impactor on the cutting tooth performance index as co-variables to construct a hierarchical variable matrix.

[0017] Specifically, the construction process of the first relational coupling model also includes:

[0018] By obtaining the historical failure risk probability and stress distribution map information of different regions in the hierarchical variable matrix, and combining the first sub-major variables in the first principal variable, factor analysis is performed in conjunction with the preset contribution rate threshold to obtain the first sub-major variable sequence and corresponding contribution degree for each region, and the first risk association connection is constructed using the contribution degree corresponding to the first sub-major variable sequence.

[0019] Based on each first sub-major variable and second major variable in the first major variable, principal component analysis is used in conjunction with a preset contribution rate threshold to obtain the second major variable sequence corresponding to each first sub-major variable and the corresponding contribution degree and the autocorrelation coefficient between each second major variable sequence. The contribution degree between each first sub-major variable and the second major variable sequence is used to construct a second performance correlation link, and the autocorrelation coefficient between each second major variable sequence is used to construct a horizontal stress correlation coefficient.

[0020] Specifically, the construction process of the first relational coupling model also includes:

[0021] Based on the first sub-main variable sequence and its corresponding contribution and the failure risk probability of the corresponding region, a BP neural network with a built-in damage accumulation function is used to construct the first risk correlation coupling function for each region of the cut-off tooth.

[0022] Specifically, the damage accumulation function is constructed by combining multiaxial equivalent stress, real-time temperature, plastic strain rate and time as input variables with nonlinear differential equations. It is used to measure the damage accumulation at different time points of the cutting tooth and correct the predicted failure risk probability, thereby guiding the real-time control of the heat treatment process.

[0023] Based on the first sub-main variable and the corresponding second main variable sequence for each region, the corresponding contribution and the autocorrelation coefficient between each second sub-main variable sequence under each second main variable, the second performance correlation coupling function for each region of the truncated tooth is obtained by BP neural network.

[0024] Based on the second sub-principal variable sequence corresponding to each region, combined with the background variables and covariates in the hierarchical variable matrix, the association analysis algorithm is used to obtain the set of association covariates and the corresponding association degree matrix corresponding to the second sub-principal variable sequence of each region.

[0025] Specifically, the construction process of the first relational coupling model also includes:

[0026] Based on the set of associated covariates corresponding to the second sub-principal variable sequence of each region and the corresponding correlation matrix, the third stress correlation coupling function corresponding to each second sub-principal variable is obtained, and the third stress collaborative correlation connection is constructed by using the correlation matrix between the second sub-principal variable sequence and the set of associated covariates of each region.

[0027] Based on the hierarchical variable matrix, combined with the first risk correlation connection, the second performance correlation connection, the transverse stress correlation coefficient, the third stress synergistic correlation connection, and the stress gradient of the transition zone, a hierarchical variable correlation mapping matrix is ​​constructed by combining the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function through the topological space algorithm.

[0028] Specifically, the construction process of the first relational coupling model also includes:

[0029] A comprehensive risk prediction function is constructed based on the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function using a weighted algorithm. At the same time, the input of the particle swarm algorithm is constructed using the weights corresponding to the comprehensive risk prediction function and the weights corresponding to the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function.

[0030] The fitness function is constructed by taking the minimum value of the comprehensive risk prediction function, and the constraint space is constructed by combining the constraint information and constraint variable values ​​of each historical scenario with the hierarchical variable association mapping matrix.

[0031] Based on the input of the particle swarm optimization algorithm, the fitness function, the constraint space, the variable information in the hierarchical variable matrix obtained by the finite element method combined with the simulation algorithm for each scenario, and the preset training period, the first relational coupling model is trained by the particle swarm optimization algorithm to obtain the first relationship coupling model after training, and outputs the first risk correlation coupling function, the second performance correlation coupling function, the third stress correlation coupling function and the minimum failure risk probability for each scenario.

[0032] Specifically, the process of obtaining the tooth-cutting partition result sequence for each scenario includes:

[0033] Based on the simulated stress distribution cloud map obtained by the first relational coupling model for each scenario, the initial region division is performed by the mesh segmentation layer of the region cloud map feature extraction model to obtain the non-transition region and transition region of the cutting tooth in each scenario.

[0034] The non-transition region is input into the non-transition feature extraction layer constructed by bidirectional temporal convolution to obtain the non-transition stress feature vector. At the same time, the transition region is input into the transition feature extraction layer with built-in stress equalization gradient function to obtain the transition stress feature vector.

[0035] The non-transitional stress feature vector and the transitional stress feature vector are input into the convolutional attention fusion layer and fused according to the position points corresponding to the stress feature vectors to obtain a complete stress distribution feature cloud map of the cutting tooth.

[0036] Based on the complete stress distribution feature cloud map of the cutting tooth and combined with the partitioning performance rules, the output layer is constructed through support vector machine to obtain the cutting tooth partitioning result sequence under each scenario;

[0037] The partition performance rules are obtained by combining the stress characteristics of each different component of the cutting tooth in historical processing with the partition process parameters, and through correlation analysis algorithms.

[0038] Specifically, obtaining the sequence of partitioned process parameters for each scenario includes:

[0039] Based on the sequence of cutting tooth partitioning results in each scenario, combined with the process template knowledge graph, the initial matching partitioning process parameter sequence is obtained through matching algorithms and partitioning performance rules.

[0040] Based on the initial matching partition process parameter sequence, a heating-cooling control command is generated by combining a temperature control fuzzy control model with a temperature-stress change function. The temperature-stress change function is constructed by using a convolutional radial kernel function based on the temperature change rate per unit time and the stress change amplitude of the corresponding area location points in the historical cutting teeth. It is used to measure the stress change state under different temperature changes, thereby adjusting the heat preservation time length of the corresponding area location points.

[0041] Real-time stress map of the cutting tooth body processed by heating-cooling control command is monitored, and the complete real-time stress distribution feature cloud map of the cutting tooth is extracted by combining the twin algorithm with the regional cloud map feature extraction model.

[0042] The real-time stress distribution feature cloud map of the cutting tooth is fed back to the first relational coupling model to predict the corresponding failure risk. The real-time stress distribution feature cloud map of the cutting tooth and the corresponding failure risk prediction are compared with the simulated stress distribution cloud map and the minimum failure risk probability in the corresponding scenario to obtain the stress feature deviation map and failure risk deviation value.

[0043] The stress characteristic deviation map and failure risk deviation value are fed back into the regional cloud map feature extraction model and twin matching model to adjust the process parameter sequence matching process of the partition in real time until the deviation between the real-time cutting tooth stress distribution characteristic cloud map and the simulated stress distribution cloud map in the corresponding scenario and the deviation between the real-time failure risk prediction and the simulated minimum failure risk probability in the corresponding scenario both meet the corresponding preset deviation thresholds.

[0044] The stress cloud map-based gradient heat treatment system for cutting tooth bodies includes: a modeling and analysis module, a twin matching module, a partition mapping module, and a monitoring and feedback module.

[0045] The modeling and analysis module is used to obtain the three-dimensional model of the initial stress of the cutting tooth. Through the finite element simulation algorithm combined with the preset first relationship coupling model and constraint space, the simulated stress distribution cloud map and the failure risk probability of the corresponding scenario are obtained in each scenario.

[0046] The partitioning mapping module, based on the simulated stress distribution cloud map and the corresponding failure risk probability in each scenario, obtains the tooth partitioning result sequence for each scenario through the regional cloud map feature extraction model combined with the preset partitioning performance mapping; the tooth partitioning result sequence for each scenario includes partitioning range information and regional heat treatment difference target performance index information; the target performance index information includes hardness, wear resistance, impact toughness and bending strength.

[0047] The twin matching module, based on the result sequence of the cutting tooth partition in each scenario and the process template knowledge graph, obtains the partition process parameter sequence for each scenario through the twin matching model; the partition process parameter sequence includes heating temperature gradient, holding time, cooling rate and performance index;

[0048] The monitoring and feedback module performs real-time heat treatment on the cutting teeth based on the process parameter sequence of each scenario and a temperature control fuzzy control model. At the same time, it monitors the heat treatment process, obtains a real-time processing stress distribution cloud map, and feeds the real-time processing stress distribution cloud map back to the first relational coupling model to simulate the failure risk probability. Based on the simulation results, it adjusts the heat treatment process in real time until the failure risk probability threshold is met.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] This invention addresses the shortcomings of existing technologies by constructing a multi-physics coupling model and a failure risk prediction mechanism to achieve precise zoned heat treatment driven by stress distribution. Its core innovation lies in establishing a fully closed-loop intelligent control system encompassing simulation, zoning, process, and twin feedback. Specifically: based on the initial stress model and operating condition constraints, a high-precision stress cloud map and failure risk probability are generated through a dynamic coupling algorithm; using feature extraction and zoning mapping techniques, heat treatment areas are automatically divided according to stress gradients, and differentiated performance targets are set; combining knowledge graphs and twin matching models, the optimal process parameter sequence is generated in real time; a temperature-controlled fuzzy control model is used to execute heat treatment, simultaneously monitoring the processing stress cloud map and feeding it back to the risk prediction module for dynamic correction. This application breaks through the limitations of traditional experience-driven methods, achieving proactive prevention of failure risks and adaptive optimization of process parameters, significantly improving the wear resistance, fatigue resistance, and life consistency of cutting tools under complex operating conditions, while reducing energy consumption and process iteration cycles, providing a new generation of intelligent solutions for the manufacturing of core components of mining equipment. Attached Figure Description

[0051] Figure 1 This is a flowchart of the gradient heat treatment method for cutting tooth body based on stress cloud diagram in Embodiment 1 of the present invention;

[0052] Figure 2 This is a diagram of the regional cloud map feature extraction model architecture in Embodiment 1 of the present invention;

[0053] Figure 3 This is a block diagram of the gradient heat treatment system for cutting teeth based on stress cloud diagram in Embodiment 2 of the present invention. Detailed Implementation

[0054] Example 1

[0055] Please see Figure 1 The present invention provides an embodiment of a gradient heat treatment method for cutting tooth bodies based on stress cloud diagrams, comprising the following steps:

[0056] Obtain the three-dimensional model of the initial stress of the cutting tooth, and obtain the simulated stress distribution cloud map and the failure risk probability of the corresponding scenario by combining the preset first relationship coupling model and constraint space through the finite element simulation algorithm.

[0057] It should be further explained that, based on the structural characteristics of the cutting tooth in this embodiment, the cutting tooth is horizontally fixed on the vibration-resistant test bench to ensure that the tooth head, tooth shank, and transition fillet are in a free vibration state. Miniature accelerometers and piezoelectric vibration sensors are attached to the surface of different areas of the tooth body (such as the welding area of ​​the tooth head and the forging area of ​​the tooth shank). The sensor array is densely arranged according to the areas where stress is easy to concentrate to ensure coverage of all parts where the initial residual stress needs to be measured. A sinusoidal vibration signal with a controllable frequency is applied to one end of the cutting tooth through a vibration excitation device. The excitation frequency is gradually swept from low frequency to high frequency, so that the tooth body generates multi-mode vibration. The sensors collect the vibration acceleration, amplitude, and vibration frequency response signals of each area in real time, and simultaneously record the time series of vibration excitation parameters and response signals.

[0058] Based on vibration theory, the residual stress value at each measuring point is obtained by analyzing the correlation between vibration frequency shift, amplitude attenuation rate and material elastic modulus in different regions, combined with a pre-established vibration characteristic-residual stress mapping model. It should be further noted that the vibration characteristic-residual stress mapping model in this embodiment is calibrated by vibration test of standard specimens with known residual stress and is constructed by those skilled in the art based on historical vibration data.

[0059] The calculated residual stress values ​​of each region are processed together with the geometric dimensions obtained by the coordinate measuring machine and the property parameters of the cutting tooth obtained by the material mechanics experiment. The results are then linked to the three-dimensional geometric model through finite element preprocessing software, and finally a three-dimensional model of the initial stress of the cutting tooth containing the initial residual stress distribution is obtained.

[0060] It should be further explained that one method for obtaining the three-dimensional model of the initial stress of the cutting tooth in this embodiment is as follows:

[0061] Based on the physical parameter sequence of the cutting tooth, the specific values ​​of the geometric dimensions of the tooth shank and the transition fillet are obtained by a coordinate measuring machine, and the elastic modulus, Poisson's ratio, yield strength and other property parameters of the tooth material are obtained by material mechanics experiments.

[0062] Based on the above geometric dimensions and material properties, a three-dimensional geometric model of the cutting tooth is constructed using CAD software to accurately reproduce the structural details of the tooth head, tooth shank, and transition fillet.

[0063] Based on the manufacturing process of the cutting tooth, the initial residual stress in different regions of the tooth body is measured by X-ray diffraction, and the residual stress data is correlated with the three-dimensional geometric model by finite element preprocessing software.

[0064] Based on the associated model, initial stress boundary conditions are set using finite element analysis software, and the measured residual stress values ​​are assigned to the corresponding regions. After mesh generation and stress field initialization calculation, a three-dimensional model containing the initial stress distribution of each part of the tooth body is obtained. It should be noted that the geometric dimensions in this embodiment include length and diameter; the manufacturing process of the cutting tooth includes forging and welding; the initial residual stress includes the tensile stress at the welded joint of the tooth head and the compressive stress after forging the tooth shank.

[0065] Based on the simulated stress distribution cloud map and the corresponding failure risk probability in each scenario, the tooth cutting partition result sequence in each scenario is obtained by combining the regional cloud map feature extraction model with the preset partition performance mapping.

[0066] Based on the sequence of cutting tooth partitioning results for each scenario and the process template knowledge graph, the partitioning process parameter sequence for each scenario is obtained through a twin matching model; the twin matching model is preferably a combination of twin simulation algorithm and matching algorithm.

[0067] Based on the process parameter sequence of each scenario, the cutting teeth are heat-treated in real time through a temperature control fuzzy control model. At the same time, the heat treatment process is monitored, and a real-time processing stress distribution cloud map is obtained. The real-time processing stress distribution cloud map is fed back to the first relational coupling model to simulate the failure risk probability. The heat treatment process is adjusted in real time according to the simulation results until the failure risk probability threshold is met.

[0068] It should be further explained that in this embodiment, the first relationship coupling model and the constraint space are constructed and trained by combining the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction and stress distribution map information of the target impact object to the cutting tooth and the corresponding failure risk probability obtained from simulations under different scenarios with a BP neural network optimized by particle swarm optimization; it should also be explained that the target impact object and the target impact object are two interacting things. If the target impact object is the cutting tooth, the target impact object can be: coal seam, rock layer or metal device;

[0069] It should be further explained that the process template knowledge graph in this embodiment is constructed by combining scene parameters, cutting tooth model, stress cloud map of each heat treatment, heating temperature gradient, holding time, cooling rate and cutting tooth performance index with graph algorithm;

[0070] It should be further noted that the tooth cutting partition result sequence for each scenario in this embodiment includes partition range information and target performance index information of regional heat treatment differences;

[0071] It should be further explained that the target performance index information for regional heat treatment differences in this embodiment includes the hardness, wear resistance, impact toughness and bending strength of the cutting teeth; the process parameter sequence for each zone includes the heating temperature gradient, holding time, cooling rate and cutting tooth performance index.

[0072] This process employs a coordinate measuring machine to acquire geometric dimensions, materials mechanics experiments to determine material properties, and vibration methods to detect initial residual stress. A three-dimensional model of initial stress is constructed by combining finite element simulation with a first-relationship coupling model to accurately locate high-stress wear areas such as the tooth tip cutting edge. The stress distribution characteristics of this area are captured through a region cloud map feature extraction model using mesh segmentation, bidirectional feature extraction, and convolutional fusion. High-hardness heat treatment process parameters are matched according to zoning performance rules. A temperature-controlled fuzzy control model combined with a temperature-stress change function is used to regulate the heating-cooling process in real time, forming a uniform and dense microstructure to improve surface hardness and wear resistance. Simultaneously, a real-time feedback mechanism compares real-time and simulated stress cloud maps to continuously optimize parameters and prevent localized wear aggravation. To address fatigue resistance, vibration methods and X-ray diffraction were used to capture the initial residual stress (e.g., forging compressive stress) in the transition fillet stress concentration zone and the alternating stress zone of the tooth shank, and incorporated into the finite element simulation. A region cloud map feature extraction model focused on extracting the stress gradient characteristics of the transition zone, setting hardness and bending strength targets based on zoning rules. The holding time and cooling rate were optimized through process template knowledge graph matching, and the temperature control model was used to smooth the stress gradient in the transition zone, reducing stress concentration. A first-relationship coupling model employed a BP neural network with particle swarm optimization to predict failure risk, providing real-time feedback to adjust heat treatment parameters, reducing the amplitude of alternating stress and the stress concentration factor in the tooth shank, and inhibiting the initiation and propagation of fatigue cracks. This comprehensive approach forms a closed loop from precise modeling, zoning feature extraction, process parameter optimization to real-time control, ensuring high wear resistance in high-stress wear zones and high fatigue resistance in stress alternating and concentration zones, thus extending service life.

[0073] It should be further explained that the construction process of the first relational coupling model in this embodiment includes:

[0074] Based on the collected data from different scenarios, including the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction and stress distribution of the cutting tooth, and the corresponding failure risk probability of the target impacted by the target with different hardness, combined with the preprocessing algorithm, we obtain the preprocessed numerical sequence and the preprocessed image sequence.

[0075] It should be further explained that the stress distribution map information in this embodiment includes the peak stress, average stress, high stress area ratio, and residual stress in the tooth head region; the stress concentration factor, alternating stress amplitude, and fatigue safety factor in the tooth shank region; the stress gradient, stress uniformity variation coefficient, and shear stress in the transition zone; and the equivalent plastic strain, fracture toughness, and stress amplitude at different coordinate points of the overall cutting tooth.

[0076] It should be further explained that, in this embodiment, the peak stress in the tooth tip region refers to the maximum instantaneous stress borne by the tooth tip, directly reflecting whether the local stress exceeds the material's yield strength, and determining whether the tooth tip will undergo plastic deformation or fracture; the average stress reflects the overall stress level of the tooth tip and affects the long-term wear rate; the high-stress area ratio is the proportion of areas where the stress exceeds the material's allowable value in the total area of ​​the tooth tip; a higher ratio indicates a larger range of local failure risk; residual stress is the internal stress remaining from the manufacturing process; tensile stress will exacerbate crack initiation, while compressive stress can improve fatigue resistance. These variables collectively focus on the evaluation of the tooth tip's impact resistance and wear resistance; the stress concentration system in the tooth shank region... The stress coefficient is the ratio of the maximum local stress in the tooth shank to the nominal stress, reflecting the amplification of local stress by the transition structure. A high coefficient indicates that this area is prone to becoming a fracture initiation point. The nominal stress refers to the average stress value calculated based on the macroscopic stress state and overall dimensional parameters of the tooth shank, without considering the details of the local structure of the tooth shank, such as the fillet radius, abrupt changes in cross-section, surface defects, etc., which lead to stress concentration. The alternating stress amplitude is the difference between the maximum and minimum stress under cyclic loading. The larger the difference, the faster the fatigue damage accumulates. The fatigue safety factor is the ratio of the material's fatigue limit to the actual alternating stress, used to assess the fatigue fracture risk of the tooth shank under repeated loading. The stress gradient in the transition zone is the amount of stress change per unit length. A large gradient indicates that the stress transition from the tooth tip to the tooth shank is abrupt, which is prone to shear force and structural delamination. The stress uniformity variation coefficient is the ratio of the stress standard deviation to the mean. A high coefficient indicates a chaotic stress distribution and increases the probability of local failure. Shear stress is the stress parallel to the cross-section of the transition zone, which directly affects whether the transition zone cracks due to shear force exceeding the material's shear strength. These variables specifically assess the stability of the transition zone as a stress transfer hub. The overall equivalent plastic strain reflects the cumulative amount of plastic deformation of the entire cutting tool material, indicating the degree of plastic damage after long-term use; excessive strain leads to loss of dimensional accuracy. Fracture toughness is the material's ability to prevent the propagation of existing cracks; a higher value indicates a stronger ability of the cutting tool to continue functioning even after microcracks appear, used to determine overall service life and fracture resistance reliability. These variables, from local to global, static to dynamic, and instantaneous to cumulative, comprehensively cover the stress characteristics of various parts, providing multi-dimensional quantitative basis for failure risk prediction, structural optimization, and the setting of zonal parameters for gradient heat treatment processes.

[0077] The failure risk probability is used as the target variable, the cutting tooth performance index is used as the first principal variable, the stress distribution map information is used as the second principal variable, the hardness of the target impactor is used as the background variable, and the frictional resistance, frictional temperature, transient impact load, wear degree, cutting direction of the cutting tooth, and influence coefficient of the target impactor on the cutting tooth performance index are used as co-variables to construct a hierarchical variable matrix.

[0078] It should be further noted that the process of obtaining the influence coefficient of the target impact object on the cutting tooth performance index in this embodiment includes:

[0079] Based on the hardness gradient sequence of the target impact object, the target impact object is simulated by setting standard specimens with different hardness levels. Combined with the benchmark value of the cutting tooth performance index, a basic dataset for calculating the influence coefficient is constructed.

[0080] Based on the basic dataset, impact wear tests on cutting teeth were carried out by controlling the single variable method to obtain the measured values ​​of performance indices such as hardness, wear resistance, impact toughness, and bending strength of cutting teeth under different hardness levels.

[0081] Based on the difference between the measured value and the benchmark value of the performance index, the initial influence of the target impact object on each performance index of the cutting tooth under different hardness levels is calculated.

[0082] Based on the monitoring data of frictional resistance, frictional temperature, transient impact load, wear degree, and cutting direction of the cutting tooth in the co-variables, the correction coefficients of the above co-variables on the initial influence magnitude are quantified by the multiple linear regression algorithm.

[0083] The initial values ​​of the influence coefficients of the target impact object on each performance index of the cutting tooth are obtained by multiplying the initial influence amplitude and the correction coefficient.

[0084] Based on the actual interaction data between the target impact object and the cutting tooth in historical tests, the initial value of the influence coefficient is iteratively optimized through the error backpropagation algorithm, and finally the influence coefficient of the target impact object on the performance index of the cutting tooth is obtained.

[0085] By obtaining the historical failure risk probability and stress distribution map information of different regions in the hierarchical variable matrix, and combining the first sub-major variables in the first principal variable, factor analysis is performed in conjunction with the preset contribution rate threshold to obtain the first sub-major variable sequence and corresponding contribution degree for each region, and the first risk association connection is constructed using the contribution degree corresponding to the first sub-major variable sequence.

[0086] It should be further explained that the specific implementation process of factor analysis in this embodiment includes:

[0087] Based on the historical failure risk probability of each region in the hierarchical variable matrix, the stress distribution map information of the corresponding region, and the initial stress index in the first principal variable, the standardized variable set is obtained by removing outliers in the initial stress index and performing standardization.

[0088] Based on this set, the loading coefficients of each initial stress index and potential common factors are calculated using factor analysis algorithm. Common factors that can summarize the common characteristics of stress are extracted, and the proportion of each factor that explains the total variation of the variable is accumulated.

[0089] Based on the preset contribution rate threshold, common factors that meet the cumulative explanation ratio are selected. According to the absolute value of the load coefficient of each initial stress index in the compliant factors, the first sub-principal variable sequence that has a significant impact on failure risk in each region is determined.

[0090] Based on the load ratio of each first sub-principal variable in its respective factor, its contribution to failure risk is calculated. The first sub-principal variable sequence is mapped to the historical failure risk probability of the corresponding region through the contribution, and the first risk correlation is obtained. The initial stress index includes the original stress data such as the peak stress of the tooth tip, the stress concentration factor of the tooth shank, and the stress gradient of the transition zone.

[0091] Based on each first sub-major variable and second major variable in the first major variable, principal component analysis is used in conjunction with a preset contribution rate threshold to obtain the second major variable sequence corresponding to each first sub-major variable and the corresponding contribution degree and the autocorrelation coefficient between each second major variable sequence. The contribution degree between each first sub-major variable and the second major variable sequence is used to construct a second performance correlation link, and the autocorrelation coefficient between each second major variable sequence is used to construct a horizontal stress correlation coefficient.

[0092] It should be further explained that the specific implementation process of principal component analysis in this embodiment includes:

[0093] Based on the first sub-major variables and the second major variables in the first major variable, by checking the integrity of performance indicators and stress data, removing extreme values ​​caused by measurement deviations, and then using standardization to convert different types of indicators into dimensionless features, we can ensure that performance indicators such as hardness and toughness are comparable with stress distribution map information on the same scale, and obtain a purified variable dataset.

[0094] Based on this dataset, the covariance matrix between each variable is calculated to reflect the correlation between performance indicators and stress distribution map information. The eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvalues ​​characterize the explanatory power of each principal component for the total variation of the variables, and the eigenvectors reflect the weight ratio of each variable in the principal components. Principal components that can summarize the core information of the variables are extracted.

[0095] Based on a preset contribution rate threshold, the proportion of eigenvalues ​​of principal components is accumulated. When the accumulated proportion reaches the threshold, the extraction stops and the corresponding principal components are retained.

[0096] Based on the eigenvectors of the retained principal components, the second principal variables are sorted according to their weight proportions in the principal components to determine the sequence of second principal variables associated with each first sub-principal variable. Based on the eigenvalue proportions of each principal component, the contribution of the corresponding second principal variable is assigned. At the same time, the autocorrelation coefficient is obtained by calculating the linear correlation coefficient between the sequences of second principal variables. Based on the contribution, a mapping relationship is established between the first sub-principal variables (i.e., the target performance index) and the corresponding second principal variable sequence (i.e., stress distribution map information), and the second performance correlation connection is constructed. The intrinsic correlation between stress distribution map information is constructed using the autocorrelation coefficient, and the horizontal stress correlation coefficient is constructed.

[0097] Based on the first sub-main variable sequence and its corresponding contribution and the failure risk probability of the corresponding region, a BP neural network with a built-in damage accumulation function is used to construct the first risk correlation coupling function for each region of the cut-off tooth.

[0098] It should be further explained that the damage accumulation function in this embodiment is constructed by combining multiaxial equivalent stress, real-time temperature, plastic strain rate and time as input variables with nonlinear differential equations. It is used to measure the damage accumulation at different time points of the cutting tooth and correct the predicted failure risk probability, thereby guiding the real-time control of the heat treatment process.

[0099] It should be further explained that the construction process of the first risk-related coupling function in this embodiment includes:

[0100] Based on finite element simulation data, multiaxial equivalent stress time series data of each region of the cutting tooth are extracted, and the temperature field distribution of the tooth surface and interior is collected in real time through a high sampling rate sensor array.

[0101] Measurement of transient response of plastic strain rate based on high-speed camera and digital image correlation method;

[0102] The cumulative action time is recorded using a load spectrum analyzer; four types of physical quantities are input into the damage accumulation function through a nonlinear differential equation framework to calculate the micro-area damage increment in real time.

[0103] Based on the BP neural network architecture, the damage increment sequence is spatiotemporally aligned with the historical failure probability database of the corresponding region; the weights of the hidden layer nodes are iteratively optimized through the backpropagation algorithm to establish a quantitative mapping relationship between damage increment and failure probability; by embedding the damage increment and failure probability into the loss function of the BP neural network, the network's prediction of failure risk probability is corrected in real time, and the first risk correlation coupling function that can reflect the material damage process is obtained.

[0104] Based on the dynamic failure risk probability output by this function, the performance of each region of the cutting tooth can be controlled in real time by adjusting parameters such as heating temperature gradient, holding time, and cooling rate in the heat treatment process.

[0105] Based on the first sub-main variable and the corresponding second main variable sequence for each region, the corresponding contribution and the autocorrelation coefficient between each second sub-main variable sequence under each second main variable, the second performance correlation coupling function for each region of the truncated tooth is obtained by BP neural network.

[0106] Based on the second sub-principal variable sequence corresponding to each region, combined with the background variables and co-variables in the hierarchical variable matrix, the association analysis algorithm is used to obtain the set of association covariates and the corresponding association degree matrix corresponding to the second sub-principal variable sequence of each region.

[0107] It should be further explained that one specific way of implementing the association analysis algorithm in this embodiment is as follows:

[0108] Based on the second sub-principal variable sequence corresponding to each region, and combined with the background variables and co-variables in the hierarchical variable matrix, such as frictional resistance and transient impact load, the synchronous change relationship between variables is quantified by constructing a variable co-occurrence matrix to obtain the initial associated dataset;

[0109] Based on the initial association dataset, the influence strength of background variables and co-variables on the second sub-principal variable is evaluated by a pre-trained Bayesian model. The chi-square test is then used to screen out significantly related variable combinations to obtain preliminary association rules.

[0110] Based on the damage accumulation function, by introducing it as a weighting factor into the correlation analysis process, the contribution of each variable to the material damage evolution is quantified, and correlation rules containing the dynamic influence of damage are obtained.

[0111] Based on association rules that include the dynamic effects of damage, the direction of the dependency relationship between variables is determined by calculating the information gain ratio, and the strength of the nonlinear association between variables is evaluated by combining the mutual information metric method, so as to obtain the set of association covariates corresponding to the second sub-principal variable sequence of each region.

[0112] Based on the set of associated covariates corresponding to the second sub-principal variable sequence in each region, a standardized correlation matrix is ​​obtained by calculating the Pearson correlation coefficient matrix and performing a significance test. The standardized correlation matrix also includes the correction information of the damage accumulation function on the correlation strength.

[0113] Based on the set of associated covariates corresponding to the second sub-principal variable sequence of each region and the corresponding correlation matrix, the third stress correlation coupling function corresponding to each second sub-principal variable is obtained, and the third stress collaborative correlation connection is constructed by using the correlation matrix between the second sub-principal variable sequence and the set of associated covariates of each region.

[0114] It should be further explained that the specific process of obtaining the third stress correlation coupling function in this embodiment includes:

[0115] Based on the sequence of second sub-principal variables corresponding to each region, and combined with the background variables and co-variables in the hierarchical variable matrix, the second sub-principal variables and the set of related covariates are weighted and fused by using the elements in the correlation matrix as weight coefficients to obtain preliminary variable correlation characteristics.

[0116] Based on the preliminary variable association characteristics, a damage accumulation function is introduced. By using the damage accumulation amount as a moderating factor, the association strength between the second sub-main variable and the covariate is modified to obtain association characteristics that include the dynamic impact of damage.

[0117] Based on the corrected association characteristics, by calculating the conditional probability and mutual information values ​​between variables, covariates that have a significant impact on the second sub-major variable are selected to form a set of association covariates;

[0118] Based on the set of associated covariates, the correlation between each covariate and the second sub-principal variable is quantified through multiple regression analysis to generate a correlation matrix;

[0119] Based on the correlation matrix, a mapping relationship is established between the second sub-major variable and the set of related covariates according to the correlation weight, and the third stress correlation coupling function corresponding to each second sub-major variable is obtained. At the same time, the correlation values ​​in the matrix are used to construct the third stress co-correlation connection between the second sub-major variable sequence and the set of related covariates.

[0120] Based on the hierarchical variable matrix, combined with the first risk correlation connection, the second performance correlation connection, the transverse stress correlation coefficient, the third stress synergistic correlation connection, and the stress gradient of the transition zone, a hierarchical variable correlation mapping matrix is ​​constructed by combining the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function through the topological space algorithm.

[0121] It should be further explained that, in this embodiment, the stress gradient of the transition zone is used as a transition parameter for the correlation between the tooth head and tooth shank regional variables. It is incorporated into the correlation construction of the hierarchical variable matrix through the topological space algorithm. This is used to quantify the rate of change of the first risk correlation connection and the second performance correlation connection at the regional transition, and to adjust the transition strength of the transverse stress correlation coefficient and the third stress synergistic correlation connection. This ensures that the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function form a continuous mapping in the topological space, and that the hierarchical variable correlation mapping matrix can accurately reflect the transition correlation characteristics between variables in different regions.

[0122] A comprehensive risk prediction function is constructed based on the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function using a weighted algorithm. At the same time, the input of the particle swarm algorithm is constructed using the weights corresponding to the comprehensive risk prediction function and the weights corresponding to the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function.

[0123] The fitness function is constructed by taking the minimum value of the comprehensive risk prediction function, and the constraint space is constructed by combining the constraint information and constraint variable values ​​of each historical scenario with the hierarchical variable association mapping matrix.

[0124] Based on the input of the particle swarm optimization algorithm, the fitness function, the constraint space, the variable information in the hierarchical variable matrix obtained by the finite element method combined with the simulation algorithm for each scenario, and the preset training period, the first relational coupling model is trained by the particle swarm optimization algorithm to obtain the first relationship coupling model after training, and outputs the first risk correlation coupling function, the second performance correlation coupling function, the third stress correlation coupling function and the minimum failure risk probability for each scenario.

[0125] It should be further explained that, in this embodiment, one specific way of implementing the particle swarm algorithm is as follows:

[0126] Based on the weights of the first risk correlation coupling function, the second performance correlation coupling function, and the third stress correlation coupling function, as well as the weight of the comprehensive risk prediction function, the initial input parameters of the particle swarm algorithm are formed by integrating them.

[0127] Based on the objective of minimizing the comprehensive risk prediction function, a fitness function is constructed by setting a rule that a smaller function output value corresponds to a higher fitness.

[0128] Based on the hierarchical variable association mapping matrix, and combined with the constraint information and constraint variable values ​​of each historical scenario, a constraint space is constructed by defining the value boundaries and association restrictions of each variable.

[0129] Based on the input parameters, fitness function, and constraint space of the particle swarm optimization algorithm, combined with the variable information in the hierarchical variable matrix obtained from finite element simulation for each scenario and the preset training period, the particle position and velocity are iteratively updated through the particle swarm optimization algorithm to search for the optimal solution. This allows the particles to gradually approach the minimum value of the fitness function within the constraint space, thereby obtaining the first relational coupling model after training. The optimized first risk correlation coupling function, second performance correlation coupling function, third stress correlation coupling function, and corresponding minimum failure risk probability are output for each scenario.

[0130] Further explanation is needed; please refer to [link / reference]. Figure 2 The regional cloud map feature extraction model in this embodiment includes a grid segmentation layer, a non-transitional feature extraction layer, a transitional feature extraction layer, a convolutional attention fusion layer, and an output layer.

[0131] It should be further explained that the process of obtaining the tooth cutting partition result sequence for each scenario in this embodiment includes:

[0132] Based on the simulated stress distribution cloud map obtained by the first relational coupling model for each scenario, the initial region division is performed by the mesh segmentation layer of the region cloud map feature extraction model to obtain the non-transition region and transition region of the cutting tooth in each scenario.

[0133] It should be further noted that one implementation method for the initial region division of the mesh segmentation layer in this embodiment includes:

[0134] Based on the simulated stress distribution cloud map obtained by the first relationship coupling model for each scenario, the spatial range and accuracy standard of the mesh segmentation are determined by extracting the pixel coordinates and corresponding stress amplitudes of each position of the cutting teeth in the cloud map.

[0135] Based on this spatial range and accuracy standard, by setting the size parameters of the mesh cells, the stress distribution cloud map is uniformly divided into mesh cells covering the entire cutting tooth area to obtain the initial mesh matrix;

[0136] Based on the stress amplitude of each grid cell, the stress gradient between cells is obtained by calculating the ratio of the stress difference between adjacent cells to the spatial distance.

[0137] Based on the magnitude of the stress gradient, mesh elements with significantly higher gradient values ​​than their surroundings are identified as potential transition region elements.

[0138] Based on the spatial relationship of potential transition region units, adjacent high-gradient units are merged using a connected component analysis algorithm to form the initial boundary of the transition region.

[0139] Based on the continuous distribution characteristics of the remaining low-gradient grid cells, the initial boundary of the non-transition region is formed by merging adjacent low-gradient cells.

[0140] Based on the boundary coherence check of the transition and non-transition regions, isolated units are eliminated by adjusting the assignment of edge mesh units, thus obtaining the non-transition and transition regions of the cut-off teeth in each scene.

[0141] The non-transition region is input into the non-transition feature extraction layer constructed by bidirectional temporal convolution to obtain the non-transition stress feature vector. At the same time, the transition region is input into the transition feature extraction layer with built-in stress equalization gradient function to obtain the transition stress feature vector.

[0142] It should be further explained that one implementation process of the non-transitional feature extraction layer constructed by bidirectional temporal convolution in this embodiment includes:

[0143] Based on stress distribution data in non-transition regions, stress values ​​are converted into feature matrices suitable for convolution operations by standardizing the data.

[0144] The parameters of the forward and backward convolutional kernels of the bidirectional temporal convolutional layer are set. The forward convolutional kernel is set with the convolution direction in the spatial order from the tooth head to the tooth shank, and the backward convolutional kernel is set with the convolution direction in the spatial order from the tooth shank to the tooth head. Both use the same kernel size and number. The stride is set to match the spatial resolution of the stress data, and the padding mode is set to maintain the integrity of edge features. The feature matrix is ​​scanned forward by the forward convolutional kernel to extract the stress change features along the tooth head to tooth shank direction. At the same time, the feature matrix is ​​scanned backward by the backward convolutional kernel to extract the stress change features along the tooth shank to tooth head direction. Based on the feature maps output by the forward and backward convolutions, the bidirectional features are fused by channel concatenation, and then the feature dimension is compressed by the pooling layer to obtain the non-transitional stress feature vector.

[0145] It should be further explained that, in this embodiment, one implementation process of the transition feature extraction layer with built-in stress equivalent gradient function includes:

[0146] Based on the stress distribution data of the transition region, stress values ​​and spatial coordinates are converted into a structured feature matrix through standardization processing;

[0147] Based on this matrix, a stress equivalent gradient function is constructed. First, the spatial neighborhood calculation step size and gradient difference window size are set. The stress change rate of each point is obtained by performing a first-order difference operation on the stress values ​​of adjacent grid points. At the same time, the gradient magnitude and direction parameters are obtained by combining the direction vector decomposition.

[0148] Then, set a gradient isopleth threshold, classify the gradient magnitude according to the threshold interval, and generate stress contour lines by tracking the continuously distributed grid points in the same interval to mark the isopleth regions where the gradient changes continuously.

[0149] When applying this function, set the size of the feature extraction window and the scanning step size, scan each isopleth region one by one, and extract the mean and extreme values ​​of the gradient change rate, the distribution density and curvature of the isopleths, and the gradient abrupt change amplitude at the region boundary.

[0150] Based on the extracted local features, they are arranged in spatial coordinate order and integrated into global features through a feature concatenation algorithm. Principal component analysis is then used for dimensionality compression to obtain the transition stress feature vector.

[0151] The non-transitional stress feature vector and the transitional stress feature vector are input into the convolutional attention fusion layer and fused according to the position points corresponding to the stress feature vectors to obtain a complete stress distribution feature cloud map of the cutting tooth.

[0152] It should be further explained that one implementation process of the convolutional attention fusion layer in this embodiment includes:

[0153] Based on the non-transition stress feature vector and the transition stress feature vector, the spatial coordinate information corresponding to each feature point is extracted by analyzing the spatial coordinate labels embedded in the vector, and the spatial distribution of each position point on the cutting tooth surface is determined.

[0154] Based on the spatial distribution of each position point on the cutting tooth surface, the feature points of the non-transition region and the transition region are matched one-to-one according to their actual physical positions using a coordinate matching algorithm. Interpolation correction is performed on the edge feature points with coordinate offset to ensure that the feature points of the two regions are completely aligned in spatial position.

[0155] Based on the aligned feature point distribution, the bidirectional convolutional kernel parameters of the convolutional attention fusion layer are set. The horizontal convolutional kernel is set along the length of the cutter to capture axial feature association, and the vertical convolutional kernel is set along the radial direction of the cutter to capture radial feature interaction. The kernel size is adjusted according to the density of position points, and the stride is kept consistent with the spacing between position points to avoid information loss. The attention weight of each position point is generated by the Sigmoid function, and the weight is positively correlated with the importance of the position point in stress transmission.

[0156] The feature association strength in the axial and radial directions is calculated by using bidirectional convolutional kernels. The association strength is then converted into attention weights and multiplied with the feature vectors at the corresponding positions to enhance the features at key positions and suppress the features at secondary positions.

[0157] Based on the weighted feature vectors, the feature values ​​of non-transition regions and transition regions are fused point by point according to spatial location. The three-dimensional spatial dimension is restored through deconvolution operation, and the feature gaps at the junction of regions are filled to obtain a complete feature cloud map of the stress distribution of the cutting tooth.

[0158] Based on the complete stress distribution feature cloud map of the cutting tooth and combined with the partitioning performance rules, the output layer is constructed through support vector machine to obtain the cutting tooth partitioning result sequence under each scenario;

[0159] It should be further explained that the partitioning performance rules in this embodiment are obtained by combining the stress characteristics of each different component of the cutting tooth in historical processing with the partitioning process parameters and through correlation analysis algorithm.

[0160] Based on the stress characteristics of each component of the cutting tooth in historical processing data, a multi-dimensional feature vector is constructed by extracting parameters such as stress amplitude, distribution area, and gradient change rate. At the same time, variables such as temperature gradient, holding time, and cooling rate in the corresponding processing parameters are sorted out.

[0161] Based on this dataset, we constructed a co-occurrence matrix to quantify the synchronous change relationship between stress characteristics and process parameters, calculated the conditional probability distribution to evaluate the influence intensity of each process parameter on the performance of different stress regions, and combined the chi-square test to screen out the parameter combinations that are significantly related.

[0162] Based on the screening results, the cutting teeth are divided into different regions by setting stress thresholds, with each region corresponding to a specific stress range. Based on the partitioning results, the mutual information value between stress characteristics and process parameters in each region is calculated to determine the direction and strength of their correlation and generate initial correlation rules.

[0163] Based on the initial association rules, the effectiveness of the rules is evaluated by introducing confidence and support indicators. Low-confidence rules are eliminated, and high-confidence rules are retained to form a preliminary mapping relationship. Based on the preliminary mapping relationship, the range and combination of process parameters are adjusted through expert knowledge calibration and actual processing effect verification to obtain the optimal matching rule between stress characteristics and process parameters in each region. Based on the optimal rules of each region, a complete zoning performance rule is formed by integration.

[0164] For example, to better illustrate the partition performance rules, an example of a setting is given as follows:

[0165] Zone I: Tooth tip cutting edge, stress range: σ≥300MPa, target hardness 85-88HRC, wear resistance improvement a% (ASTM G65 wear standard adopted); σ represents stress; In Zone I, σ≥300MPa corresponds to the tooth tip cutting edge area of ​​the cutting tooth. The stress in this area is not less than 300MPa. Based on this stress level characteristic, the performance target for this area is set as a hardness of 58-62HRC and a wear resistance improvement of a% (ASTM G65 wear standard adopted) to measure its wear resistance improvement effect. The stress level of different areas can be clearly defined by the numerical range of σ, and then the performance target of each area can be set accordingly.

[0166] Zone II: Tooth head body, stress range: 250-300MPa, hardness 50-55HRC, impact toughness ≥20J / cm²; Zone II corresponds to the tooth head body area of ​​the cutting tooth. The stress in this area is between 250 and 300MPa. Based on this stress level characteristic, the performance target for this area is set as a hardness of 50-55HRC and an impact toughness of not less than 49J / cm². This forms a corresponding match between the stress state and mechanical performance requirements of this area, ensuring that the tooth head body can stably transmit loads and resist impacts when subjected to corresponding stresses, avoiding premature damage due to insufficient performance, and providing reliable support for the overall working performance of the cutting tooth; the following have the same meaning, so they will not be repeated here.

[0167] Zone III: Transition fillet, stress range: 200-250MPa, hardness 45-50HRC, bending strength ≥220MPa;

[0168] Zone IV: Upper part of the tooth shank, stress range: 150-200MPa, hardness 35-42HRC, impact toughness ≥55J / cm²;

[0169] Zone V: Tooth shank tail, stress range: σ≤150MPa, hardness 28-32HRC, elongation ≥15%.

[0170] It should be further explained that the acquisition of the partitioning process parameter sequence for each scenario in this embodiment includes:

[0171] Based on the sequence of cutting tooth partitioning results in each scenario, combined with the process template knowledge graph, the initial matching partitioning process parameter sequence is obtained through matching algorithms and partitioning performance rules.

[0172] It should be further explained that the initial matching partition process parameter sequence acquisition process in this embodiment is as follows:

[0173] Based on the sequence of cut-tooth partitioning results for each scenario, the spatial coordinate boundaries and region type identifiers of each partition are obtained by parsing the partitioning range information in the sequence.

[0174] Based on the analytical information on the regional heat treatment differences in target performance indicators, target performance parameters such as hardness, wear resistance, impact toughness, and bending strength corresponding to each zone are extracted.

[0175] Based on the process template knowledge graph, by retrieving historical data that matches the current scenario parameters and the cutting tooth model, the heat treatment process parameters such as heating temperature gradient, holding time, and cooling rate of the corresponding partition in the historical scenario are obtained, as well as the associated stress cloud map features and cutting tooth performance index.

[0176] Based on the partitioned performance rules, by matching the target performance parameters of each partition, the process parameter constraints such as the heating temperature gradient range, holding time interval, and cooling rate threshold corresponding to each target performance parameter are obtained.

[0177] The spatial coordinate boundaries and region type identifiers of each partition are matched with the spatial features of historical partitions in the process template knowledge graph. The target performance parameters of each partition are numerically matched with the historical cutting tooth performance index. The stress distribution characteristics of each partition are similarly matched with the historical stress cloud map characteristics.

[0178] Based on the above multi-dimensional matching results, and combined with the process parameter constraints, the historical process parameters that meet the constraints and have the highest matching degree are selected as candidate process parameters for each partition.

[0179] Based on the candidate process parameters of each partition, the compatibility of process parameters of adjacent partitions is verified by simulation algorithm, parameter conflicts are eliminated, and an initial matching partition process parameter sequence is formed for each scenario.

[0180] Based on the initial matching partition process parameter sequence, a heating-cooling control command is generated by combining a temperature control fuzzy control model with a temperature-stress change function. The temperature-stress change function is constructed by using a convolutional radial kernel function based on the temperature change rate per unit time and the stress change amplitude of the corresponding area location points in the historical cutting teeth. It is used to measure the stress change state under different temperature changes, thereby adjusting the heat preservation time length of the corresponding area location points.

[0181] It should be further explained that the specific process of implementing the temperature control fuzzy control model in this embodiment includes:

[0182] Based on the initial matching partition process parameter sequence, a basic control vector is constructed by extracting parameters such as target temperature and holding time at each region location point;

[0183] Based on the basic control vector, a temperature-stress change function is constructed. First, the unit time temperature change rate and corresponding stress change amplitude of different regional locations are extracted from historical data. The data is sampled by sliding window by setting the convolution window size and stride. The radial basis kernel function is applied to calculate the temperature-stress correlation weight of each sampling point, and a function model reflecting the relationship between temperature change and stress response is constructed.

[0184] Based on the temperature-stress change function, the current temperature change rate is input into the model to calculate the predicted stress change value at the corresponding location point, and the stress deviation is generated by comparing it with the target stress state.

[0185] Based on stress deviation, a fuzzy control rule base is set up, with stress deviation and deviation change rate as input variables and heat preservation time adjustment as output variables. The membership function and fuzzy inference rules of each variable are defined. Based on the fuzzy inference results, defuzzification is performed by the centroid method to obtain the accurate heat preservation time adjustment value and generate the adjusted heating-cooling control command.

[0186] Real-time stress map of the cutting tooth body processed by heating-cooling control command is monitored, and the complete real-time stress distribution feature cloud map of the cutting tooth is extracted by combining the twin algorithm with the regional cloud map feature extraction model.

[0187] The real-time stress distribution feature cloud map of the cutting tooth is fed back to the first relational coupling model to predict the corresponding failure risk. The real-time stress distribution feature cloud map of the cutting tooth and the corresponding failure risk prediction are compared with the simulated stress distribution cloud map and the minimum failure risk probability in the corresponding scenario to obtain the stress feature deviation map and failure risk deviation value.

[0188] The stress characteristic deviation map and failure risk deviation value are fed back into the regional cloud map feature extraction model and twin matching model to adjust the process parameter sequence matching process of the partition in real time until the deviation between the real-time cutting tooth stress distribution characteristic cloud map and the simulated stress distribution cloud map in the corresponding scenario and the deviation between the real-time failure risk prediction and the simulated minimum failure risk probability in the corresponding scenario both meet the corresponding preset deviation thresholds.

[0189] This embodiment first uses measured residual stress data from coordinate measuring machine (CMM) and X-ray diffraction, combined with finite element stress field initialization technology, to accurately reproduce the tensile and compressive stress distribution caused by forging and welding in a three-dimensional model. This technology overcomes the limitations of traditional geometric modeling, using the residual stress inside the material as the benchmark input for the heat treatment process. This allows subsequent gradient heat treatment to specifically eliminate harmful tensile stress in the welded area of ​​the tooth head and utilize the beneficial compressive stress in the forging area of ​​the tooth shank. By compensating for manufacturing defects through feedforward, it avoids heat treatment distortion and early cracking caused by residual stress concentration at the source, significantly improving the initial integrity of the tooth structure. Secondly, it integrates a BP neural network with particle swarm optimization and a damage accumulation function to construct a condition-driven stress-risk mapping model. The damage function integrates the instantaneous intensity of multiaxial equivalent stress, the time-varying effect of temperature softening, the impact damage characteristics of plastic strain rate, and the time accumulation effect through nonlinear differential equations to accurately calculate the micro-area damage increment; the neural network establishes a quantitative correlation between the damage increment and the historical failure probability. This model combines factor analysis of hierarchical variable matrices to analyze the contribution weights of key parameters such as peak stress at the tooth tip, alternating stress amplitude in the tooth shank, and stress gradient in the transition zone to failure, thus realizing the transformation from "stress distribution characteristics" to probabilistic failure risk. Third, the regional cloud map feature extraction model intelligently identifies stress contour lines and divides the transition zone gradient bandwidth through grid segmentation and bidirectional convolution architecture: the non-transition zone uses bidirectional time convolution to capture the axial stress attenuation law from the tooth tip to the tooth shank, and the transition zone quantifies the shear mutation intensity at the boundary based on the stress contour gradient function. Combining support vector machines and partitioning performance rules, stress cloud map features are mapped to differentiated performance targets, such as high hardness and wear resistance of the tooth tip and high toughness and impact resistance of the tooth shank. This embodiment breaks through the rigidity of fixed threshold partitioning, realizing precise design of strengthening in high stress areas, gradual change in stress abrupt areas, and toughening in low stress areas, eliminating stress concentration caused by traditional straight boundary lines. Fourth, the process template knowledge graph integrates historical scene parameters, stress cloud maps, and process effect data through graph algorithms to construct a multi-dimensional correlation network of "stress features - performance targets - process parameters". The twin matching model, based on this graph, compares the real-time partitioning results with similar scene templates to generate an initial process parameter sequence. The temperature control fuzzy control model introduces a temperature-stress change function and dynamically adjusts the holding time and cooling intensity based on the temperature change-stress response relationship calculated by the radial basis kernel function, realizing the predictive feedback of stress and heating temperature. Fifth, the real-time monitored processing stress cloud map, after feature extraction, is fed back to the first relation coupling model for risk probability reassessment. By comparing the feature deviations and risk deviations between the simulated cloud map and the real-time cloud map, a triple feedback mechanism is triggered: the regional cloud map feature extraction model adjusts the partition boundaries to match the actual stress distribution; the twin matching model corrects the process parameter mapping relationship; and the temperature control model outputs heating / cooling compensation commands based on stress deviation fuzzy inference.This closed-loop system continuously tracks risk probability thresholds during the heat treatment process, dynamically suppressing damage accumulation at material phase transformation critical points. For example, when the tooth tip temperature rises too quickly, the oil cooling rate is immediately increased to prevent softening damage; when the residual tensile stress in the tooth shank exceeds the limit, cryogenic treatment to induce compressive stress is automatically triggered. Ultimately, it achieves full lifecycle risk management encompassing simulation prediction, process execution, monitoring feedback, and dynamic correction. In summary, this embodiment accurately reproduces manufacturing defects through the initial stress field, transforms the operating load into a probabilistic failure criterion using a multiphysics coupled risk model, achieves targeted structural performance design based on stress gradient sensing partitioning, outputs the globally optimal process through knowledge graph-driven twin matching, and finally dynamically curbs damage evolution through digital twin closed-loop control. This forms an intelligent decision-making chain with stress cloud map as the hub and failure risk as the constraint, completely transforming the traditional experience-based trial-and-error mode of heat treatment. It further realizes the optimization of material phase transformation process at the microscale to improve the wear resistance of the tooth tip, the control of stress gradient in the transition zone at the mesoscale to prevent crack propagation, and the strengthening of tooth shank toughness at the macroscale to resist impact instability, ultimately achieving a synergistic leap in the life and reliability of the cutting tooth in complex service environments.

[0190] Example 2

[0191] Please see Figure 3 Another embodiment of the present invention provides: a gradient heat treatment system for cutting tooth body based on stress cloud diagram, comprising: a modeling and analysis module, a twin matching module, a partition mapping module and a monitoring and feedback module;

[0192] The modeling and analysis module is used to obtain the three-dimensional model of the initial stress of the cutting tooth. Through the finite element simulation algorithm combined with the preset first relationship coupling model and constraint space, the simulated stress distribution cloud map and the failure risk probability of the corresponding scenario are obtained for each scenario.

[0193] The partitioning mapping module, based on the simulated stress distribution cloud map and the corresponding failure risk probability in each scenario, obtains the tooth partitioning result sequence for each scenario through the regional cloud map feature extraction model combined with the preset partitioning performance mapping; the tooth partitioning result sequence for each scenario includes partitioning range information and regional heat treatment difference target performance index information; the target performance index information includes hardness, wear resistance, impact toughness and bending strength.

[0194] The twin matching module, based on the result sequence of the cutting tooth partition in each scenario and the process template knowledge graph, obtains the partition process parameter sequence for each scenario through the twin matching model; the partition process parameter sequence includes heating temperature gradient, holding time, cooling rate and performance index;

[0195] The monitoring and feedback module performs real-time heat treatment on the cutting teeth based on the process parameter sequence of each scenario and a temperature control fuzzy control model. At the same time, it monitors the heat treatment process, obtains a real-time processing stress distribution cloud map, and feeds the real-time processing stress distribution cloud map back to the first relational coupling model to simulate the failure risk probability. Based on the simulation results, it adjusts the heat treatment process in real time until the failure risk probability threshold is met.

[0196] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art, under the guidance of the present invention, can make changes, modifications, substitutions and variations to the above embodiments without departing from the spirit and scope of the claims. All of these variations are within the protection scope of the present invention.

Claims

1. A method of heat treating a pick body gradient based on a stress cloud map, characterized by, The method comprises the following steps: Obtain the initial stress three-dimensional model of the cutting tooth, and obtain the simulated stress distribution cloud diagram under each scene and the failure risk probability under the corresponding scene by combining the preset first relationship coupling model and the constraint space through the finite element simulation algorithm; Based on the simulated stress distribution cloud diagram under each scene and the failure risk probability under the corresponding scene, obtain the cutting tooth partition result sequence under each scene by combining the preset partition performance mapping through the regional cloud diagram feature extraction model; Based on the cutting tooth partition result sequence under each scene and the process template knowledge graph, obtain the partition process parameter sequence under each scene through the twin matching model; Based on the partition process parameter sequence under each scene, perform real-time heat treatment on the cutting tooth through the temperature fuzzy control model, monitor the heat treatment process, obtain the real-time machining stress distribution cloud diagram, and feed back the real-time machining stress distribution cloud diagram to the first relationship coupling model for failure risk probability simulation, and adjust the heat treatment process in real time according to the simulation result until the failure risk probability threshold is met.

2. The stress map based cutting pick body gradient heat treatment method of claim 1, wherein, The first relationship coupling model and the constraint space are obtained by combining the particle swarm optimization BP neural network with the friction resistance, friction temperature, transient impact load, wear degree, cutting direction and stress distribution information of the target impacted object on the cutting tooth under different scenes and the corresponding failure risk probability. The process template knowledge graph is obtained by combining the scene parameters, cutting tooth model, stress cloud diagram of each heat treatment, heating temperature gradient, holding time, cooling speed and cutting tooth performance index through the graph algorithm. The cutting tooth partition result sequence under each scene includes partition range information and regional heat treatment difference target performance index information; the regional heat treatment difference target performance index information includes cutting tooth hardness, wear resistance, impact toughness and bending strength; and the partition process parameter sequence includes heating temperature gradient, holding time, cooling speed and cutting tooth performance index.

3. The stress map based cutting pick body gradient heat treatment method of claim 2, wherein, The construction process of the first relationship coupling model comprises: Based on the collected friction resistance, friction temperature, transient impact load, wear degree, cutting direction and stress distribution information of the target impacted object on the cutting tooth under different scenes and the corresponding failure risk probability, obtain the preprocessed numerical sequence and the preprocessed image sequence through the preprocessing algorithm. The stress distribution information includes the peak stress, average stress, high stress area proportion and residual stress of the tooth head region, the stress concentration coefficient, alternating stress amplitude and fatigue safety factor of the tooth handle region, the stress gradient, stress uniformity variation coefficient and shear stress of the transition region, and the equivalent plastic strain, fracture toughness and stress amplitude of different part coordinate points of the whole cutting tooth; The failure risk probability is taken as the target variable, the cutting tooth performance index is taken as the first main variable, the stress distribution information is taken as the second main variable, the hardness of the target impact object is taken as the background variable, and the influence coefficients of the friction resistance, friction temperature, transient impact load, wear degree, cutting direction and cutting tooth performance index of the target impacted object on the cutting tooth are taken as the collaborative variables to construct the hierarchical variable matrix.

4. The stress map based cutting pick body gradient heat treatment method of claim 3, wherein, The construction process of the first relationship coupling model further includes: Obtaining the historical failure risk probability corresponding to different regions in the layered variable matrix and the stress distribution information of the corresponding region, combining each first sub-main variable in the first main variable, and through factor analysis combined with a preset contribution rate threshold, obtaining a first sub-main variable sequence corresponding to each region and a corresponding contribution degree, and constructing a first risk association connection using the contribution degree of the first sub-main variable sequence corresponding to each region; Based on each first sub-main variable and the second main variable in the first main variable, through principal component analysis combined with a preset contribution rate threshold, obtaining a second main variable sequence corresponding to each first sub-main variable and a corresponding contribution degree and an autocorrelation coefficient between each second main variable sequence, and constructing a second performance association connection using the contribution degree of each first sub-main variable and the second main variable sequence, and simultaneously constructing a transverse stress association coefficient using the autocorrelation coefficient between each second main variable sequence.

5. The stress map based cutting pick body gradient heat treatment method of claim 4, wherein, The construction process of the first relationship coupling model further includes: Based on the first sub-main variable sequence and the corresponding contribution degree and the failure risk probability of the corresponding region, a BP neural network with a built-in damage accumulation function is used to construct a first risk association coupling function corresponding to each region of the cutting tooth; The damage accumulation function is constructed by taking multi-axial equivalent stress, real-time temperature, plastic strain rate and time as input variables combined with a nonlinear differential equation, for measuring the damage accumulation amount corresponding to different time points of the cutting tooth, and correcting the predicted failure risk probability, thereby guiding the real-time regulation of the heat treatment process; Based on each first sub-main variable and the corresponding second main variable sequence and the corresponding contribution degree of each region and the autocorrelation coefficient between each second sub-main variable sequence under each second main variable, a BP neural network is used to obtain a second performance association coupling function corresponding to each region of the cutting tooth; Based on each second sub-main variable sequence corresponding to each region, the background variable and the cooperative variable in the layered variable matrix are combined, and through association analysis algorithm, an associated covariate set corresponding to each second sub-main variable sequence of each region and a corresponding association degree matrix are obtained.

6. The stress map based cutting pick body gradient heat treatment method of claim 5, wherein, The construction process of the first relationship coupling model further includes: Based on the associated covariate set corresponding to each second sub-main variable sequence of each region combined with the corresponding association degree matrix, a third stress association coupling function corresponding to each second sub-main variable is obtained, and a third stress cooperative association connection is constructed using the association degree matrix between each second sub-main variable sequence and the associated covariate set of each region; Based on the layered variable matrix combined with the first risk association connection, the second performance association connection, the transverse stress association coefficient, the third stress cooperative association connection and the stress gradient of the transition zone, through a topological space algorithm combined with the first risk association coupling function, the second performance association coupling function and the third stress association coupling function, a layered variable association mapping matrix is constructed.

7. The stress map based cutting pick body gradient heat treatment method of claim 6 wherein, The construction process of the first relationship coupling model further includes: The comprehensive risk prediction function is constructed by a weighting algorithm based on the first risk correlation coupling function, the second performance correlation coupling function and the third stress correlation coupling function, and the particle swarm algorithm input is constructed by using the weight corresponding to the comprehensive risk prediction function and the weights corresponding to the first risk correlation coupling function, the second performance correlation coupling function and the third stress correlation coupling function. The fitness function is constructed based on the minimum value of the comprehensive risk prediction function, and the constraint space is constructed by using the hierarchical variable correlation mapping matrix in combination with the constraint information and the constraint variable value in each historical scene. Based on the particle swarm algorithm input, the fitness function, the constraint space, the variable information in the hierarchical variable matrix in each scene obtained by the finite element combined simulation algorithm and the preset training period, the first relationship coupling model is obtained by training through the particle swarm algorithm, and the trained first risk correlation coupling function, the second performance correlation coupling function, the third stress correlation coupling function and the minimum failure risk probability in each scene are output.

8. The stress map based cutting pick body gradient heat treatment method of claim 7, wherein, The acquisition process of the cutting tooth partition result sequence in each scene includes: Based on the simulated stress distribution cloud map in each scene obtained by the first relationship coupling model, initial region division is performed through the grid segmentation layer of the region cloud feature extraction model to obtain the non-transition region and the transition region of the cutting tooth in each scene; The non-transition region is input into the non-transition feature extraction layer constructed by the bidirectional time convolution to obtain the non-transition stress feature vector, and the transition region is input into the transition feature extraction layer with the stress equivalent gradient function built-in to obtain the transition stress feature vector; The non-transition stress feature vector and the transition stress feature vector are input into the convolution attention fusion layer for fusion according to the position points corresponding to the stress feature vectors to obtain a complete cutting tooth stress distribution feature cloud map; Based on the complete cutting tooth stress distribution feature cloud map and the partition performance rule, an output layer is constructed by using a support vector machine to obtain the cutting tooth partition result sequence in each scene. The partition performance rule is obtained by analyzing the stress features corresponding to each different component of the cutting tooth in historical machining and the partition process parameters through an association analysis algorithm.

9. The stress map based cutting pick body gradient heat treatment method of claim 8, wherein, The acquisition of the partition process parameter sequence in each scene includes: Based on the cutting tooth partition result sequence in each scene and the process template knowledge graph, an initial matching partition process parameter sequence is obtained by matching algorithm and partition performance rule; Based on the initial matching partition process parameter sequence, a heating-cooling control instruction is generated by using a temperature control fuzzy control model combined with a temperature-stress change function; the temperature-stress change function is constructed by using a convolution radial kernel function according to the temperature change rate per unit time of different region position points of the cutting tooth and the stress change amplitude of the corresponding region position points, and is used to measure the stress change state under different temperature changes to adjust the holding time length of the corresponding region position points; The real-time stress map of the cutting tooth body processed by the heating-cooling control instruction is monitored in real time, and a complete real-time cutting tooth stress distribution feature cloud map of the real-time stress map is extracted by using a twin algorithm combined with a region cloud feature extraction model. The real-time cutting tooth stress distribution characteristic cloud map is fed back to the first relationship coupling model for corresponding failure risk prediction, and the real-time cutting tooth stress distribution characteristic cloud map and the corresponding failure risk prediction and the stress distribution cloud map simulated under the corresponding scene and the minimum failure risk probability under the corresponding scene are compared to obtain a stress characteristic deviation map and a failure risk deviation value; The stress characteristic deviation map and the failure risk deviation value are fed back to the regional cloud map feature extraction model and the twin matching model to real-time adjust the partition process parameter sequence matching process until the real-time cutting tooth stress distribution characteristic cloud map and the stress distribution cloud map simulated under the corresponding scene correspond to the deviation and the real-time failure risk prediction and the minimum failure risk probability simulated under the corresponding scene correspond to the deviation value that meet the corresponding preset deviation threshold.

10. A pick tooth body gradient heat treatment system based on stress contouring, implemented based on the pick tooth body gradient heat treatment method based on stress contouring of any one of claims 1-9, characterized in that, Comprise: The modeling analysis module, the twin matching module, the partition mapping module and the monitoring feedback module; The modeling analysis module is used for obtaining a cutting tooth initial stress three-dimensional model, and obtaining a simulated stress distribution cloud map under each scene and a failure risk probability under the corresponding scene through a finite element simulation algorithm combined with a preset first relationship coupling model and a constraint space; The partition mapping module obtains a cutting tooth partition result sequence under each scene through a regional cloud map feature extraction model combined with a preset partition performance mapping based on the simulated stress distribution cloud map under each scene and the failure risk probability under the corresponding scene; the cutting tooth partition result sequence under each scene comprises partition range information and regional heat treatment difference target performance index information; the target performance index information comprises hardness, wear resistance, impact toughness and bending strength; The twin matching module obtains a partition process parameter sequence under each scene through a twin matching model based on the cutting tooth partition result sequence under each scene combined with a process template knowledge graph; The partition process parameter sequence comprises a heating temperature gradient, a holding time, a cooling speed and a performance index; The monitoring feedback module obtains a real-time processing stress distribution cloud map by monitoring a heat treatment process through a temperature control fuzzy control model for real-time heat treatment of the cutting tooth based on the partition process parameter sequence under each scene, feeds the real-time processing stress distribution cloud map back to the first relationship coupling model for failure risk probability simulation, and adjusts the heat treatment process in real time according to the simulation result until the failure risk probability threshold is met.

Images