Intelligent control method and system for temperature gradient of steel ball heat treatment

Abstract

The application discloses a steel ball heat treatment temperature gradient intelligent regulation method and system, and relates to the technical field of temperature regulation.The steel ball heat treatment temperature gradient intelligent regulation method and system comprise the following steps: S1, obtaining steel ball heat treatment data and preprocessing the steel ball heat treatment data; S2, constructing a space-time derivative relationship between thermal disturbance residual and surface temperature, generating a thermal disturbance coupling evaluation value, and identifying a candidate micro-disturbance region through threshold judgment and clustering operation; S3, constructing a power disturbance joint data set, analyzing the coupling strength of disturbance trend response lag and spatial thermal inhomogeneity, and using the coupling strength to adjust a heating power setting value, drive point temperature matrix updating and thermal disturbance coupling evaluation value recalculation; and S4, generating a target number set, evaluating a thermal disturbance response offset value of the target number, constructing a path branch regulation structure, and correcting a thermal disturbance regulation driving value calculation mode.The problems that local thermal disturbance is covered by the overall thermal field and early-stage abnormalities cannot be identified are solved.

Description

[0001] This application is a divisional application of application filed on October 21, 2025, with application number 2025115062780 and invention title "Intelligent Control Method, System and Device for Heat Treatment Temperature Gradient of Steel Ball". Technical Field

[0002] This invention relates to the field of temperature control technology, specifically to a method and system for intelligent control of temperature gradient in the heat treatment of steel balls. Background Technology

[0003] In industrial heat treatment processes, temperature control systems are the core support for ensuring the stability of material properties and the consistency of finished product quality. Especially when heat-treating heat-sensitive materials or structurally complex parts, traditional single-point temperature control or overall average control modes often cannot effectively cope with the effects of local disturbances, boundary heat leakage, or furnace structural asymmetry. Therefore, more refined and dynamic temperature gradient control strategies are needed.

[0004] For example, the invention with publication number CN119536408A provides a temperature control method for the annealing process of photovoltaic glass, which relates to the field of photovoltaic glass processing technology. The method involves obtaining the composition and structure information of photovoltaic glass, retrieving frequent annealing curves based on the glass composition and structure information, determining the annealing furnace for annealing treatment, inputting the reference annealing temperature curve into the controller of the annealing furnace for annealing control, and monitoring in real time. The method compares the temperature time sequence at multiple points with the reference annealing temperature curve, identifies the transverse temperature gradient and the longitudinal temperature gradient, analyzes the impact of the transverse temperature gradient and the longitudinal temperature gradient on annealing quality, and controls the annealing temperature according to the annealing quality impact index.

[0005] For example, the invention with announcement number CN117452989A relates to the field of data processing technology, specifically a method for regulating and testing the performance of temperature control valves based on a BP neural network, including the following steps: acquiring data samples of an oil-injected screw compressor during operation; constructing a basic neural network based on a Keras sequence model; inputting the data samples into the basic neural network for training, using an improved Adam algorithm to update the gradient of the neural network to obtain the optimal neural network; and using the optimal neural network to predict the temperature control valve regulation result of the sample to be tested.

[0006] The above-mentioned technical solutions improve the accuracy and responsiveness of temperature control in specific scenarios, but they still mainly focus on matching the structural parameters of the material itself and optimizing the control model. They lack systematic solutions for early identification of local disturbances during heat treatment, feedback correction of control mismatch, and dynamic path branch adjustment mechanisms. In particular, in the case of heat treatment of multi-particle steel balls, they cannot effectively deal with the problem of regional temperature sinking caused by factors such as minor air leakage and furnace door leakage.

[0007] Therefore, in order to address the above problems, there is an urgent need for a method and system for intelligent control of the temperature gradient in the heat treatment of steel balls. Summary of the Invention

[0008] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method and system for intelligent control of temperature gradient in steel ball heat treatment, which solves the problem that local thermal disturbances are masked by the overall thermal field, leading to the inability to identify early anomalies.

[0009] Technical solution To achieve the above objectives, the present invention provides the following technical solution: a method and system for intelligent control of temperature gradient in steel ball heat treatment, comprising the following steps: S1, collecting multi-source data during the steel ball heat treatment process, obtaining steel ball heat treatment data from the multi-source data, and preprocessing the steel ball heat treatment data; S2, based on the preprocessed steel ball heat treatment data, constructing the spatiotemporal derivative relationship between thermal disturbance residual and surface temperature, generating a thermal disturbance coupling evaluation value, and identifying candidate perturbation regions through threshold judgment and clustering operations; S3, extracting numbers based on candidate perturbation regions, constructing a power disturbance joint dataset, analyzing the coupling strength between disturbance trend response lag and spatial thermal inhomogeneity, and using it to adjust the heating power setpoint, driving the point temperature matrix update and recalculating the thermal disturbance coupling evaluation value; S4, re-extracting residual mutation regions based on the updated point temperature matrix and thermal disturbance coupling evaluation value, comparing them with the candidate perturbation regions of the previous frame to generate a target number set, evaluating the thermal disturbance response offset value of the target number and constructing a path branch control structure, and correcting the calculation method of the thermal disturbance control driving value.

[0010] Furthermore, the specific steps for collecting multi-source data during the heat treatment process of steel balls, obtaining heat treatment data from the multi-source data, and preprocessing the heat treatment data are as follows: By deploying thermocouple temperature probes, infrared temperature measurement arrays, furnace door edge pressure sensors, furnace gas flow rate monitoring devices, multi-layer embedded thermocouple modules, and heating circuit current detection units, the local temperature rise response, thermal field distribution, and heating energy consumption behavior of the steel balls during the heat treatment process are dynamically collected throughout the entire process, obtaining heat treatment data for the steel balls. This data includes surface point temperature, surface temperature rise rate, gas flow rate, multi-layer temperature difference, heating power, and point temperature matrix. Based on the thermal field topology... A consistent interpolation repair algorithm spatially reconstructs the heat treatment data of steel balls, filling in the temperature matrix discontinuities caused by short-term infrared occlusion and sensor hysteresis. A local derivative-limited sliding filter algorithm performs first-order time-domain smoothing on the temperature rise rate data, suppressing temperature fluctuations caused by uneven steel ball arrangement and unstable thermal radiation. A multi-layer residual focusing abrupt change detection algorithm identifies and removes short-term temperature imbalance regions in the point temperature matrix, filtering out instantaneous thermal disturbance signals caused by furnace door opening, hot gas layer interference, and boundary overheating. A component normalization weighted normalization algorithm performs unit consistency transformation on the heat treatment data of steel balls, achieving normalization of the heat treatment data.

[0011] Furthermore, based on the preprocessed heat treatment data of the steel ball, the specific steps for constructing the spatiotemporal derivative relationship between the thermal perturbation residual and the surface temperature to generate the thermal perturbation coupling evaluation value are as follows: Based on the preprocessed heat treatment data of the steel ball, the point temperature matrix at each time point is arranged in chronological order to construct a three-dimensional temperature data structure, forming a spatial temperature evolution sequence; inter-frame differencing is performed on the point temperature matrix in the time dimension to extract the thermal perturbation residual, and the first and second derivatives of the surface temperature sequence are calculated; the time derivative of the thermal perturbation residual is multiplied by the second derivative of the surface temperature to obtain the surface temperature rise rate instability term; the spatial gradient of the thermal perturbation residual is multiplied by the multilayer temperature difference and gas flow rate to obtain the thermal perturbation spatial coupling enhancement term; the surface temperature rise rate instability term and the thermal perturbation spatial coupling enhancement term are added together to obtain the thermal perturbation coupling evaluation value.

[0012] Furthermore, the specific steps for identifying candidate perturbation regions through threshold judgment and clustering operations are as follows: Real-time comparison of the thermal perturbation coupling evaluation value and the thermal perturbation threshold; when the thermal perturbation coupling evaluation value is greater than or equal to the thermal perturbation threshold, the corresponding position in the current frame is marked as an abnormal response point; when the thermal perturbation coupling evaluation value is less than the thermal perturbation threshold, the corresponding position in the current frame is marked as a normal heating point; extract all abnormal response points to construct an abnormal point set; perform eight-neighbor clustering on the abnormal point set according to spatial location within a sliding time window; calculate the cluster area and average thermal perturbation coupling evaluation value for the clustered regions; remove regions with areas smaller than the minimum perturbation area threshold and average thermal perturbation coupling evaluation values ​​lower than the perturbation intensity threshold; extract the surface temperature sequence of the corresponding position in the point temperature matrix for the retained regions; calculate the surface temperature difference between the current frame and the previous frame; when the temperature difference of more than half of the positions in the region is negative, the current region is marked as a candidate perturbation region.

[0013] Further, based on the extracted numbers of candidate perturbation regions, a joint power perturbation dataset is constructed, and the specific steps for analyzing the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity are as follows: The spatial locations of candidate perturbation regions are mapped to the pre-divided heating power control areas in the heat treatment equipment, obtaining the region number associated with each candidate region; the heat treatment data of the steel ball under that number is called, and the corresponding region's thermal perturbation coupling evaluation value and average thermal perturbation coupling evaluation value are read synchronously to construct a joint power perturbation dataset; based on the joint power perturbation dataset, the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity is analyzed: the thermal perturbation coupling evaluation value is subtracted from the average thermal perturbation coupling evaluation value, then divided by the average thermal perturbation coupling evaluation value, plus a minimum term, to obtain the perturbation trend enhancement term; the absolute value of the time derivative of the heating power is taken, and this absolute value is subtracted from 1 to obtain the power response lag term; the multi-layer temperature difference is multiplied by the gas flow rate to obtain the spatial perturbation intensity term; the perturbation trend enhancement term, the power response lag term, and the spatial perturbation intensity term are multiplied sequentially to obtain the thermal perturbation control driving value.

[0014] Furthermore, the specific steps for adjusting the heating power setpoint, driving the update of the point temperature matrix, and recalculating the thermal disturbance coupling evaluation value are as follows: Adjust the heating power setpoint of the corresponding region number in the current time frame according to the value of the thermal disturbance control driving value; use the thermal disturbance control driving value as an incremental input and add it to the current heating power output value; set an upper limit for the heating power change rate for the region number adjacent to the candidate perturbation region to limit the power adjustment amplitude per unit time; in the time frame after the heating power adjustment, call the position data of the corresponding region number in the point temperature matrix and recalculate the thermal disturbance residual; update the thermal disturbance coupling evaluation value based on the latest thermal disturbance residual, surface temperature, and multi-layer temperature difference; re-determine the abnormal response point based on the updated thermal disturbance coupling evaluation value and identify the candidate perturbation region.

[0015] Further, the specific steps for re-extracting residual mutation regions based on the updated point temperature matrix and thermal perturbation coupling evaluation values, and comparing them with the candidate perturbation regions of the previous frame to generate a target number set are as follows: In the current time frame after the heating power adjustment, extract the surface temperature data at each position in the point temperature matrix to construct the current thermal field distribution map; using the thermal image data of the inner wall of the equipment as a reference, generate the theoretical symmetrical thermal distribution grid for the current time frame; point by point, map the current thermal field distribution map and the theoretical symmetrical thermal distribution grid in spatial coordinates, recalculate the thermal perturbation residuals at each position, and construct the residual distribution map for the current time frame; perform spatial difference on the residual distribution map, extract the positions where the gradient value is greater than the mutation threshold, and form a residual mutation point set; retain three or more consecutive horizontal and vertical coordinate points in the residual mutation point set to form a suspected perturbation point set; compare each spatial coordinate in the suspected perturbation point set with the coordinates of the candidate perturbation regions marked in the previous time frame, and when the difference between the horizontal and vertical coordinates is not greater than 1, extract the region number where it is located, and summarize them to form a target number set.

[0016] Further, the specific steps for evaluating the thermal perturbation response offset value of the target number and constructing the path branch control structure, and correcting the calculation method of the thermal perturbation control driving value are as follows: For each region number in the target number set, evaluate the degree of offset between the thermal perturbation response change and the control driving: Subtract the product of the thermal perturbation control driving value and the thermal perturbation control weight from the time derivative of the thermal perturbation coupling evaluation value, and take the absolute value of the difference to obtain the perturbation response difference term; Subtract the thermal perturbation residual between the current frame and the previous frame to obtain the perturbation residual change term; Divide the perturbation residual change term by the spatial gradient of the surface temperature and add a minimum term to obtain the perturbation normalization change term; Multiply the perturbation response difference term and the perturbation normalization change term to obtain the thermal perturbation response offset evaluation. Estimation; Real-time comparison of thermal disturbance response offset evaluation value and response offset threshold. When the thermal disturbance response offset evaluation value is less than or equal to the response offset threshold, the original calculation structure of the thermal disturbance control driving value is maintained. When the thermal disturbance response offset evaluation value is greater than the response offset threshold, in the next time frame, the difference between the real-time disturbance coupling evaluation value and the average thermal disturbance coupling evaluation value is no longer used to calculate the disturbance trend enhancement term. Instead, the thermal disturbance residual sequence with that number in the sliding time window is called, and the proportion of frames that are consistent with the direction of heating power change is counted. This proportion is then introduced into the calculation structure of the disturbance trend enhancement term as a substitute term. The corrected disturbance trend enhancement term is used to update the thermal disturbance control driving value, forming the path branch control logic under the response offset condition.

[0017] The second aspect of this invention provides an intelligent temperature gradient control system for steel ball heat treatment, comprising: a steel ball heat treatment data acquisition and preprocessing module, a thermal disturbance coupling evaluation and anomaly localization module, a regional control driving modeling and power correction module, and a response offset tracking identification and dynamic correction module. The steel ball heat treatment data acquisition and preprocessing module is used to acquire multi-source data during the steel ball heat treatment process, obtain steel ball heat treatment data from the multi-source data, and preprocess the steel ball heat treatment data. The thermal disturbance coupling evaluation and anomaly localization module is used to construct the spatiotemporal derivative relationship between the thermal disturbance residual and the surface temperature based on the preprocessed steel ball heat treatment data, generate a thermal disturbance coupling evaluation value, and apply a threshold value. The system employs a judgment and clustering operation to identify candidate perturbation regions. A region-based regulation-driven modeling and power correction module is used to extract numbers from candidate perturbation regions, construct a joint power perturbation dataset, analyze the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity, and adjust the heating power setpoint to drive point temperature matrix updates and recalculate thermal perturbation coupling evaluation values. A response offset tracking and dynamic correction module is used to re-extract residual abrupt change regions based on the updated point temperature matrix and thermal perturbation coupling evaluation values, compare them with the candidate perturbation regions of the previous frame to generate a target number set, evaluate the thermal perturbation response offset value of the target number, construct a path branch regulation structure, and correct the calculation method of the thermal perturbation regulation driving value.

[0018] The third aspect of this invention provides an intelligent temperature gradient control device for heat treatment of steel balls, characterized in that it comprises: a multi-source thermal field sensing processor, a disturbance identification and control driver, and a response offset diagnostic corrector, wherein: the multi-source thermal field sensing processor is used to integrate thermal field sensing devices and data preprocessing algorithms to collect and standardize multi-source thermal parameters during the heat treatment process of steel balls, providing data support for subsequent temperature distribution analysis and state identification; the disturbance identification and control driver is used to analyze local disturbance characteristics during the heat treatment process of steel balls based on standardized heat treatment data, identify potential abnormal areas, construct a heating control response mechanism, and dynamically adjust the power distribution; the response offset diagnostic corrector is used to evaluate the response offset behavior during the heating control process, determine the matching relationship between thermal disturbance response and power adjustment, and perform structural correction on the control path to achieve adaptive closed-loop regulation.

[0019] Beneficial effects The present invention has the following beneficial effects: (1) The intelligent control method and system for temperature gradient of steel ball heat treatment, by constructing a thermal disturbance coupling evaluation value and combining the spatiotemporal derivative of the thermal disturbance residual with multi-layer temperature difference and gas flow rate, can realize the early identification of local disturbance trend during the heating process of steel ball, and provide a quantitative basis for accurate modeling of thermal field non-uniformity. At the same time, the thermal disturbance coupling evaluation value has the characteristics of real-time calculation and spatial distribution visualization, and can be used as the core criterion for subsequent micro-disturbance region screening and heating power adjustment triggering.

[0020] (2) The intelligent control method and system for heat treatment temperature gradient of steel ball can achieve dynamic and accurate compensation of the target area of ​​disturbance by constructing a thermal disturbance control driving value coupled with the disturbance trend enhancement term, power response lag term and spatial disturbance intensity term, and correcting the heating power setting value in real time, thereby improving the timeliness and pertinence of the control response.

[0021] (3) The intelligent control method and system for heat treatment temperature gradient of steel ball introduces thermal disturbance response offset evaluation value, establishes consistency analysis mechanism between thermal disturbance drive and response, and adaptively adjusts control path when thermal disturbance response offset evaluation value exceeds response offset threshold, so as to avoid energy waste and temperature over-adjustment caused by power regulation mismatch and improve the robustness of control strategy.

[0022] (4) The intelligent control method and system for the temperature gradient of the heat treatment of steel balls, through the comparison of theoretical thermal images with symmetrical thermal grids and the extraction of thermal disturbance residual mutation maps, constructs a non-equilibrium diagnostic model of micro-perturbation thermal field. It can realize early perception and active intervention of fine-grained problems such as micro-air leakage and abnormal furnace door sealing before the total thermal field triggers the system alarm. Attached Figure Description

[0023] Figure 1 Flowchart of intelligent temperature gradient control method for heat treatment of steel balls; Figure 2 This is a structural diagram of an intelligent temperature gradient control system for heat treatment of steel balls. Figure 3 Distribution map of regional thermal disturbance coupling assessment values; Figure 4 This is a line graph showing the offset determination of regional thermal disturbance response. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1-4This invention provides a technical solution: a method and system for intelligent control of temperature gradient in steel ball heat treatment, comprising the following steps: S1, collecting multi-source data during the steel ball heat treatment process, obtaining steel ball heat treatment data from the multi-source data, and preprocessing the steel ball heat treatment data; S2, based on the preprocessed steel ball heat treatment data, constructing the spatiotemporal derivative relationship between thermal disturbance residual and surface temperature, generating a thermal disturbance coupling evaluation value, and identifying candidate perturbation regions through threshold judgment and clustering operations; S3, extracting numbers based on candidate perturbation regions, constructing a power perturbation joint dataset, analyzing the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity, and using it to adjust the heating power setpoint, driving the point temperature matrix update and recalculating the thermal perturbation coupling evaluation value; S4, re-extracting residual mutation regions based on the updated point temperature matrix and thermal perturbation coupling evaluation value, comparing them with the candidate perturbation regions of the previous frame to generate a target number set, evaluating the thermal perturbation response offset value of the target number and constructing a path branch control structure, and correcting the calculation method of the thermal perturbation control driving value.

[0026] Specifically, the following steps are taken to collect multi-source data during the heat treatment of steel balls, obtain heat treatment data from the multi-source data, and preprocess the heat treatment data: By deploying thermocouple temperature probes, infrared temperature arrays, furnace door edge pressure sensors, furnace gas flow rate monitoring devices, multi-layer embedded thermocouple modules, and heating circuit current detection units, and based on preset measurement points covering the entire heating area, the local temperature rise response, thermal field distribution, and heating energy consumption behavior of the steel balls during the heat treatment process are dynamically collected throughout the entire process. This ensures the continuity and accuracy of the heat treatment data in both time and space dimensions, obtaining heat treatment data for the steel balls. This heat treatment data includes surface point temperature, surface temperature rise rate, gas flow rate, multi-layer temperature difference, heating power, and point temperature matrix. The steel ball heat treatment data is then processed using an interpolation repair algorithm based on thermal field topology consistency. Spatial reconstruction of the ball heat treatment data is performed to fill in the temperature matrix discontinuities caused by short-term infrared occlusion and sensor hysteresis, enhancing the spatial continuity of the point temperature matrix. A local derivative-limited sliding filter algorithm is used to perform first-order time-domain smoothing on the temperature rise rate data, suppressing temperature fluctuations caused by uneven steel ball arrangement and unstable thermal radiation, and extracting the stable trend of temperature rise rate. A multi-layer residual focusing abrupt change detection algorithm is used to identify and remove short-term temperature imbalance regions in the point temperature matrix, filtering out instantaneous thermal disturbance signals caused by furnace door slight opening, hot gas layer interference, and boundary overheating, improving the purity of the extracted thermal disturbance residuals. A component normalization weighted normalization algorithm is used to perform unit consistency transformation on the steel ball heat treatment data, achieving normalization of the data and providing a unified measurement basis for the subsequent evaluation model in terms of input feature dimensions.

[0027] In this implementation scheme, the acquisition and preprocessing of steel ball heat treatment data enables high-frequency and high-precision acquisition of surface point temperature, surface temperature rise rate, gas flow rate, multi-layer temperature difference, heating power, and point temperature matrix. In terms of data quality improvement, interpolation repair enhances the spatial integrity of the point temperature matrix, sliding filtering improves the stability of the temperature rise rate, residual mutation detection eliminates unstructured thermal disturbance noise, and normalization processing establishes a unified dimensional standard. This provides an accurate, continuous, and clean input data foundation for subsequent calculation of thermal disturbance coupling evaluation values, abnormal area identification, and control-driven construction.

[0028] Specifically, based on the preprocessed heat treatment data of the steel ball, the spatiotemporal derivative relationship between the thermal perturbation residual and the surface temperature is constructed, and the specific steps for generating the thermal perturbation coupling evaluation value are as follows: Based on the preprocessed heat treatment data of the steel ball, the point temperature matrix at each time point is arranged in chronological order to construct a three-dimensional temperature data structure, forming a spatial temperature evolution sequence, so that the point temperature matrix has continuous evolution characteristics on the time axis; inter-frame differencing is performed on the point temperature matrix in the time dimension to extract the thermal perturbation residual, and the significance of short-term thermal field changes is further enhanced by combining the differencing results; and the first and second derivatives of the surface temperature sequence are calculated to capture the surface temperature. The surface temperature change trend and acceleration information are analyzed. The time derivative of the thermal disturbance residual is multiplied by the second derivative of the surface temperature to obtain the surface temperature rise rate instability term, which reflects the degree of nonlinear fluctuation in the local thermal response of the steel ball surface. The spatial gradient of the thermal disturbance residual is multiplied by the multilayer temperature difference and gas velocity to obtain the thermal disturbance spatial coupling enhancement term, which reflects the cumulative enhancement effect of thermal disturbance under vertical interlayer conduction and convection disturbance. The surface temperature rise rate instability term and the thermal disturbance spatial coupling enhancement term are added to obtain the thermal disturbance coupling evaluation value, which is used to characterize the comprehensive response intensity of the thermal field disturbance in the current heating process on the spatiotemporal scale.

[0029] The specific formula for calculating the thermal disturbance coupling evaluation value is as follows: ; In the formula, S represents the thermal disturbance coupling evaluation value, A represents the surface temperature, C represents the thermal disturbance residual, D represents the multilayer temperature difference, and V represents the gas flow rate.

[0030] In this embodiment, Table 1 is a data table of thermal disturbance coupling evaluation values. The variables included in the table are: surface temperature, thermal disturbance residual, multi-layer temperature difference, gas flow rate, and thermal disturbance coupling evaluation values ​​calculated based on the thermal disturbance evaluation model. Specifically, region B1 has a surface temperature of 720°C, a thermal disturbance residual of 0.16, a multi-layer temperature difference of 28, a gas flow rate of 2.2, and a calculated thermal disturbance coupling evaluation value of 99.06; region B2 has a surface temperature of 710°C, a thermal disturbance residual of 0.15, a multi-layer temperature difference of 29, a gas flow rate of 2.0, and a calculated thermal disturbance coupling evaluation value of 87.44; region B3 has a surface temperature of 730°C, a thermal disturbance residual of 0.17, a multi-layer temperature difference of 2... 7. With a gas flow rate of 2.4, the calculated thermal disturbance coupling evaluation value is 110.66; the surface temperature of region B4 is 725, the thermal disturbance residual is 0.155, the multi-layer temperature difference is 28, and the gas flow rate is 2.1, the calculated thermal disturbance coupling evaluation value is 91.64; the surface temperature of region B5 is 715, the thermal disturbance residual is 0.16, the multi-layer temperature difference is 30, and the gas flow rate is 2.3, the calculated thermal disturbance coupling evaluation value is 110.92.

[0031] Table 1. Thermal Disturbance Coupling Evaluation Values

[0032] like Figure 3 The diagram shows the distribution of regional thermal disturbance coupling assessment values, displaying the thermal disturbance coupling assessment values ​​for five regional locations. Each bar corresponds to the local thermal disturbance intensity of a specific region. The blue dashed line represents the system-set thermal disturbance threshold, used to distinguish between abnormal response points and normal heating points. Green bars indicate that the thermal disturbance coupling assessment value for that region is below the thermal disturbance threshold, and is judged as a normal heating point. Red bars indicate that the assessment value exceeds the thermal disturbance threshold, and is judged as an abnormal response point, indicating that there may be problems such as enhanced local thermal disturbance, delayed control response, and uneven spatial temperature distribution in that region. Each bar is labeled with a judgment label and the actual assessment value, facilitating intuitive identification of the regional status and providing a reference for subsequent drive control and response correction.

[0033] In this implementation scheme, by constructing a thermal perturbation coupling evaluation value based on the relationship between thermal perturbation residual and the spatiotemporal derivative of surface temperature, it is possible to integrate the instability of surface temperature rise rate and the coupling enhancement characteristics of spatial thermal perturbation. Considering the linkage between the time derivative of thermal perturbation residual and the second derivative of surface temperature, the influence of multi-layer temperature difference and gas flow velocity on the spatial thermal field perturbation intensity is superimposed, thereby realizing the quantitative expression of the local perturbation evolution trend during the heat treatment of steel balls. This provides accurate, dynamic, and structurally clear key input indicators for subsequent abnormal response point identification, candidate micro-perturbation region extraction, and control driving value construction.

[0034] Specifically, the steps for identifying candidate perturbation regions through threshold judgment and clustering are as follows: Real-time comparison of the thermal perturbation coupling evaluation value and the thermal perturbation threshold. When the thermal perturbation coupling evaluation value is greater than or equal to the thermal perturbation threshold, the corresponding position in the current frame is marked as an abnormal response point, serving as the initial criterion for potential perturbation-sensitive regions. When the thermal perturbation coupling evaluation value is less than the thermal perturbation threshold, the corresponding position in the current frame is marked as a normal heating point, used to distinguish the distribution characteristics of stable heating regions and local abnormal regions in the thermal field. All abnormal response points are extracted to construct an abnormal point set. Within a sliding time window, eight-neighbor clustering is performed on the abnormal point set according to its spatial location. Based on the adjacency relationships of each abnormal response point in the up, down, left, right, and four diagonal directions, a connected structure is constructed, forming... Local spatial clustering regions are used to enhance the identification of aggregation features in thermal perturbation response. The cluster area and average thermal perturbation coupling evaluation value are calculated for the clustered regions to comprehensively evaluate the perturbation scale and intensity of the clustered regions. Regions with an area smaller than the minimum perturbation area threshold and an average thermal perturbation coupling evaluation value lower than the perturbation intensity threshold are removed, filtering out isolated perturbation pixels and low-amplitude response regions. For the retained regions, the surface temperature sequence of the corresponding position is extracted from the point temperature matrix, and the surface temperature difference between the current frame and the previous frame is calculated to further identify the actual response change direction of the heating trend. When the temperature difference of more than half of the positions in the region is negative, the current region is marked as a candidate micro-perturbation region, which serves as a key positioning basis for subsequent regulation-driven construction and dynamic compensation adjustment.

[0035] In this implementation scheme, by using threshold judgment based on thermal perturbation coupling evaluation value and spatial clustering operation, it is possible to accurately identify and regionalize abnormal response points. While ensuring that abnormal response points are marked in a timely manner when the thermal perturbation coupling evaluation value exceeds the thermal perturbation threshold, the scheme combines eight-neighbor clustering within a sliding time window with the dual constraints of cluster area and average thermal perturbation coupling evaluation value to eliminate local invalid perturbation regions. Furthermore, the scheme verifies the temperature change trend within the region through inter-frame difference of surface temperature sequence, thereby improving the accuracy and physical consistency of candidate perturbation region screening. This provides a more accurate perturbation region input source for subsequent control region numbering and dynamic adjustment of heating power.

[0036] Specifically, the steps for constructing a power disturbance joint dataset based on candidate disturbance regions and extracting their numbers, and analyzing the coupling strength between disturbance trend response lag and spatial thermal inhomogeneity, are as follows: The spatial locations of candidate disturbance regions are mapped to the pre-divided heating power control areas in the heat treatment equipment, obtaining the region number associated with each candidate region to form a mapping relationship between spatial disturbance response and control unit; the heat treatment data of the steel ball under that number is retrieved, and the thermal disturbance coupling evaluation value and average thermal disturbance coupling evaluation value of the corresponding region are read simultaneously. Combined with the original collected data and intermediate calculation indicators, a power disturbance joint dataset is constructed to describe the multidimensional coupling state between disturbance response and control behavior; based on the power disturbance joint dataset, the coupling strength between disturbance trend response lag and spatial thermal inhomogeneity is analyzed. Uniform coupling strength: Subtracting the average thermal disturbance coupling evaluation value from the thermal disturbance coupling evaluation value and dividing by the average thermal disturbance coupling evaluation value, plus a minimum term, yields the disturbance trend enhancement term, which quantifies the relative surge in thermal disturbance in the current region compared to the historical average. Taking the absolute value of the time derivative of the heating power and subtracting this absolute value from 1 yields the power response lag term, reflecting the slowness of the heating power adjustment per unit time. Multiplying the multi-layer temperature difference by the gas flow velocity yields the spatial disturbance strength term, characterizing the degree of spatial thermal imbalance under the combined effects of inter-layer thermal field conduction intensity and airflow disturbance. Multiplying the disturbance trend enhancement term, the power response lag term, and the spatial disturbance strength term in sequence yields the thermal disturbance control driving value, which serves as the dynamic control input benchmark for guiding the correction of the heating power setpoint for the region number.

[0037] The specific formula for calculating the thermal disturbance control driving value is as follows: ; In the formula, R represents the thermal disturbance control driving value, and S represents the thermal disturbance coupling evaluation value. The average thermal disturbance coupling assessment value is represented by Q, where Q represents heating power, D represents multi-layer temperature difference, and V represents gas flow rate. Indicates a minus term.

[0038] In this implementation scheme, by constructing a joint dataset of power disturbances and analyzing the coupling strength between disturbance trend response lag and spatial thermal inhomogeneity, the candidate disturbance region number can be accurately identified. Based on key variables such as thermal disturbance coupling evaluation value, average thermal disturbance coupling evaluation value, multi-layer temperature difference, gas flow rate and heating power time derivative, disturbance trend enhancement term, power response lag term and spatial disturbance intensity term are extracted to form thermal disturbance regulation driving value as the basis for dynamic adjustment of temperature control system, thereby realizing intelligent correction of local heating power and enhancing the rapid response capability and spatial adaptability of temperature control system under disturbance intervention scenario.

[0039] Specifically, the steps for adjusting the heating power setpoint, driving the point temperature matrix update, and recalculating the thermal disturbance coupling evaluation value are as follows: Based on the value of the thermal disturbance control driving value, adjust the heating power setpoint of the corresponding region number in the current time frame; implement parameter injection by mapping the index relationship between the region number and the heating power output channel; use the thermal disturbance control driving value as an incremental input, superimpose it onto the current heating power output value to form a new heating power command signal, and transmit it to the temperature control execution port; for region numbers adjacent to the candidate perturbation region, set an upper limit for the heating power change rate to limit the power adjustment amplitude per unit time and prevent step power changes from causing regional disturbances. The temperature fluctuates drastically. In the time frame after the heating power is adjusted, the location data of the corresponding region number in the point temperature matrix is ​​called, and the corresponding surface point temperature sequence and historical temperature rise trend data are extracted. The thermal perturbation residual is recalculated, and the residual distribution map under the current frame is constructed. Based on the latest thermal perturbation residual, surface temperature and multi-layer temperature difference, the thermal perturbation coupling evaluation value is updated to maintain the synchronization consistency between the model output and the actual thermal response on site. Based on the updated thermal perturbation coupling evaluation value, abnormal response points are re-determined, the abnormal point set reconstruction process is executed, and candidate micro-perturbation regions are identified to provide updated regional response basis for the calculation of thermal perturbation control driving value in the next time frame.

[0040] In this implementation scheme, by dynamically adjusting the heating power setpoint based on the thermal disturbance control driving value and superimposing a constraint strategy that limits the rate of power change, it is possible to achieve refined power correction operation for region numbering, thereby improving the stability and specificity of power output response. At the same time, in each time frame, the thermal disturbance residual is recalculated based on the updated heating power command, and the thermal disturbance coupling evaluation value is updated in combination with surface temperature and multi-layer temperature difference, ensuring a high degree of consistency between the evaluation model and the actual thermal field state. This further enables real-time re-judgment of abnormal response points and continuous tracking of candidate micro-disturbance regions, thereby enhancing the temperature control system's closed-loop perception capability of the disturbance evolution process and the accuracy of regional adaptive control.

[0041] Specifically, the steps for re-extracting residual abrupt change regions based on the updated point temperature matrix and thermal perturbation coupling evaluation values, and comparing them with the candidate perturbation regions of the previous frame to generate a target number set are as follows: In the current time frame after the heating power adjustment, surface temperature data at each position in the point temperature matrix are extracted to construct the current thermal field distribution map, reflecting the actual temperature response state of each region; using the thermal image data of the inner wall of the equipment as a reference, a theoretical symmetrical thermal distribution grid for the current time frame is generated. The theoretical symmetrical thermal distribution grid is a two-dimensional thermal field reference matrix constructed based on the geometric symmetry of the furnace cavity structure and the assumption of symmetrical temperature distribution at each point under ideal heating conditions. It is used to characterize the ideal state distribution of the point temperature matrix under undisturbed conditions and to establish an ideal thermal field model based on the symmetry of the furnace cavity; the current thermal field distribution map and the theoretical symmetrical thermal distribution grid are mapped point-by-point in spatial coordinates. The thermal perturbation residuals at each location are recalculated, and the perturbation deviation under the ideal thermal distribution is removed to construct the residual distribution map for the current time frame. Spatial difference is performed on this residual distribution map to extract locations where the gradient value is greater than the abrupt change threshold, thereby enhancing the response sensitivity to the boundary of severe perturbation and forming a set of residual abrupt change points. Three or more consecutive coordinate points in the horizontal and vertical directions are retained in the set of residual abrupt change points, and isolated points and noise signals are filtered out to form a set of suspected perturbation points, thus constructing a perturbation response region with spatial structural characteristics. Each spatial coordinate in the set of suspected perturbation points is compared with the coordinates of the candidate perturbation regions marked in the previous time frame. When the difference between the horizontal and vertical coordinates is not greater than 1, the region number is extracted to ensure the spatial continuity identification mechanism. The results are then summarized to form a set of target numbers, providing region number input for subsequent offset analysis and control path correction.

[0042] In this implementation scheme, residual abrupt change regions are re-extracted based on the updated point temperature matrix and thermal disturbance coupling evaluation values. A target number set is generated by comparing the coordinates of candidate micro-disturbance regions from the previous time frame. This allows for timely identification of new disturbance response locations after thermal field state adjustments. Under the spatial correspondence between the theoretical symmetrical thermal distribution grid and the current thermal field distribution map, thermal disturbance residuals are extracted and a residual distribution map is constructed. Spatial continuity constraints are introduced into the residual abrupt change point set to enhance the stability of disturbance region boundary identification. Furthermore, the continuity relationship of disturbances between time frames is established through coordinate comparison, ensuring that the temperature control system has the ability to continuously track micro-disturbance sources and recursively identify paths. This provides a dynamically updated number input basis for subsequent thermal disturbance response offset evaluation and control logic branch switching.

[0043] Specifically, the specific steps for evaluating the thermal disturbance response offset value of the target number and constructing the path branch control structure, and correcting the calculation method of the thermal disturbance control driving value are as follows: For each region number in the target number set, evaluate the degree of offset between the thermal disturbance response change and the control driving: Subtract the product of the thermal disturbance control driving value and the thermal disturbance control weight from the time derivative of the thermal disturbance coupling evaluation value, and take the absolute value of the difference to obtain the disturbance response difference term; wherein, by statistically analyzing the proportion of frames in the same direction as the change in heating power and the change in thermal disturbance coupling evaluation value within the sliding time window, and combining the disturbance residual change amplitude of the corresponding region number, the thermal disturbance control weight is calculated using a proportional weighting method, and the value range of the thermal disturbance control weight is [0,1]; Subtract the thermal disturbance residual between the current frame and the previous frame to obtain the disturbance residual change term; Divide the disturbance residual change term by the spatial gradient of the surface temperature. Adding a minimum term to the degree yields the normalized change term for the disturbance. Multiplying the disturbance response difference term by the normalized change term yields the thermal disturbance response offset evaluation value. The thermal disturbance response offset evaluation value and the response offset threshold are compared in real time. When the thermal disturbance response offset evaluation value is less than or equal to the response offset threshold, the original calculation structure of the thermal disturbance control driving value is maintained. When the thermal disturbance response offset evaluation value is greater than the response offset threshold, the difference between the real-time disturbance coupling evaluation value and the average thermal disturbance coupling evaluation value is no longer used to calculate the disturbance trend enhancement term in the next time frame. Instead, the thermal disturbance residual sequence with that number in the sliding time window is called, and the proportion of frames that are consistent with the direction of heating power change is counted. This proportion is then introduced into the calculation structure of the disturbance trend enhancement term as a substitute term. The corrected disturbance trend enhancement term is used to update the thermal disturbance control driving value, forming the path branch control logic under the response offset condition.

[0044] The specific formula for calculating the thermal disturbance response offset evaluation value is as follows: In the formula, M represents the thermal disturbance response offset evaluation value, R represents the thermal disturbance control driving value, S represents the thermal disturbance coupling evaluation value, A represents the surface temperature, and C represents the thermal disturbance residual. This represents the change term of the disturbance residual. Indicates minterms, This indicates the thermal disturbance control weight.

[0045] In this embodiment, Table 2 is a data table of thermal disturbance response offset evaluation values. The variables included in the data table are: surface temperature, thermal disturbance residual, disturbance residual variation term, thermal disturbance control driving value, thermal disturbance coupling evaluation value, thermal disturbance control weight, and thermal disturbance response offset evaluation value calculated according to the response offset evaluation model. Specifically, region Y1 has a surface temperature of 720°C, a thermal perturbation residual of 9.0, a perturbation residual variation term of 0.9, a thermal perturbation control driving value of 1.9, a thermal perturbation coupling evaluation value of 98, a thermal perturbation control weight of 0.68, and a calculated thermal perturbation response offset evaluation value of 0.770; region Y2 has a surface temperature of 716°C, a thermal perturbation residual of 9.0, a perturbation residual variation term of 0.9, a thermal perturbation control driving value of 1.7, a thermal perturbation coupling evaluation value of 95, a thermal perturbation control weight of 0.68, and a calculated thermal perturbation response offset evaluation value of 0.777; and region Y3 has a surface temperature of 730°C, a thermal perturbation residual of 10.0, a perturbation residual variation term of 1.0, and a thermal perturbation control driving value of 2. 2. The thermal disturbance coupling evaluation value is 109, the thermal disturbance control weight is 0.68, and the calculated thermal disturbance response offset evaluation value is 0.967; the surface temperature of region Y4 is 725, the thermal disturbance residual is 8.5, the disturbance residual variation term is 0.8, the thermal disturbance control driving value is 1.8, the thermal disturbance coupling evaluation value is 95, the thermal disturbance control weight is 0.68, and the calculated thermal disturbance response offset evaluation value is 0.831; the surface temperature of region Y5 is 719, the thermal disturbance residual is 9.6, the disturbance residual variation term is 0.96, the thermal disturbance control driving value is 1.9, the thermal disturbance coupling evaluation value is 108, the thermal disturbance control weight is 0.68, and the calculated thermal disturbance response offset evaluation value is 0.898.

[0046] Table 2 Data table of thermal disturbance response offset evaluation values

[0047] like Figure 4 The figure shows a line graph illustrating the regional thermal disturbance response offset determination. It displays the thermal disturbance response offset determination for five region numbers, used to analyze the consistency between the control drive and the thermal disturbance response. The blue dashed line represents the response offset threshold, used to distinguish response offset states; red squares indicate offset regions where the thermal disturbance response offset assessment value is higher than the response offset threshold; green dots represent stable regions within the response offset threshold. The connecting black lines reflect the continuous change in the degree of offset between regions, helping to identify the clustering trend and spatial distribution anomalies of the response offset, providing support for subsequent thermal control path correction and feedback strategy selection.

[0048] In this implementation scheme, by evaluating the thermal disturbance response offset value of the target number and constructing a path branch control structure, a response consistency verification mechanism can be established between the thermal disturbance coupling evaluation value and the thermal disturbance control driving value. Using the thermal disturbance response offset evaluation value composed of the disturbance response difference term and the disturbance normalization change term, the response lag state in the control execution process can be dynamically identified. When the response offset evaluation value exceeds the response offset threshold, the original disturbance trend enhancement term composition method is replaced by the proportion of frames in the historical thermal disturbance residual sequence that are consistent with the direction of heating power change. This modifies the calculation structure of the thermal disturbance control driving value and forms a path branch control logic for the offset state, thereby enhancing the steady-state maintenance capability and strategy adaptation capability of the temperature control system under complex disturbance conditions.

[0049] like Figure 2 As shown, the second aspect of the present invention provides an intelligent control system for the temperature gradient of steel ball heat treatment, comprising: a steel ball heat treatment data acquisition and preprocessing module, a thermal disturbance coupling evaluation and anomaly localization module, a regional control driving modeling and power correction module, and a response offset tracking identification and dynamic correction module, wherein: the steel ball heat treatment data acquisition and preprocessing module is used to acquire multi-source data during the steel ball heat treatment process, obtain steel ball heat treatment data from the multi-source data, and preprocess the steel ball heat treatment data; the thermal disturbance coupling evaluation and anomaly localization module is used to construct the spatiotemporal derivative relationship between the thermal disturbance residual and the surface temperature based on the preprocessed steel ball heat treatment data, generate a thermal disturbance coupling evaluation value, and pass a threshold... The module identifies candidate perturbation regions through value judgment and clustering operations; the regional regulation-driven modeling and power correction module is used to extract numbers based on candidate perturbation regions, construct a joint power perturbation dataset, analyze the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity, and adjust the heating power setpoint to drive the update of the point temperature matrix and recalculation of the thermal perturbation coupling evaluation value; the response offset tracking identification and dynamic correction module is used to re-extract residual abrupt change regions based on the updated point temperature matrix and thermal perturbation coupling evaluation value, compare them with the candidate perturbation regions of the previous frame to generate a target number set, evaluate the thermal perturbation response offset value of the target number and construct a path branch regulation structure to correct the calculation method of the thermal perturbation regulation driving value.

[0050] In this implementation scheme, by constructing a data acquisition and preprocessing module for steel ball heat treatment, a thermal disturbance coupling evaluation and anomaly localization module, a regional regulation-driven modeling and power correction module, and a response offset tracking and identification and dynamic correction module, it is possible to achieve the following: from multi-source physical quantity data acquisition and normalization processing, to thermal disturbance residual extraction and coupling evaluation value generation, to dynamic correction of heating power setpoints based on candidate micro-disturbance regions and continuous updating of point temperature matrix, and finally to complete the offset identification and adaptive correction of the regulation path structure for consistency between regulation execution and disturbance response. This forms a full-process temperature regulation capability for early response to local disturbances, high-resolution diagnosis and closed-loop dynamic intervention, significantly improving the thermal field stability and response accuracy during the steel ball heat treatment process.

[0051] The third aspect of this invention provides an intelligent temperature gradient control device for heat treatment of steel balls, characterized in that it comprises: a multi-source thermal field sensing processor, a disturbance identification and control driver, and a response offset diagnostic corrector, wherein: the multi-source thermal field sensing processor is used to integrate thermal field sensing devices and data preprocessing algorithms to collect and standardize multi-source thermal parameters during the heat treatment process of steel balls, providing data support for subsequent temperature distribution analysis and state identification; the disturbance identification and control driver is used to analyze local disturbance characteristics during the heat treatment process of steel balls based on standardized heat treatment data, identify potential abnormal areas, construct a heating control response mechanism, and dynamically adjust the power distribution; the response offset diagnostic corrector is used to evaluate the response offset behavior during the heating control process, determine the matching relationship between thermal disturbance response and power adjustment, and perform structural correction on the control path to achieve adaptive closed-loop regulation.

[0052] In this implementation scheme, by integrating a multi-source thermal field sensing processor, a disturbance identification and control driver, and a response offset diagnostic corrector, it is possible to achieve real-time acquisition and unified standardization of multi-source thermal parameters during the heat treatment process of steel balls. It supports the identification and control drive construction of local anomalies based on thermal disturbance coupling evaluation values, and dynamically adjusts the control path structure in combination with response offset evaluation results, forming a closed-loop temperature control mechanism based on actual thermal field response as feedback, thereby improving control accuracy, intervention timeliness and system adaptability, and ensuring the consistency of thermal field distribution and the uniformity of steel ball surface hardness during the heat treatment process.

[0053] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0054] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent control of temperature gradient during heat treatment of steel balls, characterized in that, Includes the following steps: S1. Collect multi-source data during the heat treatment process of steel balls, obtain heat treatment data of steel balls from the multi-source data, and preprocess the heat treatment data of steel balls. S2, Based on the preprocessed heat treatment data of the steel ball, the spatiotemporal derivative relationship between the thermal perturbation residual and the surface temperature is constructed to generate the thermal perturbation coupling evaluation value, and candidate perturbation regions are identified through threshold judgment and clustering operation; S3. Based on the candidate perturbation region, extract the number and construct a power perturbation joint dataset. Analyze the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity, and use it to adjust the heating power setpoint, drive the point temperature matrix update and recalculate the thermal perturbation coupling evaluation value. S4. Based on the updated point temperature matrix and thermal perturbation coupling evaluation value, the residual mutation region is re-extracted, compared with the candidate perturbation region of the previous frame to generate a target number set, the thermal perturbation response offset value of the target number is evaluated and the path branch control structure is constructed, and the calculation method of thermal perturbation control driving value is corrected. The specific steps for collecting multi-source data during the heat treatment process of steel balls, obtaining heat treatment data of steel balls from the multi-source data, and preprocessing the heat treatment data of steel balls are as follows: By deploying thermocouple temperature probes, infrared temperature measurement arrays, furnace door edge pressure sensors, furnace gas flow rate monitoring devices, multi-layer embedded thermocouple modules, and heating circuit current detection units, the local heating response, thermal field distribution, and heating energy consumption behavior of steel balls during the heat treatment process are dynamically collected throughout the entire process, and steel ball heat treatment data is obtained. The steel ball heat treatment data includes surface point temperature, surface temperature rise rate, gas flow rate, multi-layer temperature difference, heating power, and point temperature matrix. Spatial reconstruction of steel ball heat treatment data is performed using an interpolation repair algorithm based on thermal field topology consistency to fill in temperature matrix discontinuities caused by short-term infrared occlusion and sensor hysteresis. First-order time-domain smoothing of temperature rise rate data is performed using a local derivative-limited sliding filter algorithm to suppress temperature fluctuations caused by uneven steel ball arrangement and unstable thermal radiation. A multi-layer residual focusing abrupt change detection algorithm identifies and removes short-term temperature imbalance regions in the point temperature matrix, filtering out instantaneous thermal disturbance signals caused by furnace door opening, hot gas layer interference, and boundary overheating. Unit consistency transformation of the steel ball heat treatment data is performed using a component normalization weighted normalization algorithm to achieve data normalization. The specific steps for constructing the spatiotemporal derivative relationship between thermal perturbation residuals and surface temperature based on the preprocessed heat treatment data of the steel ball, and generating thermal perturbation coupling evaluation values, are as follows: Based on the preprocessed heat treatment data of the steel ball, the point temperature matrix at each time point is arranged in chronological order to construct a three-dimensional temperature data structure and form a spatial temperature evolution sequence. Inter-frame difference is performed on the point temperature matrix in the time dimension to extract the thermal perturbation residual, and the first and second derivatives of the surface temperature sequence are calculated. The time derivative of the thermal disturbance residual is multiplied by the second derivative of the surface temperature to obtain the surface temperature rise rate instability term; the spatial gradient of the thermal disturbance residual is multiplied by the multilayer temperature difference and gas flow rate to obtain the thermal disturbance spatial coupling enhancement term; the surface temperature rise rate instability term and the thermal disturbance spatial coupling enhancement term are added together to obtain the thermal disturbance coupling evaluation value.

2. The intelligent control method for temperature gradient of steel ball heat treatment according to claim 1, characterized in that: The specific steps for identifying candidate perturbation regions through threshold judgment and clustering operations are as follows: The thermal disturbance coupling evaluation value and the thermal disturbance threshold are compared in real time. When the thermal disturbance coupling evaluation value is greater than or equal to the thermal disturbance threshold, the corresponding position in the current frame is marked as an abnormal response point. When the thermal disturbance coupling evaluation value is less than the thermal disturbance threshold, the corresponding position in the current frame is marked as a normal heating point; Extract all abnormal response points to construct an abnormal point set. Perform eight-neighbor clustering on the abnormal point set according to spatial location within a sliding time window. Calculate the cluster area and average thermal perturbation coupling evaluation value for the clustered regions. Remove regions with an area smaller than the minimum perturbation area threshold and an average thermal perturbation coupling evaluation value lower than the perturbation intensity threshold. For the preserved region, extract the surface temperature sequence of the corresponding position in the point temperature matrix, calculate the surface temperature difference between the current frame and the previous frame, and mark the current region as a candidate perturbation region when the temperature difference of more than half of the positions in the region is negative.

3. The intelligent control method for temperature gradient of steel ball heat treatment according to claim 1, characterized in that: The specific steps for extracting numbers based on candidate perturbation regions, constructing a joint power perturbation dataset, and analyzing the coupling strength between perturbation trend response hysteresis and spatial thermal inhomogeneity are as follows: The spatial location of the candidate perturbation region is mapped to the heating power control region that has been divided in the heat treatment equipment, and the region number associated with each candidate region is obtained; the heat treatment data of the steel ball under the number is called, and the thermal perturbation coupling evaluation value and the average thermal perturbation coupling evaluation value of the corresponding region are read simultaneously to construct a power perturbation joint dataset; Based on the joint dataset of power disturbances, the coupling strength between disturbance trend response lag and spatial thermal inhomogeneity is analyzed: the disturbance trend enhancement term is obtained by subtracting the average thermal disturbance coupling evaluation value from the average thermal disturbance coupling evaluation value and then dividing by the average thermal disturbance coupling evaluation value and adding a minimum term; the absolute value of the time derivative of the heating power is taken and subtracted from 1 to obtain the power response lag term; the spatial disturbance intensity term is obtained by multiplying the multilayer temperature difference with the gas flow rate; the disturbance trend enhancement term, the power response lag term, and the spatial disturbance intensity term are multiplied in sequence to obtain the thermal disturbance regulation driving value.

4. The intelligent control method for temperature gradient of steel ball heat treatment according to claim 1, characterized in that: The specific steps for adjusting the heating power setpoint, driving the point temperature matrix update, and recalculating the thermal disturbance coupling evaluation value are as follows: Based on the value of the thermal disturbance control driving value, adjust the heating power setting value of the corresponding region number in the current time frame: use the thermal disturbance control driving value as an incremental input and add it to the current heating power output value; for the region number adjacent to the candidate micro-disturbance region, set an upper limit for the heating power change rate to limit the power adjustment range per unit time; In the time frame after the heating power is adjusted, the location data of the corresponding area number in the point temperature matrix is ​​called to recalculate the thermal disturbance residual; Update the thermal disturbance coupling assessment values ​​based on the latest thermal disturbance residuals, surface temperatures, and multilayer temperature differences. Based on the updated thermal perturbation coupling evaluation value, abnormal response points are re-determined, and candidate perturbation regions are identified.

5. The intelligent control method for heat treatment temperature gradient of steel balls according to claim 1, characterized in that: The specific steps for re-extracting residual abrupt change regions based on the updated point temperature matrix and thermal perturbation coupling evaluation values, and comparing them with the candidate perturbation regions of the previous frame to generate a target number set are as follows: In the current time frame after the heating power is adjusted, surface temperature data at each position in the point temperature matrix are extracted to construct the current thermal field distribution map; using the thermal image data of the inner wall of the equipment as a reference, a theoretical symmetrical thermal distribution grid for the current time frame is generated; the current thermal field distribution map and the theoretical symmetrical thermal distribution grid are mapped point by point in the spatial coordinates, and the thermal disturbance residuals at each position are recalculated to construct the residual distribution map for the current time frame; spatial difference is performed on the residual distribution map to extract the positions where the gradient value is greater than the mutation threshold, forming a set of residual mutation points; three or more consecutive coordinate points in the horizontal and vertical directions are retained in the set of residual mutation points to form a set of suspected disturbance points; Each spatial coordinate in the suspected disturbance point set is compared with the coordinates of the candidate disturbance region marked in the previous time frame. When the difference between the horizontal and vertical coordinates is not greater than 1, the region number is extracted and summarized to form a target number set.

6. The intelligent control method for temperature gradient of steel ball heat treatment according to claim 1, characterized in that: The specific steps for evaluating the thermal disturbance response offset value of the target number, constructing the path branch control structure, and correcting the calculation method of the thermal disturbance control driving value are as follows: For each region in the target number set, evaluate the offset between the thermal perturbation response change and the control drive: subtract the product of the thermal perturbation control drive value and the thermal perturbation control weight from the time derivative of the thermal perturbation coupling evaluation value, and take the absolute value of the difference to obtain the perturbation response difference term; subtract the thermal perturbation residual between the current frame and the previous frame to obtain the perturbation residual change term; divide the perturbation residual change term by the spatial gradient of the surface temperature and add a minimum term to obtain the perturbation normalization change term; Multiply the disturbance response difference term by the disturbance normalization change term to obtain the thermal disturbance response offset evaluation value; The thermal disturbance response offset evaluation value and the response offset threshold are compared in real time. When the thermal disturbance response offset evaluation value is less than or equal to the response offset threshold, the original calculation structure of the thermal disturbance control driving value is maintained. When the thermal disturbance response offset evaluation value is greater than the response offset threshold, the difference between the real-time disturbance coupling evaluation value and the average thermal disturbance coupling evaluation value is no longer used to calculate the disturbance trend enhancement term in the next time frame. Instead, the thermal disturbance residual sequence with that number in the sliding time window is called, the proportion of frames that are consistent with the direction of heating power change is counted, and this proportion is introduced into the calculation structure of the disturbance trend enhancement term as a substitute term. The modified perturbation trend enhancement term is used to update the thermal perturbation control driving value, forming the path branch control logic under the response offset condition.

7. A smart temperature gradient control system for heat treatment of steel balls, employing the smart temperature gradient control method for heat treatment of steel balls as described in any one of claims 1-6, characterized in that: include: The system includes a steel ball heat treatment data acquisition and preprocessing module, a thermal disturbance coupling assessment and anomaly localization module, a regional control-driven modeling and power correction module, and a response offset tracking, identification, and dynamic correction module, among which: The steel ball heat treatment data acquisition and preprocessing module is used to acquire multi-source data during the steel ball heat treatment process, obtain steel ball heat treatment data from the multi-source data, and preprocess the steel ball heat treatment data. The thermal disturbance coupling evaluation and anomaly localization module is used to construct the spatiotemporal derivative relationship between thermal disturbance residual and surface temperature based on the preprocessed steel ball heat treatment data, generate thermal disturbance coupling evaluation value, and identify candidate micro-disturbance regions through threshold judgment and clustering operation. The regional regulation-driven modeling and power correction module is used to extract numbers based on candidate perturbation regions, construct a joint power perturbation dataset, analyze the coupling strength between perturbation trend response lag and spatial thermal inhomogeneity, and adjust the heating power setpoint to drive the update of the point temperature matrix and recalculation of the thermal perturbation coupling evaluation value. The response offset tracking and identification and dynamic correction module is used to re-extract the residual mutation region based on the updated point temperature matrix and thermal perturbation coupling evaluation value, compare it with the candidate perturbation region of the previous frame to generate a target number set, evaluate the thermal perturbation response offset value of the target number and construct the path branch control structure, and correct the calculation method of the thermal perturbation control driving value.

Images