Intelligent management and control equipment and method for heat treatment process of large casting

Abstract

The invention relates to the technical field of equipment for heat treatment, in particular to intelligent management and control equipment and method for the heat treatment process of a large casting, data are collected through a thermocouple sensor and an atmosphere concentration sensor which are installed in an annealing furnace, and annealing data of historical qualified products and unqualified products are analyzed; and identifying a heat influence transfer relationship between different thermocouple regions, and setting a temperature change threshold value for each region by combining an accumulative effect of abnormal influence of different regions in different stages. According to the scheme, precise control over the local temperature is achieved, and the uniformity and process stability of the overall temperature field in the annealing process of the large casting can be improved by compensating thermal interference between the areas.

Description

Technical Field

[0001] This invention relates to the field of heat treatment equipment technology, specifically to an intelligent control device and method for the heat treatment process of large castings. Background Technology

[0002] Annealing, a key heat treatment process in the manufacture of large castings, involves heating the castings to above their recrystallization temperature and holding them at that temperature. This allows the internal microstructure of the material to recover and recrystallize, thereby eliminating residual stress and improving machinability. Due to the large size and complex structure of large castings, dead zones in airflow circulation and areas of heat radiation shielding easily form within the annealing furnace, leading to uneven temperature distribution and gradient fluctuations in the protective atmosphere concentration. This spatial heterogeneity of thermal parameters directly causes uneven heating of the castings and asynchronous carburizing / nitriding processes, ultimately resulting in quality problems such as workpiece deformation, hardness dispersion, and batch stability deterioration. Therefore, achieving precise and coordinated control of thermal parameters in multiple regions within the furnace is a core technical aspect for ensuring the quality of annealing large castings.

[0003] Existing problems: Current technologies mainly employ a holistic feedback control strategy based on the average furnace temperature, achieving macroscopic temperature control by adjusting global heating power and circulating air volume. This control method neglects the thermal conduction coupling and thermal radiation interference effects between different regions within a large annealing furnace. This allows temperature anomalies in a localized area to trigger cascading temperature fluctuations in related areas via nonlinear heat transfer paths, resulting in an overall temperature field imbalance. When the average furnace temperature or the temperature at individual points exceeds a threshold, the controlled object is already in an abnormal state of localized overheating or underheating. The control system struggles to capture early, weak anomaly signals in time, leading to irreversible quality loss. In other words, the traditional holistic control model lacks understanding and quantitative analysis of the dynamic thermal coupling relationships between regions, failing to achieve proactive identification and rapid response to localized temperature anomalies. This severely restricts the process stability and product quality consistency of the annealing process for large castings. Summary of the Invention

[0004] This invention provides an intelligent control device and method for the heat treatment process of large castings to solve existing problems.

[0005] The intelligent control equipment and method for heat treatment of large castings of the present invention adopts the following technical solution: In a first aspect, one embodiment of the present invention provides an intelligent control device for the heat treatment process of large castings, including an annealing furnace body. The device further includes an intelligent control system, which includes a multi-area sensor array and a controller signal-connected to the multi-area sensor array, wherein: The multi-region sensor array is deployed along the furnace spatial coordinate system to collect temperature time-series data and atmosphere concentration data at multiple measuring points inside the furnace during the heat treatment process. The controller is used to acquire temperature time-series data from multiple measuring points within the heat treatment furnace and corresponding historical samples of product quality grades; divide the temperature time-series data into multiple process stages and extract the stage-specific temperature anomaly characteristics of each measuring point; based on the historical samples, identify temperature anomaly pattern clusters using a clustering algorithm, and calculate the quality sensitivity quantification index for each measuring point according to the proportion of defective products in each cluster; wherein, the quality sensitivity quantification index is used to quantify the contribution intensity of the temperature anomaly at the measuring point to product quality defects; based on the historical samples, analyze the lag and correlation of temperature changes between different measuring points using a time-series matching algorithm, identify the causal chain of thermal coupling between regions, and calculate the regional thermal coupling quantification index for each measuring point; wherein, the regional thermal coupling quantification index... This is used to quantify the correlation and conduction strength of temperature changes at a measuring point due to thermal disturbances at other measuring points. Based on the historical samples, the correlation between abnormal characteristics of previous process stages and product quality in subsequent stages is analyzed to obtain a cross-stage thermal state memory effect correction coefficient. Combining the quality sensitivity quantification index, the regional thermal coupling quantification index, and the cross-stage thermal state memory effect correction coefficient, a differentiated temperature change threshold for each measuring point is dynamically generated. The differentiated temperature change threshold is negatively correlated with the quality sensitivity quantification index and positively correlated with the regional thermal coupling quantification index. The temperature data of each measuring point is monitored in real time. When the temperature change parameter of any measuring point exceeds the differentiated temperature change threshold, a corresponding regional coordinated control strategy is executed according to the current process stage.

[0006] Secondly, one embodiment of the present invention provides an intelligent control method for the heat treatment process of large castings, the method comprising: Acquire temperature time-series data from multiple measuring points inside the heat treatment furnace and corresponding historical samples of product quality grades; The temperature time series data is divided into multiple process stages and the stage-specific temperature anomaly features of each measuring point are extracted. Based on the historical samples, temperature anomaly pattern clusters are identified using a clustering algorithm, and a quality sensitivity quantification index for each measuring point is calculated according to the proportion of defective products in each cluster; wherein, the quality sensitivity quantification index is used to quantify the contribution intensity of temperature anomaly at the measuring point to product quality defects. Based on the historical samples, the lag and correlation of temperature changes between different measuring points are analyzed by time-series matching algorithm to identify the causal chain of thermal coupling between regions and to calculate the regional thermal coupling quantification index for each measuring point; wherein, the regional thermal coupling quantification index is used to quantify the correlation conduction strength of temperature changes at measuring points caused by thermal disturbances at other measuring points. Based on the historical samples, the correlation between abnormal characteristics of the preceding process stage and product quality in the subsequent stage is analyzed to obtain the cross-stage thermal state memory effect correction coefficient. By combining the quality-sensitive quantitative index, the regional thermal coupling quantitative index, and the cross-stage thermal state memory effect correction coefficient, a differentiated temperature change threshold for each measuring point is dynamically generated; wherein, the differentiated temperature change threshold is negatively correlated with the quality-sensitive quantitative index and positively correlated with the regional thermal coupling quantitative index. Real-time monitoring of temperature data at each measuring point; when the temperature change parameter at any measuring point exceeds the differentiated temperature change threshold, the corresponding regional coordinated control strategy is executed according to the current process stage.

[0007] Furthermore, the extraction of staged temperature anomaly features at each measuring point includes: According to the characteristics of the annealing process, the temperature timing data is divided into a heating stage, a holding stage, and a cooling stage. The actual temperature curves of each measuring point at each stage and the corresponding preset standard process curves for each stage are obtained respectively. The temperature anomaly coefficient for each stage is quantified by calculating the morphological similarity index between the actual temperature curve and the standard process curve in terms of temperature change trend; wherein, the lower the morphological similarity index, the larger the temperature anomaly coefficient. Based on the temperature anomaly coefficients at each stage, the stage-specific temperature anomaly characteristics of each measuring point are determined.

[0008] Furthermore, based on the historical samples, the process of identifying temperature anomaly pattern clusters using a clustering algorithm and calculating quality sensitivity metrics based on the proportion of non-conforming products in each cluster includes: Obtain the phased temperature anomaly characteristics of each measuring point and their corresponding product quality grades; Cluster the data with similar temperature anomaly characteristics in the historical samples to generate multiple temperature anomaly pattern clusters; Calculate the percentage of non-conforming products in each temperature anomaly pattern cluster, and identify the clusters whose percentage exceeds the preset risk threshold as critical risk clusters; Calculate the average value of temperature anomaly characteristics of each measuring point within each critical risk cluster, and then weight and accumulate the values ​​according to the proportion of non-conforming products in each critical risk cluster to obtain the contribution intensity of temperature anomaly at that measuring point to product quality defects. Use the contribution intensity as a quantitative indicator of quality sensitivity.

[0009] Furthermore, the step of analyzing the hysteresis and correlation of temperature changes between different measuring points using a time-series matching algorithm, identifying causal chains of thermal coupling between regions, and calculating quantifiable indicators of regional thermal coupling includes: The temperature time series data of each measuring point are paired up, and the degree of synchronization of temperature change trends between each pair of measuring points is analyzed. The measuring point pairs with consistent temperature change trends and a degree of synchronization exceeding the preset correlation threshold are selected. For the selected measurement point pairs, a correspondence of temperature change characteristics is established through dynamic pattern matching method to determine the temporal order of temperature changes between the two measurement points. The measurement point with the earlier temperature change is regarded as the source of influence, and the measurement point with the later temperature change is regarded as the object of influence, thus constructing a causal chain of thermal coupling between regions. Based on the aforementioned thermally coupled causal chain, the unidirectional influence intensity of each affected object's measuring point on the temperature disturbance of the corresponding source measuring point is quantified; The distribution characteristics of the intensity of all unidirectional influences on each measuring point when it is affected are statistically analyzed. The overall influence of temperature disturbances from other measuring points on the measuring point is calculated, and the overall influence is used as a quantitative index of regional thermal coupling.

[0010] Furthermore, the quantification of the unidirectional influence intensity of the temperature disturbance of the influence source on each affected object based on the thermally coupled causal chain includes: Acquire temperature time-series data of the source measurement point and the affected object measurement point, identify the extreme points of fluctuation in the temperature change process, and divide the temperature change process into multiple continuous temperature fluctuation intervals based on the extreme points of fluctuation. For each temperature fluctuation range, calculate the temperature change rate of the influencing source and the temperature change response rate of the affected object. Based on the degree of difference between the temperature change rate and the temperature change response rate, the local influence intensity value of the influence source on the affected object within the current temperature fluctuation range is determined. The distribution pattern of local influence intensity values ​​corresponding to all temperature fluctuation intervals is statistically analyzed, and the unidirectional influence intensity is determined based on the distribution pattern.

[0011] Furthermore, based on the historical samples, the method of obtaining a cross-stage thermal state memory effect correction coefficient by analyzing the correlation between abnormal characteristics of previous process stages and product quality in subsequent stages includes: The sample data that exceed the preset abnormal threshold in the temperature anomaly characteristics of the preceding process stage are selected from the historical samples, and the temperature performance data of the sample data in the subsequent process stage and the corresponding product quality level are obtained. The sample data is classified into multiple thermal state abnormality patterns, and the first proportion of unqualified products in each thermal state abnormality pattern is calculated. For each thermal anomaly mode, obtain its temperature performance data in subsequent process stages, and count the second proportion of temperature anomaly features exceeding the preset anomaly threshold. Based on the product of the first proportion and the second proportion, calculate the cross-stage influence factor of the preceding process stage on the subsequent process stage. Iterative analysis was performed on multiple combinations of preceding and subsequent process stages, and cross-stage influence factors of each combination were integrated to generate a cross-stage thermal state memory effect correction coefficient.

[0012] Furthermore, the generation of the cross-stage thermal state memory effect correction coefficient by comprehensively considering the cross-stage influence factors of each of the preceding-subsequent process stages includes: Normalized factors are obtained by normalizing the cross-stage influence factors of each combination of preceding and subsequent process stages. The fusion influence factor is obtained by weighted fusion processing based on the normalization factor, and the fusion influence factor is used as the correction coefficient for cross-stage thermal state memory effect.

[0013] Furthermore, the step of dynamically generating differentiated temperature change thresholds for each measuring point by combining the quality-sensitive quantification index, the regional thermal coupling quantification index, and the cross-stage thermal state memory effect correction coefficient includes: A basic threshold coefficient is generated by nonlinearly combining the quality-sensitive quantification index and the regional thermal coupling quantification index; the higher the quality-sensitive quantification index, the smaller the basic threshold coefficient, and the higher the regional thermal coupling quantification index, the larger the basic threshold coefficient. The basic threshold coefficient is coupled and corrected with the cross-stage thermal state memory effect correction coefficient to obtain the final threshold coefficient; wherein, the larger the cross-stage thermal state memory effect correction coefficient is, the smaller the final threshold coefficient is. The final threshold coefficient is applied to a preset benchmark threshold to obtain the differentiated temperature change threshold for each measuring point.

[0014] Furthermore, the real-time monitoring of temperature data at each of the aforementioned measuring points, when the temperature change parameter at any of the measuring points exceeds the differentiated temperature change threshold, executes a corresponding regional coordinated control strategy based on the current process stage, including: The temperature change parameters of each measuring point are acquired in real time and compared with the differential temperature change threshold. When it is determined that the temperature change parameter at any measuring point exceeds the differentiated temperature change threshold, the current process stage is identified. If it is in the heating stage, perform at least one of the following operations on the area where the measurement point exceeds the threshold: heating power adjustment and circulating airflow distribution optimization. If it is in the heat preservation stage, perform at least one of the following operations on the area where the measurement point exceeds the threshold: micro-power compensation and atmosphere stability maintenance. If the temperature is dropping, at least one of the following operations should be performed on the area where the measurement point exceeds the threshold: cooling medium flow rate regulation and heat compensation.

[0015] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, data is collected using thermocouple sensors and atmosphere concentration sensors installed in the annealing furnace. By analyzing the annealing data of historical qualified and unqualified products, the heat transfer relationship between different thermocouple regions is identified. Combined with the cumulative effect of the abnormal influence at different stages in different regions, a temperature change threshold is set for each region.

[0016] This invention achieves intelligent control of the annealing process—from local temperature monitoring to global quality prediction, and from single-point feedback to multi-zone linkage—by constructing a three-dimensional regional collaborative intelligent control system driven by quality sensitivity, thermal coupling, and cross-stage memory effect. Furthermore, through dynamic time warping matching and extreme point segmentation quantization technology, it accurately identifies the causal links and hysteresis characteristics of heat transfer between regions, establishing a differentiated threshold generation mechanism based on coupling strength. This mechanism can trigger control actions in the early stages of temperature anomalies, improving temperature field uniformity and process stability. Moreover, by introducing a cross-stage thermal state memory effect correction coefficient to couple and correct the threshold, it effectively eliminates the path dependence between the heating, holding, and cooling stages, solving the cumulative effect of quality fluctuations caused by isolated setting of traditional process parameters and ensuring the quality consistency of large castings throughout the annealing process. Finally, based on a multi-physics distributed sensor array and a staged adaptive control strategy, it can perform precise control operations such as micro-compensation of heating power, redistribution of circulating airflow, and optimization of cooling medium flow rate according to the real-time process status. This reduces energy consumption, shortens the process cycle, and improves batch production stability and equipment operating efficiency while ensuring product quality. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the intelligent control equipment for the heat treatment process of large castings provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the intelligent control method for the heat treatment process of large castings provided in an embodiment of the present invention.

[0019] The numbers in the diagram are: 100, Intelligent control equipment for heat treatment of large castings; 110, Intelligent control system; 120, Furnace cover device; 130, Annealing furnace body; 140, Circulating fan; 150, Cold air mixing valve; 160, Heat exhaust valve. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent control device and method for the heat treatment process of large castings proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] The following description, in conjunction with the accompanying drawings, details the specific solution of the intelligent control equipment and method for the heat treatment process of large castings provided by this invention.

[0023] Please see Figure 1 This illustration shows an intelligent control device 100 for the heat treatment process of large castings provided by an embodiment of the present invention, comprising: The intelligent control system 110 serves as the central control unit of the aforementioned intelligent control equipment 100 for the heat treatment process of large castings. It includes an industrial computer, a controller (which may be a programmable logic controller), a PID controller, and a power regulator. The industrial computer manages the human-machine interface, enabling process parameter settings and visualized monitoring of equipment operation status. The controller executes discrete logic control functions such as valve on / off and fan start / stop. The PID controller achieves precise furnace temperature control and thermal parameter stability through a closed-loop feedback mechanism. The power regulator dynamically adjusts the output power of the furnace wall heating elements according to control commands to cooperate with the temperature control system in completing precise heating during the heating, holding, and cooling processes. The intelligent control system 110, through the synergistic effect of a multi-level control architecture, enables intelligent decision-making and execution throughout the entire annealing process of large castings.

[0024] The furnace cover device 120 is located on the top of the annealing furnace body and adopts a lifting sealing structure or a hinged opening and closing structure. Its core function is to maintain the airtightness of the protective atmosphere inside the furnace and prevent the leakage of protective gases such as nitrogen and hydrogen, which would cause oxidation or decarburization on the surface of the castings. The device can be equipped with an integrated circulating fan assembly to enhance the flow of gas inside the furnace through mechanical forced convection. At the same time, it serves as a bearing platform for loading and unloading workpieces, enabling the vertical suspension of large castings into and out of the furnace, ensuring that the workpieces maintain geometric stability during the heat treatment process.

[0025] The annealing furnace body 130 can be a horizontal or vertical cylindrical sealed container, with an internal lining of a multi-layer heat insulation structure composed of aluminosilicate fiber or refractory bricks, which effectively reduces heat dissipation to the environment and maintains the furnace thermal efficiency. Electric heating elements such as resistance strips or silicon carbide rods are evenly arranged on the inner side of the furnace wall as heat sources. Under the control of the power regulator of the intelligent control system 110, electrical energy is converted into radiant heat and convective heat, providing the required temperature field environment for the annealing process.

[0026] The circulating fan 140 is deployed on the furnace cover or side wall of the furnace chamber. It adopts a variable frequency speed-regulating centrifugal fan or axial flow fan. By forcibly driving the gas in the furnace to form a directional circulating flow field, the hot air is evenly distributed in space during the heating stage or the cold air is evenly distributed during the cooling stage. This eliminates the dead airflow caused by natural convection, controls the temperature uniformity of the effective working area in the furnace within the target range, and improves the homogeneity of the microstructure properties of different parts of large castings.

[0027] The cold air valve 150 is installed in the cold air inlet pipe at the bottom or side of the annealing furnace body 130. It is an electric regulating valve or a pneumatic butterfly valve. According to the flow control signal issued by the intelligent control system 110 during the cooling stage, it precisely adjusts the amount of ambient cold air introduced. It works in conjunction with the hot air pipeline to achieve precise control of the temperature gradient and cooling rate in the furnace, preventing secondary thermal stress from being generated in the casting during rapid cooling.

[0028] The exhaust air valve 160 is located at the top or upper part of the hot air outlet pipe of the annealing furnace body 130. It is a high-temperature resistant regulating valve with multiple functions such as furnace pressure balancing, exhaust gas discharge and accelerated cooling. During the heating and heat preservation stages, the opening degree is adjusted to maintain a slight positive pressure in the furnace to prevent external air from seeping in. During the cooling stage, it works with the cold air mixing valve 150 to increase the opening degree to achieve forced heat exhaust, shortening the process cycle while ensuring the purity of the atmosphere.

[0029] In addition, the hot air / cold air pipelines form the fluid channels for heat transfer inside the furnace. They consist of heat-resistant stainless steel pipes, insulation layers, and flow distribution devices. During the heating stage, hot air pressurized by the circulating fan 140 is delivered to the bottom of the furnace to achieve uniform temperature. During the cooling stage, cold air introduced by the cold air mixing valve 150 is directionally delivered to the high-temperature area. The pipeline layout follows the principle of fluid dynamics optimization to ensure that the flow resistance of the medium is minimized and the temperature field adjustment response is rapid.

[0030] When the intelligent control equipment 100 for the heat treatment process of large castings is used, the internal chamber of the annealing furnace body 130 serves as the heat treatment reaction space, capable of accommodating various large castings. The equipment is designed for workpieces primarily including long shaft parts, heavy structural components, and other alloy steel castings with large geometric dimensions and complex cross-sectional shapes. To prevent bending deformation of the workpieces due to the coupling effect of their own weight and thermal stress during heating, holding, and cooling, the workpieces can be vertically suspended at the central axis of the furnace using a special hanger, ensuring a certain non-contact safety distance between each part of the workpiece and the heating elements on the furnace wall. The vertical suspension method, combined with the sealed structure of the furnace cover device 120, effectively suppresses creep deformation of the workpieces during prolonged high-temperature annealing. Simultaneously, the protective atmosphere (such as a nitrogen-hydrogen mixture) driven by the circulating fan 140 forces circulation, forming a stable dynamic gas protective layer on the workpiece surface. This reduces the kinetic rate of oxidation and decarburization reactions on the workpiece surface, achieving the goal of precision annealing heat treatment with minimal oxidation and no decarburization.

[0031] The aforementioned controller, based on a heterogeneous computing architecture of embedded industrial computers and programmable logic controllers, can achieve fully automated execution of multi-dimensional thermal parameter analysis and decision-making control tasks by running the intelligent control method for the heat treatment process of large castings provided in this embodiment of the invention. The multi-physics raw data required by the aforementioned intelligent control method for the heat treatment process of large castings is acquired in real time through a multi-region sensor array deployed inside the annealing furnace body 130. This sensor array establishes a bidirectional communication link with the controller of the intelligent control system 110 via an industrial fieldbus to achieve sensing. decision making The system employs a closed-loop architecture. The temperature sensing layer utilizes a six-point distributed thermocouple sensor network, arranged axially in three layers (upper, middle, and lower) along the furnace wall. Each layer has two measuring points evenly distributed at a 120-degree phase angle in the circumferential direction, forming a three-dimensional temperature monitoring grid covering the entire furnace. The sensor probe end face maintains a non-contact safety distance of 10 to 15 centimeters from the casting surface, ensuring thermal response sensitivity while avoiding mechanical interference. Its goal is to control the temperature uniformity of the effective working area within the furnace within ±5℃ and accurately capture the dynamic temperature difference between dead zones in airflow circulation and high-temperature areas shielded by thermal radiation. The atmosphere sensing layer consists of an oxygen probe installed in the lower middle section and a hydrogen probe deployed at the upper airflow circulation outlet. The oxygen probe is inserted to a depth of one-third of the furnace diameter to avoid areas of stagnant gas flow, monitoring the oxygen concentration in the furnace in real time and keeping it below 50 ppm. The hydrogen probe simultaneously detects the uniformity of hydrogen concentration distribution under hydrogen-based protective atmosphere conditions. The two work together to prevent oxidation and decarburization reactions on the casting surface caused by localized atmosphere depletion. The multi-region sensor array provides the controller with a high-quality spatiotemporal resolution thermal parameter data stream through a spatiotemporal synchronous sampling mechanism, which constitutes the basic input for the execution of intelligent control algorithms.

[0032] For the intelligent control method of the large casting heat treatment process executed by the controller in the intelligent control system 110, please refer to [link to relevant documentation]. Figure 2 The diagram illustrates a flowchart of an intelligent control method for the heat treatment process of large castings provided by an embodiment of the present invention. The method may include: Step S210: Obtain temperature time-series data from multiple measuring points inside the heat treatment furnace and corresponding historical samples of product quality grades.

[0033] The aforementioned temperature time-series data from multiple measuring points within the heat treatment furnace refers to a sequence of discrete temperature observations continuously collected throughout the entire annealing process by a multi-regional sensor array deployed along the furnace's spatial coordinate system. This data stream is indexed by time, recording the temperature values ​​sampled sequentially by each thermocouple sensor during the three stages of heating, holding, and cooling. This can be achieved through a six-point distributed thermocouple sensor network. This network is arranged in three layers (upper, middle, and lower) along the axial direction of the furnace wall, with two measuring points evenly distributed at a 120-degree phase angle in each layer, forming a three-dimensional monitoring grid covering the entire furnace area. The sensor probe end face maintains a non-contact safety distance of 10 to 15 centimeters from the casting surface. The analog temperature signal is converted into digital time-series data via an industrial fieldbus at a preset sampling period and transmitted to the controller. The data characteristics are manifested as a structured temperature curve with clear timestamps, spatial coordinate identifiers, and stage markers.

[0034] The aforementioned historical samples of product quality grades refer to labeled datasets obtained from batches of large castings that have completed the annealing process, through offline non-destructive testing and material performance testing. This dataset includes multi-point temperature time-series data for each batch of castings, as well as product quality grade labels determined by quality inspection experts based on criteria such as hardness uniformity, grain size, and residual stress level. The quality grades are binarized into qualified and unqualified products. These samples can be extracted from the historical database of a Manufacturing Execution System (MES) or manually entered into a laboratory quality traceability system. After data cleaning, a one-to-one mapping relationship is established with the temperature time-series data. This data serves as the benchmark ground value for training clustering models, calculating quality sensitivity metrics, and verifying the effectiveness of regional thermal coupling metrics, supporting the controller's supervised learning process.

[0035] Step S220: Divide the temperature time series data into multiple process stages and extract the stage-specific temperature anomaly characteristics of each measuring point.

[0036] It is understandable that the annealing process for large castings exhibits strong staged characteristics in physical metallurgy: the heating stage involves rapid heating of the material from room temperature to recrystallization temperature, where thermal stress dominates and the temperature gradient is large; the control objective is to constrain the heating rate to prevent the initiation of thermal shock cracks. The holding stage is at a constant high-temperature plateau, where atomic diffusion and dislocation climb drive recovery recrystallization; thermal stress relaxation and microstructure homogenization are the core objectives, and the control objective shifts to maintaining temperature fluctuations within a certain range to ensure sufficient microstructure transformation. The cooling stage involves controlled slow cooling to avoid secondary phase transformation and residual stress regeneration; the cooling rate is strictly limited. The failure modes and causes of quality defects in these three stages are drastically different. Excessive heating leads to grain boundary cracking, uneven holding causes hardness dispersion, and improper cooling introduces phase transformation stress. Using a uniform threshold model cannot capture the specific abnormal modes of each stage. Furthermore, the annealing process exhibits a certain path dependence; the temperature field distribution and microstructure formed in the preceding stages serve as initial conditions that profoundly alter the thermal conductivity and phase transformation kinetics of subsequent stages, constituting a thermal state memory effect. Therefore, after segmenting and analyzing the temperature time series data according to process characteristics, the temperature anomaly characteristics of each stage can be accurately extracted, the abnormal pattern clusters with physical significance can be identified, and a structured data basis can be provided for the subsequent stage-specific calculation of quality sensitivity and thermal coupling. Otherwise, cross-stage data mixing will cause the abnormal signals to be submerged in the aliasing effect of different dynamic behaviors, making it impossible to effectively identify and intervene in early weak quality risks.

[0037] Optionally, step S220 may include: dividing the temperature time series data into a heating stage, a holding stage, and a cooling stage according to the characteristics of the annealing process; obtaining the actual temperature curve of each measuring point in each stage and the corresponding preset standard process curve; quantifying the temperature anomaly coefficient of each stage by calculating the morphological similarity index between the actual temperature curve and the standard process curve in terms of temperature change trend; wherein, the lower the morphological similarity index, the larger the temperature anomaly coefficient; and determining the stage-specific temperature anomaly characteristics of each measuring point based on the temperature anomaly coefficient of each stage.

[0038] Based on the characteristics of the annealing process, the temperature timing data can be divided into three process windows with independent thermodynamic behaviors and metallurgical objectives: the heating stage, the holding stage, and the cooling stage. The heating stage begins with room temperature heating. At this stage, the intelligent control system sends coordinated start commands to the circulating fan and power regulator. Under the constraint of the PID controller, the heating element outputs a preset low power gradient, raising the casting temperature above the target recrystallization temperature through combined radiation and convection heat transfer. The core control point of this stage is to limit the heating rate within the material's thermal stress tolerance threshold to prevent the initiation of thermal shock cracks. The holding stage is triggered when the furnace temperature stabilizes and reaches the set plateau temperature (usually determined to be between 600℃ and 900℃ based on the material phase diagram). The PLC timer accurately calculates the process duration based on the casting wall thickness, with a standard of 4 hours of holding time per 100 mm. In this isothermal plateau, the atomic diffusion activation energy is fully satisfied, and dislocation climb and grain boundary migration drive microstructure recovery and recrystallization. Macroscopically, this manifests as an increase in the metal creep rate and the elimination of residual stress gradient relaxation. Temperature fluctuations must be strictly suppressed within ±5℃ to ensure uniform microstructure transformation. The cooling phase begins at the end of the holding time. The control system cuts off the power to the heating elements while maintaining forced convection operation of the circulating fan. Slow cooling with the furnace is achieved through the coordinated adjustment of the cold air mixing valve and the hot air exhaust valve. The cooling rate is constrained by the process specifications to not exceed 50℃ / hour, aiming to avoid excessively rapid cooling that could induce secondary martensitic phase transformation and the formation of new residual stress fields. The fundamental differences in the physical mechanisms of the three phases determine that the statistical characteristics and anomaly patterns of the temperature data are non-homogeneous. Mixed analysis would lead to aliasing of thermodynamic behaviors; therefore, segmented extraction is a necessary data preprocessing method for achieving accurate anomaly identification.

[0039] The actual temperature curves at each measuring point in each stage can be acquired in real time during the annealing process using a multi-region sensor array. This array continuously records the temperature digital signals output by each thermocouple sensor at a preset sampling period. After filtering and calibration, it forms temperature time-series data with timestamps and measuring point identifiers. This data is then segmented and reconstructed into actual temperature curves for the heating, holding, and cooling stages based on the time points of the process stages. The preset standard process curves for each stage can be predefined by the heat treatment process specifications. Their data originates from material phase diagrams, creep kinetic models, and engineering experience databases, representing the temperature-time target trajectory that each measuring point should follow under ideal conditions. This curve can serve as a benchmark for anomaly quantification. Historical samples are obtained by extracting data records from completed batches through the production execution system or quality traceability platform. Each annealing data set includes the full temperature time series of each measuring point, sensor spatial coordinates, atmosphere concentration monitoring values, and product quality grade labels marked after expert system or physicochemical testing. These labels are binarized into qualified and unqualified products, providing the truth value basis required for supervised learning in subsequent cluster analysis and influence calculation.

[0040] The above scheme can measure the cosine similarity between the actual temperature curve constructed at each measuring point within a specific stage and the corresponding preset standard process curve. This similarity index characterizes the degree of convergence of the two curves in the direction of change in the temperature-time coordinate system, with a value ranging between 0 and 1. Then, a temperature anomaly coefficient is generated by subtracting this cosine similarity from 1. The lower the morphological similarity index, the larger the temperature anomaly coefficient. That is, when the consistency between the actual temperature curve and the standard process curve decreases, the anomaly coefficient monotonically increases to amplify the degree of deviation. This calculation process traverses all measuring points and the three stages of heating, holding, and cooling, ultimately generating a set of temperature anomaly coefficients for each measuring point at each stage, which serves as the basic input for subsequent pattern clustering and quality sensitivity analysis.

[0041] Step S230: Based on historical samples, identify temperature anomaly pattern clusters using a clustering algorithm, and calculate the quality sensitivity quantification index for each measuring point according to the proportion of defective products in each cluster; wherein, the quality sensitivity quantification index is used to quantify the contribution intensity of temperature anomaly at the measuring point to product quality defects.

[0042] Optionally, step S230 may include: acquiring the phased temperature anomaly characteristics of each measuring point and their corresponding product quality levels; clustering data with similar temperature anomaly characteristics in historical samples to generate multiple temperature anomaly pattern clusters; calculating the proportion of non-conforming products in each temperature anomaly pattern cluster, and identifying clusters with proportions exceeding a preset risk threshold as key risk clusters; calculating the average temperature anomaly characteristics of each measuring point within each key risk cluster, and weighting and accumulating them according to the proportion of non-conforming products in each key risk cluster to obtain the contribution intensity of the temperature anomaly at that measuring point to product quality defects, and using the contribution intensity as a quality sensitivity quantification indicator.

[0043] The staged temperature anomaly characteristics of each measuring point are obtained by calling the calculation results of step S220. These results are based on temperature time-series data collected by a multi-region sensor array. After dividing the annealing process into heating, holding, and cooling stages according to its characteristics, the morphological similarity index between the actual temperature curve and the preset standard process curve is calculated for each measuring point within each stage. This quantifies and generates the temperature anomaly coefficient for each stage, forming structured anomaly feature data with measuring point identification, stage markers, and timestamps. The corresponding historical samples of product quality grades are extracted from the database of the production execution system or quality traceability platform, containing quality inspection records of completed annealing batches. These records include labels indicating whether each batch of castings is qualified or unqualified after physical and chemical inspection and expert system comprehensive judgment. A one-to-one mapping relationship is established between these records and the staged temperature anomaly feature data of each measuring point in the corresponding batch, using batch identification codes. This constructs the feature-label pairing dataset required for supervised learning, providing a sample basis with quality labels for subsequent cluster analysis.

[0044] The clustering process can employ density-based spatial clustering algorithms, such as DBSCAN (Density-Based Spatial Clustering of Applications with Noise). This algorithm performs unsupervised pattern mining on all anomaly coefficient sequences in historical samples, aggregating data with similar temperature anomaly characteristics into multiple temperature anomaly pattern clusters. By scanning the sample space, this algorithm automatically identifies high-density regions composed of core points and reachable points and assigns them to the same cluster. It effectively captures the distribution patterns and local aggregation characteristics of anomaly coefficients without requiring a pre-defined number of clusters, thereby generating a set of clusters representing different anomaly pattern types. Subsequently, for each temperature anomaly pattern cluster, the product quality grades corresponding to its member samples are statistically analyzed, and the proportion of defective products is calculated. A higher proportion indicates a stronger correlation between the anomaly pattern represented by the cluster and quality defects; that is, temperature changes at measurement points with higher anomaly levels within the cluster have a stronger driving effect on product quality deterioration. To identify high-risk anomaly patterns, a preset risk threshold is set to 0.3. Only clusters with a non-conforming product ratio exceeding this threshold are retained as key risk clusters for subsequent weighted calculation of quality sensitivity metrics. This process eliminates low-confidence noise clusters and focuses on typical anomaly patterns that substantially contribute to quality defects. It is understood that the DBSCAN algorithm described above is a relatively mature algorithm in the existing technology. For its specific implementation and working principle, please refer to relevant technologies; this embodiment will not elaborate further.

[0045] The above scheme calculates the arithmetic mean of the temperature anomaly coefficients corresponding to all historical samples within each critical risk cluster for each measuring point, obtaining the cluster-wide average anomaly characteristic that characterizes the typical deviation of the measuring point under a specific anomaly pattern. Then, the cluster-wide average anomaly characteristic of each measuring point is weighted and multiplied by the proportion of non-conforming products in that critical risk cluster. The proportion of non-conforming products serves as a risk weight coefficient, reflecting the driving force of the anomaly pattern on product quality deterioration. The weighted product results corresponding to all critical risk clusters are summed, and the resulting sum represents the contribution intensity of the temperature anomaly at that measuring point to product quality defects. This contribution intensity is directly used as a quality sensitivity quantification index to quantify the sensitivity of temperature anomaly fluctuations at that measuring point to the final product quality defects.

[0046] Step S240: Based on historical samples, analyze the lag and correlation of temperature changes between different measuring points using a time-series matching algorithm, identify the causal chain of thermal coupling between regions, and calculate the regional thermal coupling quantification index for each measuring point; wherein, the regional thermal coupling quantification index is used to quantify the correlation conduction strength of temperature changes at measuring points caused by thermal disturbances at other measuring points.

[0047] The aforementioned time-series matching algorithm is a dynamic analysis method for multivariate time series data. Its core lies in quantitatively identifying the conduction path and lag characteristics of thermal disturbances within the furnace space by measuring the degree of synchronicity and temporal order between temperature time-series data from different measuring points. The algorithm first pairs all measuring points together, using sliding window cross-correlation analysis or dynamic time warping techniques to calculate the morphological similarity and time delay parameters of the paired sequences. It then filters out measuring point pairs with consistent temperature change trends and correlation strength exceeding a preset threshold. Furthermore, by establishing corresponding matching relationships for temperature change characteristics, it accurately determines the lead-lag temporal relationship of temperature fluctuations between two measuring points, thereby identifying measuring points with leading temperature changes as influencing sources and those with lagging temperature changes as affected objects. Finally, it outputs a thermally coupled topology structure containing directional and intensity weights. It is understood that the time-series matching algorithm is a relatively mature technology in the existing field. For its implementation and working principle, please refer to relevant technologies; the embodiments of this invention will not elaborate further.

[0048] The aforementioned inter-regional thermal coupling causal chain is a furnace spatial thermodynamic correlation network constructed based on the identification results of the time-series matching algorithm. Its essence is a directed acyclic graph structure, where nodes represent the spatial regions corresponding to each temperature measuring point, and directed edges characterize the unidirectional transmission relationship of thermal disturbances from the source region to the affected region. This causal chain quantifies the distribution characteristics of the intensity of all unidirectional influences on each affected measuring point, comprehensively calculates the overall degree of influence from temperature disturbances at other measuring points, thereby capturing the nonlinear thermal coupling effect driven by heat conduction, heat radiation, and airflow circulation within the furnace, providing a quantitative basis for the subsequent generation of differentiated thresholds.

[0049] Optionally, step S240 may include: pairing the temperature time series data of each measuring point, analyzing the synchronicity of temperature change trends between each pair of measuring points, and selecting measuring point pairs whose temperature change trends are consistent and whose synchronicity exceeds a preset association threshold; for the selected measuring point pairs, establishing a correspondence of temperature change characteristics through a dynamic pattern matching method, determining the temporal order of temperature changes between the two measuring points, taking the measuring point with the earlier temperature change as the source of influence, and the measuring point with the later temperature change as the affected object, and constructing a regional thermal coupling causal chain; based on the thermal coupling causal chain, quantifying the unidirectional influence intensity of each affected object measuring point on the temperature disturbance of the corresponding source measuring point; statistically analyzing the distribution characteristics of all unidirectional influence intensities experienced by each measuring point when it is an affected object, comprehensively calculating the overall influence degree of the measuring point on the temperature disturbance of other measuring points, and using the overall influence degree as a quantitative index of regional thermal coupling.

[0050] Optionally, the above-mentioned method, based on thermally coupled causal chains, quantifies the unidirectional influence intensity of the temperature disturbance of the source on each affected object. This includes: acquiring time-series temperature data of the measuring points of the source and the affected object; identifying the extreme points of fluctuation during the temperature change process; dividing the temperature change process into multiple continuous temperature fluctuation intervals based on these extreme points; calculating the temperature change rate of the source and the temperature change response rate of the affected object for each temperature fluctuation interval; determining the local influence intensity value of the source on the affected object within the current temperature fluctuation interval based on the degree of difference between the temperature change rate and the temperature change response rate; statistically analyzing the distribution pattern of the local influence intensity values ​​corresponding to all temperature fluctuation intervals; and determining the unidirectional influence intensity based on the distribution pattern.

[0051] The calculation of the above-mentioned regional thermal coupling quantitative index first organizes the temperature anomaly characteristics of all measuring points in the historical samples into a structured data matrix. Each row of this matrix corresponds to a historical data record of an annealing process, each column corresponds to a measuring point, and the matrix elements are the temperature anomaly coefficients of each measuring point in the three stages of heating, holding, and cooling. Its typical structure is shown in Table 1: Table 1 Historical Temperature Anomaly Feature Matrix

[0052] Based on this matrix, Pearson correlation analysis is performed on the column vector containing the target measurement point and the column vectors of all other measurement points. Measurement points with a correlation coefficient greater than 0.7 are selected as influencing columns, forming a candidate set of influencing sources. To identify truly effective influencing sources with causal temporal relationships from the candidate set, the Dynamic Time Warping (DTW) algorithm is used to perform sequence matching between the target measurement point and each influencing column, obtaining one-to-one matching pairs between elements and calculating the time difference between each pair. Affected The source of influence is only retained among all matching pairs. The influence column that is always positive, i.e., the column where temperature changes always precede changes at the target measuring point, is used to construct a causal chain of thermal coupling between regions. It can be understood that the above Pearson correlation analysis is a relatively mature technique in the prior art; its specific implementation and working principle can be found in related technologies, and will not be elaborated further in this embodiment.

[0053] For each affected object's measuring point and its source of influence's measuring point, the intensity of the unidirectional influence needs to be further quantified. First, the extreme points of temperature fluctuations in the two sequences are identified, and the time-series data are then divided based on these extreme points. For each of the continuous temperature fluctuation intervals, the rate of temperature change of the influencing source is calculated separately. Response rate to temperature changes of the affected object And calculate the intensity of local influence. Statistically calculate the set of local influence intensity values ​​for all intervals. The frequency distribution will show the highest frequency. This is used as the unidirectional influence strength of the source of influence on the affected object. If multiple modes with the same highest frequency appear in the set of local influence strength values, the arithmetic mean of these multiple modes is taken as the unidirectional influence strength to ensure the certainty and representativeness of the calculation results.

[0054] For each affected object measurement point, the intensity of all unidirectional influences it receives is arranged in descending order to form an influence intensity sequence. ,in, To effectively influence the number of sources, the regional thermal coupling quantification index of this measuring point is established. It can be calculated using two-factor normalization: ,in, The total number of measurement points. To influence the mean intensity; This represents the maximum value of the individual influence intensity among all measuring points. The regional thermal coupling quantification index reflects the correlation and conduction intensity of thermal disturbances from other measuring points at each measuring point. A larger value indicates that the temperature change at that measuring point is more susceptible to external thermal coupling.

[0055] Step S250: Based on historical samples, obtain the cross-stage thermal state memory effect correction coefficient by analyzing the correlation between abnormal characteristics of previous process stages and product quality of subsequent stages.

[0056] Step S260: Combine the quality-sensitive quantitative index, the regional thermal coupling quantitative index, and the cross-stage thermal state memory effect correction coefficient to dynamically generate the differentiated temperature change threshold for each measuring point; wherein, the differentiated temperature change threshold is negatively correlated with the quality-sensitive quantitative index and positively correlated with the regional thermal coupling quantitative index.

[0057] Optionally, step S250 may include: screening sample data from historical samples where the temperature anomaly characteristics of the preceding process stage exceed a preset anomaly threshold, and obtaining the temperature performance data of the sample data in the subsequent process stage and its corresponding product quality grade; classifying the sample data into patterns to obtain multiple thermal state anomaly patterns, and calculating the first proportion of unqualified products in each thermal state anomaly pattern; for each thermal state anomaly pattern, obtaining its temperature performance data in the subsequent process stage, and calculating the second proportion where the temperature anomaly characteristics exceed the preset anomaly threshold; calculating the cross-stage influence factor of the preceding process stage on the subsequent process stage based on the product of the first proportion and the second proportion; performing iterative analysis on multiple preceding-subsequent process stage combinations, and generating a cross-stage thermal state memory effect correction coefficient by comprehensively considering the cross-stage influence factors of each combination.

[0058] Optionally, the above-mentioned cross-stage influence factors of each combination of preceding and subsequent process stages are integrated to generate cross-stage thermal state memory effect correction coefficients, including: normalizing the cross-stage influence factors of each combination of preceding and subsequent process stages to obtain normalized factors; performing weighted fusion processing based on the normalized factors to obtain fused influence factors, and using the fused influence factors as cross-stage thermal state memory effect correction coefficients.

[0059] Optionally, step S260 may include: generating a basic threshold coefficient by nonlinearly combining a mass-sensitive quantitative index and a regional thermal coupling quantitative index; wherein, the higher the mass-sensitive quantitative index, the smaller the basic threshold coefficient, and the higher the regional thermal coupling quantitative index, the larger the basic threshold coefficient; coupling the basic threshold coefficient with a cross-stage thermal state memory effect correction coefficient to obtain a final threshold coefficient; wherein, the larger the cross-stage thermal state memory effect correction coefficient, the smaller the final threshold coefficient; applying the final threshold coefficient to a preset benchmark threshold to obtain the differentiated temperature change threshold for each measuring point.

[0060] The aforementioned cross-stage thermal state memory effect correction coefficient is generated based on the path dependence principle of the heat treatment process, aiming to quantify the cross-stage transmission effect of thermal state anomalies in preceding processes on product quality in subsequent stages. This effect stems from the continuous physical nature of the annealing process, where the temperature distribution and microstructure formed in the previous stage serve as initial or boundary conditions, altering the thermal conductivity and phase transformation kinetics in subsequent stages, thus jointly determining the final quality. The calculation process of the cross-stage thermal state memory effect correction coefficient mainly includes: firstly, screening sample data from historical samples where the temperature anomaly characteristics of preceding processes exceed a preset anomaly threshold, and obtaining their temperature performance data and corresponding product quality grades in subsequent processes. Secondly, classifying the screened sample data into patterns, using density clustering algorithms to generate multiple thermal state anomaly patterns, and statistically analyzing the percentage of non-conforming products in each pattern. A higher ratio indicates a stronger driving force of the anomalous mode in the preceding stage on quality defects. For each anomalous mode, its temperature performance data in subsequent process stages is obtained, and the proportion of temperature anomalies exceeding a preset anomaly threshold is statistically analyzed. This ratio reflects the persistence and amplification effect of preceding anomalies in subsequent stages. The product of the first and second proportions is taken as the cross-stage influence factor of the preceding process stage on the subsequent process stage, i.e. The above analysis was iteratively performed on the two combinations of the preceding-following process stages, namely heating-holding and holding-cooling, to obtain the degree of influence of heating on cooling. The degree of influence of insulation on cooling Take the arithmetic mean of the two. As the overall influence of the preceding stage on the subsequent stage in this region, the larger the value, the more susceptible the subsequent stage is to the abnormal effects of the preceding stage, resulting in poor product quality.

[0061] Base threshold coefficient Through quality sensitivity metrics The nonlinear combination of the quantification index s with regional thermal coupling is generated, and the calculation formula is as follows: This design embodies the logic of physical control: the higher the quality sensitivity quantification index (the greater the contribution of temperature anomalies at this measuring point to product quality defects), and the lower the regional thermal coupling quantification index (the less susceptible this measuring point is to thermal disturbances from other measuring points, and the slower the control response), then... The larger the value, the more stringent the threshold needs to be set for that area to achieve early anomaly detection.

[0062] The final threshold coefficient is This mechanism ensures that: The larger the value of y (higher quality sensitivity and lower regional thermal coupling), and the larger the value of y (stronger cross-stage memory effect), the smaller the final threshold coefficient, and the more stringent the generated differentiated temperature change threshold. Multiplying this threshold coefficient by the preset benchmark threshold yields the dynamic threshold for each measuring point, realizing an adaptive threshold generation strategy of "low tolerance in quality-sensitive areas, high tolerance in strongly coupled areas, and strict control over large cross-stage influence areas".

[0063] Finally, the final threshold coefficient is applied to the preset baseline threshold, and the differentiated temperature change threshold for each measuring point is obtained through multiplication: Tdifferentiation = final threshold coefficient × Tbaseline. Here, the preset baseline threshold Tbaseline is the initial temperature change upper limit specified in the process specification, that is, the basic threshold that the temperature anomaly in each area cannot exceed; otherwise, control measures must be triggered. This three-level calculation framework realizes a comprehensive consideration from quality sensitivity, spatial coupling, to time dependence, making the differentiated threshold negatively correlated with quality sensitivity and positively correlated with regional thermal coupling, ensuring that more stringent temperature control standards are implemented in high-risk areas.

[0064] Step S270: Monitor the temperature data of each measuring point in real time. When the temperature change parameter of any measuring point exceeds the differential temperature change threshold, execute the corresponding regional coordinated control strategy according to the current process stage.

[0065] Optionally, step S270 may include: acquiring temperature change parameters at each measuring point in real time and comparing them with a differentiated temperature change threshold; identifying the current process stage when it is determined that the temperature change parameter at any measuring point exceeds the differentiated temperature change threshold; if it is in the heating stage, performing at least one operation of heating power adjustment and circulating airflow distribution optimization on the area where the measuring point exceeds the threshold; if it is in the heat preservation stage, performing at least one operation of micro-power compensation and atmosphere stability maintenance on the area where the measuring point exceeds the threshold; if it is in the cooling stage, performing at least one operation of cooling medium flow rate adjustment and heat compensation on the area where the measuring point exceeds the threshold.

[0066] The aforementioned regional coordinated control strategy is based on phased adaptive control logic. By monitoring the temperature change parameters at each measuring point in real time and dynamically comparing them with differentiated temperature change thresholds, it achieves early identification and targeted intervention for threshold-exceeding anomalies. When the controller determines that the temperature change parameter at any measuring point exceeds its corresponding differentiated threshold, it first identifies the current annealing process stage and then triggers differentiated control actions that match the thermal characteristics of that stage, ensuring that the control measures accurately respond to the staged failure mode.

[0067] During the heating phase, the core of the control lies in constraining the rate of temperature rise to avoid thermal shock and stress concentration in the structure. The controller continuously calculates the real-time heating rate of each measuring point. When the heating rate of a measuring point approaches or exceeds its differential threshold, the heating power of the area where the measuring point is located is immediately reduced. The power output of the local heating element is reduced through the power regulator to suppress the overheating trend. At the same time, the circulation airflow distribution optimization is initiated, increasing the speed of the nearby circulating fan or adjusting the damper opening to enhance the forced convection heat dissipation capacity. The protective atmosphere concentration in the area is monitored simultaneously. If the atmosphere is diluted due to a sudden temperature rise, the incremental adjustment of the protective gas replenishment valve is triggered to maintain the purity of the atmosphere. Conversely, if the heating rate of a measuring point lags behind the process setting, the heating power of the area is appropriately increased without affecting the thermal balance of the adjacent areas. The hot air flow field in the furnace is redistributed through the damper to enhance the heat transfer efficiency of the area and ensure the synchronization and uniformity of the overall temperature field.

[0068] During the heat preservation stage, the control objective shifts to maintaining a slight stability of the temperature field and the uniformity of microstructure transformation. The controller monitors whether the temperature at each measuring point fluctuates within the allowable tolerance range of the differential threshold. When a temperature drift trend is detected, slight power compensation is implemented in the area where the measuring point exceeds the threshold. A PI control algorithm is used to calculate a small heating power correction amount and apply it to the power regulator to achieve precise temperature correction at the sub-Celsius level. During this stage, the circulating fan maintains high-speed and stable operation to enhance the gas homogenization effect in the furnace and prevent the regeneration of local temperature gradients. All heating adjustments prioritize atmosphere stability to avoid oxidation or decarburization of the casting surface due to power fluctuations disrupting the atmosphere balance, thus ensuring the coordinated stability of thermal parameters during the heat preservation process.

[0069] During the cooling phase, the key to control is precisely regulating the cooling rate to suppress secondary phase transformation and residual stress regeneration. The controller calculates the cooling rate at each measuring point in real time. When the cooling rate at a measuring point exceeds the upper limit of the differentiated threshold, the cooling medium flow rate is immediately adjusted, reducing or closing the opening of the corresponding cooling vent to limit the amount of cold air entering. If necessary, the heating element is briefly pulsed and heated by the power regulator to actively slow down the cooling process. If the cooling rate at a measuring point is below the lower limit of the threshold, the cooling medium flow rate is increased, the opening of the exhaust vent valve is increased, and the total speed of the circulating fan is increased to enhance the heat dissipation capacity. At the same time, protective gas is continuously introduced and its flow rate is adaptively reduced as the temperature decreases to prevent oxidation at low temperatures. This phased control strategy, through the dynamic constraints of differentiated thresholds and the coordinated execution of regional actions, achieves precise quality assurance throughout the entire annealing process of large castings.

[0070] This invention is now complete.

[0071] In summary, in this embodiment of the invention, a multi-dimensional quantitative model is constructed to achieve intelligent dynamic generation of temperature control thresholds for large casting annealing processes. First, the quality sensitivity quantification index characterizes the contribution intensity of a single measuring point temperature anomaly to product quality defects. The larger this index, the higher the risk of quality deterioration caused by temperature fluctuations at that measuring point. Therefore, a smaller differential temperature change threshold needs to be set to achieve strict control over this area and ensure the annealing quality of the casting. Second, the regional thermal coupling quantification index characterizes the correlation and transmission strength of temperature changes at a measuring point to thermal disturbances at other measuring points. For a given area, the less it is affected by other areas (i.e., the lower the thermal coupling quantification index), the stronger its thermal independence. This makes it more difficult to effectively control temperature anomalies in that area through airflow circulation formed by adjusting fan speed, because the blockage of the thermal disturbance transmission path makes it difficult for the control effect to reach that area. Therefore, a lower threshold must be set to achieve rapid capture and early intervention of anomalies in this area, avoiding irreversible impacts on product annealing quality when anomalies accumulate to a large extent. Finally, the cross-stage thermal state memory effect correction coefficient is used to quantify the cross-stage quality impact caused by the memory effect and path dependence in the heat treatment process. The larger the coefficient, the more susceptible the subsequent stage is to the influence of abnormal thermal states in the preceding stage, leading to product quality defects. Therefore, it is necessary to combine the abnormal conduction characteristics of the three stages of heating, holding, and cooling to set a lower threshold for areas with significant cross-stage effects, thereby eliminating the cumulative effect of previous abnormalities on the final quality through a feedforward compensation mechanism. The above three factors jointly determine the differentiated temperature change threshold at each measuring point through nonlinear coupling, enabling the system to adaptively adjust the threshold stringency according to the dynamic changes in quality risk sensitivity, spatial thermal coupling characteristics, and the strength of time path dependence. Ultimately, this achieves precise regional collaborative control and quality consistency assurance throughout the entire annealing process of large castings.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent control device for the heat treatment process of large castings, comprising an annealing furnace body, characterized in that, The device also includes an intelligent control system, which comprises a multi-area sensor array and a controller connected to the multi-area sensor array via signals, wherein: The multi-region sensor array is deployed along the furnace spatial coordinate system to collect temperature time-series data and atmosphere concentration data at multiple measuring points inside the furnace during the heat treatment process. The controller is used to acquire temperature time-series data from multiple measuring points within the heat treatment furnace and corresponding historical samples of product quality grades; divide the temperature time-series data into multiple process stages and extract the stage-specific temperature anomaly characteristics of each measuring point; based on the historical samples, identify temperature anomaly pattern clusters using a clustering algorithm, and calculate the quality sensitivity quantification index for each measuring point according to the proportion of defective products in each cluster; wherein, the quality sensitivity quantification index is used to quantify the contribution intensity of the temperature anomaly at the measuring point to product quality defects; based on the historical samples, analyze the lag and correlation of temperature changes between different measuring points using a time-series matching algorithm, identify the causal chain of thermal coupling between regions, and calculate the regional thermal coupling quantification index for each measuring point; wherein, the regional thermal coupling quantification index... This is used to quantify the correlation and conduction strength of temperature changes at a measuring point due to thermal disturbances at other measuring points. Based on the historical samples, the correlation between abnormal characteristics of previous process stages and product quality in subsequent stages is analyzed to obtain a cross-stage thermal state memory effect correction coefficient. Combining the quality sensitivity quantification index, the regional thermal coupling quantification index, and the cross-stage thermal state memory effect correction coefficient, a differentiated temperature change threshold for each measuring point is dynamically generated. The differentiated temperature change threshold is negatively correlated with the quality sensitivity quantification index and positively correlated with the regional thermal coupling quantification index. The temperature data of each measuring point is monitored in real time. When the temperature change parameter of any measuring point exceeds the differentiated temperature change threshold, a corresponding regional coordinated control strategy is executed according to the current process stage.

2. A method for intelligent control of the heat treatment process of large castings, characterized in that, The method includes: Acquire temperature time-series data from multiple measuring points inside the heat treatment furnace and corresponding historical samples of product quality grades; The temperature time series data is divided into multiple process stages and the stage-specific temperature anomaly features of each measuring point are extracted. Based on the historical samples, temperature anomaly pattern clusters are identified using a clustering algorithm, and a quality sensitivity quantification index for each measuring point is calculated according to the proportion of defective products in each cluster; wherein, the quality sensitivity quantification index is used to quantify the contribution intensity of temperature anomaly at the measuring point to product quality defects. Based on the historical samples, the lag and correlation of temperature changes between different measuring points are analyzed by time-series matching algorithm to identify the causal chain of thermal coupling between regions and to calculate the regional thermal coupling quantification index for each measuring point; wherein, the regional thermal coupling quantification index is used to quantify the correlation conduction strength of temperature changes at measuring points caused by thermal disturbances at other measuring points. Based on the historical samples, the correlation between abnormal characteristics of the preceding process stage and product quality in the subsequent stage is analyzed to obtain the cross-stage thermal state memory effect correction coefficient. By combining the quality-sensitive quantitative index, the regional thermal coupling quantitative index, and the cross-stage thermal state memory effect correction coefficient, a differentiated temperature change threshold for each measuring point is dynamically generated; wherein, the differentiated temperature change threshold is negatively correlated with the quality-sensitive quantitative index and positively correlated with the regional thermal coupling quantitative index. Real-time monitoring of temperature data at each measuring point; when the temperature change parameter at any measuring point exceeds the differentiated temperature change threshold, the corresponding regional coordinated control strategy is executed according to the current process stage.

3. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, The extraction of phased temperature anomaly characteristics at each measuring point includes: According to the characteristics of the annealing process, the temperature timing data is divided into a heating stage, a holding stage, and a cooling stage. The actual temperature curves of each measuring point at each stage and the corresponding preset standard process curves for each stage are obtained respectively. The temperature anomaly coefficient for each stage is quantified by calculating the morphological similarity index between the actual temperature curve and the standard process curve in terms of temperature change trend; wherein, the lower the morphological similarity index, the larger the temperature anomaly coefficient. Based on the temperature anomaly coefficients at each stage, the stage-specific temperature anomaly characteristics of each measuring point are determined.

4. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, Based on the historical samples, a clustering algorithm is used to identify temperature anomaly pattern clusters, and a quality sensitivity quantification index is calculated according to the proportion of non-conforming products in each cluster, including: Obtain the phased temperature anomaly characteristics of each measuring point and their corresponding product quality grades; Cluster the data with similar temperature anomaly characteristics in the historical samples to generate multiple temperature anomaly pattern clusters; Calculate the percentage of non-conforming products in each temperature anomaly pattern cluster, and identify the clusters whose percentage exceeds the preset risk threshold as critical risk clusters; Calculate the average value of temperature anomaly characteristics of each measuring point within each critical risk cluster, and then weight and accumulate the values ​​according to the proportion of non-conforming products in each critical risk cluster to obtain the contribution intensity of temperature anomaly at that measuring point to product quality defects. Use the contribution intensity as a quantitative indicator of quality sensitivity.

5. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, The process involves analyzing the hysteresis and correlation of temperature changes between different measuring points using a time-series matching algorithm, identifying causal chains of thermal coupling between regions, and calculating quantitative indices for regional thermal coupling, including: The temperature time series data of each measuring point are paired up, and the degree of synchronization of temperature change trends between each pair of measuring points is analyzed. The measuring point pairs with consistent temperature change trends and a degree of synchronization exceeding the preset correlation threshold are selected. For the selected measurement point pairs, a correspondence of temperature change characteristics is established through dynamic pattern matching method to determine the temporal order of temperature changes between the two measurement points. The measurement point with the earlier temperature change is regarded as the source of influence, and the measurement point with the later temperature change is regarded as the object of influence, thus constructing a causal chain of thermal coupling between regions. Based on the aforementioned thermally coupled causal chain, the unidirectional influence intensity of each affected object's measuring point on the temperature disturbance of the corresponding source measuring point is quantified; The distribution characteristics of the intensity of all unidirectional influences on each measuring point when it is affected are statistically analyzed. The overall influence of temperature disturbances from other measuring points on the measuring point is calculated, and the overall influence is used as a quantitative index of regional thermal coupling.

6. The intelligent control method for the heat treatment process of large castings according to claim 5, characterized in that, The quantification of the unidirectional influence intensity of the temperature disturbance of the influence source on each affected object based on the thermally coupled causal chain includes: Acquire temperature time-series data of the source measurement point and the affected object measurement point, identify the extreme points of fluctuation in the temperature change process, and divide the temperature change process into multiple continuous temperature fluctuation intervals based on the extreme points of fluctuation. For each temperature fluctuation range, calculate the temperature change rate of the influencing source and the temperature change response rate of the affected object. Based on the degree of difference between the temperature change rate and the temperature change response rate, the local influence intensity value of the influence source on the affected object within the current temperature fluctuation range is determined. The distribution pattern of local influence intensity values ​​corresponding to all temperature fluctuation intervals is statistically analyzed, and the unidirectional influence intensity is determined based on the distribution pattern.

7. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, Based on the historical samples, the correlation between abnormal characteristics of previous process stages and product quality in subsequent stages is analyzed to obtain the cross-stage thermal state memory effect correction coefficient, including: The sample data that exceed the preset abnormal threshold in the temperature anomaly characteristics of the preceding process stage are selected from the historical samples, and the temperature performance data of the sample data in the subsequent process stage and the corresponding product quality level are obtained. The sample data is classified into multiple thermal state abnormality patterns, and the first proportion of unqualified products in each thermal state abnormality pattern is calculated. For each abnormal thermal state mode, obtain its temperature performance data in subsequent process stages, and count the second proportion of temperature abnormality features exceeding the preset abnormality threshold. Based on the product of the first proportion and the second proportion, calculate the cross-stage influence factor of the preceding process stage on the subsequent process stage. Iterative analysis was performed on multiple combinations of preceding and subsequent process stages, and cross-stage influence factors of each combination were integrated to generate a cross-stage thermal state memory effect correction coefficient.

8. The intelligent control method for the heat treatment process of large castings according to claim 7, characterized in that, The cross-stage influence factors, which are synthesized from the combinations of the preceding and subsequent process stages, are used to generate the cross-stage thermal state memory effect correction coefficient, including: Normalized factors are obtained by normalizing the cross-stage influence factors of each combination of preceding and subsequent process stages. The fusion influence factor is obtained by weighted fusion processing based on the normalization factor, and the fusion influence factor is used as the correction coefficient for cross-stage thermal state memory effect.

9. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, The method of dynamically generating differentiated temperature change thresholds for each measuring point by combining the quality-sensitive quantification index, the regional thermal coupling quantification index, and the cross-stage thermal state memory effect correction coefficient includes: A basic threshold coefficient is generated by nonlinearly combining the quality-sensitive quantification index and the regional thermal coupling quantification index; the higher the quality-sensitive quantification index, the smaller the basic threshold coefficient, and the higher the regional thermal coupling quantification index, the larger the basic threshold coefficient. The basic threshold coefficient is coupled and corrected with the cross-stage thermal state memory effect correction coefficient to obtain the final threshold coefficient; wherein, the larger the cross-stage thermal state memory effect correction coefficient is, the smaller the final threshold coefficient is. The final threshold coefficient is applied to a preset benchmark threshold to obtain the differentiated temperature change threshold for each measuring point.

10. The intelligent control method for the heat treatment process of large castings according to claim 2, characterized in that, The real-time monitoring of temperature data at each of the aforementioned measuring points, when the temperature change parameter at any of the measuring points exceeds the differentiated temperature change threshold, executes a corresponding regional coordinated control strategy based on the current process stage, including: The temperature change parameters of each measuring point are acquired in real time and compared with the differential temperature change threshold. When it is determined that the temperature change parameter at any measuring point exceeds the differentiated temperature change threshold, the current process stage is identified. If it is in the heating stage, perform at least one of the following operations on the area where the measurement point exceeds the threshold: heating power adjustment and circulating airflow distribution optimization. If it is in the heat preservation stage, perform at least one of the following operations on the area where the measurement point exceeds the threshold: micro-power compensation and atmosphere stability maintenance. If the temperature is dropping, at least one of the following operations should be performed on the area where the measurement point exceeds the threshold: cooling medium flow rate regulation and heat compensation.

Images