A heating control method and system for a thread rolling machine based on hot state trajectory prediction

Abstract

The application discloses a thread rolling machine heating control method and system based on hot state trajectory prediction, and relates to the technical field of intelligent control of fastener hot working. Data such as workpiece outlet temperature, heater power, feeding speed and residence time are collected to construct a hot state trajectory; a trajectory is scanned to identify a shape section type and extract shape parameters, and a shape feature vector is coded; a trajectory shape evolution trend is predicted according to current and historical feature vectors to generate a trajectory prediction sequence; a trajectory credible domain corresponding to workpiece specifications and materials is constructed according to historical data; a trajectory offset index is obtained by comparing the prediction sequence with the credible domain to generate a heating control strategy; whether a predicted hot state node is located in the credible domain is determined to decide whether to discharge or reject the workpiece; and related parameters are corrected according to an actual trajectory. A forward-looking control strategy is generated by comparing the trajectory prediction with the credible domain, combined with predicted discharging and online self-correction, so that feedback lag, discharging unable to predict hot state offset defects and the like are overcome, and the processing stability is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology for hot working of fasteners, specifically a heating control method and system for thread rolling machines based on thermal trajectory prediction. Background Technology

[0002] Thread rolling machines, as crucial equipment in fastener production for thread forming, largely depend on the heating state of the workpiece before it enters the thread rolling station for processing quality. For high-strength bolts, stainless steel screws, and irregularly shaped fasteners, induction heating is typically used to preheat the workpiece to a suitable hot-working temperature before it enters the thread rolling station, reducing forming resistance, improving material plasticity, and enhancing thread forming quality. Currently, industrial sites commonly use induction heaters in conjunction with infrared thermometers for heating process control. During production, a target temperature range is first set based on product specifications, material type, and production process. Simultaneously, the thread rolling machine's operating speed is determined based on the production cycle, and the initial heating power is calculated based on parameters such as workpiece mass, material specific heat capacity, target temperature rise, and induction heating efficiency. During heating, the workpiece temperature at the heating coil outlet is continuously monitored by an infrared thermometer. When the detected temperature deviates from the target temperature range, the operator manually fine-tunes the output power of the induction heater based on the temperature changes. When the detected temperature consistently exceeds the allowable range, the system executes control operations such as alarms, stopping material feeding, or rejecting overheated workpieces to prevent them from entering the thread rolling station and causing product quality abnormalities.

[0003] With the increasing variety of fastener product specifications and the continuous improvement of production cycle time, the dwell time of the workpiece in the induction coil is constantly shortening. The heating process is affected by a variety of factors, including workpiece size, material batch, feeding speed, coil thermal inertia, and equipment operating status. Since induction heating is a dynamic and continuous heating process, the thermal state of the workpiece is not solely determined by the current detected temperature, but rather by a dynamic thermal change process that gradually evolves throughout the heating process. In actual production, even with the same current detected temperature, the internal thermal state formed by different heating paths may still differ significantly, leading to fluctuations in plastic deformation capacity and thread forming quality during subsequent thread rolling. However, existing control methods still mainly rely on feedback adjustment based on the temperature detected at a single moment, using the current workpiece temperature as the sole control basis. This lacks the ability to describe the continuous thermal change process and cannot reflect the future development trend of the thermal state. When process conditions change or equipment operating status is disturbed, relying solely on the current temperature feedback for control often has a certain lag, easily leading to frequent adjustments to the control strategy, increased fluctuations in the heating state, and inaccurate judgment of the material release timing, ultimately affecting the stability of the thread rolling process and product consistency. Summary of the Invention

[0004] The purpose of this application is to provide a heating control method and system for a wire rolling machine based on thermal trajectory prediction, so as to solve the problems mentioned in the background art.

[0005] In a first aspect, one embodiment of this application provides a heating control method for a thread rolling machine based on thermal trajectory prediction. The method includes: synchronously collecting multi-source process data during the workpiece heating process according to a preset sampling period. The multi-source process data includes at least the workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time. The multi-source process data is fused in chronological order to construct a thermal trajectory that describes the entire workpiece heating process. The thermal trajectory is composed of multiple consecutive thermal nodes arranged in chronological order, with each thermal node formed by a combination of the workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time at the corresponding time. The thermal trajectory is scanned along the time axis according to a preset sliding window, identifying the morphological segment type to which the sliding window belongs, and extracting multiple morphological parameters characterizing the morphological segment type. These multiple morphological parameters are encoded to obtain a thermal trajectory morphological feature vector. Based on the thermal trajectory morphological feature vectors of the current and historical times, the evolution trend of the workpiece's thermal trajectory morphology over a future period is predicted to obtain a future thermal trajectory morphological feature vector prediction sequence. The prediction sequence is then used to determine the future thermal trajectory morphological feature vector. The process involves several steps: First, a predictive sequence of the thermal trajectory is reconstructed from the predicted sequence of the thermal feature vector. Based on historical thermal trajectory data generated during production, a corresponding thermal trajectory confidence domain is constructed for each workpiece specification and material type. This domain includes the upper and lower boundaries of the trajectory, which dynamically change with time and morphological segment type. The predicted sequence of the future thermal trajectory is compared with this confidence domain, and the offset of the predicted sequence relative to the confidence domain is calculated. A trajectory offset index is then obtained based on this offset. A heating control strategy is generated based on the trajectory offset index, which includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed. Based on the predicted sequence of the future thermal trajectory, it is determined whether the predicted thermal node corresponding to the expected entry time of the workpiece into the toothed plate is within the thermal trajectory confidence domain. If so, material feeding is performed; otherwise, delayed feeding or rejection is performed. Finally, the actual thermal trajectory of the workpiece after thread rolling is collected. The actual thermal trajectory is compared with the corresponding predicted sequence of the future thermal trajectory to obtain the morphological deviation. The parameters used to construct the thermal trajectory confidence domain and generate the predicted sequence of the future thermal trajectory are then corrected based on the morphological deviation.

[0006] In conjunction with the first aspect, in some implementations of the first aspect, multi-source process data are fused in chronological order to construct a thermal trajectory that can describe the entire heating process of the workpiece. This includes: performing time alignment processing on the multi-source process data according to a unified time axis to eliminate time sequence deviations caused by different data acquisition sources, resulting in a set of process data arranged in time sequence; based on the process data set, at each sampling moment, the workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time at the corresponding moment are combined into thermal nodes, and the thermal nodes at multiple consecutive moments are arranged in chronological order to obtain the thermal trajectory.

[0007] In conjunction with the first aspect, in some implementations of the first aspect, the thermal trajectory is scanned along the time axis according to a preset sliding window, the morphological segment type to which the sliding window belongs is identified, and multiple morphological parameters characterizing the morphological segment type are extracted. The multiple morphological parameters are encoded to obtain the thermal trajectory morphological feature vector, including: identifying the morphological segment type to which the preset sliding window belongs based on the slope direction, rate of change, and fluctuation characteristics of the temperature change with time within the preset sliding window. The morphological segment type includes at least one of heating segment, stable segment, drift segment, and oscillation segment; extracting morphological parameters for the identified morphological segment type. The morphological parameters include at least one of heating slope, curvature, peak density, and fluctuation frequency; and encoding the morphological parameters in a preset order to obtain the thermal trajectory morphological feature vector.

[0008] In conjunction with the first aspect, in certain implementations of the first aspect, based on the thermal trajectory morphology feature vectors of the current moment and historical moments, the evolution trend of the thermal trajectory morphology of the workpiece over a future period is predicted to obtain a future thermal trajectory morphology feature vector prediction sequence. Then, a future thermal trajectory prediction sequence is reconstructed based on this prediction sequence. This includes: using the thermal trajectory morphology feature vectors of the current moment and historical moments within a preset backtracking time as input, predicting the morphological segment type and corresponding thermal trajectory morphology feature vectors of the workpiece within a preset future time step to obtain a future thermal trajectory morphology feature vector prediction sequence; mapping the future thermal trajectory morphology feature vector prediction sequence to future thermal nodes according to a preset correspondence between the thermal trajectory morphology feature vectors and thermal nodes, and arranging multiple consecutive future thermal nodes in chronological order to obtain the future thermal trajectory prediction sequence.

[0009] In conjunction with the first aspect, in some implementations of the first aspect, based on the thermal trajectory data formed during historical production processes, a corresponding thermal trajectory credibility domain is constructed for workpiece specifications and material types. This includes: acquiring thermal trajectory data formed during historical stable production processes; statistically analyzing the allowable location range of morphological segment types in the thermal trajectory for different workpiece specifications and different material types; and constructing a thermal trajectory credibility domain corresponding to workpiece specifications and material types based on the statistically obtained location range. The thermal trajectory credibility domain also includes at least one of platform boundary, drift tolerance, and fluctuation tolerance.

[0010] In conjunction with the first aspect, in some implementations of the first aspect, the predicted sequence of the future thermal trajectory is compared with the trusted domain of the thermal trajectory, the offset of the predicted sequence of the future thermal trajectory relative to the trusted domain of the thermal trajectory is calculated, and a trajectory offset index is obtained based on the offset. This includes: calculating the positional offset between the predicted sequence of the future thermal trajectory and the center line of the trusted domain of the thermal trajectory, the morphological offset of the morphological features of the predicted sequence of the future thermal trajectory relative to the morphological features of the target thermal trajectory, the plateau offset of the duration and temperature level of the segment in the predicted sequence of the future thermal trajectory where the temperature change amplitude is less than a preset amplitude threshold relative to the trusted domain of the thermal trajectory, the oscillation offset of the predicted sequence of the future thermal trajectory relative to the fluctuation tolerance, and the drift direction offset of the predicted sequence of the future thermal trajectory relative to the drift tolerance, as offsets; and weighting and fusing the positional offset, morphological offset, plateau offset, oscillation offset, and drift direction offset according to preset weights to obtain a trajectory offset index that characterizes the overall deviation of the predicted sequence of the future thermal trajectory relative to the trusted domain of the thermal trajectory.

[0011] In conjunction with the first aspect, in certain implementations of the first aspect, a heating control strategy is generated based on the trajectory offset index. The heating control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed, including: if the predicted future hot trajectory sequence leaves the upper boundary of the hot trajectory confidence domain earlier than the preset target crossing time, then the output power of the induction heater is reduced; if the position offset is negative and its absolute value exceeds a preset threshold, then the output power of the induction heater is increased; if the platform offset exceeds a preset threshold, then the workpiece feeding speed is reduced and the workpiece online dwell time is extended; if the oscillation offset exceeds a preset threshold, then the filtering compensation amount is increased; if the drift direction offset exceeds a preset threshold, then the heating cycle is recalculated and the heating sequence of subsequent workpieces is adjusted.

[0012] In conjunction with the first aspect, in certain implementations of the first aspect, based on the future thermal trajectory prediction sequence, it is determined whether the predicted thermal node corresponding to the expected entry time of the workpiece into the toothed plate is within the thermal trajectory confidence domain. If so, material feeding is performed; otherwise, delayed feeding or rejection is performed. This includes: if the predicted thermal node is within the thermal trajectory confidence domain, the workpiece is determined to meet the processing requirements, and material feeding is performed; if the predicted thermal node deviates from the thermal trajectory confidence domain but the deviation does not exceed the preset rejection threshold, material feeding is temporarily suspended, and the updated future thermal trajectory prediction sequence is obtained in the next sampling period for re-judgment; if the predicted thermal node exceeds the upper boundary of the thermal trajectory confidence domain, the workpiece is determined not to meet the processing requirements, and the workpiece is removed from the production process.

[0013] In conjunction with the first aspect, in some implementations of the first aspect, the actual hot trajectory of the workpiece after the thread rolling process is collected, and the actual hot trajectory is compared with the corresponding future hot trajectory prediction sequence to obtain the morphological deviation. Based on the morphological deviation, the parameters used to construct the credibility domain of the hot trajectory and generate the future hot trajectory prediction sequence are corrected, including: calculating the difference in morphological parameters between the actual hot trajectory and the future hot trajectory prediction sequence to obtain the morphological deviation; correcting the length of the preset sliding window and the segment judgment threshold used to identify the morphological segment type based on the morphological deviation; correcting the upper boundary, lower boundary, platform boundary, drift tolerance, and fluctuation tolerance of the trajectory based on the morphological deviation; and correcting the weighting coefficients used to predict the morphological feature vector prediction sequence of the future hot trajectory and generate the heating control strategy based on the morphological deviation and position offset, morphological offset, platform offset, oscillation offset, and drift direction offset.

[0014] Secondly, one embodiment of this application also provides a heating control system for a thread rolling machine based on thermal trajectory prediction. This system includes: a data acquisition module, used to synchronously acquire multi-source process data during the workpiece heating process according to a preset sampling period. The multi-source process data includes at least the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time; and a trajectory construction module, used to fuse the multi-source process data in chronological order to construct a thermal trajectory that describes the entire workpiece heating process. The thermal trajectory is composed of multiple consecutive thermal nodes arranged in chronological order, with each thermal node represented by the workpiece outlet temperature at the corresponding time. The system is composed of several parameters: the output power of the induction heater, the workpiece feeding speed, and the workpiece's online dwell time; a morphological feature extraction module, used to scan the thermal trajectory along the time axis according to a preset sliding window, identify the morphological segment type to which the sliding window belongs, and extract multiple morphological parameters characterizing the morphological segment type. These parameters are then encoded to obtain the thermal trajectory morphological feature vector; a trajectory prediction module, used to predict the evolution trend of the workpiece's thermal trajectory morphology over a future period based on the current and historical thermal trajectory morphological feature vectors, obtaining a future thermal trajectory morphological feature vector prediction sequence, and then using this prediction sequence... The system reconstructs the predicted sequence to generate a future thermal trajectory prediction sequence; a credibility domain construction module is used to construct a corresponding thermal trajectory credibility domain based on the thermal trajectory data generated during historical production processes, taking into account workpiece specifications and material types. The thermal trajectory credibility domain includes the upper and lower boundaries of the trajectory, which dynamically change with time and morphological segment types; an offset evaluation module is used to compare the future thermal trajectory prediction sequence with the thermal trajectory credibility domain, calculate the offset of the future thermal trajectory prediction sequence relative to the thermal trajectory credibility domain, and obtain the trajectory offset index based on the offset; a control strategy generation module is used to generate a heating control strategy based on the trajectory offset index. The control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed; the unloading decision module is used to determine whether the predicted hot node corresponding to the expected entry time of the workpiece into the toothed plate is within the reliable domain of the hot trajectory, based on the future hot trajectory prediction sequence; if so, unloading is performed, otherwise delayed unloading or rejection is performed; the online self-correction module is used to collect the actual hot trajectory of the workpiece after the thread rolling process, compare the actual hot trajectory with the corresponding future hot trajectory prediction sequence to obtain the shape deviation, and correct the parameters used to construct the reliable domain of the hot trajectory and generate the future hot trajectory prediction sequence based on the shape deviation.

[0015] Compared with the prior art, the beneficial effects of this application are: 1. This application synchronously collects multi-source process data such as workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time according to a preset sampling period, and integrates them in time sequence to construct a thermal trajectory that can describe the entire heating process of the workpiece. Then, it performs morphological segment identification and morphological parameter extraction on the thermal trajectory to obtain a thermal trajectory morphological feature vector. This overcomes the shortcomings of the existing technology, which only uses temperature detection at a single moment as the control basis and lacks the ability to describe the continuous evolution process of the workpiece's thermal state. This enables the control system to truly and completely reflect the dynamic thermal change law of the workpiece throughout the entire heating process.

[0016] 2. This application predicts the evolution trend of the workpiece's thermal trajectory morphology over a future period based on the thermal trajectory morphology feature vectors of the current and historical moments. It then compares the predicted future thermal trajectory sequence with the thermal trajectory credibility domain to obtain a trajectory offset index, thereby generating a forward-looking heating control strategy. Simultaneously, it determines whether to release, delay, or reject the material based on whether the predicted thermal nodes are within the thermal trajectory credibility domain. Furthermore, it uses the actual thermal trajectory of the workpiece after processing to perform online self-correction of relevant model parameters. This overcomes the shortcomings of existing feedback control methods, such as delayed response and the inability to predict subsequent thermal state offsets due to reliance on the current detected temperature for material release decisions. It achieves a shift from post-processing correction to proactive intervention in control decisions, while also possessing continuous self-learning optimization capabilities, thus significantly improving the stability of the thread rolling process and the consistency of thread forming product quality. Attached Figure Description

[0017] Figure 1 A schematic flowchart of a wire rolling machine heating control method based on thermal trajectory prediction provided in an embodiment of this application; Figure 2 This is a schematic diagram of the heating control system for a wire rolling machine based on thermal trajectory prediction, provided as an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Figure 1 This is a flowchart illustrating a heating control method for a wire rolling machine based on thermal trajectory prediction, provided as an embodiment of this application. Figure 1 As shown, an embodiment of this application provides a heating control method for a wire rolling machine based on thermal trajectory prediction, which includes the following steps: Step 101: Collect multi-source process data of the workpiece heating process synchronously according to the preset sampling period.

[0020] It should be understood that multi-source process data refers to a collection of multiple process parameters that can jointly reflect the heating state of the workpiece at the same sampling time, specifically including at least the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time.

[0021] For example, the preset sampling period can be set to a fixed value within the range of 10ms to 100ms, such as 20ms, which can be adjusted according to the running speed of the wire rolling machine and the required control response speed.

[0022] Step 102: The multi-source process data are fused in chronological order to construct a thermal trajectory that can describe the entire heating process of the workpiece. The thermal trajectory is composed of multiple consecutive thermal nodes arranged in chronological order. Each thermal node is formed by the combination of the workpiece outlet temperature, induction heater output power, workpiece feeding speed and workpiece online dwell time at the corresponding time.

[0023] It should be understood that a hot node refers to a basic descriptive unit formed by the combination of various process parameters at a certain sampling moment.

[0024] A thermal trajectory is a trajectory consisting of multiple consecutive thermal nodes arranged in chronological order, which can continuously describe the entire heating process of a workpiece.

[0025] Step 103: Scan the thermal trajectory along the time axis according to the preset sliding window, identify the morphological segment type to which the sliding window belongs, extract multiple morphological parameters that characterize the morphological segment type, encode the multiple morphological parameters, and obtain the thermal trajectory morphological feature vector.

[0026] It should be understood that a preset sliding window refers to a fixed-length window that extracts a local segment of the thermal trajectory along the time axis for analysis.

[0027] Morphological segment type refers to the trajectory morphology category classified based on the local change characteristics of the thermal trajectory.

[0028] Morphological parameters are parameters used to quantitatively characterize the changes in a certain morphological segment.

[0029] The thermal trajectory morphology feature vector is a vector formed by encoding and combining multiple morphological parameters in a preset order, used to represent the trajectory morphology of a certain morphological segment as a whole.

[0030] For example, the length of the preset sliding window can be set to 1 second, corresponding to 50 consecutive sampling times of hot nodes when the preset sampling period is 20ms. The specific window length can be adjusted according to the dwell time of the workpiece in the induction coil and the requirements for the accuracy of shape segment recognition.

[0031] Step 104: Based on the thermal trajectory morphology feature vectors of the current time and historical time, predict the thermal trajectory morphology evolution trend of the workpiece in the future period, obtain the future thermal trajectory morphology feature vector prediction sequence, and reconstruct the future thermal trajectory prediction sequence based on the future thermal trajectory morphology feature vector prediction sequence.

[0032] It should be understood that the future thermal trajectory morphology feature vector prediction sequence refers to the sequence obtained by predicting the thermal trajectory morphology feature vectors corresponding to each morphological segment within a preset time step of the workpiece in the future.

[0033] The future hot trajectory prediction sequence refers to the trajectory sequence formed by arranging future hot nodes in chronological order, which is obtained by mapping the future hot trajectory morphological feature vector prediction sequence.

[0034] For example, the number of historical moments within the preset backtracking time can be set to 10, that is, the historical thermal trajectory morphology feature vectors corresponding to the current moment and the previous 10 sliding window cycles are used as prediction inputs; the number of preset time steps can be set to 5, that is, the thermal trajectory morphology evolution trend corresponding to the next 5 sliding window cycles of the workpiece is predicted. The above values ​​can be adjusted according to the prediction accuracy and real-time calculation requirements.

[0035] Step 105: Based on the thermal trajectory data generated during the historical production process, construct the corresponding thermal trajectory confidence domain for workpiece specifications and material types. The thermal trajectory confidence domain includes the upper and lower boundaries of the trajectory that dynamically change with time and morphological segment type.

[0036] It should be understood that the credibility region of a thermal trajectory refers to the dynamic envelope interval obtained by statistically analyzing thermal trajectory data formed during historical stable production processes, used to define the reasonable fluctuation range of thermal trajectories under normal production conditions.

[0037] The upper and lower boundaries of the trajectory refer to the upper and lower limits of temperature in the credibility domain of the thermal trajectory, which change dynamically with time and morphological segment type, respectively.

[0038] For example, for a high-strength bolt workpiece of a certain specification, the upper and lower boundaries of the thermal trajectory confidence domain can be set as the envelope interval formed by the target temperature trajectory fluctuating by 5℃ above and below; the platform boundary can be set as the target temperature ±5℃; the drift tolerance can be set as the maximum allowable drift rate of 0.5℃ / s; and the fluctuation tolerance can be set as the maximum allowable fluctuation amplitude ±4℃. The above values ​​can be determined separately based on historical data statistics according to different workpiece specifications and material types.

[0039] Step 106: Compare the future hot trajectory prediction sequence with the hot trajectory confidence domain, calculate the offset of the future hot trajectory prediction sequence relative to the hot trajectory confidence domain, and obtain the trajectory offset index based on the offset.

[0040] It should be understood that offset refers to the degree of deviation of the predicted sequence of future hot trajectory from the credibility domain of the hot trajectory in multiple dimensions such as position, shape, plateau, oscillation and drift direction.

[0041] The trajectory offset index is a quantitative indicator that comprehensively represents the degree of deviation of the predicted future hot trajectory sequence from the overall credibility domain of the hot trajectory, obtained by weighted fusion of the offsets of the above-mentioned dimensions.

[0042] Step 107: Generate a heating control strategy based on the trajectory offset index. The heating control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed.

[0043] It should be understood that the heating control strategy refers to a forward-looking control scheme generated based on the deviation reflected by the trajectory deviation index, used to adjust at least one of the output power of the induction heater and the workpiece feeding speed.

[0044] Step 108: Based on the future thermal trajectory prediction sequence, determine whether the predicted thermal node corresponding to the expected entry time of the workpiece into the toothed plate is within the thermal trajectory confidence domain. If so, execute the feeding; otherwise, execute delayed feeding or rejection.

[0045] It should be understood that the expected entry time into the thread rolling machine refers to the expected time when the workpiece enters the thread rolling machine's thread rolling plate station for processing.

[0046] Predicted hot nodes refer to the hot nodes in the future hot trajectory prediction sequence that correspond to the expected entry time into the tooth plate. They are used to predict whether the thermal state of the workpiece at that time will meet the processing requirements before unloading.

[0047] Step 109: Collect the actual hot trajectory of the workpiece after the thread rolling process, compare the actual hot trajectory with the corresponding future hot trajectory prediction sequence to obtain the morphological deviation, and correct the parameters used to construct the credibility domain of the hot trajectory and generate the future hot trajectory prediction sequence based on the morphological deviation.

[0048] It should be understood that the actual thermal trajectory refers to the thermal trajectory obtained from actual test data after the workpiece has completed the thread rolling process.

[0049] Morphological deviation refers to the difference in morphological parameters between the actual thermal trajectory and the corresponding predicted future thermal trajectory sequence. It is used to measure the degree of deviation between the prediction result and the actual result, and serves as the basis for online correction of relevant model parameters.

[0050] By constructing a thermal trajectory that can continuously describe the entire heating process of a workpiece and predict its morphological evolution trend, the heating control and material feeding decision-making are transformed from a hysteretic feedback adjustment that relies on a single current detection temperature to a forward-looking adjustment based on the comparison results of future thermal trajectories and the confidence domain. Furthermore, the prediction accuracy is continuously improved through an online self-correction mechanism, thereby effectively improving the stability of the thread rolling machine heating process and the consistency of the quality of thread forming products.

[0051] Furthermore, according to a preset sampling period, multi-source process data during the workpiece heating process are collected synchronously. This multi-source process data includes at least the workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time. Specifically, the workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online dwell time are collected synchronously along a unified time axis, resulting in a raw data set arranged in a time series, as follows: ; In the formula, for Multi-source process data collected in real time; for The workpiece exit temperature at that moment; for The output power of the induction heater at any given time; for The workpiece feeding speed at any given time; for The online dwell time of the workpiece at any given moment.

[0052] Furthermore, the multi-source process data are time-aligned according to a unified timeline to eliminate timing deviations caused by different data acquisition sources, resulting in a set of process data arranged in time sequence.

[0053] Based on the process data set, at each sampling time, the corresponding workpiece outlet temperature, induction heater output power, workpiece feeding speed, and workpiece online residence time are combined into a thermal node, represented as: ; In the formula, for The hot state node corresponding to the given moment.

[0054] The hot-state nodes at multiple consecutive moments are arranged in chronological order to obtain the hot-state trajectory, which is represented as: ; In the formula, The constructed thermal trajectory; They are respectively The hot state node corresponding to the given moment.

[0055] Furthermore, based on the slope direction, rate of change, and fluctuation characteristics of the temperature change over time within a preset sliding window, the morphological segment type of the preset sliding window is identified. The morphological segment type includes at least one of the following: rapid heating segment, stable holding segment, thermal drift segment, fluctuating oscillation segment, thermal recovery segment, and thermal decay segment. Specifically, the rapid heating segment refers to a segment where the temperature within the window exhibits an approximately linear or approximately exponential growth trend over time, with the heating slope remaining at a consistently high level; the stable holding segment refers to a segment where the temperature change amplitude within the window is small, and the temperature curve tends to be horizontal; the thermal drift segment refers to a segment where the temperature within the window exhibits a slow, monotonous deviation from the target value, and the rate of change is lower than that of the rapid heating segment; the fluctuating oscillation segment refers to a segment where the temperature within the window exhibits high-frequency reciprocating fluctuations, with the fluctuation amplitude exceeding a preset threshold; the thermal recovery segment refers to a segment where the temperature within the window gradually returns to the target value after a brief deviation; and the thermal decay segment refers to a segment where the temperature within the window exhibits a continuous downward trend.

[0056] For the identified morphological segment types, morphological parameters are extracted. These parameters include at least one of the following: heating slope, curvature, rate of change of curvature, peak density, plateau duration, fluctuation frequency, thermal inertia response time, and thermal drift velocity. Specifically, the heating slope characterizes the rate of temperature change over time within the morphological segment; curvature characterizes the degree of curvature curvature within the segment; the rate of change of curvature characterizes the speed of curvature change over time, used to capture the transition trend of the trajectory morphology from one segment to another; peak density characterizes the number of local extrema per unit time within the segment, used to quantify the intensity of fluctuations and oscillations; plateau duration characterizes the duration of the stable holding segment in the time dimension, used to determine whether the workpiece has received sufficient heat treatment time; fluctuation frequency characterizes the frequency of temperature cyclical changes in the oscillating segment; thermal inertia response time characterizes the time required for the temperature response to reach a new steady state after a step change in heating power, used to reflect the magnitude of thermal inertia of the equipment and workpiece system; and thermal drift velocity characterizes the rate at which the temperature deviates from the target value in the thermal drift segment, used to quantify the severity of the drift trend.

[0057] The morphological parameters are encoded according to a preset order to obtain the thermal trajectory morphological feature vector, which is represented as follows: ; In the formula, This is the feature vector of the thermal trajectory morphology; The slope of the heating process; For curvature; The rate of change of curvature; Peak density; For the duration of the platform; The fluctuation frequency; This refers to the thermal inertia response time. This represents the thermal drift velocity.

[0058] Furthermore, using the thermal trajectory morphology feature vectors of the current moment and historical moments within a preset backtracking time as input, the morphological segment types and corresponding thermal trajectory morphology feature vectors of the workpiece within a preset future time step are predicted, resulting in a future thermal trajectory morphology feature vector prediction sequence, expressed as: ; In the formula, The number of historical moments corresponding to the preset backtracking duration; These are the thermal trajectory morphology feature vectors for the current moment and historical moments within a preset backtracking time, respectively. Preset number of time steps This refers to the predicted thermal trajectory morphology feature vector within a preset time step in the future, which is the future thermal trajectory morphology feature vector prediction sequence.

[0059] For example, the preset threshold for identifying the stable holding segment can be set to 2°C, that is, when the temperature change within the sliding window does not exceed 2°C, it is determined to be the stable holding segment; the preset threshold for identifying the fluctuating oscillation segment can be set to 3°C, that is, when the temperature fluctuation within the sliding window exceeds 3°C, it is determined to be the fluctuating oscillation segment. The above thresholds can be set according to the workpiece specifications and production cycle.

[0060] Furthermore, weighted sub-models are established for different morphological segments. Before weighted fusion, the physical quantities used in each morphological segment are converted into dimensionless intermediate variables according to a preset normalization method. Specifically, for physical quantities characterizing heat input features, normalization is performed according to the ratio of the physical quantity to a preset heat input feature reference value; for physical quantities characterizing temperature levels, normalization is performed according to the ratio of the physical quantity to a preset temperature reference value; for physical quantities characterizing time and rate, normalization is performed according to the corresponding preset time reference value and preset rate reference value, respectively. The above preset reference values ​​are all determined statistically based on thermal trajectory data formed during historical stable production processes and are updated online as production data continues to accumulate. The dimensionless intermediate variables obtained after normalization are weighted and fused according to preset weight coefficients to obtain the predicted output component of the corresponding morphological segment. This predicted output component is also a dimensionless quantity, used to characterize the predicted value of the corresponding component in the thermal trajectory morphological feature vector of the corresponding morphological segment, rather than directly adding physical quantities with different dimensions. For the rapid heating segment, the predicted output component is obtained by weighted fusion of the normalized dimensionless value corresponding to the cumulative heat input characteristics of this segment and the normalized dimensionless value corresponding to the temperature level in the middle of the trajectory, expressed as: ; In the formula, This refers to the predicted output component corresponding to the rapid heating phase. The cumulative heat input within this section is obtained by integrating the output power of the induction heater over the residence time. This is the characteristic time of the thermal inertial response in this section; This represents the median temperature level within that section. , These are the minimum and maximum temperatures within that section, respectively. , These are the preset weighting coefficients.

[0061] For the plateau duration, the predicted output component is obtained by weighted fusion of the normalized dimensionless value corresponding to the plateau duration and the normalized dimensionless value corresponding to the average temperature level of the plateau segment, expressed as: ; In the formula, The platform retains the predicted output components corresponding to the segment. For the duration of the platform; This refers to the average temperature within the platform section; , These are the preset weighting coefficients.

[0062] For the oscillating segment, the predicted output component is obtained by weighted fusion of the oscillation frequency and the inverse of the peak density, expressed as: ; In the formula, This is the predicted output component corresponding to the oscillation segment. The fluctuation frequency; Peak density; , These are the preset weighting coefficients.

[0063] For the thermal drift segment, the predicted output component is obtained by weighted fusion of the normalized dimensionless value corresponding to the drift velocity and the normalized dimensionless value corresponding to the thermal inertia response time, expressed as: ; In the formula, This is the predicted output component corresponding to the thermal drift segment; This refers to thermal drift velocity; This refers to the thermal inertia response time. , These are preset weighting coefficients. All of these weighting coefficients are obtained through regression analysis of thermal trajectory data accumulated under historical stable production conditions, and are updated online iteratively as production data continues to accumulate. The aforementioned preset reference values ​​used for normalization—namely, heat input characteristic reference values, temperature reference values, time reference values, and rate reference values—are taken as the statistical averages of the corresponding physical quantities formed during historical stable production processes. These are updated online as production data continues to accumulate, thereby ensuring that all intermediate variables after normalization are within the same order of magnitude of the dimensionless range, avoiding the direct addition of physical quantities with different dimensions.

[0064] For example, by performing regression analysis on historical stable production data, , The values ​​can be 0.6 and 0.4 respectively. , You can take 0.5 and 0.5 respectively. , You can take 0.5 and 0.5 respectively. , The values ​​can be 0.6 and 0.4 respectively. The sum of the weighting coefficients corresponding to the same predicted output component is usually 1. The specific values ​​are updated online through continuous accumulation of production data.

[0065] Based on the pre-defined correspondence between the feature vectors of the hot trajectory morphology and the hot nodes, the predicted sequence of the future hot trajectory morphology feature vectors is mapped to future hot nodes. Multiple consecutive future hot nodes are then arranged in chronological order to obtain the predicted sequence of the future hot trajectory, as follows: ; In the formula, Predicted sequences for future thermal trajectories; These are the future hot nodes obtained by mapping the future hot trajectory morphology feature vectors at the corresponding time according to a preset correspondence.

[0066] Furthermore, thermal trajectory data generated during historical stable production processes are obtained, and the range of possible locations for different workpiece specifications and material types in the thermal trajectory is statistically analyzed.

[0067] Based on the statistically obtained location range, a reliable domain for the thermal trajectory corresponding to the workpiece specifications and material types is constructed. The reliable domain for the thermal trajectory also includes at least one of the following: platform boundary, drift tolerance, and fluctuation tolerance. The upper boundary and lower boundary of the trajectory are time-varying envelope curves that dynamically change with time and morphological segment type, used to characterize the upper and lower temperature limits that the thermal trajectory can reach at each moment under normal production conditions for the corresponding workpiece specifications and material types. The drift tolerance is used to characterize the maximum allowable deviation rate in the drift segment. The fluctuation tolerance is used to characterize the maximum allowable fluctuation amplitude in the oscillation segment.

[0068] Specifically, for the same workpiece specification and material type, at the same morphological segment type and relative time position, the temperature values ​​at corresponding moments are extracted from multiple thermal trajectories formed during historical stable production processes, resulting in a historical temperature sample set at that moment. The mean and standard deviation of this historical temperature sample set are calculated. The mean plus a preset multiple of the standard deviation is used as the upper boundary value of the trajectory at that moment, and the mean minus a preset multiple of the standard deviation is used as the lower boundary value of the trajectory at that moment. The above statistical process is repeated sequentially for different morphological segment types and different relative time positions to obtain the trajectory upper and lower boundaries that dynamically change with time and morphological segment type, thus completing the construction of the trajectory upper and lower boundaries in the reliability domain of the thermal trajectory. The platform boundary is obtained by adding and subtracting a preset multiple of the standard deviation from the mean of the historical temperature samples corresponding to the stable holding segment; the drift tolerance is obtained by the preset quantile of the thermal drift velocity samples corresponding to the thermal drift segment during historical stable production processes; and the fluctuation tolerance is obtained by the preset quantile of the temperature fluctuation amplitude samples corresponding to the fluctuating oscillation segment during historical stable production processes. For example, the preset multiple can be 3, that is, the boundary is the mean plus or minus 3 times the standard deviation; the preset quantiles corresponding to the drift tolerance and fluctuation tolerance can be the 95th quantile. The specific values ​​can be set according to the workpiece specifications, material types and production process requirements, and are periodically updated as historical data is continuously accumulated.

[0069] Furthermore, the following offsets are calculated: the positional offset between the predicted future thermal trajectory sequence and the center line of the thermal trajectory confidence domain; the morphological offset of the predicted future thermal trajectory sequence relative to the morphological features of the target thermal trajectory; the plateau offset of the duration and temperature level of the segment in the predicted future thermal trajectory sequence where the temperature change amplitude is less than a preset amplitude threshold relative to the thermal trajectory confidence domain; the oscillation offset of the predicted future thermal trajectory sequence relative to the fluctuation tolerance; and the drift direction offset of the predicted future thermal trajectory sequence relative to the drift tolerance.

[0070] In addition, it also includes the time when the future thermal trajectory prediction sequence enters the thermal trajectory confidence domain and the time when it leaves the thermal trajectory confidence domain, which is used to determine the crossing time of the future thermal trajectory prediction sequence relative to the boundary of the thermal trajectory confidence domain in the future time period, and to provide a basis for generating the intervention timing of the heating control strategy.

[0071] The position offset, shape offset, platform offset, oscillation offset, and drift direction offset are weighted and fused according to preset weights to obtain the trajectory offset index, which represents the degree of deviation of the future hot trajectory prediction sequence from the overall confidence domain of the hot trajectory. It is expressed as: ; In the formula, This is the trajectory offset index; This is the position offset; This is the shape offset. This is the platform offset; This is the oscillation offset; This represents the drift direction offset. , , , , These are preset weights corresponding to position offset, shape offset, platform offset, oscillation offset, and drift direction offset, respectively. For example, the preset weights for position offset, shape offset, platform offset, oscillation offset, and drift direction offset can be 0.3, 0.25, 0.2, 0.15, and 0.1, respectively, with the sum of each weight being 1. The specific values ​​can be adjusted according to the degree of attention paid to each offset dimension in different process scenarios.

[0072] Furthermore, if the predicted sequence of the future thermal trajectory leaves the upper boundary of the thermal trajectory confidence domain earlier than the preset target crossing time, the output power of the induction heater is reduced.

[0073] If the position offset is negative and its absolute value exceeds the preset threshold, the output power of the induction heater will be increased.

[0074] If the platform offset exceeds the preset threshold, the workpiece feeding speed will be reduced and the workpiece online dwell time will be extended.

[0075] If the oscillation offset exceeds the preset threshold, the filter compensation amount will be increased.

[0076] If the drift direction offset exceeds the preset threshold, the heating cycle is recalculated, and the heating sequence of subsequent workpieces is adjusted.

[0077] For example, the preset target crossing time can be set to 2 seconds before the workpiece is expected to enter the toothed plate; the preset threshold for judging the position offset can be set to 5℃; the preset threshold for judging the platform offset can be set to 3℃; the preset threshold for judging the oscillation offset can be set to 4℃; and the preset threshold for judging the drift direction offset can be set to 0.5℃ / s. Each of the above thresholds can be set according to the workpiece specifications, material type, and production process requirements.

[0078] The adjustment amount of the workpiece feeding speed is determined by the feeding speed correction prediction component. The feeding speed correction prediction component is obtained by weighted fusion of the normalized dimensionless value corresponding to the maximum value of the peak density in the corresponding morphological segment and the normalized dimensionless value corresponding to the reciprocal of the target feeding speed. Similarly, the two are converted into dimensionless intermediate variables according to the aforementioned normalization method before being weighted and fused, as expressed as: ; In the formula, Correct the predicted component for the feeding speed; This represents the maximum value of the peak density within the corresponding morphological segment; Target feeding speed; , These are preset weighting coefficients. For example, , You can take 0.5 and 0.5 respectively.

[0079] The adjustment amount of the induction heater output power is determined by the heating power correction prediction component. This component is obtained by weighted fusion of the normalized dimensionless value corresponding to the position offset and the ratio of the current induction heater output power to the rated power. The ratio of the current induction heater output power to the rated power is itself a dimensionless quantity. The position offset is converted to a dimensionless intermediate variable using the aforementioned normalization method before being weighted and fused with the variable, expressed as: ; In the formula, Correct the predicted component for heating power; This is the position offset; This represents the current output power of the induction heater. This refers to the rated power of the induction heater. , These are preset weighting coefficients. For example, , The values ​​can be 0.6 and 0.4 respectively.

[0080] Furthermore, if the predicted hot node is within the confidence domain of the hot trajectory, the workpiece is determined to meet the processing requirements, and the feeding is executed.

[0081] If the predicted hot node deviates from the credibility domain of the hot trajectory but the degree of deviation does not exceed the preset rejection threshold, the workpiece feeding is temporarily suspended, and the updated future hot trajectory prediction sequence is obtained in the next sampling period for re-judgment.

[0082] If the predicted hot node exceeds the upper boundary of the hot trajectory confidence domain, the workpiece is determined not to meet the processing requirements and is removed from the production process.

[0083] For example, the preset rejection threshold can be set to 10°C. That is, when the predicted hot node deviates from the upper or lower boundary of the hot trajectory confidence domain by no more than 10°C, the workpiece feeding is temporarily suspended and re-judged in the next sampling period; when the deviation exceeds 10°C, the workpiece is determined not to meet the processing requirements and is rejected.

[0084] Furthermore, the difference in morphological parameters between the actual thermal trajectory and the predicted sequence of future thermal trajectories is calculated to obtain the morphological deviation, which is expressed as: ; In the formula, This is due to morphological deviation; This is the morphological feature vector corresponding to the actual thermal trajectory; The predicted value of the morphological feature vector corresponding to the future thermal trajectory prediction sequence.

[0085] Based on the morphological deviation, the length of the preset sliding window and the segment determination threshold used to identify the morphological segment type are corrected.

[0086] For example, the segment determination threshold can be adjusted within ±20% based on the size of the morphological deviation from the original threshold, and the length of the preset sliding window can be adjusted accordingly within the range of 0.5s to 2s to adapt to changes in workpiece specifications, material batches and equipment operating status.

[0087] Based on the morphological deviation, the upper boundary of the trajectory, the lower boundary of the trajectory, the platform boundary, the drift tolerance, and the fluctuation tolerance are corrected.

[0088] Based on morphological deviation and positional offset, morphological offset, plateau offset, oscillation offset, and drift direction offset, the weighting coefficients used to predict the future thermal trajectory morphological feature vector prediction sequence and generate the heating control strategy are corrected. The weighting coefficients used must include at least the weighting coefficients used to predict the future thermal trajectory morphological feature vector prediction sequence. , , , , , , , The weighting coefficients used for the feed speed correction prediction component , And the weighting coefficients used for the heating power correction prediction component. , .

[0089] Figure 2 This is a schematic diagram of the heating control system for a wire rolling machine based on thermal trajectory prediction, provided as an embodiment of this application. Figure 2 As shown in the figure, an embodiment of this application provides a heating control system for a wire rolling machine based on thermal trajectory prediction. The system includes: a data acquisition module 201, used to synchronously acquire multi-source process data during the workpiece heating process according to a preset sampling period. The multi-source process data includes at least the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time.

[0090] The trajectory construction module 202 is connected to the data acquisition module 201 and is used to fuse multi-source process data in chronological order to construct a thermal trajectory that can describe the entire heating process of the workpiece. The thermal trajectory is composed of multiple consecutive thermal nodes arranged in chronological order. Each thermal node is formed by the combination of the workpiece outlet temperature, induction heater output power, workpiece feeding speed and workpiece online dwell time at the corresponding time.

[0091] The morphological feature extraction module 203 is connected to the trajectory construction module 202. It is used to scan the thermal trajectory along the time axis according to the preset sliding window, identify the morphological segment type to which the sliding window belongs, extract multiple morphological parameters that characterize the morphological segment type, encode the multiple morphological parameters, and obtain the thermal trajectory morphological feature vector.

[0092] The trajectory prediction module 204, connected to the morphological feature extraction module 203, is used to predict the evolution trend of the workpiece's thermal trajectory morphology over a future period based on the thermal trajectory morphological feature vectors of the current and historical moments, obtaining a future thermal trajectory morphological feature vector prediction sequence, and reconstructing and generating a future thermal trajectory prediction sequence based on the future thermal trajectory morphological feature vector prediction sequence. Specifically, the trajectory prediction module 204 establishes prediction sub-models with weighted coefficients for the rapid heating segment, the platform holding segment, the fluctuation oscillation segment, and the thermal drift segment, respectively. It converts physical quantities representing heat input characteristics, temperature level, platform duration, fluctuation frequency, peak density, drift velocity, and thermal inertia response time within the corresponding morphological segment into dimensionless intermediate variables according to a preset normalization method, and then performs weighted fusion to obtain the prediction output components corresponding to each morphological segment, thereby obtaining the future thermal trajectory morphological feature vector prediction sequence. It then maps the thermal trajectory morphological feature vectors to future thermal nodes according to a preset correspondence between the thermal trajectory morphological feature vectors and thermal nodes, obtaining the future thermal trajectory prediction sequence. The specific calculation method is consistent with the specific implementation method of the corresponding steps in the method section of this application.

[0093] The trusted domain construction module 205 is used to construct a corresponding trusted domain for thermal trajectories based on thermal trajectory data formed during historical production processes, taking into account workpiece specifications and material types. The trusted domain includes the upper and lower boundaries of the trajectory, which dynamically change with time and morphological segment type. Specifically, for the same workpiece specification, material type, morphological segment type, and relative time position, the trusted domain construction module 205 extracts the temperature values ​​at corresponding moments from multiple thermal trajectories formed during historical stable production processes to obtain a historical temperature sample set. The upper and lower boundaries of the trajectory at that moment are determined by adding or subtracting a preset multiple standard deviation from the mean of this historical temperature sample set. Drift tolerance and fluctuation tolerance are determined based on the preset quantiles of the corresponding historical rate and amplitude samples. The specific calculation method is consistent with the specific implementation method of the corresponding steps in the method section of this application.

[0094] The offset evaluation module 206 is connected to the trajectory prediction module 204 and the credibility domain construction module 205 respectively. It is used to compare the future hot trajectory prediction sequence with the hot trajectory credibility domain, calculate the offset of the future hot trajectory prediction sequence relative to the hot trajectory credibility domain, and obtain the trajectory offset index based on the offset.

[0095] The control strategy generation module 207 is connected to the offset evaluation module 206 and is used to generate a heating control strategy based on the trajectory offset index. The heating control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed. The control strategy generation module 207 is also connected to the induction heater and the feeding device respectively and is used to adjust at least one of the output power of the induction heater and the workpiece feeding speed according to the heating control strategy.

[0096] The material feeding decision module 208 is connected to the trajectory prediction module 204 and the credibility domain construction module 205 respectively. It is used to determine whether the predicted hot node corresponding to the expected entry time of the workpiece into the tooth plate is within the credibility domain of the hot trajectory based on the future hot trajectory prediction sequence. If it is, the material feeding is executed; otherwise, the material feeding is delayed or the material is rejected.

[0097] The online self-correction module 209 is connected to the morphological feature extraction module 203, the trust domain construction module 205, the trajectory prediction module 204, and the control strategy generation module 207, respectively. It is used to collect the actual hot trajectory of the workpiece after the thread rolling process, compare the actual hot trajectory with the corresponding future hot trajectory prediction sequence to obtain the morphological deviation, and correct the parameters used to construct the trust domain of the hot trajectory and generate the future hot trajectory prediction sequence based on the morphological deviation.

[0098] By working together through the above modules, the construction, prediction, confidence domain comparison and online self-correction of the thermal trajectory are realized, so that the heating control and material feeding decision are transformed from a lagging feedback adjustment to a forward predictive adjustment, thereby improving the stability of the wire rolling process and the consistency of product quality.

[0099] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A heating control method for a wire rolling machine based on thermal trajectory prediction, characterized in that, include: According to the preset sampling period, multi-source process data during the workpiece heating process are collected synchronously. The multi-source process data includes at least the workpiece outlet temperature, induction heater output power, workpiece feeding speed and workpiece online residence time. The multi-source process data are fused in chronological order to construct a thermal trajectory that can describe the entire heating process of the workpiece. The thermal trajectory is composed of multiple consecutive thermal nodes arranged in chronological order. Each thermal node is formed by a combination of the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time at the corresponding time. The thermal trajectory is scanned along the time axis according to a preset sliding window, the morphological segment type to which the sliding window belongs is identified, and multiple morphological parameters characterizing the morphological segment type are extracted. The multiple morphological parameters are encoded to obtain the thermal trajectory morphological feature vector. Based on the thermal trajectory morphology feature vectors at the current and historical times, the evolution trend of the thermal trajectory morphology of the workpiece in the future period is predicted to obtain the future thermal trajectory morphology feature vector prediction sequence, and the future thermal trajectory prediction sequence is reconstructed based on the future thermal trajectory morphology feature vector prediction sequence. Based on the thermal trajectory data generated during historical production, a corresponding thermal trajectory confidence domain is constructed for workpiece specifications and material types. The thermal trajectory confidence domain includes the upper and lower boundaries of the trajectory that dynamically change with time and the type of the morphological segment. The predicted future hot trajectory sequence is compared with the confidence region of the hot trajectory, the offset of the predicted future hot trajectory sequence relative to the confidence region of the hot trajectory is calculated, and the trajectory offset index is obtained based on the offset. Based on the trajectory offset index, a heating control strategy is generated, which includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed. Based on the predicted future thermal trajectory sequence, determine whether the predicted thermal node corresponding to the expected entry time of the workpiece into the toothed plate is within the range of the thermal trajectory confidence domain. If so, execute feeding; otherwise, execute delayed feeding or rejection. The actual thermal trajectory of the workpiece after the thread rolling process is collected, and the actual thermal trajectory is compared with the corresponding future thermal trajectory prediction sequence to obtain the morphological deviation. Based on the morphological deviation, the parameters used to construct the credibility domain of the thermal trajectory and generate the future thermal trajectory prediction sequence are corrected.

2. The method according to claim 1, characterized in that, The step of fusing the multi-source process data in chronological order to construct a thermal trajectory that can describe the entire workpiece heating process includes: The multi-source process data is time-aligned according to a unified timeline to eliminate timing deviations caused by different data acquisition sources, resulting in a set of process data arranged in time sequence. Based on the process data set, at each sampling time, the workpiece outlet temperature, the induction heater output power, the workpiece feeding speed, and the workpiece online dwell time at the corresponding time are combined into the hot state node, and the hot state nodes at multiple consecutive times are arranged in chronological order to obtain the hot state trajectory.

3. The method according to claim 1, characterized in that, The process involves scanning the thermal trajectory along the time axis using a preset sliding window, identifying the morphological segment type to which the sliding window belongs, extracting multiple morphological parameters characterizing the morphological segment type, and encoding these multiple morphological parameters to obtain a thermal trajectory morphological feature vector, including: Based on the slope direction, rate of change and fluctuation characteristics of the temperature change over time within the preset sliding window, the morphological segment type to which the preset sliding window belongs is identified. The morphological segment type includes at least one of the following: heating segment, stable segment, drift segment and oscillation segment. For the identified morphological segment type, the morphological parameters are extracted, and the morphological parameters include at least one of heating slope, curvature, peak density and fluctuation frequency. The morphological parameters are encoded in a preset order to obtain the thermal trajectory morphological feature vector.

4. The method according to claim 1, characterized in that, The process of predicting the evolution trend of the workpiece's thermal trajectory morphology over a future period based on the thermal trajectory morphology feature vectors at the current and historical times, obtaining a future thermal trajectory morphology feature vector prediction sequence, and reconstructing a future thermal trajectory prediction sequence based on the future thermal trajectory morphology feature vector prediction sequence includes: Using the thermal trajectory morphology feature vectors of the current time and the historical time within the preset backtracking time as input, the morphology segment type and the corresponding thermal trajectory morphology feature vectors of the workpiece within the future preset time step are predicted to obtain the future thermal trajectory morphology feature vector prediction sequence. According to the preset correspondence between the hot trajectory morphology feature vector and the hot node, the future hot trajectory morphology feature vector prediction sequence is mapped to the future hot node, and multiple consecutive future hot nodes are arranged in chronological order to obtain the future hot trajectory prediction sequence.

5. The method according to claim 1, characterized in that, Based on the thermal trajectory data generated during historical production processes, a corresponding thermal trajectory reliability domain is constructed for each workpiece specification and material type, including: Acquire the thermal trajectory data formed during historical stable production processes, and for different workpiece specifications and different material types, statistically analyze the allowable position range of the morphological segment type in the thermal trajectory; Based on the statistically obtained location range, a thermal trajectory confidence domain corresponding to the workpiece specifications and material type is constructed. The thermal trajectory confidence domain also includes at least one of platform boundary, drift tolerance, and fluctuation tolerance.

6. The method according to claim 5, characterized in that, The step of comparing the predicted future hot trajectory sequence with the confidence region of the hot trajectory, calculating the offset of the predicted future hot trajectory sequence relative to the confidence region of the hot trajectory, and obtaining the trajectory offset index based on the offset includes: The offset is calculated as follows: the positional offset between the future thermal trajectory prediction sequence and the center line of the thermal trajectory confidence domain; the morphological offset of the future thermal trajectory prediction sequence relative to the target thermal trajectory morphological features; the duration and temperature level of the segment in the future thermal trajectory prediction sequence where the temperature change amplitude is less than a preset amplitude threshold relative to the thermal trajectory confidence domain; the oscillation offset of the future thermal trajectory prediction sequence relative to the fluctuation tolerance; and the drift direction offset of the future thermal trajectory prediction sequence relative to the drift tolerance. The position offset, shape offset, platform offset, oscillation offset, and drift direction offset are weighted and fused according to preset weights to obtain the trajectory offset index, which characterizes the degree of deviation of the future hot trajectory prediction sequence from the overall credibility domain of the hot trajectory.

7. The method according to claim 6, characterized in that, The heating control strategy is generated based on the trajectory offset index. The heating control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed, including: If the predicted future thermal trajectory sequence leaves the upper boundary of the thermal trajectory confidence domain earlier than the preset target crossing time, then the output power of the induction heater is reduced. If the position offset is negative and its absolute value exceeds a preset threshold, then the output power of the induction heater is increased. If the platform offset exceeds a preset threshold, the workpiece feeding speed is reduced and the workpiece online dwell time is extended. If the oscillation offset exceeds a preset threshold, the filter compensation amount is increased; If the drift direction offset exceeds a preset threshold, the heating cycle is recalculated, and the heating sequence of subsequent workpieces is adjusted.

8. The method according to claim 1, characterized in that, Based on the predicted future thermal trajectory sequence, the step of determining whether the predicted thermal node corresponding to the expected entry time of the workpiece into the toothed plate is within the confidence region of the thermal trajectory is performed. If so, feeding is executed; otherwise, delayed feeding or rejection is performed. If the predicted hot state node is located within the confidence domain of the hot state trajectory, the workpiece is determined to meet the processing requirements, and the material feeding is executed. If the predicted hot node deviates from the credibility region of the hot trajectory but the degree of deviation does not exceed the preset rejection threshold, the workpiece feeding is temporarily suspended, and the updated future hot trajectory prediction sequence is obtained in the next sampling period for re-judgment. If the predicted hot node exceeds the upper boundary of the trajectory of the hot trajectory confidence domain, it is determined that the workpiece does not meet the processing requirements and the workpiece is removed from the production process.

9. The method according to claim 6, characterized in that, The actual hot trajectory of the workpiece after thread rolling is collected, and the actual hot trajectory is compared with the corresponding future hot trajectory prediction sequence to obtain the shape deviation. Based on the shape deviation, the parameters used to construct the confidence region of the hot trajectory and generate the future hot trajectory prediction sequence are corrected, including: The morphological deviation is obtained by calculating the difference between the actual thermal trajectory and the predicted future thermal trajectory sequence on the morphological parameters. Based on the morphological deviation, the length of the preset sliding window and the segment determination threshold used to identify the morphological segment type are corrected; Based on the morphological deviation, the upper boundary of the trajectory, the lower boundary of the trajectory, the platform boundary, the drift tolerance, and the fluctuation tolerance are corrected; Based on the morphological deviation, the position offset, the morphological offset, the platform offset, the oscillation offset, and the drift direction offset, the prediction sequence of the future thermal trajectory morphological feature vector and the weighting coefficients used to generate the heating control strategy are corrected.

10. A heating control system for a wire rolling machine based on thermal trajectory prediction, characterized in that, include: The data acquisition module is used to synchronously acquire multi-source process data during the workpiece heating process according to a preset sampling period. The multi-source process data includes at least the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time. The trajectory construction module is used to fuse the multi-source process data in chronological order to construct a thermal trajectory that can describe the entire heating process of the workpiece. The thermal trajectory is composed of thermal nodes at multiple consecutive moments arranged in chronological order. Each thermal node is formed by a combination of the workpiece outlet temperature, the output power of the induction heater, the workpiece feeding speed, and the workpiece online dwell time at the corresponding moment. The morphological feature extraction module is used to scan the thermal trajectory along the time axis according to a preset sliding window, identify the morphological segment type to which the sliding window belongs, extract multiple morphological parameters that characterize the morphological segment type, encode the multiple morphological parameters, and obtain the thermal trajectory morphological feature vector. The trajectory prediction module is used to predict the evolution trend of the thermal trajectory shape of the workpiece in the future period based on the thermal trajectory shape feature vector at the current time and the historical time, to obtain the future thermal trajectory shape feature vector prediction sequence, and to reconstruct and generate the future thermal trajectory prediction sequence based on the future thermal trajectory shape feature vector prediction sequence. The trusted domain construction module is used to construct a corresponding trusted domain for thermal trajectory based on the thermal trajectory data formed during the historical production process, and for the workpiece specifications and material types. The trusted domain for thermal trajectory includes the upper boundary and lower boundary of the trajectory that dynamically change with time and the type of the morphological segment. The offset evaluation module is used to compare the future hot trajectory prediction sequence with the hot trajectory confidence region, calculate the offset of the future hot trajectory prediction sequence relative to the hot trajectory confidence region, and obtain the trajectory offset index based on the offset. A control strategy generation module is used to generate a heating control strategy based on the trajectory offset index. The heating control strategy includes adjusting at least one of the output power of the induction heater and the workpiece feeding speed. The material feeding decision module is used to determine whether the predicted hot node corresponding to the expected entry time of the workpiece into the toothed plate is within the range of the hot trajectory confidence domain, based on the future hot trajectory prediction sequence. If it is, the material feeding is executed; otherwise, the material feeding is delayed or the material is rejected. The online self-correction module is used to collect the actual hot trajectory of the workpiece after the thread rolling process, compare the actual hot trajectory with the corresponding future hot trajectory prediction sequence to obtain the morphological deviation, and correct the parameters used to construct the confidence domain of the hot trajectory and generate the future hot trajectory prediction sequence based on the morphological deviation.

Images