Temperature correction method and correction system for hot press

Abstract

The invention discloses a temperature correction method and system for a hot press, and particularly relates to the technical field of hot press temperature control, and the method comprises the steps: obtaining and preprocessing multi-dimensional operation data, including a temperature time sequence, pressure setting and material attributes; constructing a coupling model based on the multilayer neural network, and predicting temperature distribution of different process stages; performing sensor drift compensation, analyzing temperature deviation characteristics, judging whether sensor drift or material thermal diffusion abnormity exists or not, and generating a local distortion mark; determining a compensation mode according to the distortion mark, and starting a local compensation mode or a global balance control mode if the distortion is recognized as over-limit distortion; and comparing with the output of the coupling model, verifying the temperature correction effect, and if the deviation is reduced and the distortion mark is converged, outputting a stable temperature correction result. The problems of temperature distortion and sensor drift are effectively solved, and the long-term precision and stability of the hot press under complex working conditions are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature control of a hot press, and more specifically, to a temperature correction method and correction system for a hot press. Background Art

[0002] Hot presses are widely used in the manufacturing process of multi-layer composite materials. Their key function is to achieve uniform heating and tight bonding of materials by precisely controlling the temperature, pressure and pressing time of the heating plate to meet the strength, dimensional accuracy and performance requirements of different processes. In actual production, the heating area of ​​the hot press is equipped with multiple temperature sensors to monitor the temperature distribution on the surface of the heating plate in real time, and dynamically adjust the heating power or cooling through a feedback mechanism to achieve temperature control. However, with the increase in process complexity and the diversification of material types, higher requirements are placed on the accuracy and response speed of temperature control. Traditional single sensor monitoring and static adjustment strategies are difficult to adapt to dynamic needs in complex scenarios.

[0003] Existing hot press temperature control technology has multiple problems in long-term operation or complex working conditions. On the one hand, sensor drift may lead to the accumulation of measurement errors, which in turn affects the accuracy of temperature feedback; on the other hand, in the hot pressing process of multi-layer composite materials, there is a high degree of coupling between temperature, pressure and material properties. If only simple constant parameter control is used, it is often difficult to deal with local temperature distortion and sensor drift problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a temperature calibration method and a calibration system for a hot press to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a temperature calibration method for a hot press, comprising the following steps: Step 1, obtaining multidimensional operation data to be corrected: obtaining multidimensional operation data generated by the hot press in the process of pressing the multi-layer composite material, the multidimensional operation data including the hot press temperature time series data, pressure setting parameters and material property information; denoising and interpolating the multidimensional operation data, and outputting a preprocessed data set; Step 2: Construct a coupling model: Identify the coupling relationship between temperature, pressure, and material in the preprocessing data set, and construct a coupling model based on a multi-layer neural network. The coupling model outputs the predicted temperature distribution at different process stages. Step 3: Perform sensor drift compensation and obtain local distortion mark: According to the difference between the temperature prediction distribution of the coupling model and the actual temperature distribution, extract the deviation characteristics, and judge whether there is sensor drift or material thermal diffusion anomaly by analyzing the deviation characteristics. Feed the compensated temperature correction value back to the coupling model to correct the sensor drift, and perform local distortion mark based on the sensor drift compensation result. The local distortion mark includes position coordinates, duration and deviation amplitude. Step 4. Determine the temperature correction mode according to the local distortion mark: Analyze the local distortion mark to determine whether it is necessary to start the compensation mode or the global balance control mode; if it is identified as over-limit distortion, start the compensation mode and implement power adjustment or auxiliary cooling in the corresponding heating section; otherwise, enter the global balance control mode to coordinately fine-tune the pressure and temperature parameters of the hot press; Step 5. Verification after temperature correction: After the local compensation mode or the global balance control mode, a new temperature measurement value is obtained and compared with the coupling model output. If the deviation is reduced and the distortion mark is in a convergence state, the compensation in step 4 is considered to be effective and the temperature correction result is output; otherwise, the compensation is invalid.

[0006] Preferably, the coupling model is constructed as follows: The pressure, material characteristic vectors and start-stop time series are selected as input to build a coupling model of the hot press; Based on the mutual information theory, key variables with high coupling to temperature changes are screened; A multi-layer neural network is used to perform nonlinear fitting on key variables and output the predicted temperature; Improve the accuracy of temperature prediction and suppress overfitting by minimizing the loss function; The loss function L(θ) of the coupling model satisfies the following formula: Among them, θ is the coupling model parameter, specifically the set of weights or coefficients to be trained inside the coupling model, including the weight matrix of the neural network and the kernel function coefficients of the support vector machine. θ is updated through an iterative training algorithm (such as gradient descent or Adam); M is the number of data points, indicating the number of samples or measurement moments included in the training set, and i is the sequence number; x i refers to the input vector at the ith sample or measurement moment; Represents the predicted temperature, given an input vector x i and the output results under the coupling model parameter θ; represents the reference temperature. Given an input vector x i The output results below come from temperature sensor networks, thermal imager measurements, or experimental calibration values; is a balance factor, balancing the model fitting error and model complexity, and controlling the regularization term the degree of importance attached to is a regularization term to suppress overfitting.

[0007] Preferably, the coupling model compares the predicted temperature with the measured temperature point by point during the training phase. If the prediction error or mutual information deviation in the local area is continuously higher than a preset threshold, it is marked as a deviation and the deviation feature is extracted; the preset threshold represents the standard for determining whether the temperature deviation exceeds the allowable range; Analyze the deviation characteristics. If the cumulative time that the temperature prediction distribution of the coupling model and the measured temperature distribution show deviation at a single type of interface exceeds a threshold, it can be determined that there is an abnormal thermal diffusion problem at the material interface. If there is abnormal thermal diffusion at the material interface, the user will be warned of process abnormalities and prompted to adjust process parameters; If the difference between the temperature reading of a specific sensor and the predicted temperature value of the adjacent normal sensor and the coupled model is found to be continuously beyond the normal variation range through local temperature deviation analysis and continuous monitoring, it is recorded as an abnormal drift sensor; Calculate the weighted difference between the adjacent normal sensor readings and the coupling model prediction value to obtain the compensation coefficient k of the jth sensor j , the compensation coefficient range is between 0 and 1; then k j Applied to abnormal drift sensor data: in, Represents the corrected temperature value of the jth sensor, which is used to replace the original measured temperature; represents the original measured temperature of the jth sensor; represents the temperature predicted by the coupled model; Indicates that k j The corresponding complementary factor is used to balance the weighted ratio of the measured temperature and the predicted temperature.

[0008] Preferably, the process of analyzing the local distortion mark and determining whether to start the compensation or global balance control mode includes: Set dynamic distortion threshold: Define a dynamic distortion threshold that decays over time based on the deviation amplitude provided by the local distortion marker , in order to distinguish short-term fluctuations from continuous distortion, the dynamic distortion threshold Satisfies the following formula: in, is the initial threshold, represents the decay rate, is the accumulated time since the last compensation was completed; if the deviation of the heating area exceeds , it is judged as local distortion; Identify excessive distortion: If the local distortion persists within the time window and the spatial range exceeds the preset proportion, it is marked as excessive distortion; Determine the temperature correction mode: If over-limit distortion is detected, the compensation mode is activated to implement rapid power adjustment or auxiliary cooling in the corresponding heating section; otherwise, the global balance control mode is activated to make linkage fine adjustments to the overall pressure and temperature of the hot press.

[0009] Preferably, the local compensation mode includes: Obtaining a local compensation control signal: defining a proportional gain coefficient and an integral gain coefficient for the region where the over-limit distortion mark is located, calculating the local compensation control signal through a local compensation control signal model, using s to represent the sequence number of the over-limit distortion region, and the local compensation control signal model satisfies the following formula; in, A local compensation control signal representing a heating or cooling adjustment amount performed on the s-th over-limit distortion region at time k; The temperature deviation of the sth out-of-limit distortion region at time k is If it is positive, it means that the actual temperature is higher than the target and cooling is required. If it is negative, it means that the actual temperature is lower than the target and heating is required; is the proportional gain coefficient of the sth over-limit distortion region, is the integral gain coefficient of the sth over-limit distortion region, In the time interval The integral of the temperature deviation is τ, and τ is the integral variable.

[0010] Preferably, the starting of the global balance control mode includes: Coupling input correction: Based on the initial input of the coupling model, increase the temperature correction , the temperature correction value satisfies the following formula: in, is the balance coefficient, which controls the influence of the corresponding global correction on the overall heating or cooling of the system; is the temperature deviation at the jth sensor; N is the number of sensors, and η is the regularization constant to avoid numerical instability caused by the denominator tending to zero; Use temperature correction to adjust power and pressure in conjunction: The temperature correction is decomposed into the adjustment signals for the heating power W and the pressure P, ΔW is the adjustment of the heating power, ΔP is the adjustment of the pressure; Among them, k W , k P , are the correction coefficients for heating power and pressure, respectively, which are adjusted according to the experimental or process requirements; if the temperature correction , increase the heating power, increase the pressure at the same time, accelerate the temperature rise and improve the thermal contact efficiency; if the temperature correction amount , reduce heating power, reduce pressure, delay temperature rise and reduce heat accumulation; After the above adjustments are completed, the real-time temperature distribution is obtained again through the sensor, and the new temperature is input into the coupling model for verification. If the deviation is not significantly reduced, the correction coefficients of heating power and pressure are further increased in the next iteration cycle.

[0011] Preferably, when the hot press is in a state of frequent start-stop, the iteration cycle of the local compensation mode and the global balance control is automatically shortened, so that the global adjustment and the local compensation are performed alternately at a higher frequency to adapt to sudden temperature fluctuations or material changes.

[0012] Preferably, the method also includes a step of iterative correction based on adaptive optimization, if the local deviation characteristics continue to exceed the dynamic distortion threshold after compensation, the regularization term and compensation gain parameters of the coupling model are adjusted, and the compensation gain parameters include a proportional gain coefficient and an integral gain coefficient; adaptive optimization is performed to update the coupling model; based on the optimized coupling model, the effectiveness of the compensation strategy is re-determined; if no over-limit distortion occurs after multiple rounds of iterations, a stable temperature correction result is output.

[0013] Preferably, the coupling model is trained based on the adjusted loss function, the optimized coupling model is output, and the proportional gain coefficient and the integral gain coefficient are reset according to the test data; The optimized coupling model and compensation gain parameters will affect the next step of temperature prediction and compensation operations.

[0014] To achieve the purpose of the present invention, a temperature correction system for a hot press is provided, comprising: The multi-dimensional data acquisition and pre-processing module is used to obtain the multi-dimensional operation data of the hot press, denoise the acquired data, use the interpolation algorithm to fill the missing values, format the processed data into a time series, and output the pre-processed data set; Coupling model building module: Based on the preprocessing data set, identify the coupling relationship between temperature, pressure, and material properties; train the coupling model through a multi-layer neural network to generate a coupling model that can predict the temperature distribution at different process stages, input the preprocessing data set into the coupling model, and output the corresponding temperature prediction distribution; Deviation feature extraction and marking module: Calculate the difference and extract local deviation features, including deviation amplitude, duration and spatial distribution, based on the predicted temperature distribution and the actual measured temperature distribution; determine whether the source of distortion is related to sensor drift or abnormal material thermal diffusion. If it is sensor drift, correct the sensor output by calculating the compensation coefficient; generate local distortion marks, which include position coordinates and key information of deviation attributes; Compensation strategy determination and execution module: analyzes local distortion marks. If it is identified as over-limit distortion, the compensation mode is activated to adjust the heating power or auxiliary cooling of the distortion area; otherwise, the global balance control mode is entered to make linkage fine-tuning of the pressure and temperature parameters of the hot press; Correction result verification and iteration module: record the temperature measurement value after compensation and compare it with the predicted value of the coupling model; if the deviation is reduced and the local distortion mark converges, output a stable temperature correction result; otherwise, return to the deviation feature extraction and marking module to regenerate the local distortion mark, and iteratively optimize the coupling model and compensation strategy.

[0015] Technical effects and advantages of the present invention: (1) The temperature correction method for a hot press provided by the present invention is based on the coupling relationship between temperature, pressure and material properties. By adopting a multi-layer neural network, a corresponding coupling model is constructed, which can predict the temperature distribution in real time and extract local deviation characteristics in combination with the measured temperature data, accurately determine the sensor drift and the abnormal thermal diffusion of the material interface, and significantly improve the data reliability and deviation traceability capability; the compensation strategy based on the combination of the dynamic distortion threshold and the compensation mode can realize local rapid temperature rise and fall and global coordinated balance control, effectively improve the temperature uniformity and production efficiency, and effectively solve the problem of inaccurate temperature control caused by sensor drift and abnormal thermal diffusion of materials in traditional hot presses.

[0016] (2) The temperature correction method for the hot press provided by the present invention enhances the adaptability to sudden temperature fluctuations and material changes under frequent start-stop conditions by shortening the control iteration cycle and optimizing real-time feedback. The long-term stability of the correction result is ensured through sensor drift correction and inertia correction strategies, thus providing reliable guarantee for high-precision temperature control of the hot press in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a flow chart of the temperature correction method of the hot press of the present invention.

[0018] Figure 2 This is a flow chart of the adaptive optimization iterative correction of the present invention. DETAILED DESCRIPTION

[0019] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0020] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0021] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.

[0022] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0023] Example 1, see Figure 1 The present invention provides a method for calibrating the temperature of a hot press, comprising the following steps: Step 1, obtaining multidimensional operation data to be corrected: obtaining multidimensional operation data generated by the hot press during the pressing of multi-layer composite materials, the multidimensional operation data including hot press temperature time series data, pressure setting parameters and material property information; denoising and interpolating the multidimensional operation data to form a preprocessed data set for temperature coupling analysis; Step 2: Construct a coupling model: Identify the coupling relationship between temperature, pressure, and material in the preprocessing data set, and construct a coupling model based on a multi-layer neural network. The coupling model outputs the predicted temperature distribution at different process stages. Step 3: Perform sensor drift compensation and obtain local distortion mark: According to the difference between the temperature prediction distribution of the coupling model and the actual temperature distribution, extract the deviation characteristics, and judge whether there is sensor drift or material thermal diffusion anomaly by analyzing the deviation characteristics. Feed the compensated temperature correction value back to the coupling model to correct the sensor drift, and perform local distortion mark based on the sensor drift compensation result. The local distortion mark includes position coordinates, duration and deviation amplitude. Explanation: The deviation characteristics refer to the multi-dimensional characteristics of the difference between the predicted temperature distribution and the actual temperature distribution, which include not only the numerical value of the deviation (i.e., the deviation amplitude), but also the persistence of the deviation (e.g., whether the deviation exists for a long time or changes continuously), spatial distribution (e.g., whether the deviation is concentrated in a certain heating area or presents a specific spatial pattern), and temporal changes; Step 4. Determine the temperature correction mode according to the local distortion mark: Analyze the local distortion mark to determine whether it is necessary to start the compensation mode or the global balance control mode; if it is identified as over-limit distortion, start the compensation mode and implement power adjustment or auxiliary cooling in the corresponding heating section; otherwise, enter the global balance control mode to coordinately fine-tune the pressure and temperature parameters of the hot press; Step 5. Verification after temperature correction: After the local compensation mode or the global balance control mode, a new temperature measurement value Tnew is obtained, and Tnew is compared with the coupling model output. If the deviation is reduced and the distortion mark is in a convergence state, the compensation in step 4 is considered to be effective, and the temperature correction result is output; otherwise, it means that the compensation is invalid.

[0024] It needs to be further explained in the embodiment of the present invention that the coupling model is constructed in the following manner: The pressure, material characteristic vector (including thermal conductivity, specific heat capacity, thickness of each layer, obtained based on material property information) and start-stop time series were selected as input to construct a coupled model of the hot press; Based on the mutual information theory, key variables with high coupling to temperature changes are screened; Explanation: Screening key variables with high coupling to temperature changes based on mutual information theory means that when analyzing the relationship between temperature changes and various operating parameters (such as pressure, material properties, start-stop cycles), the degree of dependence or correlation strength of each operating parameter on temperature is first quantified through mutual information; when the mutual information value between a variable and temperature is large, it means that the corresponding variable has a higher degree of influence or coupling on temperature changes; through this method, the key variables with the most decisive or significant impact on temperature are first screened out from many potential influencing factors, and the key variables will be focused on and processed in subsequent coupling modeling or compensation analysis, thereby improving the efficiency and accuracy of the model; A multi-layer neural network is used to perform nonlinear fitting on key variables and output the predicted temperature; Improve the accuracy of temperature prediction and suppress overfitting by minimizing the loss function; The loss function L(θ) of the coupling model satisfies the following formula: Among them, θ is the coupling model parameter, specifically the set of weights or coefficients to be trained inside the coupling model, including the weight matrix of the neural network and the kernel function coefficients of the support vector machine. θ is updated through an iterative training algorithm (such as gradient descent or Adam); M is the number of data points, indicating the number of samples or measurement moments included in the training set, and i is the sequence number; x i Refers to the input vector at the ith sample or measurement moment, such as pressure, time, material properties (such as thermal conductivity, specific heat capacity), and other temperature-related characteristics (such as ambient temperature, start-stop status); Represents the predicted temperature, given an input vector x i and the output results under the coupling model parameter θ; represents the reference temperature. Given an input vector x iThe output results below come from temperature sensor networks, thermal imager measurements, or experimental calibration values; is a balance factor, balancing the model fitting error and model complexity, and controlling the regularization term the degree of importance attached to The regularization term for suppressing overfitting is the sum of squares of penalty weights in the embodiment of the present invention.

[0025] It is necessary to further explain in the embodiment of the present invention that the coupling model compares the predicted temperature with the measured temperature point by point during the training phase. If the prediction error or mutual information deviation in the local area is continuously higher than the preset threshold, it is marked as a deviation and the deviation feature is extracted; the preset threshold represents the standard for determining whether the temperature deviation exceeds the allowable range; Analyze the deviation characteristics. If the cumulative time that the temperature prediction distribution of the coupling model and the measured temperature distribution show deviation at a single type of interface exceeds a threshold, it can be determined that there is an abnormal thermal diffusion problem at the material interface. If there is an abnormal thermal diffusion problem at the material interface, the user will be warned of process abnormalities and prompted to adjust process parameters. If the difference between the temperature reading of a specific sensor and the predicted temperature value of the adjacent normal sensor and the coupled model is found to be continuously beyond the normal variation range through local temperature deviation analysis and continuous monitoring, it is recorded as an abnormal drift sensor; Calculate the weighted difference between the adjacent normal sensor readings and the coupling model prediction value to obtain the compensation coefficient k of the jth sensor j , the compensation coefficient ranges from 0 to 1; the larger the value, the more trust is placed on the measured data, and the smaller the value, the more reliance is placed on the model prediction; k j Applied to abnormal drift sensor data: in, Represents the corrected temperature value of the jth sensor, which is used to replace the original measured temperature; represents the original measured temperature of the jth sensor; represents the temperature predicted by the coupled model; Indicates that k j The corresponding complementary factor is used to balance the weighted ratio of the measured temperature and the predicted temperature.

[0026] Explanation: The compensation coefficient k of the jth sensor is j The way to obtain is: Calculate the difference between the actual temperature of the jth sensor and the temperature predicted by the coupling model, denoted as ; Calculate the difference between the temperature of the jth sensor and the reading of its adjacent normal sensor, denoted as ; The compensation coefficient is calculated by the following formula: in, and is a weighting factor that controls the relative influence of the difference between the actual temperature and the temperature predicted by the coupled model.

[0027] It is necessary to further explain in the embodiment of the present invention that in the step 4, the process of analyzing the local distortion mark and determining whether to start the compensation or global balance control mode includes: Set dynamic distortion threshold: Define a dynamic distortion threshold that decays over time based on the deviation amplitude provided by the local distortion marker , in order to distinguish short-term fluctuations from continuous distortion, the dynamic distortion threshold Satisfies the following formula: in, is the initial threshold, Indicates the decay rate, which is initially set to the empirical coefficient. Δtl is the accumulated time since the last compensation was completed. If the deviation of the heating area exceeds , it is judged as local distortion; Identify excessive distortion: If the local distortion persists within the time window and the spatial range exceeds the preset proportion (the spatial distribution is concentrated and the area is larger than the preset value), it is marked as excessive distortion. That is, only when the corresponding area meets the preset conditions in amplitude, time and spatial range, it is identified as being in the "excessive distortion" state; Determine the temperature correction mode: If over-limit distortion is detected, the compensation mode is activated to implement rapid power adjustment or auxiliary cooling in the corresponding heating section; otherwise, the global balance control mode is activated to make linkage fine adjustments to the overall pressure and temperature of the hot press.

[0028] The compensation mode selection information will be synchronously transmitted to the subsequent calculation unit to calculate the correction amount and update the coupling model.

[0029] It needs to be further explained in the embodiment of the present invention that the startup compensation mode includes: Obtaining a local compensation control signal: for the area where the over-limit distortion mark is located, define a proportional gain coefficient and an integral gain coefficient, calculate the local compensation control signal through a local compensation control signal model, and use s to represent the sequence number of the over-limit distortion area. The local compensation control signal model satisfies the following formula: in, A local compensation control signal representing a heating or cooling adjustment amount performed on the s-th over-limit distortion region at time k; The temperature deviation of the sth out-of-limit distortion region at time k is If it is positive, it means that the actual temperature is higher than the target and cooling is required. If it is negative, it means that the actual temperature is lower than the target and heating is required; is the proportional gain coefficient of the sth over-limit distortion region, which is used to amplify the temperature deviation at the current moment and determine the impact of the instantaneous deviation on the compensation control; is the integral gain coefficient of the sth over-limit distortion region, which determines the impact of the accumulated deviation on the compensation control; In the time interval The integral of the temperature deviation is used to eliminate persistent or slowly accumulated temperature errors; τ is the integral variable.

[0030] It needs to be further explained in the embodiment of the present invention that starting the global balance control mode includes: Coupling input correction: Based on the initial input of the coupling model, increase the temperature correction , the temperature correction value satisfies the following formula: in, is the balance coefficient, which controls the influence of the corresponding global correction on the overall heating or cooling of the system; is the temperature deviation at the jth sensor; N is the number of sensors, and η is the regularization constant to avoid numerical instability caused by the denominator tending to zero; Use temperature correction to adjust power and pressure in conjunction: The temperature correction is decomposed into the adjustment signals for the heating power W and the pressure P, ΔW is the adjustment of the heating power, ΔP is the adjustment of the pressure; Among them, k W , k P , are the correction coefficients for heating power and pressure, respectively, which are adjusted according to the experimental or process requirements; if the temperature correction , increase the heating power, increase the pressure at the same time, accelerate the temperature rise and improve the thermal contact efficiency; if the temperature correction amount , reduce heating power, reduce pressure, delay temperature rise and reduce heat accumulation; Real-time feedback and iterative optimization: After the above adjustments are completed, the real-time temperature distribution is obtained again through the sensor, and the new temperature input is verified in the coupling model. If the deviation is not significantly reduced, the correction coefficients of heating power and pressure are further increased in the next iteration cycle.

[0031] It needs to be further explained in the embodiment of the present invention that starting the global balance control mode includes: When the thermal conductivity of the material is poor (such as polymer composite materials): give priority to adjusting the heating power to reduce the potential impact of pressure changes on the material structure; When the material has high thermal sensitivity (such as thin metal): it is preferred to adjust the pressure to affect the thermal contact efficiency to avoid local overheating caused by excessive power changes.

[0032] What needs to be further explained in the embodiment of the present invention is that when the hot press is in a frequent start-stop state, the iteration cycle of the local compensation mode and the global balance control is automatically shortened, so that the global adjustment and local compensation are alternately performed at a higher frequency to adapt to sudden temperature fluctuations or material changes.

[0033] Summary: The temperature correction method for a hot press provided by the embodiment of the present invention adopts a multi-layer neural network and a coupling model based on mutual information theory, which can accurately analyze and correct the temperature deviation in the operation of the hot press. By real-time monitoring of temperature, pressure and material property data, combined with a dynamic compensation strategy, it can effectively identify and handle problems such as local distortion, sensor drift, and abnormal thermal diffusion at the material interface, which not only improves the real-time performance of temperature correction, but also enhances the system's adaptability to different materials and production processes.

[0034] Embodiment 2: The difference between the embodiment of the present invention and embodiment 1 is that the method further includes a step of iterative correction based on adaptive optimization, specifically including: See also Figure 2 The method further includes a step of iterative correction based on adaptive optimization. If the local deviation feature continues to exceed the dynamic distortion threshold after compensation, the regularization term and compensation gain parameter of the coupling model are adjusted, and the compensation gain parameter includes a proportional gain coefficient and an integral gain coefficient; adaptive optimization is performed to update the coupling model; the effectiveness of the compensation strategy is re-determined according to the optimized coupling model; if no over-limit distortion occurs after multiple rounds of iterations, a stable temperature correction result is output to further ensure the long-term accuracy and stability of the hot press in a high-complexity scenario. Specifically, the following steps are included: Step 101, compensation effect evaluation and local distortion mark update: after each compensation is completed, the temperature measurement value after compensation is recorded and compared with the predicted value of the coupling model; if the deviation amplitude is reduced and the local distortion mark converges, a stable temperature correction result is output; if the deviation is not significantly reduced, proceed to the next step; Step 102, deviation feature analysis and compensation strategy adjustment: Analyze the deviation features of the local distortion area and determine whether the compensation effect meets the following requirements: In k consecutive sampling periods, the deviation value of the local area is less than the set dynamic distortion threshold; the area ratio of the distorted area is lower than the set threshold; If the conditions are not met and the deviation characteristics still exist after multiple compensations, the corresponding area is judged to be a long-term over-limit distortion area, and the coupling model needs to be further optimized; Step 103, optimizing the coupling model and adjusting the compensation gain: retraining the coupling model by adjusting the regularization factor and adjusting the compensation gain parameter; Step 104, dynamically correct the dynamic distortion threshold and the final temperature correction: adjust the attenuation rate of the dynamic distortion threshold to ensure adaptability to temperature changes; if after multiple rounds of continuous optimization, no long-term over-limit distortion area is detected and the difference between the predicted temperature distribution and the actual temperature is significantly reduced, then output the final stable temperature correction result.

[0035] It needs to be further explained in the embodiment of the present invention that the coupling model is trained based on the adjusted loss function, the optimized coupling model is output, and the proportional gain coefficient and the integral gain coefficient are reset according to the test data; The optimized coupling model and compensation gain parameters will affect the next step of temperature prediction and compensation operations.

[0036] It is explained that the adjustment method of the loss function includes adding the sum of the squares of the temperature deviations of all local distortion areas to the loss function. By penalizing the local distortion temperature deviation, the coupling model is prompted to not only pay attention to the overall temperature error, but also to minimize the temperature deviation of the local distortion area, thereby ensuring more precise temperature control in the local area.

[0037] Furthermore, after optimizing the coupling model and adjusting the compensation gain, if the local distortion still exists and manifests as a lag in temperature rise or a lag in temperature change; The inertia correction factor μ is used as a part of the compensation gain to correct the hysteresis effect of temperature change, and the optimized local compensation control signal is output through the following formula; When correcting the dynamic distortion threshold, the inertia correction factor μ will help further adjust the compensation strategy to make the temperature change smoother and avoid over-compensation or delayed compensation due to thermal inertia.

[0038] Explanation: The inertia correction factor μ takes values ​​between 0 and 1. The smaller the value, the lighter the inertia effect correction for temperature changes. The larger the value, the more the system relies on inertia correction to cope with slower temperature increases or delayed responses.

[0039] Summary: The embodiment of the present invention can achieve accurate correction of temperature control of the hot press under complex working conditions by continuously optimizing the coupling model, compensation gain parameters and dynamic distortion threshold; through multiple rounds of compensation and optimization, the system can analyze local distortion characteristics in real time, dynamically adjust the compensation strategy, and continuously improve the correction accuracy in each round of iteration. The adjusted coupling model can effectively deal with changes in material properties, sensor drift and other temperature distortion problems, ensuring the long-term stability and accuracy of the temperature control system.

[0040] Embodiment 3, the embodiment of the present invention provides a temperature correction system for a hot press, comprising: The multi-dimensional data acquisition and preprocessing module is used to obtain the multi-dimensional operation data of the hot press, including temperature time series data, pressure setting parameters and material property information; denoise the collected data, use the interpolation algorithm to fill the missing values, and format the processed data into a time series, and output the preprocessed data set; Coupling model building module: Based on the preprocessing data set, identify the coupling relationship between temperature, pressure, and material properties; train the coupling model through a multi-layer neural network to generate a coupling model that can predict the temperature distribution at different process stages, input the preprocessing data set into the coupling model, and output the corresponding temperature prediction distribution; Deviation feature extraction and marking module: Calculate the difference and extract local deviation features, including deviation amplitude, duration and spatial distribution, based on the predicted temperature distribution and the actual measured temperature distribution; determine whether the source of distortion is related to sensor drift or abnormal material thermal diffusion. If it is sensor drift, correct the sensor output by calculating the compensation coefficient; generate local distortion marks, which include position coordinates and key information of deviation attributes; Compensation strategy determination and execution module: analyzes local distortion marks. If it is identified as over-limit distortion, the compensation mode is activated to adjust the heating power or auxiliary cooling of the distortion area; otherwise, the global balance control mode is entered to make linkage fine-tuning of the pressure and temperature parameters of the hot press; Correction result verification and iteration module: record the temperature measurement value after compensation and compare it with the predicted value of the coupling model; if the deviation is reduced and the local distortion mark converges, output a stable temperature correction result; otherwise, return to the deviation feature extraction and marking module to regenerate the local distortion mark, and iteratively optimize the coupling model and compensation strategy.

[0041] Finally: 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 spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A temperature calibration method for a hot press, characterized in that: The following steps are involved: Step 1, obtaining multidimensional operation data to be corrected: obtaining multidimensional operation data generated by the hot press in the process of pressing the multi-layer composite material, denoising and interpolating the multidimensional operation data, and outputting a preprocessed data set; Step 2: Construct a coupling model: Identify the coupling relationship between temperature, pressure, and material in the preprocessing data set, and construct a coupling model based on a multi-layer neural network. The coupling model outputs the predicted temperature distribution at different process stages. Step 3: Perform sensor drift compensation and obtain local distortion mark: According to the difference between the temperature prediction distribution of the coupling model and the actual temperature distribution, extract the deviation characteristics, and judge whether there is sensor drift or material thermal diffusion anomaly by analyzing the deviation characteristics. Feed the compensated temperature correction value back to the coupling model to correct the sensor drift, and perform local distortion mark based on the sensor drift compensation result. The local distortion mark includes position coordinates, duration and deviation amplitude. Step 4. Determine the temperature correction mode according to the local distortion mark: Analyze the local distortion mark to determine whether it is necessary to start the compensation mode or the global balance control mode; if it is identified as over-limit distortion, start the compensation mode and implement power adjustment or auxiliary cooling in the corresponding heating section; otherwise, enter the global balance control mode to coordinately fine-tune the pressure and temperature parameters of the hot press; Step 5. Verification after temperature correction: After the local compensation mode or the global balance control mode, a new temperature measurement value is obtained and compared with the coupling model output. If the deviation is reduced and the distortion mark is in a convergence state, the compensation in step 4 is considered to be effective and the temperature correction result is output; otherwise, the compensation is invalid.

2. The temperature calibration method of a hot press according to claim 1, characterized in that: The coupling model is constructed as follows: The pressure, material characteristic vectors and start-stop time series are selected as input to build a coupling model of the hot press; Based on the mutual information theory, the key variables with high coupling degree to temperature change are selected; a multi-layer neural network is used to perform nonlinear fitting on the key variables and output the predicted temperature; The accuracy of temperature prediction is improved and overfitting is suppressed by minimizing the loss function; the loss function L(θ) of the coupling model satisfies the following formula: Among them, θ is the coupling model parameter, specifically the set of weights or coefficients to be trained inside the coupling model, including the weight matrix of the neural network and the kernel function coefficients of the support vector machine. θ is updated through an iterative training algorithm (such as gradient descent or Adam); M is the number of data points, indicating the number of samples or measurement moments included in the training set, and i is the sequence number; x i refers to the input vector at the ith sample or measurement moment; Represents the predicted temperature, given an input vector x i and the output results under the coupling model parameter θ; represents the reference temperature. Given an input vector x i The output results below come from temperature sensor networks, thermal imager measurements, or experimental calibration values; is the balance factor, controlling the regularization term the degree of importance attached to is a regularization term to suppress overfitting.

3. The temperature calibration method of a hot press according to claim 1, characterized in that: The coupling model compares the predicted temperature with the measured temperature point by point during the training phase. If the prediction error or mutual information deviation in a local area is continuously higher than a preset threshold, it is marked as a deviation and the deviation feature is extracted. The preset threshold represents the standard for determining whether the temperature deviation exceeds the allowable range. Analyze the deviation characteristics. If the cumulative time that the temperature prediction distribution of the coupling model and the measured temperature distribution show deviation at a single type of interface exceeds a threshold, it can be determined that there is an abnormal thermal diffusion problem at the material interface. If there is abnormal thermal diffusion at the material interface, the user will be warned of process abnormalities and prompted to adjust process parameters; If the difference between the temperature reading of a specific sensor and the predicted temperature value of the adjacent normal sensor and the coupled model is found to be continuously beyond the normal variation range through local temperature deviation analysis and continuous monitoring, it is recorded as an abnormal drift sensor; Calculate the weighted difference between the adjacent normal sensor readings and the coupling model prediction value to obtain the compensation coefficient k of the jth sensor j , the compensation coefficient range is between 0 and 1; then k j Applied to abnormal drift sensor data: in, Represents the corrected temperature value of the jth sensor, which is used to replace the original measured temperature; represents the original measured temperature of the jth sensor; represents the temperature predicted by the coupled model; Indicates that k j The corresponding complementary factor is used to balance the weighted ratio of the measured temperature and the predicted temperature.

4. The temperature calibration method of the hot press according to claim 3, characterized in that: The compensation coefficient k of the jth sensor j The way to obtain is: Calculate the difference between the actual temperature of the jth sensor and the temperature predicted by the coupling model, denoted as ; Calculate the difference between the temperature of the jth sensor and the reading of its adjacent normal sensor, denoted as ; The compensation coefficient is calculated by the following formula: in, and is a weighting factor that controls the relative influence of the difference between the actual temperature and the temperature predicted by the coupled model.

5. The temperature calibration method of a hot press according to claim 3, characterized in that: The process of analyzing the local distortion mark and determining whether to start the compensation or global balance control mode includes: Set dynamic distortion threshold: Define a dynamic distortion threshold that decays over time based on the deviation amplitude provided by the local distortion marker , in order to distinguish short-term fluctuations from continuous distortion, the dynamic distortion threshold Satisfies the following formula: in, is the initial threshold, represents the decay rate, is the accumulated time since the last compensation was completed; if the deviation of the heating area exceeds , it is judged as local distortion; Identify excessive distortion: If the local distortion persists within the time window and the spatial range exceeds the preset proportion, it is marked as excessive distortion; Determine the temperature correction mode: If over-limit distortion is detected, the compensation mode is activated to implement rapid power adjustment or auxiliary cooling in the corresponding heating section; otherwise, the global balance control mode is activated to make linkage fine adjustments to the overall pressure and temperature of the hot press.

6. The temperature calibration method of a hot press according to claim 5, characterized in that: The local compensation mode includes: Obtaining a local compensation control signal: defining a proportional gain coefficient and an integral gain coefficient for the region where the over-limit distortion mark is located, calculating the local compensation control signal through a local compensation control signal model, using s to represent the sequence number of the over-limit distortion region, and the local compensation control signal model satisfies the following formula; in, A local compensation control signal representing a heating or cooling adjustment amount performed on the s-th over-limit distortion region at time k; The temperature deviation of the sth out-of-limit distortion region at time k is If it is positive, it means that the actual temperature is higher than the target and cooling is required. If it is negative, it means that the actual temperature is lower than the target and heating is required; is the proportional gain coefficient of the sth over-limit distortion region, is the integral gain coefficient of the sth over-limit distortion region, In the time interval The integral of the temperature deviation is τ, and τ is the integral variable.

7. The temperature calibration method of a hot press according to claim 5, characterized in that: The starting global balance control mode comprises: Coupling input correction: Based on the initial input of the coupling model, increase the temperature correction , the temperature correction value satisfies the following formula: in, is the balance coefficient, which controls the influence of the corresponding global correction on the overall heating or cooling of the system; is the temperature deviation at the jth sensor; N is the number of sensors, and η is the regularization constant to avoid numerical instability caused by the denominator tending to zero; Use temperature correction to adjust power and pressure in conjunction: The temperature correction is decomposed into the adjustment signals for the heating power W and the pressure P, ΔW is the adjustment of the heating power, ΔP is the adjustment of the pressure; Among them, k W , k P , are the correction coefficients for heating power and pressure, respectively, which are adjusted according to the experimental or process requirements; if the temperature correction , increase the heating power, increase the pressure at the same time, accelerate the temperature rise and improve the thermal contact efficiency; if the temperature correction amount , reducing heating power while reducing pressure, delaying temperature rise and reducing heat accumulation.

8. The temperature calibration method of a hot press according to claim 1, characterized in that: When the hot press is in a state of frequent start and stop, the iteration cycle of the local compensation mode and the global balance control is automatically shortened, so that the global adjustment and local compensation are alternately performed at a higher frequency to adapt to sudden temperature fluctuations or material changes.

9. The temperature calibration method of a hot press according to claim 1, characterized in that: The method further includes a step of iterative correction based on adaptive optimization, if the local deviation feature continues to exceed the dynamic distortion threshold after compensation, adjusting the regularization term and compensation gain parameters of the coupling model, the compensation gain parameters including a proportional gain coefficient and an integral gain coefficient; performing adaptive optimization to update the coupling model; and re-determining the effectiveness of the compensation strategy based on the optimized coupling model; If no over-limit distortion occurs after multiple rounds of iterations, a stable temperature correction result is output.

10. The temperature correction system of the hot press is characterized by: include: The multi-dimensional data acquisition and pre-processing module is used to obtain the multi-dimensional operation data of the hot press, denoise the acquired data, use the interpolation algorithm to fill the missing values, format the processed data into a time series, and output the pre-processed data set; Coupling model building module: Based on the preprocessing data set, identify the coupling relationship between temperature, pressure, and material properties; train the coupling model through a multi-layer neural network to generate a coupling model that can predict the temperature distribution at different process stages, input the preprocessing data set into the coupling model, and output the corresponding temperature prediction distribution; Deviation feature extraction and marking module: Calculate the difference and extract local deviation features, including deviation amplitude, duration and spatial distribution, based on the predicted temperature distribution and the actual measured temperature distribution; determine whether the source of distortion is related to sensor drift or abnormal material thermal diffusion. If it is sensor drift, the sensor output is corrected by calculating the compensation coefficient; Generate local distortion marks, which include position coordinates and key information of deviation attributes; Compensation strategy determination and execution module: analyzes local distortion marks, and if it is identified as excessive distortion, activates the compensation mode to adjust the heating power or auxiliary cooling of the distorted area; Otherwise, enter the global balance control mode and make linkage fine adjustments to the pressure and temperature parameters of the hot press; Correction result verification and iteration module: record the temperature measurement value after compensation and compare it with the predicted value of the coupling model; If the deviation is reduced and the local distortion mark converges, a stable temperature correction result is output; otherwise, the deviation feature extraction and marking module is returned to regenerate the local distortion mark, and the coupling model and compensation strategy are iteratively optimized.

Images