Aluminum profile heat treatment system based on multi-parameter feedback and LSTM

Abstract

The invention relates to the technical field of aluminum profile heat treatment control, in particular to an aluminum profile heat treatment system based on multi-parameter feedback and LSTM, which comprises a server and an execution terminal, and the server comprises a multi-dimensional parameter acquisition module, a feature fusion preprocessing module, an LSTM prediction optimization module, a deviation hierarchical control module and a dynamic iterative storage module. The multi-dimensional parameter acquisition module acquires real-time heat treatment temperature, profile thickness, cooling medium flow velocity and finished product tensile strength; the feature fusion preprocessing module purifies and optimizes the data and then outputs a state parameter vector; the LSTM prediction optimization module predicts the tensile strength based on the vector, and outputs the parameter adjustment amount in combination with the deviation value and the profile thickness; the deviation grading control module executes differential adjustment according to deviation grades; and the dynamic iteration storage module stores data, regularly and incrementally trains the model, and optimizes a parameter mapping relation. According to the method, through multi-parameter cooperation, LSTM nonlinear adaptation and dynamic iteration, the heat treatment quality stability and self-adaptability of the aluminum profile are improved.

Description

Technical Field

[0001] This invention relates to the field of aluminum profile heat treatment control technology, and specifically to an aluminum profile heat treatment system based on multi-parameter feedback and LSTM. Background Technology

[0002] Aluminum profiles are commonly used materials in industrial production, and the quality of their heat treatment directly affects the mechanical properties and service life of the products. Existing aluminum profile heat treatment technologies mainly rely on traditional fixed parameter control or single-parameter feedback adjustment, which has the following shortcomings: Using only temperature as the core control parameter ignores key influencing factors such as profile thickness and cooling rate, leading to large deviations in tensile strength within the same batch of aluminum profiles and poor product quality stability; the use of traditional PID control algorithms makes it difficult to adapt to the nonlinear characteristics of the heat treatment process, resulting in a lag in response to changes in process parameters and limited adjustment accuracy; a lack of adaptability necessitates manual readjustment of process parameters for different batches and specifications of aluminum profiles, which is cumbersome and inefficient; and the absence of data accumulation and model iteration mechanisms makes it impossible to optimize control strategies using historical production data, leading to a decline in control accuracy over long-term use.

[0003] Therefore, there is an urgent need for an intelligent heat treatment system that combines multi-parameter feedback and intelligent algorithms to solve the problems of single parameter control, poor nonlinearity adaptation, and insufficient adaptability in the existing technology, and to improve the accuracy and intelligence level of aluminum profile heat treatment. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention aims to provide an aluminum profile heat treatment system based on multi-parameter feedback and LSTM, which solves the problems of weak adaptability to nonlinear operating conditions, lack of adaptability, and low control accuracy of existing aluminum profile heat treatment systems. It can achieve dynamic optimization of process parameters and improve the accuracy of aluminum profile heat treatment.

[0005] The basic solution provided by this invention is an aluminum profile heat treatment system based on multi-parameter feedback and LSTM, including a server and an execution terminal. The server includes a multi-dimensional parameter acquisition module, a feature fusion preprocessing module, an LSTM prediction optimization module, a deviation grading control module, and a dynamic iterative storage module. The multi-dimensional parameter acquisition module is used to control the multi-sensor array to acquire the original state parameters of the aluminum profile heat treatment process and the tensile strength of the finished product after heat treatment. The original state parameters include real-time heat treatment temperature, profile thickness and cooling medium flow rate. The feature fusion preprocessing module is used to remove noise, extract key features, perform feature cross-fusion and standardization on the original state parameters, and then output a state parameter vector. The LSTM prediction optimization module has a built-in LSTM deep learning model, which is used to output the target tensile strength prediction value of aluminum profile after heat treatment by inputting the state parameter vector. Based on the deviation value between the target tensile strength prediction value and the preset target tensile strength, and combined with the profile thickness adaptability, it outputs the original state parameter adjustment amount, including the real-time heat treatment temperature adjustment amount and the cooling medium flow rate adjustment amount. The deviation grading control module is used to classify deviation levels according to the deviation value and execute a differentiated dynamic adjustment strategy accordingly. The adjustment strategy is related to the deviation level, the adjustment amount of the original state parameters, and the profile thickness adaptability. The dynamic iterative storage module is used to store the original state parameters, finished product tensile strength, state parameter vector, target tensile strength prediction value, state parameter adjustment instructions and deviation level; The dynamic iterative storage module has a built-in incremental training unit. When the amount of stored data reaches the preset storage threshold, it combines historical core parameter data to perform incremental iterative training on the LSTM deep learning model, and updates the mapping relationship between the state parameter vector and the target tensile strength prediction value, the original state parameter adjustment amount, and the model parameters. The execution terminal includes a heating control unit and a cooling control unit; the heating control unit adjusts the temperature to achieve real-time heat treatment temperature adjustment, and the cooling control unit adjusts the rotation speed to achieve cooling medium flow rate adjustment.

[0006] The principle of this invention is as follows: A multi-sensor array comprehensively captures key raw state parameters (real-time heat treatment temperature, profile thickness, cooling medium flow rate) and core performance indicators (tensile strength) of the finished product during the heat treatment process, ensuring data coverage of the core dimensions affecting heat treatment quality. Subsequently, a feature fusion preprocessing module purifies and optimizes the raw data, removing noise interference and strengthening the correlation of key features to generate a standardized state parameter vector adapted to the LSTM model input. Then, utilizing the temporal modeling and nonlinear fitting capabilities of the LSTM deep learning model, the tensile strength of the finished product is accurately predicted based on the input vector. By calculating the deviation between the predicted value and the preset target value, and combining the influence of profile thickness on heat conduction, targeted process parameter adjustments are derived in reverse. A deviation grading control module matches differentiated adjustment strategies according to the magnitude of the deviation, avoiding insufficient accuracy caused by a single adjustment mode. Finally, a dynamic iterative storage module completely retains the entire process data. When the data volume reaches a preset threshold, incremental model training is initiated to continuously optimize the parameter mapping relationship, achieving dynamic improvement in system control accuracy.

[0007] The beneficial effects of this invention are as follows: it achieves multi-parameter collaborative control of three core parameters, including temperature, thickness, and cooling flow rate, comprehensively covering factors affecting heat treatment quality, effectively reducing tensile strength deviations in aluminum profiles of the same batch, and improving product consistency; utilizing the temporal learning and nonlinear fitting advantages of the LSTM deep learning model, compared with traditional PID control, it can more accurately adapt to the characteristics of nonlinear heat conduction and strong parameter coupling in the heat treatment process of aluminum profiles, with faster parameter adjustment response speed and higher accuracy; by accumulating production data through a dynamic iterative storage module and regularly updating the model, it can adapt to the heat treatment requirements of different batches and specifications of aluminum profiles without manual intervention, reducing manual debugging costs and improving production efficiency; based on the deviation value, it classifies levels and performs differentiated adjustments to avoid over-adjustment or under-adjustment, balancing control accuracy and process stability.

[0008] Furthermore, the original state parameters also include the initial temperature of the aluminum profile; the multi-sensor array includes a temperature sensor for acquiring real-time heat treatment temperature, a thickness sensor for acquiring profile thickness, a flow rate sensor for acquiring cooling medium flow rate, a strength detection sensor for acquiring finished tensile strength, and an infrared thermometer for acquiring the initial temperature of the aluminum profile; the execution terminal also includes an initial temperature adjustment unit for adjusting the initial temperature of the aluminum profile.

[0009] The initial temperature of aluminum profiles directly affects the temperature rise rate and thermal stress distribution during heat treatment. Collecting and adjusting the initial temperature parameters ensures that parameter control covers the entire process of preheating, heat treatment, and cooling, avoiding inconsistent heat treatment effects due to differences in initial temperature. The matching infrared thermometer and initial temperature adjustment unit enable precise detection and control of the initial temperature, providing more comprehensive data support for subsequent process parameter optimization and further enhancing the system's adaptability to aluminum profiles in different initial states.

[0010] Furthermore, the processing flow of the feature fusion preprocessing module includes: removing random noise from the original state parameters through Kalman filtering, extracting key features using principal component analysis, fusing feature vectors of different dimensional parameters into a high-dimensional feature matrix through a feature cross-fusion unit, and finally standardizing the vectors to a preset interval through normalization to generate a state parameter vector; in the LSTM prediction optimization module, the calculation of the adjustment amount of the original state parameters is positively correlated with the deviation value, and the sub-intervals are divided according to the profile thickness, with different sub-intervals corresponding to different adjustment amount adaptation coefficients to adapt to the thermal conductivity differences of aluminum profiles of different thicknesses.

[0011] Kalman filtering effectively removes random noise during sensor acquisition, principal component analysis accurately extracts core features and reduces data dimensionality, feature cross-fusion strengthens the correlation between different parameters, and normalization eliminates parameter magnitude differences. This multi-step collaborative approach ensures the accuracy and reliability of the input model data. Addressing the significant differences in thermal conductivity of aluminum profiles with varying thicknesses, the approach divides thickness sub-intervals and matches them with dedicated adjustment coefficients. This makes parameter adjustments more closely reflect actual thermal conductivity patterns, avoiding issues such as insufficient heating of thick profiles and overheating of thin profiles caused by a uniform adjustment mode, thus improving the consistency of heat treatment quality for aluminum profiles of different specifications.

[0012] Furthermore, the deviation classification control module divides the deviation levels into three levels: the first level deviation corresponds to a deviation value not greater than the first threshold, the second level deviation corresponds to a deviation value greater than the first threshold but not greater than the second threshold, and the third level deviation corresponds to a deviation value greater than the second threshold. The corresponding adjustment strategies are as follows: for the first level deviation, the current original state parameters are maintained and relevant data is recorded; for the second level deviation, the original state parameter adjustment amount is executed first and the thickness adjustment coefficient is adapted; for the third level deviation, the emergency adjustment mode is activated and the adjustment amount is executed at the upper limit, the thickness adjustment coefficient is adapted, and the equipment self-test is activated.

[0013] Differential adjustment balances accuracy and stability. The first level of deviation (small deviation) maintains parameters to avoid over-adjustment, the second level of deviation (medium deviation) precisely executes the adjustment amount, and the third level of deviation (large deviation) performs emergency adjustment and self-check. For large deviation scenarios, the equipment self-check is initiated to promptly identify potential problems such as sensor failure and execution unit abnormality, avoiding batch product quality defects or safety risks caused by equipment failure, and improving the reliability and safety of system operation.

[0014] Furthermore, the dynamic iterative storage module also includes a feature weight analysis unit, which is used to statistically analyze the weight ratio of the core original state parameters and the optimal adjustment coefficient of different aluminum profile series, forming a mapping table of material, thickness and adjustment coefficient, providing an initial parameter matching basis for the heat treatment of new materials and new specifications of aluminum profiles.

[0015] By using the feature weight analysis unit to explore the parameter influence patterns of different aluminum profile series, a standardized mapping table is formed. When dealing with new materials and specifications of aluminum profiles, the initial parameters can be directly matched based on the mapping table without having to start debugging from scratch. This significantly shortens the process adaptation cycle of new products, reduces trial production costs, and improves production flexibility and market response speed. Attached Figure Description

[0016] Figure 1 This is a system block diagram of an embodiment of the aluminum profile heat treatment system based on multi-parameter feedback and LSTM of the present invention. Detailed Implementation

[0017] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown: A heat treatment system for aluminum profiles based on multi-parameter feedback and LSTM, including a server and an execution terminal. The server includes a multi-dimensional parameter acquisition module, a feature fusion preprocessing module, an LSTM prediction optimization module, a deviation grading control module, and a dynamic iterative storage module. The multi-dimensional parameter acquisition module is used to control the multi-sensor array to acquire the original state parameters of the aluminum profile heat treatment process and the tensile strength of the finished product after heat treatment. The original state parameters include real-time heat treatment temperature, profile thickness and cooling medium flow rate. The feature fusion preprocessing module is used to remove noise, extract key features, perform feature cross-fusion and standardization on the original state parameters, and then output a state parameter vector. The LSTM prediction optimization module has a built-in LSTM deep learning model, which is used to output the target tensile strength prediction value of aluminum profile after heat treatment by inputting the state parameter vector. Based on the deviation value between the target tensile strength prediction value and the preset target tensile strength, and combined with the profile thickness adaptability, it outputs the original state parameter adjustment amount, including the real-time heat treatment temperature adjustment amount and the cooling medium flow rate adjustment amount. The deviation grading control module is used to classify deviation levels according to the deviation value and execute a differentiated dynamic adjustment strategy accordingly. The adjustment strategy is related to the deviation level, the adjustment amount of the original state parameters, and the profile thickness adaptability. The dynamic iterative storage module is used to store the original state parameters, finished product tensile strength, state parameter vector, target tensile strength prediction value, state parameter adjustment instructions and deviation level; The dynamic iterative storage module has a built-in incremental training unit. When the amount of stored data reaches the preset storage threshold, it combines historical core parameter data to perform incremental iterative training on the LSTM deep learning model, and updates the mapping relationship between the state parameter vector and the target tensile strength prediction value, the original state parameter adjustment amount, and the model parameters. The execution terminal includes a heating control unit and a cooling control unit; the heating control unit adjusts the temperature to achieve real-time heat treatment temperature adjustment, and the cooling control unit adjusts the rotation speed to achieve cooling medium flow rate adjustment.

[0018] Furthermore, the original state parameters also include the initial temperature of the aluminum profile; the multi-sensor array includes a temperature sensor for acquiring real-time heat treatment temperature, a thickness sensor for acquiring profile thickness, a flow rate sensor for acquiring cooling medium flow rate, a strength detection sensor for acquiring finished tensile strength, and an infrared thermometer for acquiring the initial temperature of the aluminum profile; the execution terminal also includes an initial temperature adjustment unit for adjusting the initial temperature of the aluminum profile.

[0019] Furthermore, the processing flow of the feature fusion preprocessing module includes: removing random noise from the original state parameters through Kalman filtering, extracting key features using principal component analysis, fusing feature vectors of different dimensional parameters into a high-dimensional feature matrix through a feature cross-fusion unit, and finally standardizing the vectors to a preset interval through normalization to generate a state parameter vector; in the LSTM prediction optimization module, the calculation of the adjustment amount of the original state parameters is positively correlated with the deviation value, and the sub-intervals are divided according to the profile thickness, with different sub-intervals corresponding to different adjustment amount adaptation coefficients to adapt to the thermal conductivity differences of aluminum profiles of different thicknesses.

[0020] Furthermore, the deviation classification control module divides the deviation levels into three levels: the first level deviation corresponds to a deviation value not greater than the first threshold, the second level deviation corresponds to a deviation value greater than the first threshold but not greater than the second threshold, and the third level deviation corresponds to a deviation value greater than the second threshold. The corresponding adjustment strategies are as follows: for the first level deviation, the current original state parameters are maintained and relevant data is recorded; for the second level deviation, the original state parameter adjustment amount is executed first and the thickness adjustment coefficient is adapted; for the third level deviation, the emergency adjustment mode is activated and the adjustment amount is executed at the upper limit, the thickness adjustment coefficient is adapted, and the equipment self-test is activated.

[0021] Furthermore, the dynamic iterative storage module also includes a feature weight analysis unit, which is used to statistically analyze the weight ratio of the core original state parameters and the optimal adjustment coefficient of different aluminum profile series, forming a mapping table of material, thickness and adjustment coefficient, providing an initial parameter matching basis for the heat treatment of new materials and new specifications of aluminum profiles.

[0022] In this embodiment, the multi-sensor array is selected as follows: real-time heat treatment temperature is acquired via a K-type thermocouple; profile thickness is acquired via a laser displacement sensor; cooling medium flow rate is acquired via an electromagnetic flow rate sensor; finished product tensile strength is acquired via a portable electronic tensile gauge; initial aluminum profile temperature is acquired via an infrared thermometer; the execution terminal is configured as follows: the heating control unit uses a combination of a solid-state relay and an electric heating tube (1kW power); the cooling control unit uses a combination of a variable frequency water pump and cooling pipes (frequency range 5-50Hz); the initial temperature adjustment unit uses an infrared heating preheating device (adjustable power 0.5-2kW).

[0023] The server and model parameters are as follows: The server is equipped with an Intel Core i7 processor and 16GB of memory, and supports the TensorFlow framework; the LSTM deep learning model training uses the Adam optimizer, with the loss function being the mean squared error (MSE), and training stops when the MSE is less than 0.01; storage and iteration parameters: the dynamic iteration storage module adopts a dual backup mode of local hard drive and cloud storage, with a preset storage threshold of 200 sets of valid data; during incremental training, the learning rate is set to 0.001, the number of iterations is 100, and the model parameters and mapping relationships are automatically updated after training is completed; bias threshold settings: the first threshold is set to 5MPa, and the second threshold is set to 15MPa.

[0024] Profile thickness sub-range division: The profile thickness is clearly divided into two core sub-ranges, namely, a preset thickness threshold of 4mm. Sub-range one is for thicknesses less than 4mm (thin profiles), and sub-range two is for thicknesses not less than 4mm (thick profiles). The adjustment adjustment coefficients for sub-range one are: 0.8-0.9 times the base calculated value for real-time heat treatment temperature adjustment and 1.1-1.2 times the base calculated value for cooling medium flow rate adjustment. The adjustment adjustment coefficients for sub-range two are: 1.0-1.1 times the base calculated value for real-time heat treatment temperature adjustment and 0.9-1.0 times the base calculated value for cooling medium flow rate adjustment, to adapt to the differences in heat conduction efficiency of aluminum profiles of different thicknesses (thin profiles have faster heat conduction, requiring a reduced temperature adjustment range and an increased cooling adjustment range; thick profiles have slower heat conduction, requiring an increased temperature adjustment range and a reduced cooling adjustment range).

[0025] Taking the processing of 6061 series aluminum profiles (target tensile strength 260MPa) as an example, the specific workflow is as follows: Parameter acquisition: The multi-dimensional parameter acquisition module controls the sensor array to acquire data in real time. The real-time heat treatment temperature is acquired at a frequency of 1Hz, the profile thickness is acquired once before heat treatment, the cooling medium flow rate is acquired at a frequency of 0.5Hz, and the tensile strength of the finished product is acquired once after heat treatment. All data are uploaded to the server synchronously.

[0026] Feature preprocessing: The feature fusion preprocessing module first removes data noise through Kalman filtering (process noise covariance is 0.01, observation noise covariance is 0.1); then it extracts the first 3 principal components (cumulative contribution rate exceeds 95%) through principal component analysis; subsequently, it calculates the product term and ratio term between parameters through the feature cross-fusion unit to construct a 6-dimensional high-dimensional feature matrix; finally, it normalizes the data to the [0,1] interval through min-max normalization to generate a state parameter vector.

[0027] Prediction and Adjustment Calculation: The LSTM prediction optimization module takes the state parameter vector as input and outputs the predicted target tensile strength value; it calculates the deviation value. If the deviation value is 8MPa (corresponding to the second-level deviation), then based on the positive correlation between the deviation value and the adjustment amount (adjustment coefficient is 0.05), it calculates the real-time heat treatment temperature adjustment. Cooling medium flow rate adjustment amount If the profile thickness is 3mm (belonging to sub-interval one), then the adaptation coefficients are taken as 0.8 and 1.2 respectively, and the final adjustment amount is... , This generates a state parameter adjustment instruction.

[0028] Graded control execution: The deviation graded control module determines that the deviation value of 8 MPa belongs to the second level of deviation and sends an adjustment command to the execution terminal; the heating control unit adjusts the electric heating tube voltage through a solid-state relay to increase the real-time heat treatment temperature by 0.32℃; the cooling control unit adjusts the water pump speed through a frequency converter to increase the cooling medium flow rate by 0.48 m / s.

[0029] Data storage and iteration: The dynamic iteration storage module stores the raw data collected this time, the preprocessed state parameter vector, the predicted value, the adjustment instructions and the deviation level; when the amount of stored data reaches 200 sets, the incremental training unit calls the historical core parameter data (valid data with tensile strength deviation not exceeding 5MPa) to perform incremental training on the LSTM model and update the parameter mapping relationship and model weights.

[0030] When processing 7075 series aluminum profiles, the feature weight analysis unit calls the pre-stored mapping table of material, thickness and adjustment coefficients to match the core parameter weight ratio (temperature 65%, cooling flow rate 25%, initial temperature 10%) and optimal adjustment coefficient of the 7075 series aluminum profile (thickness 5mm), which are directly input into the system as initial parameters without the need for manual debugging.

[0031] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A heat treatment system for aluminum profiles based on multi-parameter feedback and LSTM, characterized in that: It includes a server and an execution terminal. The server includes a multi-dimensional parameter acquisition module, a feature fusion preprocessing module, an LSTM prediction optimization module, a deviation grading control module, and a dynamic iterative storage module. The multi-dimensional parameter acquisition module is used to control the multi-sensor array to acquire the original state parameters of the aluminum profile heat treatment process and the tensile strength of the finished product after heat treatment. The original state parameters include real-time heat treatment temperature, profile thickness and cooling medium flow rate. The feature fusion preprocessing module is used to remove noise, extract key features, perform feature cross-fusion and standardization on the original state parameters, and then output a state parameter vector. The LSTM prediction optimization module has a built-in LSTM deep learning model, which is used to output the target tensile strength prediction value of aluminum profile after heat treatment by inputting the state parameter vector. Based on the deviation value between the target tensile strength prediction value and the preset target tensile strength, and combined with the profile thickness adaptability, it outputs the original state parameter adjustment amount, including the real-time heat treatment temperature adjustment amount and the cooling medium flow rate adjustment amount. The deviation grading control module is used to classify deviation levels according to the deviation value and execute a differentiated dynamic adjustment strategy accordingly. The adjustment strategy is related to the deviation level, the adjustment amount of the original state parameters, and the profile thickness adaptability. The dynamic iterative storage module is used to store the original state parameters, finished product tensile strength, state parameter vector, target tensile strength prediction value, state parameter adjustment instructions and deviation level; The dynamic iterative storage module has a built-in incremental training unit. When the amount of stored data reaches the preset storage threshold, it combines historical core parameter data to perform incremental iterative training on the LSTM deep learning model, and updates the mapping relationship between the state parameter vector and the target tensile strength prediction value, the original state parameter adjustment amount, and the model parameters. The execution terminal includes a heating control unit and a cooling control unit; the heating control unit adjusts the temperature to achieve real-time heat treatment temperature adjustment, and the cooling control unit adjusts the rotation speed to achieve cooling medium flow rate adjustment.

2. The aluminum profile heat treatment system based on multi-parameter feedback and LSTM according to claim 1, characterized in that: The original state parameters also include the initial temperature of the aluminum profile; the multi-sensor array includes a temperature sensor for acquiring real-time heat treatment temperature, a thickness sensor for acquiring profile thickness, a flow rate sensor for acquiring cooling medium flow rate, a strength detection sensor for acquiring finished tensile strength, and an infrared thermometer for acquiring the initial temperature of the aluminum profile; the execution terminal also includes an initial temperature adjustment unit for adjusting the initial temperature of the aluminum profile.

3. The aluminum profile heat treatment system based on multi-parameter feedback and LSTM according to claim 2, characterized in that: The processing flow of the feature fusion preprocessing module includes: removing random noise from the original state parameters through Kalman filtering, extracting key features through principal component analysis, fusing feature vectors of different dimensional parameters into a high-dimensional feature matrix through a feature cross-fusion unit, and finally normalizing the vectors to a preset interval to generate a state parameter vector. In the LSTM prediction optimization module, the calculation of the adjustment amount of the original state parameters is positively correlated with the deviation value, and the module is divided into sub-intervals according to the profile thickness. Different sub-intervals correspond to different adjustment amount adaptation coefficients to adapt to the thermal conductivity differences of aluminum profiles with different thicknesses.

4. The aluminum profile heat treatment system based on multi-parameter feedback and LSTM according to claim 3, characterized in that: The deviation classification control module divides the deviation levels into three levels: Level 1 deviation corresponds to a deviation value not greater than a first threshold, Level 2 deviation corresponds to a deviation value greater than the first threshold but not greater than the second threshold, and Level 3 deviation corresponds to a deviation value greater than the second threshold. The corresponding adjustment strategies are as follows: Level 1 deviation maintains the current original state parameters and records relevant data; Level 2 deviation prioritizes the adjustment of the original state parameters and adapts to the thickness adjustment coefficient; Level 3 deviation initiates an emergency adjustment mode and executes the adjustment amount at the upper limit, adapts to the thickness adjustment coefficient, and initiates equipment self-test.

5. The aluminum profile heat treatment system based on multi-parameter feedback and LSTM according to claim 4, characterized in that: The dynamic iterative storage module also includes a feature weight analysis unit, which is used to statistically analyze the weight ratio of core original state parameters and the optimal adjustment coefficient of different aluminum profile series, forming a mapping table of material, thickness and adjustment coefficient, and providing an initial parameter matching basis for the heat treatment of new materials and new specifications of aluminum profiles.

Images