Aluminum alloy plate process parameter dynamic regulation method and system

Abstract

The application belongs to the technical field of industrial control, and discloses an aluminum alloy plate process parameter dynamic regulation method and system. Through the introduction of model optimization in the test stage and online adjustment mechanism in the formal straightening stage, the model library established offline is combined with real-time data collected online, so that dynamic and self-adaptive regulation of straightening process parameters of different types and specifications of aluminum alloy plates is realized. The limitations caused by the dependence of the traditional method on a fixed parameter table or an inaccurate model are overcome, and the straightening precision and production efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial control, in particular to an aluminum alloy plate process parameter dynamic regulation method and system. BACKGROUND

[0002] In the continuous production line of aluminum alloy plates, rolling is one of the key processes. After rolling, the plate usually has various plate shape defects, such as bending, waving, twisting, etc., and internal residual stress is also generated. In order to eliminate these defects and improve the plate shape quality, the plate needs to be sent to the straightening process. The multi-roll straightening machine is a commonly used straightening equipment, which repeatedly deforms the plate between the upper and lower staggered rows of straightening rollers, so as to eliminate plate shape defects and reduce residual stress. The process parameters of the straightening machine, such as the relative vertical position (i.e. the reduction) between the upper and lower straightening rollers of each row, the straightening speed (plate conveying speed), etc., play a decisive role in the straightening effect. By accurately controlling these parameters, the plastic deformation of the plate can be regulated, and finally a flat and low-stress plate can be obtained.

[0003] The same production line often needs to process aluminum alloy plates of different alloy compositions, different heat treatment states and different specifications (thickness, width). In the face of plates of different alloys and different specifications, there are essential differences in their plastic deformation behavior in the multi-roll straightening machine and the response characteristics to the straightening parameters (especially the roller reduction combination).

[0004] In order to effectively straighten different plates, it is necessary to set the corresponding straightening process parameters according to the type and specification of the plate being produced. The traditional straightening parameter setting method mainly relies on two ways: one is to pre-establish a set of fixed process parameter table for each common alloy-specification combination. This method requires that, before actual production, a large number of offline experiments or complex numerical simulations based on precise physical models are needed to determine the best set of straightening parameters for each target plate combination. It requires extremely huge time, manpower and computing resources, which is costly and inefficient, and in fact it is difficult to cover all potential production needs, especially for uncommon alloys or special specifications.

[0005] The second is to calculate the parameters based on simplified physical models. These models are usually based on idealized assumptions and have large deviations from the actual situation, which will cause deviations between the preset model parameters and the actual situation, resulting in differences between the straightening effect predicted by the model and the actual obtained plate shape, affecting the straightening accuracy.

[0006] When the production line needs to switch to a new alloy type or specification combination, there is often no ready-made accurate model or reliable process parameters to refer to. At this time, it is usually necessary to rely on experienced operators for debugging, but the accuracy and efficiency of such debugging are highly dependent on the experience level of the operators, and it is difficult to ensure consistency.

[0007] Therefore, there is an urgent need for an advanced method that can quickly evaluate or identify the actual physical and mechanical properties of the current plate when passing through the straightening machine based on the type, specification information of the current plate and real-time online detection data. Based on this identification result, the straightening process parameters are quickly determined and dynamically adjusted to improve straightening accuracy and efficiency, reduce trial and error and waste products to adapt to different plates and production conditions.

[0008] In view of the above problems, the prior art needs to be improved. SUMMARY

[0009] The purpose of the present application is to provide an aluminum alloy plate process parameter dynamic regulation method and system, which can quickly determine and dynamically adjust the straightening process parameters according to the basic information and real-time data of the plate, improve the straightening accuracy and efficiency, reduce trial and error and waste products, and adapt to different plates and production conditions.

[0010] In a first aspect, the present application provides an aluminum alloy plate process parameter dynamic regulation method for adjusting the process parameters of a multi-roll straightening machine when straightening an aluminum alloy plate. The steps of the method include:

[0011] A1. Obtain the basic information of the aluminum alloy plate and the real-time online detection data in the test stage; use the preset process parameters to straighten the aluminum alloy plate in the test stage;

[0012] A2. According to the basic information, determine a group of candidate prediction models from a preset prediction model library; each prediction model in the prediction model library is used to predict the deformation response of the aluminum alloy plate according to the process parameters and real-time online detection data;

[0013] A3. According to the real-time online detection data in the test stage, determine the optimal prediction model with the optimal prediction accuracy of the actual deformation response of the aluminum alloy plate from the candidate prediction models;

[0014] A4. Use the optimal prediction model to perform formal straightening on the aluminum alloy plate, and perform online adjustment on the optimal prediction model based on the real-time online detection data during formal straightening;

[0015] A5. According to the real-time online detection data at the current time, adjust the process parameters online using the online adjusted optimal prediction model.

[0016] Preferably, the basic information comprises an alloy grade, a specification, a thickness, a width, a length, and a surface state.

[0017] The real-time online detection data comprises input data and response data; the input data comprises pre-straightening plate shape data, thickness data, and temperature data; and the response data comprises post-straightening plate shape data.

[0018] The process parameters comprise a reduction and a straightening speed; and the preset process parameters comprise a preset reduction and a preset straightening speed.

[0019] Preferably, step A1 comprises:

[0020] A101. Obtaining the basic information of the aluminum alloy plate;

[0021] A102. Testing straightening the aluminum alloy plate using the preset process parameters, and continuously collecting the real-time online detection data to form a real-time online detection data sequence during the testing straightening.

[0022] A103. Calculating a deformation response characteristic parameter of the aluminum alloy plate within a current testing length range according to the real-time online detection data sequence.

[0023] A104. If the deformation response characteristic parameter does not satisfy a preset testing end threshold condition, continuously collecting the real-time online detection data and updating the real-time online detection data sequence, and returning to step A103, otherwise, ending the testing phase, and taking the current real-time online detection data sequence as the real-time online detection data of the testing phase.

[0024] Preferably, step A2 comprises:

[0025] A201. According to the basic information, navigating along a hierarchical index structure in the preset prediction model library to locate a prediction model subset related to the basic information; the hierarchical index structure is constructed in advance according to multiple information dimensions of the basic information.

[0026] A202. Based on a preset matching rule of the basic information and each prediction model in the prediction model subset, determining a group of prediction models with a matching degree higher than a preset matching degree threshold as the candidate prediction models.

[0027] Preferably, step A202 comprises:

[0028] Based on a preset matching rule of the basic information and each prediction model in the prediction model subset, determining a basic matching degree between the basic information and each prediction model in the prediction model subset.

[0029] According to the real-time online detection data of the test stage, analyze the actual deformation response characteristics of the aluminum alloy plate;

[0030] For each of the prediction model subset, based on the actual deformation response characteristics, calculate the prediction matching degree of the actual deformation response of the prediction model;

[0031] For each of the prediction model subset, based on the actual deformation response characteristics, calculate the prediction matching degree of the actual deformation response of the prediction model;

[0032] Determine a group of prediction models whose comprehensive matching degree is higher than the preset matching degree threshold as the candidate prediction models.

[0033] Preferably, step A3 comprises:

[0034] A301. Based on the input data in the real-time online detection data of the test stage and the preset process parameters used in the test stage, use each of the candidate prediction models to predict the deformation response of the aluminum alloy plate;

[0035] A302. Compare the deformation response predicted by each of the candidate prediction models with the actual deformation response reflected by the real-time online detection data of the test stage, and calculate the prediction error of each of the candidate prediction models;

[0036] A303. Determine the candidate prediction model with the smallest prediction error as the optimal prediction model.

[0037] Preferably, step A4 comprises:

[0038] A401. After starting the formal straightening stage, based on the input data detected first, use the optimal prediction model to determine the initial process parameters, and collect real-time online detection data during the formal straightening to obtain real-time online detection data at the formal straightening;

[0039] A402. Extract the real-time online detection data at the current time from the real-time online detection data at the formal straightening to determine the actual deformation response of the aluminum alloy plate at the current time;

[0040] A403. Based on the current process parameters and the input data in the real-time online detection data at the current time, use the current optimal prediction model to predict the deformation response of the aluminum alloy plate to obtain the predicted deformation response at the current time;

[0041] A404. Compare the actual deformation response at the current time with the predicted deformation response at the current time to calculate the prediction deviation at the current time;

[0042] A405. calculating an adjustment amount for the internal parameters of the current optimal prediction model based on the prediction deviation at the current time and a preset adjustment rule;

[0043] A406. updating the internal parameters of the current optimal prediction model according to the adjustment amount to obtain an online-adjusted optimal prediction model.

[0044] Preferably, step A405 comprises:

[0045] judging a current production stage according to the prediction deviation at the current time and the real-time online detection data at the formal straightening; the production stage comprises an initial transient stage and a steady state stage;

[0046] selecting a corresponding adjustment rule from a preset adjustment rule library according to the judged production stage; the adjustment rule library comprises at least one adjustment rule applicable to the initial transient stage and at least one adjustment rule applicable to the steady state stage;

[0047] calculating a synergistic adjustment amount for multiple related parameters of the current optimal prediction model based on the prediction deviation at the current time and the selected adjustment rule as the adjustment amount.

[0048] Preferably, step A5 comprises:

[0049] A501. determining a target deformation response according to the real-time online detection data at the current time and a preset straightening requirement;

[0050] A502. using the online-adjusted optimal prediction model to solve a process parameter that can make the predicted deformation response of the online-adjusted optimal prediction model reach the target deformation response under the input data condition in the real-time online detection data at the current time as a target process parameter;

[0051] A503. adjusting the process parameter according to the target process parameter.

[0052] In a second aspect, the application provides an aluminum alloy plate process parameter dynamic regulation system for adjusting process parameters of a multi-roll straightening machine when straightening an aluminum alloy plate, the system comprising:

[0053] an information acquisition module for acquiring basic information of an aluminum alloy plate and real-time online detection data in a test stage; the aluminum alloy plate is straightened using a preset process parameter in the test stage;

[0054] a candidate model screening module for determining a group of candidate prediction models from a preset prediction model library according to the basic information; each prediction model in the prediction model library is used to predict a deformation response of an aluminum alloy plate according to a process parameter and real-time online detection data;

[0055] a most optimal model determination module configured to determine, from the real-time online detection data in the test stage, a most optimal prediction model with the most optimal prediction accuracy for the actual deformation response of the aluminum alloy plate among the candidate prediction models;

[0056] a model adjustment module configured to perform formal straightening on the aluminum alloy plate by using the most optimal prediction model, and to perform online adjustment on the most optimal prediction model based on real-time online detection data in the formal straightening;

[0057] a process parameter adjustment module configured to perform online adjustment on process parameters by using the most optimal prediction model after online adjustment, according to real-time online detection data at the current time.

[0058] Beneficial effects: The aluminum alloy plate process parameter dynamic regulation method and system provided by the present application introduces a model optimization mechanism in the test stage and a model online adjustment mechanism in the formal straightening stage, combines the model library established offline with real-time data collected online, and thus realizes dynamic and self-adaptive regulation of straightening process parameters for different types and specifications of aluminum alloy plates, overcomes the limitations caused by the dependence of the traditional method on a fixed parameter table or an inaccurate model, and improves straightening accuracy and production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 A flowchart of the aluminum alloy plate process parameter dynamic regulation method provided by the embodiments of the present application.

[0060] Figure 2 A structural schematic diagram of the aluminum alloy plate process parameter dynamic regulation system provided by the embodiments of the present application.

[0061] Label explanation: 1, information acquisition module; 2, candidate model screening module; 3, most optimal model determination module; 4, model adjustment module; 5, process parameter adjustment module. DETAILED DESCRIPTION

[0062] The technical model in the present application will be described clearly and completely in combination with the drawings in the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The components of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0063] It should be noted that similar reference numerals and letters refer to like items in the accompanying drawings, and once an item is defined in one drawing, it is not necessary to further define and explain it in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", and the like are only used to distinguish description, and cannot be understood as indicating or implying relative importance.

[0064] Reference Figure 1 The present application proposes an aluminum alloy plate process parameter dynamic regulation method for adjusting process parameters when a multi-roll straightening machine straightens an aluminum alloy plate. The steps of the method include:

[0065] A1. Obtain basic information of the aluminum alloy plate and real-time online detection data in the test stage; use preset process parameters to straighten the aluminum alloy plate in the test stage;

[0066] A2. According to the basic information, determine a group of candidate prediction models from a preset prediction model library; each prediction model in the prediction model library is used to predict the deformation response of the aluminum alloy plate according to the process parameters and the real-time online detection data;

[0067] A3. According to the real-time online detection data in the test stage, determine the optimal prediction model with the optimal prediction accuracy of the actual deformation response of the aluminum alloy plate from the candidate prediction models;

[0068] A4. Use the optimal prediction model to perform formal straightening on the aluminum alloy plate, and perform online adjustment on the optimal prediction model based on the real-time online detection data during the formal straightening;

[0069] A5. According to the real-time online detection data at the current time, adjust the process parameters online by using the online adjusted optimal prediction model.

[0070] Wherein, the basic information refers to the inherent properties of the aluminum alloy plate, such as alloy grade, specification, thickness, width, length and surface state, etc., which can be obtained by manual input, reading from the production management system or automatic identification by sensors, etc., which is mainly to provide preliminary basis for subsequent model selection.

[0071] Wherein, the real-time online detection data refers to the data reflecting the state of the plate and the straightening effect collected in real time during the straightening process, such as including the plate shape data before straightening, thickness data, temperature data (as input data) and plate shape data after straightening (as response data), which can be continuously collected by online plate shape instrument, thickness gauge, infrared temperature measuring instrument and other equipment, which is mainly to obtain data reflecting the real deformation characteristics of the current plate, and provide objective basis for model evaluation, selection and adjustment.

[0072] Among them, process parameters refer to the set values ​​used to control the straightening process, including the amount of reduction (relative position of each roller) and the straightening speed (sheet material conveying speed). These can be set and adjusted through the straightening machine control system. The main purpose is to achieve plastic deformation by regulating the stress state of the sheet material.

[0073] The prediction model library refers to a pre-established collection of prediction models of various types. Each prediction model is used to predict the deformation response of aluminum alloy plates based on process parameters and real-time online detection data. It can include models based on physical principles, data-driven machine learning models (such as neural networks and support vector machines), or a combination of both. Its main purpose is to cover the deformation laws under different plate properties and straightening conditions.

[0074] Among them, the candidate prediction model refers to a set of prediction models that may be applicable to the current board material, which are initially screened from the prediction model library based on basic information. They are determined from the model library through indexing or matching rules based on basic information. Their main purpose is to narrow down the scope of model evaluation and improve the efficiency of model selection.

[0075] Among them, the optimal prediction model refers to the model with the highest prediction accuracy for the actual deformation response of aluminum alloy plates after verification with actual data among the candidate prediction models. It is determined by comparing the error between the prediction results of each candidate model in the testing phase and the actual detection data. Its main purpose is to select the model that can most accurately describe the current deformation behavior of the plate.

[0076] Among them, deformation response refers to the change in plate shape or residual stress state of aluminum alloy plate under the action of specific process parameters. It is usually measured by plate shape data after straightening, and it is the core target that needs to be predicted and controlled in the straightening process.

[0077] Among them, online adjustment of the optimal prediction model refers to the dynamic correction of the internal parameters of the optimal prediction model based on the prediction deviation fed back from real-time detection data during the formal straightening process. It can be implemented by using algorithms based on error backpropagation, Kalman filtering, or other adaptive control algorithms. Its main purpose is to enable the model to adapt to small fluctuations and changes in the production process and improve the real-time accuracy and robustness of the model.

[0078] Among them, online adjustment of process parameters refers to using the optimal prediction model after online adjustment to solve the process parameters for straightening aluminum alloy plates during the formal straightening process.

[0079] The core innovation of the present application is that by introducing a model optimization in the test stage and an online adjustment mechanism in the formal straightening stage, the model library established offline is combined with the real-time data collected online, so that the dynamic and self-adaptive regulation of the straightening process parameters of aluminum alloy plates of different types and specifications is realized, the limitations caused by the dependence on fixed parameter table or inaccurate model in the traditional method are overcome, and the straightening precision and production efficiency are improved.

[0080] Specifically, the method first acquires the basic information of the aluminum alloy plate to be straightened, and uses the preset process parameters to test straighten the plate, during which real-time online detection data is continuously collected, which reflects the real deformation characteristics of the current plate under the actual straightening conditions. Based on the acquired basic information, the system preliminarily selects a group of candidate prediction models from the preset prediction model library, which are pre-constructed according to different plate characteristics. Then, the real-time online detection data collected in the test stage is used to evaluate the prediction accuracy of each candidate model, and the model with the highest prediction accuracy is selected as the optimal prediction model. In the formal straightening stage, the system uses the optimal prediction model to straighten the aluminum alloy plate, and continuously collects real-time online detection data. More importantly, based on the real-time data collected in the formal straightening process, the system adjusts the internal parameters of the optimal prediction model online, so that it can adapt to the actual changes in the production process. Finally, according to the real-time online detection data at the current time, the optimal prediction model after online adjustment is used to dynamically calculate and adjust the current process parameters, so as to realize accurate straightening control. The whole process forms a closed-loop regulation process of data-driven and model-adaptive.

[0081] Through the above scheme, the present application can dynamically select and adjust the prediction model according to the actual characteristics of the aluminum alloy plate and the real-time state in the production process, and then realize accurate and self-adaptive regulation of the straightening process parameters. This significantly improves the straightening accuracy and adaptability of plates of different alloys and specifications, reduces the dependence on a large number of offline experiments and empirical trial-and-error, shortens the specification switching time, reduces the scrap rate and production cost, and improves the flexibility and efficiency of the production line.

[0082] In some embodiments, the basic information includes alloy grade, specification, thickness, width, length, and surface state;

[0083] The real-time online detection data includes input data and response data; the input data includes pre-straightening plate shape data, thickness data, and temperature data; and the response data includes post-straightening plate shape data.

[0084] The process parameters include reduction and straightening speed; and the preset process parameters include preset reduction and preset straightening speed.

[0085] Among them, the alloy grade, specification, thickness, width, length and surface state are inherent properties of aluminum alloy plates, which affect the mechanical behavior and deformation response of the plate during straightening. Clarifying these basic information makes it possible to select the candidate model related to the current plate characteristics from the prediction model library in the subsequent steps, laying the foundation for determining the optimal model and making predictions.

[0086] Among them, the pre-straightening plate shape data, thickness data and temperature data are input variables that affect the straightening process and results. The pre-straightening plate shape is the initial state that needs to be corrected; the thickness is the basic geometric parameter of the plate; the temperature will affect the mechanical properties of the material. Obtaining these input data and using them as input for the prediction model in the subsequent steps enables the model to make predictions and adjustments based on the current plate state and environmental conditions. The post-straightening plate shape data is the result of the straightening process, reflecting the deformation response and straightening effect of the plate. Obtaining these response data and comparing them with the model prediction results enables the evaluation of the prediction accuracy of the model and the determination of the optimal model.

[0087] Among them, the reduction and straightening speed are the main control parameters of the multi-roll straightening machine, which determine the bending deformation degree and deformation rate of the plate. The prediction model is used to predict the plate shape response under given reduction and straightening speed, or to solve the reduction and straightening speed that can achieve the target plate shape. Clarifying these process parameters enables the method to directly act on the control variables of the straightening machine.

[0088] These clear information and parameters provide an accurate data basis for the establishment, selection, evaluation and online adjustment of the prediction model, enabling the prediction model to more accurately reflect the actual deformation response of the aluminum alloy plate. This helps to select the model with the optimal prediction accuracy for the actual deformation response, and enables effective adjustment of the model based on actual data, thereby improving the accuracy of the final process parameter adjustment and overcoming the limitations of dynamic regulation caused by incomplete information.

[0089] In some embodiments, step A1 comprises:

[0090] A101. Obtain the basic information of the aluminum alloy plate;

[0091] A102. Test straightening of the aluminum alloy plate using the preset process parameters, and continuously collect the real-time online detection data during the test straightening process to form a real-time online detection data sequence;

[0092] A103. Calculate the deformation response characteristic parameters of the aluminum alloy plate within the current test length range according to the real-time online detection data sequence;

[0093] A104. If the deformation response characteristic parameter does not satisfy the preset test end threshold condition, continue to collect the real-time online detection data and update the real-time online detection data sequence, and return to step A103, otherwise, end the test phase, and take the current real-time online detection data sequence as the real-time online detection data of the test phase.

[0094] wherein the deformation response characteristic parameter refers to a key index capable of characterizing the actual deformation behavior of the aluminum alloy plate under the current preset process parameters, which is extracted or calculated from the real-time online detection data sequence. These parameters can be statistical quantities, change rates, recognition results of specific patterns, etc. For example, the waviness, residual stress level, plate shape change trend, etc. of the plate can be calculated. These parameters are an abstraction and refinement of the original data sequence, and are used to more essentially reflect the mechanical response characteristics of the plate. The preset test end threshold condition refers to a standard preset for judging whether the test straightening phase can end. This condition is usually based on the deformation response characteristic parameter, for example, when the average value of the waviness is stable within a certain range, or the estimated value of the residual stress is lower than a certain threshold, it is considered that the test can end. This condition aims to ensure that the test is stopped in time after sufficient data reflecting the stable deformation characteristics of the plate is obtained. Based on the characteristic parameter to judge the end of the test and to collect in a loop refers to comparing the calculated deformation response characteristic parameter with the preset test end threshold condition. If the condition is not met, the system automatically controls to continue the test straightening and data collection, updates the data sequence, then recalculates the characteristic parameter and judges again, forming a feedback loop. Only when the characteristic parameter satisfies the end condition, the test phase is terminated. This mechanism enables the test process to adaptively determine the test length according to the actual deformation response of the plate, avoiding the shortcomings that may be caused by fixed length test.

[0095] In the present application, first, the basic information of the plate is acquired, which provides the basis for subsequent testing and model selection. Then, the plate is tested and straightened using the preset process parameters. During the testing process, the key is to continuously collect real-time online detection data and organize these data into a data sequence in order. This continuous collection and sequence formation enables the system to capture the dynamic deformation process of the plate within a certain length range through the straightening machine, rather than just the instantaneous state at a certain point. Next, the system calculates deformation response characteristic parameters that can reflect the actual deformation characteristics of the plate within the current testing length range based on this data sequence. These characteristic parameters are an effective extraction of the original data sequence and can more essentially depict the behavior of the plate under the preset parameters. Subsequently, the system enters a judgment cycle based on data feedback. It compares the calculated deformation response characteristic parameters with the preset test end threshold conditions. If the characteristic parameters have not yet reached the preset stable or convergent state, it indicates that the currently collected data may not be sufficient to fully reflect the stable deformation characteristics of the plate, and the system will control to continue testing and straightening and data collection, update the data sequence, and return to step A103 to recalculate the characteristic parameters and judge again. This cycle continues until the deformation response characteristic parameters meet the preset threshold conditions. Once the conditions are met, the system considers that sufficient stable and representative data has been acquired, and ends the testing phase, and takes the current complete data sequence as the real-time online detection data of the testing phase.

[0096] This adaptive test end mechanism based on deformation response characteristic parameters makes the testing process no longer dependent on fixed testing length or time, but dynamically adjusts the testing range according to the actual deformation response of the plate. This ensures that the testing time is as short as possible while obtaining effective and stable data, improving the testing efficiency. At the same time, the real-time online detection data sequence containing dynamic processes and stable states obtained, as well as the deformation response characteristic parameters extracted therefrom, can more accurately represent the actual behavior of the current plate under the preset process parameters. These high-quality data as inputs for subsequent steps, especially for model selection, optimal model determination, and subsequent online model adjustment, significantly improve the accuracy and robustness of the entire process parameter dynamic regulation method.

[0097] In some embodiments, step A2 comprises:

[0098] A201. According to the basic information, navigate in the preset prediction model library along a hierarchical index structure to locate a subset of prediction models related to the basic information; the hierarchical index structure is constructed in advance according to multiple information dimensions of the basic information;

[0099] A202. Based on the preset matching rules of the basic information and each prediction model in the prediction model subset, determine a set of prediction models with a matching degree higher than a preset matching degree threshold as the candidate prediction models.

[0100] Wherein, the hierarchical index structure refers to a tree-like or hierarchical data structure, whose nodes and levels are pre-constructed according to different dimensions of basic information (such as alloy grade, thickness range, width range, etc.). Navigation refers to searching along the path of the hierarchical index structure layer by layer downward according to the specific basic information of the aluminum alloy plate to be straightened until the leaf node or node set corresponding to these basic information is located. The prediction model subset refers to the set of prediction models located by navigating along the hierarchical index structure. The models in this subset have a higher relevance to the basic information of the current aluminum alloy plate.

[0101] Wherein, the preset matching rule refers to a standard or algorithm used to evaluate the relevance between the basic information of the aluminum alloy plate and each prediction model in the prediction model subset. The matching rule can be based on the data characteristics, model type, model complexity, and other factors related to the basic information used during model training.

[0102] Wherein, the matching degree refers to a numerical value calculated according to the preset matching rule, reflecting the relevance or applicability between the basic information of the current aluminum alloy plate and a certain prediction model. The higher the matching degree value, the more relevant the model is to the current plate.

[0103] Wherein, the preset matching degree threshold is a pre-set numerical limit. Only when the calculated matching degree is higher than this threshold, the corresponding prediction model will be considered as sufficiently relevant to the current plate, and thus be selected into the candidate prediction model set.

[0104] Specifically, the present scheme provides a method for determining candidate prediction models from a preset prediction model library. First, the basic information of the aluminum alloy plate to be straightened is obtained, which describes the inherent characteristics and specifications of the plate. Then, using these basic information, navigation is performed in the hierarchical index structure constructed in advance according to the dimension of the basic information. The hierarchical index structure organizes the huge model library according to alloy type, specification range, etc. Along the path corresponding to the current plate basic information, a subset of prediction models related to the characteristics of the current plate can be quickly located. This process avoids traversing the entire model library, significantly improving the efficiency of model positioning. Then, in the subset of prediction models located, according to the preset matching rule, the matching degree between the basic information of the current plate and each prediction model in the subset is evaluated, and the matching degree value is calculated. The matching rule considers factors such as the similarity between model training data and the basic information of the current plate. Finally, a preset matching degree threshold is set, and only those prediction models with a matching degree higher than the threshold are selected as candidate prediction models. These candidate models are considered to have a high correlation with the actual characteristics of the current plate and can more accurately predict its deformation response.

[0105] Through the combination of hierarchical navigation and matching degree screening, the scheme solves the problem of low screening efficiency directly from a huge model library, while ensuring the high relevance of the screened models to the current plate, laying a foundation for subsequent accurate prediction and dynamic control of process parameters.

[0106] Preferably, step A202 can include:

[0107] Based on the preset matching rule between the basic information and each prediction model in the prediction model subset, determine the basic matching degree between the basic information and each prediction model in the prediction model subset;

[0108] According to the real-time online detection data of the test stage, analyze the actual deformation response characteristics of the aluminum alloy plate;

[0109] For each prediction model in the prediction model subset, based on the actual deformation response characteristics, calculate the prediction matching degree of the prediction model to the actual deformation response;

[0110] For each prediction model in the prediction model subset, the corresponding basic matching degree and the corresponding prediction matching degree are fused to obtain a comprehensive matching degree;

[0111] Determine a group of prediction models whose comprehensive matching degree is higher than a preset matching degree threshold as the candidate prediction models.

[0112] The preset matching rule can be rule-based matching, such as setting matching conditions according to alloy type, thickness range, width range, and other dimensions; or a similarity calculation method, such as encoding basic information as a vector and calculating the distance or similarity between the vector and the model associated vector.

[0113] The basic matching degree refers to the matching degree value between the basic information of the aluminum alloy plate and the prediction model calculated according to the preset matching rule. It can be represented by percentage, score or other numerical forms, and the higher the value, the higher the matching degree.

[0114] The actual deformation response feature refers to the characteristics of the actual deformation behavior of the aluminum alloy plate in the test straightening process extracted or calculated from the real-time online detection data in the test stage. It can be represented by calculating the plate shape change, residual stress distribution, load-displacement curve characteristic parameters, etc. For example, the difference between the plate shape data before and after straightening, the curvature change at a specific position, or the stress-strain data can be analyzed.

[0115] The prediction matching degree refers to the value of evaluating the prediction ability of the prediction model for the actual deformation response characteristics of the aluminum alloy plate. It can be obtained by applying the prediction model to the input data in the test stage to obtain the predicted deformation response, then comparing the predicted result with the actual deformation response characteristics, calculating the prediction error or similarity, and converting it into a matching degree value. For example, the mean square error between the predicted plate shape and the actual plate shape can be calculated, and the smaller the mean square error, the higher the prediction matching degree.

[0116] The fusion refers to combining the basic matching degree and the prediction matching degree to form a comprehensive evaluation index. It can be obtained by weighted summation, product, fuzzy logic-based fusion or other multi-criteria decision methods. For example, weights can be set, the basic matching degree is multiplied by one weight, the prediction matching degree is multiplied by another weight, and the results are added to obtain the comprehensive matching degree. The weights can be determined by experience or training.

[0117] The preset matching degree threshold refers to the minimum comprehensive matching degree requirement for screening candidate prediction models. Only models with a comprehensive matching degree higher than this threshold will be selected as candidate prediction models. The threshold can be determined by experience or offline testing.

[0118] Specifically, in the test straightening stage of the aluminum alloy plate, first, the basic information of the plate such as alloy grade and specification is obtained. According to these basic information, in the preset prediction model library, a relevant prediction model subset is quickly located through a hierarchical index structure. For each prediction model in this subset, first, according to the preset rules or methods, the matching degree of the applicable range of the model with the current plate basic information is calculated to obtain the basic matching degree. At the same time, in the test straightening process, the real-time online detection data of the plate is continuously collected, such as the plate shape data before and after straightening. Using these real-time data, the actual deformation response characteristics of the current aluminum alloy plate under the test conditions are analyzed and extracted, such as the plate shape change amount, residual stress characteristics, etc. Then, for each model in the prediction model subset, the input data in the test stage is used to predict the deformation response of the plate through the model, and the prediction result is compared with the actual deformation response characteristics to calculate the prediction accuracy or similarity of the model to the actual deformation, and the prediction matching degree is obtained. Next, the basic matching degree and the prediction matching degree are fused, for example, a weighted average method is used to obtain the comprehensive matching degree of each model. Finally, a comprehensive matching degree threshold is set, and all prediction models with a comprehensive matching degree higher than the threshold are selected to form a candidate prediction model set. The models in this set are not only related to the basic properties of the plate, but also have good prediction ability in actual tests, so they are more likely to contain the most suitable prediction model for the current plate. In this way, the limitation of relying only on static basic information for matching is overcome, and the quality of the candidate model set is improved, laying a foundation for subsequent selection of the optimal model and accurate process parameter control.

[0119] Through the above technical solutions, the present application solves the problem that relying only on plate basic information for matching cannot fully reflect the actual deformation characteristics, resulting in an inaccurate candidate model set. By introducing the analysis of the actual deformation response characteristics in the test stage and fusing it with the matching degree based on the basic information, the applicability of the prediction model can be more comprehensively and accurately evaluated. As a result, the selected candidate prediction model set can more effectively reflect the prediction ability of the model to the actual deformation behavior of the current plate, improving the quality of the candidate model and providing a more reliable foundation for the selection of the optimal model, thereby improving the precision and robustness of the entire process parameter dynamic control method.

[0120] In some embodiments, step A3 comprises:

[0121] A301. Based on the input data in the real-time online detection data of the test stage and the preset process parameters used in the test stage, the deformation response of the aluminum alloy plate is predicted using each of the candidate prediction models;

[0122] A302. Compare the deformation response predicted by each of the candidate prediction models with the actual deformation response reflected by the real-time online detection data in the test stage, and calculate the prediction error of each of the candidate prediction models;

[0123] A303. Determine the candidate prediction model with the minimum prediction error as the optimal prediction model.

[0124] Wherein, the deformation response of the aluminum alloy plate predicted refers to the state of the aluminum alloy plate after straightening or the change of the state calculated by the prediction model according to the input process parameters and plate state data, for example, the predicted plate shape data after straightening or the change of the plate shape data after straightening relative to the plate shape data before straightening.

[0125] Wherein, the calculation of the prediction error refers to the process of quantifying the difference between the model prediction result and the actual observation result, which can use root mean square error, mean absolute error, maximum absolute error, etc.

[0126] The specific implementation steps for determining the optimal prediction model are provided, and the problem of how to objectively select the optimal model based on test data from multiple candidate prediction models is solved. Specifically, step A301 inputs the actual test conditions, i.e., the pre-set process parameters used in the test stage and the collected input data, into each candidate prediction model, so that each model can output its predicted deformation response for the same actual working condition. This step provides a unified prediction result set based on actual test data for evaluating the prediction ability of each model. Step A302 compares the deformation response predicted by each candidate model in step A301 with the actual deformation response of the aluminum alloy plate reflected in the real-time online detection data in the test stage. Through this comparison, the difference between the prediction result of each model and the actual observation result can be quantified, i.e., the prediction error of each model is calculated. This step provides an objective index for evaluating the prediction accuracy of the model, so that the performance of different models can be directly compared. Step A303 selects the candidate prediction model with the minimum prediction error as the optimal prediction model based on the prediction error calculated in step A302. The minimum prediction error means that the model has the highest prediction accuracy for the actual deformation response of the current aluminum alloy plate in the test stage. By selecting the model with the highest prediction accuracy, it can be ensured that the model used in the subsequent formal straightening stage can best reflect the actual deformation characteristics of the current plate, laying a foundation for accurate process parameter regulation.

[0127] In combination with the test stage environment and candidate model set provided by the prior scheme, the present scheme can find the model that can most accurately predict the deformation behavior of the specific plate being processed from the preselected model library, overcoming the problem of the traditional method relying on a fixed parameter table or a simplified model that is not accurate, improving the adaptability of the model to the actual plate characteristics, and thus providing a more reliable basis for subsequent online regulation. This model evaluation and selection mechanism based on actual test data is the basis for the entire dynamic regulation method to effectively cope with different plate characteristics and production fluctuations.

[0128] In one embodiment, the step of determining the optimal prediction model can be implemented as follows: first, in step A301, the preset reduction, preset straightening speed, pre-straightening shape data, thickness data, temperature data, etc. collected in the test stage are input as inputs into each candidate prediction model. Each model outputs a predicted post-straightening shape data according to its internal algorithm and parameters. Then, in step A302, the predicted post-straightening shape data output by each candidate model is compared point by point with the actual post-straightening shape data detected in the test stage, and the difference between the two is calculated. The root mean square error (RMSE) can be used as a measure of prediction error, i.e., the square root of the average of the squares of the differences between the predicted value and the actual value. Finally, in step A303, after calculating the RMSE values of all candidate prediction models, the model with the smallest RMSE value is found, which is determined as the optimal prediction model.

[0129] Through the above steps, the present scheme can objectively and quantitatively evaluate the prediction accuracy of each candidate prediction model under actual test conditions. Selecting the model with the smallest prediction error ensures that the selected model can most accurately reflect the deformation characteristics of the current plate under actual working conditions. This provides a more accurate prediction basis for subsequent online regulation of process parameters based on the model in the formal straightening stage, thereby improving the straightening effect and product quality.

[0130] In some embodiments, step A4 comprises:

[0131] A401. After starting the formal straightening stage, an initial process parameter is determined based on the first detected input data using the optimal prediction model, and real-time online detection data is collected during the formal straightening to obtain real-time online detection data during the formal straightening;

[0132] A402. The real-time online detection data at the current time is extracted from the real-time online detection data during the formal straightening to determine the actual deformation response of the aluminum alloy plate at the current time;

[0133] A403. Based on the input data in the real-time online detection data and the process parameters at the current time, the deformation response of the aluminum alloy plate is predicted using the current optimal prediction model to obtain a predicted deformation response at the current time;

[0134] A404. The actual deformation response at the current time is compared with the predicted deformation response at the current time, and a prediction deviation at the current time is calculated;

[0135] A405. Based on the prediction deviation at the current time and a preset adjustment rule, an adjustment amount for the internal parameters of the current optimal prediction model is calculated;

[0136] A406. According to the adjustment amount, the internal parameters of the current optimal prediction model are updated to obtain an online-adjusted optimal prediction model.

[0137] wherein the internal parameters of the optimal prediction model refer to variable values or weights in a mathematical expression or algorithm structure of the optimal prediction model, which determine how the model maps the input data and the process parameters to the predicted deformation response, and can be implemented by connection weights and biases of a neural network, coefficients of a regression model, or material property parameters in a physical model.

[0138] wherein the adjustment rule refers to a preset logic or algorithm for guiding the system how to calculate the adjustment amount for the internal parameters of the model according to the prediction deviation, which can be implemented by an optimization algorithm based on error gradient (such as gradient descent), a rule-based expert system, or an adaptive control law. The adjustment amount refers to a specific numerical correction for the internal parameters of the optimal prediction model, which is calculated according to the adjustment rule, and can be represented in the form of a vector or a matrix corresponding to each parameter to be updated in the model.

[0139] The present scheme elaborates how to adjust the optimal prediction model determined in the test phase online in the formal straightening phase, to ensure that the model can accurately predict the deformation response of aluminum alloy plates in the actual production process, so as to maintain or improve the straightening effect. Specifically, after starting the formal straightening phase, the system first calculates the initial process parameters based on the first detected input data using the optimal prediction model determined in the early test phase, providing a starting point for formal production based on the preferred model. At the same time, real-time online detection data is continuously collected during the entire formal straightening process, which is the basis for subsequent model online adjustment and dynamic process parameter control. The system extracts real-time online detection data at the current time from the continuously collected formal straightening data stream, which contains the plate shape information after straightening at the current time, directly reflecting the actual deformation result of the aluminum alloy plate under the current process parameters, i.e. the actual deformation response at the current time. Then, the system uses the current process parameters actually applied and the input data at the current time (such as plate shape before straightening, thickness, temperature, etc.) to input these information into the current version of the optimal prediction model (which may have been adjusted), and the prediction model believes that the deformation response of the plate should be generated under these conditions, obtaining the model predicted deformation response at the current time. By comparing the actual deformation response at the current time with the model predicted deformation response at the current time, the system calculates the difference between the two, i.e. the prediction deviation at the current time. This deviation quantifies the prediction error of the current optimal prediction model under the current production state. Based on the calculated prediction deviation at the current time, and combined with the preset adjustment rule, the system calculates the specific adjustment amount required for the internal parameters of the optimal prediction model. The adjustment rule here is a pre-set strategy that guides the system on how to correct the model according to the prediction error, for example, the greater the error, the greater the adjustment amount, or different adjustment methods are used according to the nature of the error (such as systematic deviation or random fluctuation). This step converts the prediction error into a correction instruction for the model parameters. Finally, the system updates the internal parameters of the optimal prediction model according to the calculated adjustment amount. By updating the parameters of the model, the model can better fit the actual deformation response at the current time, thereby reducing the future prediction deviation. The updated model becomes the "online adjusted optimal prediction model", which will be used for prediction and process parameter calculation at subsequent times, forming a closed loop of continuous learning and adaptation.

[0140] Through the cyclic execution of the above steps, the scheme realizes an online model adjustment mechanism based on real-time prediction deviation, so that the optimal prediction model can dynamically adapt to various uncertainties and changes in the actual production process, maintain high prediction accuracy, and thus provide more accurate prediction basis for subsequent model-based dynamic process parameter regulation. The combination of this online adjustment mechanism and the process of selecting the optimal model through test data makes the entire process parameter regulation method take into account both the generalization ability of offline optimization and the real-time performance of online adaptation, thereby improving the system's adaptability to different plates and changing working conditions.

[0141] Preferably, step A405 can include:

[0142] According to the prediction deviation of the current moment and the real-time online detection data at the time of formal straightening, the current production stage is determined; the production stage includes an initial transient stage and a steady state stage.

[0143] According to the determined production stage, the corresponding adjustment rule is selected from the preset adjustment rule library; the adjustment rule library includes at least one adjustment rule suitable for the initial transient stage and at least one adjustment rule suitable for the steady state stage.

[0144] Based on the prediction deviation of the current moment and the selected adjustment rule, the adjustment amount of the internal multiple related parameters of the current optimal prediction model is calculated as the adjustment amount.

[0145] Wherein, the production stage refers to the different time or space intervals during the formal straightening process of the aluminum alloy plate in the multi-roll straightening machine, which is divided according to the plate deformation behavior and the model prediction deviation characteristics. It can be defined according to the processing length after straightening, processing time, prediction deviation size or change rate, etc. The initial transient stage refers to the early stage when the straightening process just starts, the plate shape changes rapidly, and the prediction deviation may be large. The steady state stage refers to the later stage after the straightening process has been carried out for a period of time, the plate shape tends to be stable, and the prediction deviation fluctuation is small.

[0146] Wherein, the adjustment rule library refers to a series of algorithms, formulas or lookup table sets pre-stored for calculating the model parameter adjustment amount according to the prediction deviation, which can include multiple adjustment strategies optimized for different production stages. The adjustment rule refers to a specific strategy or algorithm in the adjustment rule library, which is used to guide how to convert the prediction deviation into the adjustment amount of the model parameters, which can be a mathematical function, a set of parameters, a decision tree or a small controller.

[0147] The synergistic adjustment amount refers to the adjustment of multiple interrelated parameters in the optimal prediction model. The adjustment amount of these parameters is not calculated in isolation, but considers their mutual influence to achieve an overall optimal adjustment effect. It can be a vector or a set of related values.

[0148] The present scheme addresses the problem that the adjustment rules used to calculate the model parameter adjustment amount may not adapt to different production stages when adjusting the optimal prediction model online. Specifically, first, the current production stage is determined based on the prediction deviation at the current time and the real-time online detection data during the formal straightening. By analyzing the size, trend of the prediction deviation, and the plate shape changes reflected by the real-time online detection data, the system can identify whether the current straightening process is in the initial transient stage of just starting straightening and rapid convergence of plate shape, or in the steady stage of already stable and small plate shape fluctuations. This judgment is necessary because the plate deformation behavior and model prediction deviation characteristics are different in different stages, and different model adjustment strategies are needed. Then, the corresponding adjustment rule is selected from the pre-set adjustment rule library according to the determined production stage. The adjustment rule library pre-stores adjustment rules optimized for different production stages. For example, the rule suitable for the initial transient stage may focus more on quickly reducing large prediction deviations, while the rule suitable for the steady stage may focus more on maintaining the stability of the model and suppressing small fluctuations. By selecting the most appropriate adjustment rule according to the current stage, the effectiveness and pertinence of the model adjustment strategy can be ensured. Finally, based on the prediction deviation at the current time and the selected adjustment rule, the synergistic adjustment amount of multiple related parameters in the optimal prediction model is calculated as the adjustment amount. Here, the prediction deviation at the current time is used as the basis for adjustment, combined with the adjustment rule selected for the current production stage, to calculate the adjustment amount needed for the internal parameters of the optimal prediction model. The emphasis on synergistic adjustment of multiple related parameters means that the adjustment process considers the mutual influence between different parameters in the model, and performs overall optimization adjustment rather than isolated adjustment of individual parameters.

[0149] The current production stage is judged according to the prediction deviation at the current moment and the real-time online detection data at the formal straightening time, a corresponding adjustment rule is selected from the preset adjustment rule library according to the judged production stage, and a coordinated adjustment amount of a plurality of related parameters in the optimal prediction model is calculated based on the prediction deviation at the current moment and the selected adjustment rule, so that the online adjustment strategy of the optimal prediction model can dynamically adapt to different stages of the straightening process. This overcomes the problem that the fixed adjustment rule cannot effectively cope with the deformation characteristics of the plate and the model deviation characteristics in different production stages, improves the accuracy, stability and adaptability of the model online adjustment, so that the optimal prediction model after online adjustment can more accurately predict the deformation response of the plate, and provides a more reliable basis for subsequent online adjustment of process parameters.

[0150] In some embodiments, step A5 comprises:

[0151] A501. determining a target deformation response according to the real-time online detection data at the current moment and a preset straightening requirement;

[0152] A502. solving a process parameter that can make the predicted deformation response of the online adjusted optimal prediction model reach the target deformation response under the input data condition in the real-time online detection data at the current moment, as a target process parameter, by using the online adjusted optimal prediction model;

[0153] A503. adjusting the process parameters according to the target process parameters.

[0154] Wherein, the target deformation response refers to the deformation state or the plate shape quality index that the aluminum alloy plate is expected to reach after straightening, which is determined according to the current state of the plate reflected by the real-time online detection data at the current moment and the preset straightening requirement, and can adopt target plate shape flatness, target residual stress level or specific plate shape curve as a representation.

[0155] Wherein, solving a process parameter that can make the predicted deformation response of the online adjusted optimal prediction model reach the target deformation response refers to using mathematical or calculation methods, based on the online adjusted optimal prediction model, taking the input data in the real-time online detection data at the current moment as part of the model input, and setting the target deformation response as the expected output of the model, to reversely calculate the process parameter combination that can make the model output reach the expected value, which can be realized by using numerical optimization algorithm, model inversion technology or method based on lookup table. The target process parameter refers to the process parameter value calculated by the above solving process, which is used to guide the parameter setting of the straightening machine at the current moment, and represents the accurate setting value of the reduction amount and the straightening speed and other parameters required to achieve the target deformation response.

[0156] The scheme accurately determines the process parameters required to achieve the target by introducing the concept of target deformation response and using an online-adjusted prediction model for inverse solving. Specifically, first, according to the real-time online detection data at the current time, combined with the preset straightening requirements, a specific expected deformation state reflecting the current actual situation and the final straightening target, i.e., the target deformation response, is determined. This step provides a clear control target for subsequent parameter calculation. Determining the target deformation response according to the real-time online detection data at the current time makes the target more in line with the current actual production state, rather than relying solely on the preset ideal requirements. Then, using an online-adjusted optimal prediction model that can more accurately reflect the current plate characteristics and equipment state, and taking the input data in the real-time online detection data at the current time as the input condition of the model, inverse solving is performed. The prediction model is usually used to predict the deformation response based on the process parameters and input data, while this step is to solve the process parameters that can produce the expected response (target deformation response) and the current input data. Using the online-adjusted optimal prediction model for solving ensures that the calculated target process parameters are based on the most accurate model at the current time, thereby improving the reliability of the parameters. Solving under the input data condition in the real-time online detection data at the current time ensures that the calculation result is applicable to the current actual production state. The process parameters obtained by solving are the target process parameters. Finally, according to the calculated target process parameters, the process parameters of the straightening machine are actually adjusted. This is the process of converting theoretical calculation results into actual control instructions. Adjusting according to the target process parameters means that the adjustment is based on accurate calculation results, rather than simple empirical adjustment or fixed rules, thereby being able to more effectively achieve the set target deformation response, improve straightening accuracy and effect, and better cope with plate performance fluctuations and equipment state changes. By combining real-time detection data, online-adjusted models, and target-oriented inverse solving, the scheme can achieve fine, self-adaptive dynamic regulation and control of the aluminum alloy plate straightening process.

[0157] In one specific embodiment, according to the real-time online detection data at the current time, such as the real-time plate shape data of the current plate (e.g., wave height, edge wave, or mid-wave degree) and the preset straightening requirements (e.g., final plate shape flatness level), a target deformation response can be determined, such as reducing the current wave height to a certain specific value or achieving a certain target residual stress distribution. Then, using the online adjusted optimal prediction model, which can be a neural network model corrected by real-time data, under the input data conditions in the real-time online detection data at the current time, such as the thickness, width, temperature of the current plate, and the plate shape data before straightening, by running an optimization algorithm, such as a gradient-based optimization method, the process parameter inputs of the model (such as the roll reduction and straightening speed) are iteratively adjusted until the error between the predicted deformation response of the model (such as the plate shape data after straightening) and the set target deformation response is less than the preset threshold. At this time, the process parameter input is determined as the target process parameter. Finally, the calculated target reduction and straightening speed are sent to the control system of the straightening machine, and the control system drives the actuator (such as a hydraulic cylinder or a motor) to accurately adjust the relative position of the straightening roll and the conveying speed of the plate, thereby realizing the adjustment of the process parameters.

[0158] Through the above technical solution, the process parameters required to achieve the target can be accurately calculated according to the real-time state of the plate and the straightening target, overcoming the limitations of traditional methods relying on fixed parameter tables or simple feedback control. Using the online adjusted prediction model for inverse solving ensures the accuracy of parameter calculation, better adapts to plate performance fluctuations and equipment state changes, improves the dynamic adaptability and control accuracy of the straightening process, and thus obtains more flat and stable quality aluminum alloy plates.

[0159] Reference Figure 2 The present application provides an aluminum alloy plate process parameter dynamic regulation system for adjusting the process parameters of a multi-roll straightening machine when straightening an aluminum alloy plate. The system comprises:

[0160] An information acquisition module 1 is used to acquire the basic information of the aluminum alloy plate and the real-time online detection data in the test phase. The aluminum alloy plate is straightened using preset process parameters in the test phase (the specific process can refer to step A1 in the foregoing description).

[0161] A candidate model screening module 2 is used to determine a group of candidate prediction models from a preset prediction model library according to the basic information. Each prediction model in the prediction model library is used to predict the deformation response of the aluminum alloy plate according to the process parameters and the real-time online detection data (the specific process can refer to step A2 in the foregoing description).

[0162] An optimal model determining module 3 is configured to determine, from the real-time online detection data in the test stage, an optimal prediction model with the optimal prediction accuracy for the actual deformation response of the aluminum alloy plate among the candidate prediction models (for details, refer to step A3 in the foregoing description);

[0163] A model adjusting module 4 is configured to use the optimal prediction model to perform formal straightening on the aluminum alloy plate, and to perform online adjustment on the optimal prediction model based on the real-time online detection data in the formal straightening (for details, refer to step A4 in the foregoing description);

[0164] A process parameter adjusting module 5 is configured to use the optimal prediction model after the online adjustment to perform online adjustment on the process parameters according to the real-time online detection data at the current time (for details, refer to step A5 in the foregoing description).

[0165] The above merely describes the embodiments of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for dynamically adjusting process parameters of an aluminum alloy plate, used for adjusting process parameters when a multi-roll straightening machine straightens an aluminum alloy plate, characterized in that, The method comprises the following steps: A1. Obtain basic information of an aluminum alloy plate and real-time online detection data in a test stage; use preset process parameters to straighten the aluminum alloy plate in the test stage; A2. According to the basic information, determine a set of candidate prediction models from a preset prediction model library; each prediction model in the prediction model library is used to predict the deformation response of the aluminum alloy plate according to process parameters and real-time online detection data; A3. According to the real-time online detection data in the test stage, determine, in each of the candidate prediction models, an optimal prediction model with the optimal prediction accuracy of the actual deformation response of the aluminum alloy plate; A4. Use the optimal prediction model to perform formal straightening on the aluminum alloy plate, and perform online adjustment on the optimal prediction model based on real-time online detection data during the formal straightening; A5. Use the online adjusted optimal prediction model to adjust process parameters online according to real-time online detection data at the current time; The basic information includes alloy grade, specification, thickness, width, length and surface state; The real-time online detection data includes input data and response data; the input data includes pre-straightening plate shape data, thickness data and temperature data; the response data includes post-straightening plate shape data; The process parameters include reduction and straightening speed; the preset process parameters include preset reduction and preset straightening speed; Step A2 comprises: A201. According to the basic information, navigate along a hierarchical index structure in the preset prediction model library to locate a prediction model subset related to the basic information; the hierarchical index structure is constructed in advance according to multiple information dimensions of the basic information; A202. Based on preset matching rules of the basic information and each prediction model in the prediction model subset, determine a set of prediction models with a matching degree higher than a preset matching degree threshold as the candidate prediction models; Step A202 comprises: Based on the preset matching rules of the basic information and each prediction model in the prediction model subset, determine the basic matching degree between the basic information and each prediction model in the prediction model subset; According to the real-time online detection data in the test stage, analyze the actual deformation response characteristics of the aluminum alloy plate; For each prediction model in the prediction model subset, calculate the prediction matching degree of the actual deformation response of the prediction model based on the actual deformation response characteristics; For each prediction model in the prediction model subset, fuse the corresponding basic matching degree and the corresponding prediction matching degree to obtain a comprehensive matching degree; Determine a set of prediction models with a comprehensive matching degree higher than a preset matching degree threshold as the candidate prediction models.

2. The method for dynamic control of process parameters of aluminum alloy plates according to claim 1, characterized in that, Step A1 comprises: A101. Obtain the basic information of the aluminum alloy plate; A102. Use the preset process parameters to test straighten the aluminum alloy plate, and continuously collect the real-time online detection data to form a real-time online detection data sequence during the test straightening. A103. According to the real-time online detection data sequence, a deformation response characteristic parameter of the aluminum alloy plate in a current test length range is calculated; A104. If the deformation response characteristic parameter does not satisfy a preset test end threshold condition, the real-time online detection data is continuously collected and the real-time online detection data sequence is updated, and the step A103 is returned, otherwise, a test phase is ended, and the current real-time online detection data sequence is taken as real-time online detection data of the test phase.

3. The method for dynamic control of process parameters of aluminum alloy plates according to claim 1, characterized in that, The step A3 comprises: A301. Based on the preset process parameters used in the test phase and input data in the real-time online detection data of the test phase, a deformation response of the aluminum alloy plate is predicted by using each candidate prediction model; A302. The deformation response predicted by each candidate prediction model is compared with an actual deformation response reflected by the real-time online detection data of the test phase, and a prediction error of each candidate prediction model is calculated; A303. The candidate prediction model with the minimum prediction error is determined as the optimal prediction model.

4. The method for dynamic control of process parameters of aluminum alloy plate according to claim 1, characterized in that, The step A4 comprises: A401. After starting a formal straightening phase, an initial process parameter is determined based on the first detected input data by using the optimal prediction model, and real-time online detection data in the formal straightening process is collected to obtain real-time online detection data in the formal straightening; A402. Real-time online detection data at a current time is extracted from the real-time online detection data in the formal straightening, and is used to determine an actual deformation response of the aluminum alloy plate at the current time; A403. Based on a current process parameter and input data in the real-time online detection data at the current time, a deformation response of the aluminum alloy plate is predicted by using the current optimal prediction model to obtain a predicted deformation response at the current time; A404. The actual deformation response at the current time is compared with the predicted deformation response at the current time, and a prediction deviation at the current time is calculated; A405. Based on the prediction deviation at the current time and a preset adjustment rule, an adjustment amount for internal parameters of the current optimal prediction model is calculated; A406. The internal parameters of the current optimal prediction model are updated according to the adjustment amount to obtain an online adjusted optimal prediction model.

5. The method for dynamic control of process parameters of aluminum alloy plate according to claim 4, characterized in that, The step A405 comprises: According to the prediction deviation at the current time and the real-time online detection data in the formal straightening, a current production phase is judged; the production phase comprises an initial transient phase and a steady state phase; According to the judged production phase, a corresponding adjustment rule is selected from a preset adjustment rule library; the adjustment rule library comprises at least one adjustment rule suitable for the initial transient phase and at least one adjustment rule suitable for the steady state phase; Based on the prediction deviation at the current time and the selected adjustment rule, a synergistic adjustment amount for multiple related parameters of the current optimal prediction model is calculated as the adjustment amount.

6. The method of claim 1, wherein the process parameters are dynamically adjusted based on the measured properties of the aluminum alloy sheet. The step A5 comprises: A501. According to the real-time online detection data at the current time and a preset straightening requirement, a target deformation response is determined; A502. Using the online adjusted optimal prediction model, under the input data condition in the real-time online detection data at the current time, solve the process parameters that can make the prediction deformation response of the online adjusted optimal prediction model reach the target deformation response as the target process parameters; A503. Perform process parameter adjustment according to the target process parameters.

7. An aluminum alloy plate process parameter dynamic regulation system for adjusting process parameters when a multi-roll straightening machine straightens an aluminum alloy plate, characterized in that, The system comprises: An information acquisition module is configured to acquire basic information of an aluminum alloy plate and real-time online detection data in a test phase; the aluminum alloy plate is straightened using preset process parameters in the test phase; A candidate model screening module is configured to determine a group of candidate prediction models from a preset prediction model library according to the basic information; each prediction model in the prediction model library is configured to predict a deformation response of the aluminum alloy plate according to process parameters and real-time online detection data; An optimal model determination module is configured to determine, among the candidate prediction models, an optimal prediction model with the optimal prediction accuracy for the actual deformation response of the aluminum alloy plate according to the real-time online detection data in the test phase; A model adjustment module is configured to perform formal straightening of the aluminum alloy plate using the optimal prediction model and perform online adjustment of the optimal prediction model based on real-time online detection data during the formal straightening; A process parameter adjustment module is configured to perform online adjustment of process parameters using the online adjusted optimal prediction model according to real-time online detection data at the current time; The basic information includes alloy grade, specification, thickness, width, length, and surface state; The real-time online detection data includes input data and response data; the input data includes pre-straightening plate shape data, thickness data, and temperature data; and the response data includes post-straightening plate shape data; The process parameters include reduction and straightening speed; and the preset process parameters include preset reduction and preset straightening speed; When the candidate model screening module determines a group of candidate prediction models from a preset prediction model library according to the basic information, the following is performed: A201. According to the basic information, navigate along a hierarchical index structure in the preset prediction model library to locate a prediction model subset related to the basic information; the hierarchical index structure is constructed in advance according to multiple information dimensions of the basic information; A202. Based on a preset matching rule between the basic information and each prediction model in the prediction model subset, determine a group of prediction models with a matching degree higher than a preset matching degree threshold as the candidate prediction models; Step A202 comprises: Based on a preset matching rule between the basic information and each prediction model in the prediction model subset, determine a basic matching degree between the basic information and each prediction model in the prediction model subset; According to the real-time online detection data in the test phase, analyze the actual deformation response characteristics of the aluminum alloy plate; For each prediction model in the prediction model subset, based on the actual deformation response characteristics, calculate a prediction matching degree of the prediction model for the actual deformation response; For each of the prediction model subsets, the corresponding basic matching degree and the corresponding prediction matching degree are fused to obtain a comprehensive matching degree; A group of prediction models whose comprehensive matching degrees are higher than a preset matching degree threshold are determined as the candidate prediction models.

Images