Duplex steel forming quality prediction and process adjustment method

Abstract

The present application provides a method for predicting the forming quality and adjusting the process of dual-phase steel, which relates to the field of dual-phase steel forming technology, including: establishing a multi-dimensional data acquisition network; obtaining a combination of process parameters; combining the process parameter combination in the temperature field simulation space and the stress field simulation space to perform mapping correlation analysis on the dual-phase steel forming quality fluctuation; dividing the process parameter combination; using the preset forming quality standard as a constraint condition, the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components are trained and adjusted. This application can solve the technical problem in the prior art that the forming quality fluctuation is difficult to accurately control due to the complex interaction of multiple process parameters in the dual-phase steel forming process. By establishing a multi-dimensional data acquisition network and a simulation space, accurate prediction of the forming quality and intelligent adjustment of the process parameters are achieved, ensuring the stability of product quality.

Description

Technical Field

[0001] The present application relates to the technical field of dual-phase steel forming, and in particular to a method for predicting the quality of dual-phase steel forming and adjusting the process. Background Art

[0002] Duplex steel is a high-strength, low-alloy steel composed primarily of ferrite and a small amount of martensite. Its excellent combination of strength and ductility makes it widely used in the automotive, construction, and manufacturing industries, particularly in the forming of automotive bodies and structural components. The forming process of duplex steel involves multiple variables, including forming temperature, pressure, time, material type, proportion, and purity. These process parameters interact in complex ways, and the impact of different process parameters on the final forming quality is non-linear, making fluctuations in forming quality difficult to precisely control. While common process parameters such as forming temperature, pressure, and time can affect steel properties within a certain range, the complex interactions between these process parameters in actual production make it difficult to account for all influencing factors, resulting in unstable forming quality. In particular, during the production process, slight changes in factors such as forming temperature and pressure often lead to fluctuations in forming quality, introducing significant uncertainty into production.

[0003] In summary, the existing technology has a technical problem that the forming quality fluctuation is difficult to accurately control due to the complex interaction of multiple process parameters in the dual-phase steel forming process. Summary of the Invention

[0004] The purpose of this application is to provide a method for predicting the forming quality and adjusting the process of dual-phase steel, so as to solve the technical problem in the prior art that the forming quality fluctuations are difficult to accurately control due to the complex interaction of multiple process parameters in the dual-phase steel forming process.

[0005] In view of the above problems, the present application provides a method for predicting the forming quality and adjusting the process of dual-phase steel, wherein the method comprises: establishing a multi-dimensional data acquisition network, monitoring the forming process of the dual-phase steel in real time, fitting the temperature field simulation space and the stress field simulation space; obtaining a combination of process parameters under standard environmental conditions, wherein the combination of process parameters includes the forming temperature, forming pressure, and forming time corresponding to the forming process indicators, and the material type, material ratio, and material purity corresponding to the material composition indicators; in the temperature field simulation space, mapping and correlating the fluctuation of the forming quality of the dual-phase steel with the combination of the process parameters, and outputting a plurality of first influencing characteristics corresponding to the forming process indicators; in the stress field simulation space, In the simulation space, the process parameter combination is combined with the dual-phase steel forming quality fluctuation to perform mapping correlation analysis, and multiple second influencing characteristics corresponding to the material composition indicators are output; according to the multiple first influencing characteristics and the multiple second influencing characteristics, the process parameter combination is divided to obtain the first type of process parameter components, the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components, the first type of process parameter components being parameter combinations whose forming quality grades meet the preset forming quality standards; with the preset forming quality standards as constraints, the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components are learned and trained through the first type of process parameter components, and process parameter adjustments are performed.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By establishing a multi-dimensional data acquisition network, the dual-phase steel forming process is monitored in real time, and the temperature field simulation space and the stress field simulation space are fitted; under standard environmental conditions, a process parameter combination is obtained, and the process parameter combination includes the forming temperature, forming pressure, and forming time corresponding to the forming process indicators, and the material type, material ratio, and material purity corresponding to the material composition indicators; in the temperature field simulation space, the process parameter combination is combined with the dual-phase steel forming quality fluctuation to perform mapping correlation analysis, and output a plurality of first influencing characteristics corresponding to the forming process indicators; in the stress field simulation space, the process parameter combination is combined with the dual-phase steel forming quality fluctuation to perform mapping correlation analysis. A mapping correlation analysis is performed to output multiple second influencing characteristics corresponding to the material composition indicators; based on the multiple first influencing characteristics and the multiple second influencing characteristics, the process parameter combination is divided to obtain a first type of process parameter component, a second type of process parameter component, a third type of process parameter component, and a fourth type of process parameter component, wherein the first type of process parameter component is a parameter combination whose forming quality level meets the preset forming quality standard; with the preset forming quality standard as a constraint condition, the second type of process parameter component, the third type of process parameter component, and the fourth type of process parameter component are trained through the first type of process parameter component, and process parameter adjustment is performed. In other words, by establishing a multi-dimensional data acquisition network, the key parameters in the dual-phase steel forming process are monitored in real time; in the temperature field and stress field simulation space, a mapping correlation analysis is performed in combination with the process parameter combination to output the influencing characteristics; the process parameter combination is divided and trained based on the influencing characteristics, and process parameter adjustment is performed, thereby achieving accurate prediction of forming quality and intelligent adjustment of process parameters, ensuring the stability of product quality.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and a person of ordinary skill in the art can obtain other drawings based on the provided drawings without creative work.

[0010] Figure 1Schematic diagram of the process of the method for predicting the forming quality and adjusting the process of duplex steel in this application;

[0011] Figure 2 This is a flow chart of loss calculation in the dual-phase steel forming quality prediction and process adjustment method of this application. DETAILED DESCRIPTION

[0012] This application solves the technical problem in the prior art of difficult-to-precise control of forming quality fluctuations due to the complex interactions between multiple process parameters in the forming process of dual-phase steel by providing a method for predicting the forming quality and adjusting the process. By establishing a multi-dimensional data acquisition network, key parameters in the forming process of dual-phase steel are monitored in real time; in the temperature field and stress field simulation space, mapping and correlation analysis are performed on the process parameter combinations to output influencing characteristics; and process parameter combinations are divided and trained based on the influencing characteristics, and process parameter adjustments are performed. This achieves accurate prediction of forming quality and intelligent adjustment of process parameters, ensuring the stability of product quality.

[0013] Below, the technical solutions in this application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited to the example embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should also be noted that, for the convenience of description, only the parts related to this application, rather than all of them, are shown in the accompanying drawings.

[0014] For examples, please see the attached Figure 1 The present application provides a method for predicting the forming quality and adjusting the process of dual-phase steel, wherein the method comprises the following steps:

[0015] Step 1: Establish a multi-dimensional data acquisition network to monitor the dual-phase steel forming process in real time, and fit the temperature field simulation space and stress field simulation space.

[0016] Specifically, a multidimensional data acquisition network is one that collects multidimensional data in real time during the molding process. Sensors installed on production equipment (such as temperature sensors, pressure sensors, and strain gauges) capture critical data from the molding process in real time and transmit it to an analysis system, ensuring accurate data on molding temperature, stress, and pressure every second. Sensors should be deployed across key areas of the molding equipment, such as the mold interior and the molding machine's working area. Multidimensional data typically includes molding temperature, pressure, stress distribution, and displacement. For example, thermocouples capable of measuring high temperatures, such as K-type thermocouples, should be selected, ensuring their measurement range covers the temperature range of the molding process. Multiple temperature sensors should be placed on the mold and material surface to monitor temperature distribution at different locations. Sensors capable of measuring pressure generated during the molding process, such as piezoelectric pressure sensors or strain gauge pressure sensors, should be selected and placed inside the mold or on the material surface to monitor pressure changes during the molding process.

[0017] Strain gauges capable of measuring material strain are selected and attached to the mold and material surface to monitor the strain distribution at various locations. Stress is inferred by measuring the strain (deformation) of the material. When the material is deformed by force, the strain gauge deforms as the material expands and contracts, causing the gauge's resistance to change. The strain on the material surface is calculated by measuring the change in resistance of the strain gauge and the known gauge factor. The strain value is converted to a stress value based on the material's stress-strain relationship (e.g., Hooke's law). This stress value is typically derived from the material's elastic modulus and the strain measured by the strain gauge. The gauge factor is a key parameter used to describe the relationship between the change in resistance and length of a material when subjected to force and deformation. It reflects the ratio between the change in resistance of the strain gauge and the actual deformation of the material. The gauge factor varies between materials. For example, the gauge factor of metals is typically between 2 and 5, while the gauge factor of semiconductors may be higher.

[0018] Continuous observation and recording of key process parameters during the duplex steel forming process. The duplex steel forming process is very sensitive to changes in temperature, pressure, and stress, so real-time monitoring of this data can help identify and control factors affecting quality, thereby ensuring that the product meets preset requirements. Based on the collected temperature data, a temperature field model is generated in the simulation software. The construction of the temperature field usually relies on computer simulation software, such as finite element analysis (FEA) or computational fluid dynamics (CFD) and other simulation technologies, to accurately simulate the temperature distribution in actual production. By establishing a temperature field model, the temperature distribution at different locations is predicted, thereby reflecting the temperature gradient of the material during the forming process. For example, there may be slight temperature differences in different areas of the sheet metal during forming, and these differences will affect the strength and performance of the material. The temperature field simulation space refers to the three-dimensional space in which computer models are used to simulate and predict the temperature distribution during the forming process. It can help engineers identify local temperature fluctuations and optimize the heating system to ensure temperature uniformity.

[0019] The collected stress data is also used to construct a stress field model to predict the stress distribution of the material during the forming process. The stress field model can intuitively display the internal stresses to which the duplex steel sheet is subjected during the forming process, such as the risk of fracture in certain areas due to stress concentration. The establishment of a stress field simulation space can help optimize mold design and forming pressure settings to ensure uniform strength of the formed material and reduce structural weaknesses. By establishing a multi-dimensional data acquisition network and simulation space, precise monitoring and control of the duplex steel forming process can be achieved, which helps to promptly identify and resolve problems in the forming process, such as uneven temperature and stress distribution, thereby improving forming quality, reducing scrap rates, and increasing production efficiency and material utilization.

[0020] Step 2: Under standard environmental conditions, obtain a process parameter combination, wherein the process parameter combination includes molding temperature, molding pressure, and molding time corresponding to molding process indicators, and material type, material ratio, and material purity corresponding to material composition indicators.

[0021] Specifically, standard environmental conditions refer to conducting experiments or production in a specific, repeatable, and controllable environment, typically including certain temperature, humidity, air pressure, and other factors to ensure that other variables in the production process are effectively controlled and that the test results are highly comparable. Under standard environmental conditions, it is first necessary to define a controllable production environment. For example, the temperature is set to 25°C and the humidity is set to 50%, and it is ensured that there are no other external interferences during the entire production process. Under these conditions, various process parameters that affect the forming quality of duplex steel are collected and recorded, including forming process indicators (such as forming temperature, forming pressure, and forming time) and material composition indicators (such as material type, material ratio, and material purity).

[0022] Forming process indicators refer to key process parameters directly related to the forming process, including but not limited to forming temperature, forming pressure, and forming time. Forming temperature refers to the temperature required to make the material (such as metal or alloy) reach a plastic state during the production process. For example, the setting of the forming temperature must ensure that the steel is in an ideal plastic state. Too high or too low will affect the quality of the final product. Usually, the forming temperature of duplex steel is between 900°C and 1200°C. Forming pressure refers to the pressure applied to the material during the forming process, which is usually applied through a mold or forming equipment. The setting of the forming pressure is related to factors such as the design of the mold and the fluidity of the steel. Excessive pressure may lead to material waste, while too low pressure may lead to incomplete forming. Forming time refers to the length of time from the start of heating or pressurizing the material to the final completion of forming. It depends on factors such as the heating rate of the material and the duration of pressure, and needs to be optimized according to specific production conditions.

[0023] Material composition indicators refer to the composition of a material, including its type, proportion, purity, and so on. Material type refers to the specific material used to manufacture a product. For example, duplex steel may be composed of elements such as iron, carbon, chromium, and nickel. Material proportion refers to the ratio of the different elements or alloys that make up a material. The performance of duplex steel depends on the proportion of its components, and its performance is usually adjusted by the alloy ratio (such as carbon content, chromium content, etc.). Material purity refers to the proportion of pure components in a material. High-purity steel has better mechanical properties, while low-purity material may contain more impurities, affecting the quality of the final product.

[0024] According to the plan, prepare duplex steel materials of different types, proportions, and purities. Conduct multiple forming tests under standard environmental conditions according to the test plan. During each test, record the forming process parameters (forming temperature, forming pressure, forming time, etc.) and material composition indicators (material type, material proportion, material purity, etc.). By obtaining forming data for different process parameter combinations, analyze the impact of forming process parameters on forming quality and identify the key process parameters that affect forming quality.

[0025] Step three: in the temperature field simulation space, mapping and correlation analysis is performed on the dual-phase steel forming quality fluctuation in combination with the process parameter combination, and a plurality of first influencing characteristics corresponding to the forming process indicators are output.

[0026] Specifically, different preliminary process parameter combinations, including molding temperature, molding pressure, and molding time, are input into the temperature field simulation space to observe their effects on the temperature field. Quality fluctuations in duplex steel molding refer to fluctuations or variations in molding quality during the molding process due to changes in process parameters. For example, slight fluctuations in molding temperature can lead to differences in the hardness or corrosion resistance of duplex steel. Mapping correlation analysis is a method for correlating and analyzing the relationships between different variables. Mapping correlation analysis is performed by measuring quality indicators such as hardness, ductility, and corrosion resistance of the finished product and combining them with fluctuations in molding quality. By establishing a mapping relationship between various parameters (such as temperature, pressure, and time) and molding quality, it is possible to clearly identify how different parameters affect the quality of the finished product, thereby optimizing the process. For example, suppose it is found that the toughness of duplex steel decreases at temperatures above 1000°C, while the strength of the material does not meet the standard below 950°C. Using mapping correlation analysis, the relationship between temperature fluctuations and quality fluctuations can be quantified. Through mapping analysis, the correlation between forming process indicators (such as temperature, pressure, and time) and finished product quality was extracted. Factors with a significant impact on quality fluctuations were defined as primary influencing characteristics, including temperature uniformity, cooling rate, and temperature hold time. Primary influencing characteristics refer to key factors directly related to forming process indicators, as determined through correlation analysis, and directly impact forming quality. By performing mapping correlation analysis in the temperature field simulation space, process parameters with a significant impact on duplex steel forming quality were identified and their optimal ranges determined, helping to optimize the forming process and improve forming quality.

[0027] Step 4: In the stress field simulation space, mapping correlation analysis is performed on the dual-phase steel forming quality fluctuation in combination with the process parameter combination, and a plurality of second influencing characteristics corresponding to the material composition indicators are output.

[0028] Specifically, similar to the above steps, different process parameter combinations are input into the stress field simulation space to observe and analyze how these parameters affect the forming quality of the duplex steel. During the simulation, quality fluctuations during the forming process are observed and recorded. For example, by analyzing the simulation results, it is possible to identify which material composition combinations lead to problems such as insufficient hardness or decreased ductility in the finished product. A correlation mapping is established between quality fluctuations and material composition to quantify the impact of material composition on stress distribution. For example, it can be determined that a certain material ratio at a specific temperature leads to insufficient tensile strength. Through mapping and correlation analysis, the key influencing factors of material composition indicators (such as type, ratio, and purity) on quality fluctuations are identified, referred to as secondary influencing characteristics. Secondary influencing characteristics are material composition indicators that have a significant impact on forming quality, as determined through correlation analysis. They directly reflect the impact of material composition on quality fluctuations, such as toughness and fatigue performance. For example, simulation analysis has shown that increasing the purity of chromium-containing duplex steel to 95% improves its toughness and stress distribution uniformity. By performing mapping and correlation analysis in the stress field simulation space, key material parameters affecting molding quality, such as material type and material ratio, are determined, thereby optimizing material selection and ratio.

[0029] Step 5: According to the multiple first influencing characteristics and the multiple second influencing characteristics, the process parameter combinations are divided to obtain the first type of process parameter components, the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components. The first type of process parameter components are parameter combinations whose molding quality levels meet the preset molding quality standards.

[0030] Specifically, through temperature and stress field simulations, multiple primary influencing characteristics of molding process indicators and multiple secondary influencing characteristics of material composition indicators are derived. Each molding process (temperature, pressure, time) and material composition (type, ratio, purity) has a different impact on molding quality. Based on the combination of primary and secondary influencing characteristics, the parameter combinations that contribute to high-quality finished products are determined. Parameter combinations are categorized into primary, secondary, tertiary, and quaternary process parameter components based on their contribution to molding quality.

[0031] The first category of process parameter components refers to parameter combinations that enable molding quality to meet or exceed pre-determined standards. These parameters typically exhibit stable and uniform conditions in terms of temperature, pressure, and material composition, ensuring that the finished product meets pre-determined quality standards and exhibits excellent hardness, strength, and ductility, representing ideal production conditions. In other words, under these combinations, key performance indicators of duplex steel, such as hardness, strength, ductility, and fatigue resistance, fully meet or exceed expected quality standards. Both process conditions (such as molding temperature, molding pressure, and molding time) and material composition (such as material type, ratio, and purity) in these parameter combinations have a positive impact on molding quality and are positively correlated with molding quality indicators. The second category of process parameter components refers to combinations of process parameters that are negatively correlated with the first influencing characteristic and result in molding quality levels that do not meet pre-determined molding quality standards. In other words, certain process conditions (such as excessive temperature or pressure) negatively impact molding quality, resulting in substandard final product quality. The process conditions in the second category of parameter components have a negative impact on the molding quality. For example, too high a temperature will cause the material grains to coarsen, affecting the hardness; or too long a molding time will cause the product surface to deform. There is often a certain deviation from the expected standards in key quality indicators.

[0032] The third category of process parameter combinations represents the lowest molding quality grade and fails to meet the pre-set molding quality standards. These combinations exhibit significant deficiencies in temperature, pressure, and material composition, resulting in significantly substandard molding quality and requiring process parameter optimization. Not only do the process conditions and material composition fail to meet requirements, but they also significantly impact molding quality. Both the primary and secondary influencing characteristics exhibit negative effects on key indicators such as hardness and ductility in the finished product, and are generally not adopted. The fourth category of process parameter combinations represents combinations that are negatively correlated with the secondary influencing characteristic and fail to meet the pre-set molding quality standards. This negative correlation primarily refers to the negative impact of material composition (such as material type, ratio, and purity) on molding quality. The material composition of the fourth category of components negatively impacts various molding quality indicators. For example, low material purity can lead to decreased hardness and ductility, or an inappropriate ratio of specific components can increase brittleness in the finished product. Optimizing material composition (such as adjusting purity or ratio) can improve molding quality. This classification clearly identifies which process parameter combinations are effective and which require improvement.

[0033] Step 6: Using the preset molding quality standard as a constraint condition, the second type of process parameter component, the third type of process parameter component, and the fourth type of process parameter component are trained through the first type of process parameter component, and the process parameters are adjusted.

[0034] Specifically, the preset forming quality standards are target performance criteria set before production based on product application requirements. They serve as the basis for determining whether the forming process meets these standards. They cover the most important performance indicators for duplex steel forming, such as hardness, tensile strength, ductility, and fatigue resistance. The entire optimization process is constrained within the preset forming quality standards. In other words, no matter how the process parameters are adjusted, the ultimate goal is to ensure that product quality meets or approaches the predetermined standards. The first category of process parameter components represents ideal process parameter combinations that meet the preset forming quality standards. Learning and training are performed based on these ideal process parameter combinations. The goal of learning and training is to further understand which process parameters are most critical to final product quality through machine learning, data analysis, or optimization algorithms, and how to optimize them to improve overall production efficiency and product quality. Feature extraction is performed on the first category of process parameter components to generate the first primary model data. Feature extraction is performed on the second, third, and fourth category of process parameter components to generate the second primary model data.

[0035] After training the first category of process parameter components, a set of optimized process parameters can be obtained that meet the preset quality standards. On this basis, adjustments are made to the second, third, and fourth categories of process parameter components that do not meet the standards. These non-standard parameter combinations are adjusted to be as close as possible to the first category of process parameter components, ensuring that the non-standard components in the production process can ultimately meet or approach the predetermined molding quality standards through optimization. Using a loss function, the first primary model data is used to calculate the loss of the second primary model data, measuring the difference from the target. By optimizing the loss function, the model is adjusted so that the output process parameter combination meets the set quality standards.

[0036] Furthermore, step five of this application also includes:

[0037] The third category of process parameter components is a parameter combination whose molding quality grade does not meet the preset molding quality standard; the molding quality grade of the first category of process parameter components, the molding quality grade of the second category of process parameter components, and the molding quality grade of the fourth category of process parameter components are all higher than the molding quality grade of the third category of process parameter components.

[0038] Specifically, a molding quality grade refers to a classification of product molding quality based on specific quality assessment criteria during the production process. This categorizes the quality level achievable by each process parameter combination. Molding quality grades typically reflect key physical and mechanical properties of the product, such as hardness, tensile strength, ductility, and fatigue resistance. Pre-set molding quality standards are target performance criteria set prior to production based on product application requirements. These standards serve as the basis for determining whether the molding process meets these standards and cover the most important performance indicators for duplex steel molding, such as hardness, tensile strength, ductility, and fatigue resistance. The third category of process parameter combinations refers to parameter combinations whose quality grades do not meet the preset molding quality standards. This indicates significant deficiencies in temperature, pressure, time, and material purity. These deficiencies may be caused by improper material selection, improper material ratios, or inappropriate molding process parameter settings. This can lead to defects during the molding process, such as cracks, deformation, and uneven microstructure, ultimately resulting in substandard molding quality. Therefore, these combinations are generally not used in actual production. While the molding quality of the second and fourth categories of process parameter combinations may not fully meet the preset standards, they are still higher than that of the third category. The quality level of the third category of process parameter components is the lowest and cannot meet the preset molding quality standards, while the molding quality of other categories (first, second and fourth categories) is higher than that of the third category.

[0039] Furthermore, the present application further comprises the following steps:

[0040] The second category of process parameter groups is a parameter combination that is negatively correlated with the multiple first influencing characteristics and whose molding quality level does not meet the preset molding quality standards; the fourth category of process parameter groups is a parameter combination that is negatively correlated with the multiple second influencing characteristics and whose molding quality level does not meet the preset molding quality standards.

[0041] Specifically, the process conditions (such as temperature, pressure, and time) in the second category of process parameter components are negatively correlated with multiple primary influencing characteristics of molding quality. This negative correlation means that as one influencing characteristic increases, molding quality actually decreases. For example, excessively high molding temperatures can cause the material to become brittle, while excessive pressure or prolonged molding time can cause structural deformation, all of which can reduce the molding quality grade. The second category of process parameter components refers to parameter combinations whose molding quality grade does not meet the preset molding quality standards, such as when key indicators such as hardness, ductility, or tensile strength fall below the specified values.

[0042] The material composition indicators (such as material type, component ratio, and purity) within the fourth category of process parameter components are negatively correlated with several secondary characteristics that influence molding quality. For example, insufficient material purity can lead to reduced strength in the finished product, while an excessive proportion of a particular material can reduce toughness. Inappropriate material selection or improper material ratios within the fourth category of process parameter components, such as the use of materials with poor performance or a material ratio that does not meet requirements, can result in the finished material failing to meet standard strength and toughness requirements.

[0043] Further, as attached Figure 2 As shown, this application also includes the following steps:

[0044] Based on the first type of process parameter components, feature extraction is performed to obtain first primary model data; based on the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components, feature extraction is performed to obtain second primary model data; a loss function is set according to the difference between the preset molding quality standard and the corresponding molding quality grade, and the loss of the second primary model data is calculated using the first primary model data.

[0045] Specifically, the first category of parameter components already meets molding quality standards. Therefore, feature extraction can capture the characteristics of the optimal process parameter combination, representing key indicators for successful production. Representative features are extracted from the first category of process parameter components to simplify the data and identify the key factors influencing molding quality. In process control, feature extraction involves analyzing data on process conditions (such as temperature and pressure) and material composition (such as ratio and purity) to identify and select the process parameters most relevant to molding quality, such as molding temperature, pressure, time, and material composition. Specifically, based on the correlation with molding quality, the parameters with the greatest impact on molding results are selected as features. Statistical methods (analysis of variance, correlation analysis) or machine learning techniques (such as principal component analysis (PCA) and feature selection algorithms) are then used to extract key features from the data. These extracted features are then combined into the first primary model data. ANOVA specifically calculates the variance of each parameter's impact on the quality indicator to assess its contribution to molding quality fluctuations. A large variance for temperature indicates a significant impact on molding quality fluctuations. Correlation analysis specifically calculates the correlation coefficient (such as the Pearson correlation coefficient) between process parameters and quality indicators, identifying positive and negative correlations between parameters such as molding temperature and pressure and molding quality. Principal component analysis converts multiple process and material parameters into a small number of principal components, thereby extracting comprehensive features that significantly influence molding quality. Principal component analysis identifies the most important components (such as the combination of temperature and material purity) as the dominant characteristics of molding quality, while ignoring minor influencing factors. Feature selection algorithms gradually remove features with minimal impact on the model to identify the parameters that contribute most to molding quality.

[0046] Similarly, features are extracted from the second, third, and fourth categories of process parameter components that do not meet quality standards to generate the second primary model data. This data contains the features of process parameters that do not meet quality standards due to improper process conditions or material composition, reflecting the adverse influencing factors. Based on the features of the second, third, and fourth categories of process parameter components, data for substandard parameter combinations is constructed, namely the second primary model data. This data is then used for subsequent comparative analysis with the first primary model data to optimize process conditions and material ratios.

[0047] Based on the preset molding quality standard and the difference in molding quality levels for different process parameter components, a loss function is designed to calculate the error between the first and second primary model data, indicating the difference between parameter combinations that do not meet the standard and those that meet it. The loss function is used to calculate the deviation, or error, between the model output and the expected target. By optimizing the loss function, the model is guided closer to the preset molding quality standard. Based on the molding quality standard and the difference in quality levels for different process parameter components, the loss function is set to calculate the quality loss for different parameter combinations. The goal of the loss function is to reduce the difference between the first and second primary model data through continuous iteration, thereby ensuring that the second primary model data (i.e., process parameter combinations) better meet the preset molding quality standard. Using the loss function, the loss of the second primary model data is calculated based on the first primary model data. Continuous optimization of the loss function identifies key factors affecting molding quality, as well as process conditions and material compositions that should be avoided. Using high-quality data, flaws in suboptimal data can be identified and corrected, allowing process parameters (such as temperature and pressure) to be adjusted to meet the standard range, thereby improving the overall quality of the finished product.

[0048] Furthermore, the present application further comprises the following steps:

[0049] The second primary model data is updated iteratively through the loss function to update the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components; the loss function: ;in, is a loss function used to measure the difference between the preset molding quality standard and the corresponding molding quality grade. is the cross entropy loss function, α is the weight coefficient, which is used to balance the consistency between the first primary model data and the second primary model data and the consistency between the parameter combination of the second primary model data and the true label. is the prediction function of the first primary model for the input data X, is the prediction function of the second primary model for the input data X, and y is the parameter combination under the true label.

[0050] Specifically, the loss function is used to measure the difference between the predicted result and the actual target. The specific formula is: Among them, L dp It is a loss function used to measure the difference between the preset molding quality standard and the corresponding molding quality grade. It is divided into two parts. Measure the difference between the first primary model data and the second primary model data, Measures the difference between the second primary model data and the true label data. L ce is the cross-entropy loss function, which measures the difference between two probability distributions, that is, the difference between the predicted value and the true value. α is a weight coefficient that balances the consistency between the first and second primary model data and the consistency between the parameter combination of the second primary model data and the true label. A larger α places more emphasis on the data consistency of the first model, while a smaller α places more emphasis on the true label consistency of the second primary model. Is the prediction function of the first primary model for the input data X, which represents the prediction function based on the input data X and The predicted output obtained is, It is the parameter of the first primary model, which predicts the process quality based on the first type of process parameter components. It aims to learn from process parameters that meet the quality standards and generate high-quality prediction results. Is the prediction function of the second primary model for the input data X, which represents the prediction function based on the input data X and The predicted output obtained is, are the parameters of the second primary model, which predicts process quality based on the second, third, and fourth process parameter components. Its goal is to learn from process parameter combinations that do not meet quality standards to optimize the production process and reduce the occurrence of substandard products. y is the true label, which is the preset target output and represents the quality standard that the process parameters should achieve in actual production.

[0051] According to the input data X, through the first primary model and the second primary model The process parameter components are predicted and the prediction results are obtained. Then the cross entropy loss function L ce Calculate the difference between the predicted value and the true label. Weigh the difference between the prediction results of the first primary model and the prediction results of the second primary model, and adjust its importance using the weight coefficient α. Weigh the difference between the prediction results of the second primary model and the true label (i.e., the process standard), and use (1-α) as the weight coefficient. In each training iteration, according to the calculated total loss L dp , update the model parameters through optimization algorithms (such as gradient descent) and Through multiple iterations of optimization, the model gradually reduces the loss, ultimately bringing the predictions closer to the true labels and optimizing the process parameter combinations. By iteratively optimizing the loss function, the second, third, and fourth categories of process parameter components are gradually adjusted. Through this iterative process, the model parameters are gradually adjusted to better match the real data, resulting in updated data for the second primary model, enabling it to more accurately predict the molding quality grade, reducing fluctuations in the production process and ultimately meeting quality standards. As the second primary model data is updated, the second, third, and fourth categories of process parameter components are reassessed and reclassified based on the new predictions. Parameter combinations originally classified as category 2 or 4 may be reclassified as category 1 if the new predictions indicate an improvement in their molding quality grade. Similarly, parameter combinations originally classified as category 3 may remain unchanged or be further downgraded if the predictions indicate a decrease in their molding quality grade. Through iterative optimization using the loss function, the parameters of the second primary model (i.e., process parameter combinations) are continuously adjusted, gradually optimizing process parameters (such as molding temperature, pressure, and material composition), thereby improving product molding quality.

[0052] Furthermore, the present application further comprises the following steps:

[0053] iteratively updating the second primary model data, ;in, is the parameter of the next update of the second primary model, is the last updated parameter of the second primary model, are the current parameters of the second primary model, is the learning rate, is the gradient of the loss function with respect to the second primary model data.

[0054] Specifically, the second primary model data is iteratively updated using the gradient descent method: ,in, It is the parameter of the next update of the second primary model, that is, the parameter after the update in this round of iteration. After each iteration, the parameters of the model will be adjusted until they reach the optimal value. are the parameters of the second primary model last updated, that is, the parameters before this iteration. The learning rate is a hyperparameter that controls the step size of each parameter update and determines the magnitude of parameter adjustments. A learning rate that is too small may lead to slow convergence, while a learning rate that is too large may cause training instability or even failure to converge. L is the gradient of the loss function with respect to the second primary model data. The gradient is a vector that represents the rate of change of the loss function at the current parameter position. In other words, it indicates how to adjust the model parameters to reduce the value of the loss function. The direction of the gradient is the direction in which the loss function decreases fastest, and the magnitude of the gradient is the magnitude of the adjustment.dp It is a loss function that measures the difference between the current model output and the actual target, that is, the loss function of the second primary model (such as cross entropy loss or mean square error, etc.), which measures the error between the predicted result and the actual target. is the current parameter of the second primary model. First, calculate the loss function L dp About model parameters The gradient of , which represents the trend of change in the loss function under the current parameter settings, indicating in which directions there is room for improvement in the current model. According to the direction and magnitude of the gradient, the model parameters are updated by subtracting the gradient and multiplying it by the learning rate. The core idea of ​​gradient descent is to adjust the parameters in the opposite direction of the gradient, thereby reducing the value of the loss function (i.e., the optimization target). In this way, the parameters of the model continue to approach the optimal solution in each iteration, and eventually a model with minimal loss can be generated. By continuously iterating this process and updating the parameters, the performance of the model gradually improves until a convergence point is reached, that is, the loss function no longer changes significantly, or the change is very small, indicating that the model has found a better parameter combination. By calculating the gradient of the loss function with respect to the model parameters and updating them according to the gradient, the parameters of the model are continuously adjusted to gradually optimize the molding quality.

[0055] Furthermore, this application also includes:

[0056] With the preset molding quality standard as a constraint condition, feature extraction is performed based on the second type of process parameter components and the third type of process parameter components to obtain third primary model data; with the preset molding quality standard as a constraint condition, feature extraction is performed based on the fourth type of process parameter components and the third type of process parameter components to obtain fourth primary model data; an upper limit on the number of iterations is set for the second primary model data, the third primary model data, and the fourth primary model data, respectively, and an iterative competition mechanism is introduced.

[0057] Specifically, based on the preset molding quality standards, the focus is on process parameters that ensure molding quality meets the standards. The third-category process parameter component and the second-category process parameter component are combined to generate third-level model data. Since the third-category process parameter component is typically a combination of parameters associated with molding quality levels that do not meet the preset standards, while the second-category process parameter component is negatively correlated with multiple first-level influencing characteristics, these parameters are combined to understand how to improve and optimize the production process under process conditions that do not meet the quality standards. Features related to molding quality, such as process stability, quality fluctuation, and molding time, are extracted from the second and third-category process parameter components to form the third-level model data. The fourth-category process parameter component is combined with the third-category process parameter component to generate the fourth-level model data. The fourth-category process parameter component is associated with characteristics that influence molding quality fluctuations and can provide valuable feedback for optimization, especially when molding quality does not meet the preset standards. Features are extracted from the fourth and third-category process parameter components, particularly those associated with molding quality levels that do not meet the standards, to further optimize the process parameters.

[0058] Setting an upper limit on the number of iterations prevents model overfitting and ensures that the optimization process is completed within a limited timeframe. Each model (the second, third, and fourth primary models) has a fixed upper limit on the number of iterations. Iterations are terminated when the model converges to a good solution before reaching this upper limit. If convergence fails within the upper limit, iterations continue until the maximum number of iterations is reached. An iterative competition mechanism compares and competes multiple primary models (the second, third, and fourth primary models) to select the best output. Specifically, the three models are trained and optimized in each iteration, and their performance is evaluated using a loss function. The goal of each model is to minimize the deviation from a preset molding quality standard, that is, to achieve the best molding quality. In each iteration, all models compete against each other, calculating their respective loss values. Ultimately, by comparing the loss values ​​of each model, the best-performing model is selected for the next optimization step. For example, if the data quality of the second primary model is better than that of the third and fourth primary models, the second primary model will be prioritized for optimization. By extracting features, setting an upper limit on the number of iterations, and introducing an iterative competition mechanism, process parameters can be efficiently optimized to ensure that the final process meets the preset molding quality standard. By combining different categories of process parameters, multiple primary model data are generated, and through competition mechanisms and optimization processes, the production process and product quality are continuously improved.

[0059] Furthermore, the present application further comprises the following steps:

[0060] A resource allocation matrix is ​​established according to the multiple first influencing characteristics and the multiple second influencing characteristics; and resource allocation is adjusted for the second primary model data, the third primary model data, and the fourth primary model data based on the resource allocation matrix.

[0061] Specifically, the first influencing characteristic is typically related to molding process indicators (such as temperature and pressure), representing the impact of core parameters in the molding process on product quality. The second influencing characteristic is typically related to material composition, describing the impact of factors such as material ratio, type, and purity on molding quality. The resource allocation matrix is ​​a mathematical tool designed to represent how resources are allocated between different parameters and features in matrix form, demonstrating the relationship between multiple influencing factors (such as influencing characteristics) and different model data. Each influencing characteristic is associated with corresponding process parameters and model data, identifying which influencing characteristics have a greater impact on different primary models (such as the second primary model, the third primary model, etc.). The resource allocation matrix is ​​constructed based on multiple first influencing characteristics and multiple second influencing characteristics, and resources are rationally allocated by analyzing how these characteristics affect the adjustment of process parameters and model optimization.

[0062] The resource allocation matrix may be a two-dimensional matrix, with each row representing an influencing characteristic and each column representing a specific data model (such as the second primary model, the third primary model, and so on). Each element in the matrix represents the resource allocation relationship (such as weight or influence) between that influencing characteristic and the model. For example, a matrix element could be the correlation, weight, or influence between an influencing characteristic and a model. If a characteristic has a greater impact on a particular model, a larger value is assigned to that position in the matrix, indicating that more resources should be allocated to that model. If certain models respond more strongly to certain influencing characteristics, more resources are allocated to those models. Conversely, fewer resources are allocated to models with less impact. Based on the results of the resource allocation matrix, the data for the second, third, and fourth primary models are adjusted. Through the appropriate allocation of resources, the model training process is optimized, ensuring that each primary model is optimally adjusted for its most important influencing characteristic, thereby improving process optimization efficiency and product quality.

[0063] In summary, the dual-phase steel forming quality prediction and process adjustment method provided in this application has the following technical effects:

[0064] By establishing a multi-dimensional data acquisition network, the dual-phase steel forming process is monitored in real time, and the temperature field simulation space and the stress field simulation space are fitted; under standard environmental conditions, a process parameter combination is obtained, and the process parameter combination includes the forming temperature, forming pressure, and forming time corresponding to the forming process indicators, and the material type, material ratio, and material purity corresponding to the material composition indicators; in the temperature field simulation space, the process parameter combination is combined with the dual-phase steel forming quality fluctuation to perform mapping correlation analysis, and output a plurality of first influencing characteristics corresponding to the forming process indicators; in the stress field simulation space, the process parameter combination is combined with the dual-phase steel forming quality fluctuation to perform mapping correlation analysis. A mapping correlation analysis is performed to output multiple second influencing characteristics corresponding to the material composition indicators; based on the multiple first influencing characteristics and the multiple second influencing characteristics, the process parameter combination is divided to obtain a first type of process parameter component, a second type of process parameter component, a third type of process parameter component, and a fourth type of process parameter component, wherein the first type of process parameter component is a parameter combination whose forming quality level meets the preset forming quality standard; with the preset forming quality standard as a constraint condition, the second type of process parameter component, the third type of process parameter component, and the fourth type of process parameter component are trained through the first type of process parameter component, and process parameter adjustment is performed. In other words, by establishing a multi-dimensional data acquisition network, the key parameters in the dual-phase steel forming process are monitored in real time; in the temperature field and stress field simulation space, a mapping correlation analysis is performed in combination with the process parameter combination to output the influencing characteristics; the process parameter combination is divided and trained based on the influencing characteristics, and process parameter adjustment is performed, thereby achieving accurate prediction of forming quality and intelligent adjustment of process parameters, ensuring the stability of product quality.

[0065] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0066] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A method for predicting the forming quality and adjusting the process of dual-phase steel, characterized in that: include: Establish a multi-dimensional data acquisition network to monitor the dual-phase steel forming process in real time, and fit the temperature field simulation space and stress field simulation space; Under standard environmental conditions, a process parameter combination is obtained, wherein the process parameter combination includes molding temperature, molding pressure, and molding time corresponding to molding process indicators, and material type, material ratio, and material purity corresponding to material composition indicators; In the temperature field simulation space, mapping and correlation analysis are performed on the dual-phase steel forming quality fluctuation in combination with the process parameter combination, and a plurality of first influencing characteristics corresponding to the forming process indicators are output; In the stress field simulation space, mapping and correlation analysis are performed on the dual-phase steel forming quality fluctuation in combination with the process parameter combination, and a plurality of second influencing characteristics corresponding to the material composition indicators are output; Dividing the process parameter combinations according to the multiple first influencing characteristics and the multiple second influencing characteristics to obtain a first type of process parameter component, a second type of process parameter component, a third type of process parameter component, and a fourth type of process parameter component, wherein the first type of process parameter component is a parameter combination whose molding quality level meets a preset molding quality standard; Taking the preset molding quality standard as a constraint condition, the second type of process parameter component, the third type of process parameter component, and the fourth type of process parameter component are trained through the first type of process parameter component, and the process parameters are adjusted; Perform feature extraction based on the first type of process parameter components to obtain first primary model data; Perform feature extraction based on the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components to obtain second primary model data; A loss function is set based on the difference between the preset molding quality standard and the corresponding molding quality grade, and the loss calculation is performed on the second primary model data using the first primary model data; The third type of process parameter group is a parameter combination whose molding quality level does not meet the preset molding quality standard; The molding quality grade of the first category of process parameter components, the molding quality grade of the second category of process parameter components, and the molding quality grade of the fourth category of process parameter components are all higher than the molding quality grade of the third category of process parameter components; The second type of process parameter group includes parameter combinations that are negatively correlated with the plurality of first influencing characteristics and whose molding quality levels do not meet the preset molding quality standards; The fourth type of process parameter group includes parameter combinations that are negatively correlated with the plurality of second influencing characteristics and whose molding quality levels do not meet the preset molding quality standards.

2. The method for predicting the forming quality and adjusting the process of dual-phase steel according to claim 1, wherein: Iteratively updating the second primary model data by using the loss function to update the second type of process parameter components, the third type of process parameter components, and the fourth type of process parameter components; The loss function: ; in, is a loss function used to measure the difference between the preset molding quality standard and the corresponding molding quality grade. is the cross entropy loss function, α is the weight coefficient, which is used to balance the consistency between the first primary model data and the second primary model data and the consistency between the parameter combination of the second primary model data and the true label. is the prediction function of the first primary model for the input data X, is the prediction function of the second primary model for the input data X, is the parameter combination under the true label.

3. The method for predicting the forming quality and adjusting the process of dual-phase steel according to claim 2, wherein: iteratively updating the second primary model data, ; in, is the parameter of the next update of the second primary model, is the last updated parameter of the second primary model, are the current parameters of the second primary model, is the learning rate, is the gradient of the loss function with respect to the second primary model data.

4. The method for predicting the forming quality and adjusting the process of dual-phase steel according to claim 3, wherein: Taking the preset molding quality standard as a constraint condition, feature extraction is performed based on the second type of process parameter components and the third type of process parameter components to obtain third primary model data; Taking the preset molding quality standard as a constraint condition, feature extraction is performed based on the fourth type of process parameter components and the third type of process parameter components to obtain fourth primary model data; An upper limit on the number of iterations is set for the second primary model data, the third primary model data, and the fourth primary model data, respectively, and an iteration competition mechanism is introduced.

5. The method for predicting the forming quality and adjusting the process of dual-phase steel according to claim 4, characterized in that: An upper limit on the number of iterations is set for each of the third primary model data, the fourth primary model data, and the first primary model data, and an iteration competition mechanism is introduced, including: Establishing a resource allocation matrix according to the plurality of first influencing characteristics and the plurality of second influencing characteristics; Based on the resource allocation matrix, resource allocation adjustment is performed on the second primary model data, the third primary model data, and the fourth primary model data.

Images