A construction method of a "hot-dip-smoothing" coupling model for surface roughness prediction of GA

Abstract

This invention relates to the field of hot-dip galvanizing technology, and a method for constructing a "hot-dip galvanizing-leveling" coupled model for predicting the surface roughness of GA strips. The method includes: Step 1, model data acquisition; Step 2, data standardization and dataset partitioning; Step 3, surface roughness prediction model construction; Step 4, surface roughness attenuation model construction for leveling work rolls; Step 5, surface roughness prediction model construction for leveling process; and Step 6, simultaneous model establishment and LM algorithm regression. This invention focuses on two key factors: the zinc-iron alloy coating formed during the hot-dip galvanizing stage and the rolling deformation during the leveling stage. It establishes the coupling relationship between them, constructs a surface roughness prediction model for the hot-dip galvanizing stage and a surface roughness prediction model for the leveling process, combines the two models, and uses the LM regression algorithm to solve the resulting nonlinear least squares problem, ultimately deriving a "hot-dip galvanizing-leveling" coupled model for predicting the surface roughness of GA strips.

Description

Technical Field

[0001] This invention belongs to the field of hot-dip galvanizing technology, specifically relating to a method for constructing a "hot-dip galvanizing-smoothing" coupled model for predicting the surface roughness of GA strips. Background Technology

[0002] Hot-dip galvanized alloyed steel strip (GA) is an important high-end product in the steel industry, widely used in automobile manufacturing, home appliances, and construction. Its unique zinc-iron alloy coating gives it excellent corrosion resistance, machinability, and coating adhesion, making it highly sought after in the market. Surface roughness is one of the important indicators for measuring the quality of GA strip, directly affecting the subsequent coating adhesion, material machinability, and the final appearance quality of the product. Therefore, most buyers have requirements for this indicator. Thus, accurately predicting the surface roughness of GA strip has become a key link in ensuring product quality.

[0003] However, the production process of GA strip steel is multi-stage, and the surface roughness is not determined by a single factor, but is affected by the combined effects of multiple processes, such as... Figure 1 As shown, the surface roughness of GA strip steel mainly goes through three stages during the production process. First, before entering the hot-dip annealing production line, the strip steel undergoes a cold rolling process, which forms a certain degree of substrate surface morphology, namely cold-rolled substrate roughness. Subsequently, in the hot-dip alloying stage, the strip steel is immersed in a molten zinc bath and undergoes alloying treatment to form a zinc-iron alloy coating. In this process, the formation, thickness, distribution of alloying elements, and the reshaping effect of the alloying process on the surface morphology significantly alter and cover the original substrate surface morphology, thus determining the surface roughness of the output strip at this stage. Finally, in the leveling stage, the roughness of the coating is affected by the work rolls of the leveling machine and the leveling rolling process, ultimately forming the surface roughness of the finished GA strip. Since the surface roughness formation process of GA strip is complex and involves many steps, previous studies have mostly focused on the roughness prediction of a single step, and the research on the surface roughness formation mechanism of alloying treatment in the hot-dip galvanizing process is not in-depth, lacking corresponding quantitative models. Therefore, developing a mathematical model that can systematically consider the interaction of key steps in the production of GA strip and accurately predict surface roughness has become an urgent problem to be solved in the current technical field. Summary of the Invention

[0004] The purpose of this invention is to provide a simple and rationally designed method for constructing a "hot-dip coating-flattening" coupled model for predicting surface roughness of GA strips in order to solve the above problems.

[0005] The present invention achieves the above objectives through the following technical solutions: A method for constructing a "hot-dip galvanizing-leveling" coupled model for predicting surface roughness of GA belts includes the following steps: Step 1, model data acquisition; Step 2, data standardization and dataset partitioning; Step 3, surface roughness prediction model construction; Step 4, surface roughness attenuation model construction for leveling work rolls; Step 5, surface roughness prediction model construction for leveling process; and Step 6, simultaneous model analysis and LM algorithm regression. In step one above, the parameters required for model construction are first collected. In step two above, all collected data features are standardized using z-score, transforming the data into a distribution with a mean of 0 and a standard deviation of 1. In step three above, a surface roughness prediction model for the hot-dip galvanizing stage is constructed: a nonlinear model in the form of a quadratic polynomial is established as follows; In step four above, a surface roughness attenuation model for the leveling work roll is constructed: the roughness of the current work roll can be expressed as a function of the original roughness of the work roll and its rolling mileage. In step five above, a surface roughness prediction model for the leveling process is constructed: the surface roughness of the strip steel exiting the leveling process consists of two parts: the inheritance of the incoming material roughness and the imprint of the work roll roughness on the strip steel. In step six above, the simultaneous model analysis and LM algorithm regression are performed: by combining the two parts of the model, the following GA strip roughness prediction model can be obtained: For multiple empirical coefficients in the coupled "hot-dip galvanizing-flattening" model ( and The regression of the model is performed using the Levenberg-Marquardt algorithm to solve the resulting nonlinear least squares problem. The regression process iteratively adjusts these model coefficients to minimize the sum of squared residuals between the predicted values ​​of the "hot-dip galvanizing-smoothing" model and the observed roughness data.

[0006] Preferably, in step one, the required parameter is: alloying temperature. Alloying speed and the iron content of the coating Inlet thickness Yield strength Elongation at flatness The original roughness of the work roll Rolling kilometers of work rolls and the surface roughness of the finished GA strip steel .

[0007] Preferably, in step two, the standardized formula is as follows: In the formula For standardized data points, For the data points before standardization, The mean of the dataset. Let $\frac{ ...

[0008] Preferably, in step three, the nonlinear model in the form of a quadratic polynomial is: in The surface roughness of the strip steel at the exit of the hot-dip galvanizing process. These are the model coefficients.

[0009] Preferably, in step four, the specific model is represented as follows: in The surface roughness of the current working roll. These are the model coefficients.

[0010] Preferably, in step five, the specific model is typically represented as follows: In the formula To improve the surface roughness of the strip steel at the exit of the leveling process; To smooth the inherited roughness of the incoming steel strip; The roughness is caused by the imprinting of the work roller.

[0011] Preferably, in step five, although the model is divided into two parts, each part is affected by the strip parameters and the leveling and rolling process parameters. The specific model can be represented as follows: In the formula These are the model coefficients.

[0012] Preferably, in step six, all the undetermined coefficients in the model are defined as a parameter vector. Then, the loss function is defined as the sum of squared residuals, i.e., SSR, whose mathematical expression is: In the formula It is the actual measured roughness value of the i-th sample. At the current parameter estimate Below, the roughness value of the i-th sample predicted by the model, after defining the parameter vector and loss function, the initial parameter estimates of the LM algorithm are set. Initial step size adjustment factor And the threshold for determining convergence The algorithm then enters the iterative process, and after the iteration is complete, it will output the optimal parameter estimates. Through iterative optimization of the LM algorithm, the coefficients in the "hot-dip galvanizing-smoothing" model can be effectively determined, thus obtaining an accurate and reliable GA strip surface roughness prediction model.

[0013] The beneficial effects of this invention are as follows: Based on an in-depth analysis of the surface roughness formation process of GA strip steel, this invention focuses on two key factors: the zinc-iron alloy coating formed in the hot-dip galvanizing stage and the rolling deformation in the leveling stage. It establishes the coupling relationship between them, constructs a surface roughness prediction model for the hot-dip galvanizing stage and a surface roughness prediction model for the leveling process, performs a simultaneous equation of the two models, and uses the LM regression algorithm to solve the resulting nonlinear least squares problem. The regression process iteratively adjusts the coefficients of these models, and finally obtains the "hot-dip galvanizing-leveling" coupled model for predicting the surface roughness of GA strip steel, thereby achieving accurate prediction of the surface roughness of the final product. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the layout of the continuous galvanizing annealing production line of the present invention; Figure 2 This is the overall flowchart of the "hot-dip galvanizing-smoothing" coupled model for predicting the surface roughness of GA strip steel according to the present invention. Figure 3 This is a flowchart of the LM algorithm of the present invention; Figure 4 This is a schematic diagram of the prediction results of the model training set and test set in Embodiment 1 of the present invention; Figure 5 This refers to the accuracy of steel grade classification prediction and the number of steel coils in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the prediction results of the model training set and test set in Embodiment 2 of the present invention; Figure 7 This refers to the accuracy of steel grade classification prediction and the number of steel coils in Embodiment 2 of the present invention. Detailed Implementation

[0015] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0016] Example 1: Please refer to Figures 1 to 7A method for constructing a "hot-dip galvanizing-leveling" coupled model for predicting surface roughness of GA belts includes the following steps: Step 1, model data acquisition; Step 2, data standardization and dataset partitioning; Step 3, surface roughness prediction model construction; Step 4, surface roughness attenuation model construction for leveling work rolls; Step 5, surface roughness prediction model construction for leveling process; and Step 6, simultaneous model analysis and LM algorithm regression. In step one above, the parameters required for model construction are first collected. A total of 6551 industrial data points were collected, and each data point includes the surface roughness of the GA strip steel. The corresponding process parameters are as follows: Hot-dip galvanizing process parameters: alloying temperature Alloying speed and the iron content of the coating ; Leveling process parameters: Inlet thickness Yield strength Elongation at flatness The original roughness of the work roll Rolling kilometers of work rolls ; In step two above, all collected data features are standardized using z-scores, transforming the data into a distribution with a mean of 0 and a standard deviation of 1. The standardization formula is as follows: In the formula For standardized data points, For the data points before standardization, The mean of the dataset. The standard deviation of the dataset is given. To effectively evaluate the model's performance and generalization ability, the standardized dataset is divided into a training set and a test set, using an 8:2 ratio. That is, 80% of the data is used for model training and 20% of the data is used for model testing. This division strategy helps to fully fit the model during training and objectively evaluate its performance on an independent test set. In step three above, by summarizing experience from the production site and through various types of correlation analysis, the alloying temperature was determined. Alloying speed and the iron content of the coating Surface roughness exhibits a strong correlation with surface roughness. To capture the necessary nonlinearity and feature interactions, a surface roughness prediction model for the hot-dip galvanizing stage is constructed: A nonlinear model in the form of a quadratic polynomial is established as follows: in The surface roughness of the strip steel at the exit of the hot-dip galvanizing process. These are the model coefficients; In step four above, a surface roughness attenuation model for the leveling work rolls is constructed. During the leveling rolling process, the roughness of the work rolls is a key factor affecting the surface roughness of the strip. However, as the rolling process progresses, the surface roughness of the work rolls will attenuate. The current roughness of the work rolls can be expressed as a function of the original roughness of the work rolls and the rolling mileage. The specific model is as follows: in The surface roughness of the current working roll. These are the model coefficients; In step five above, a surface roughness prediction model for the leveling process is constructed: the surface roughness of the strip steel exiting the leveling process consists of two parts: the inheritance of the incoming material roughness and the imprint of the work roll roughness on the strip steel. The specific model is usually expressed as follows: In the formula To improve the surface roughness of the strip steel at the exit of the leveling process; To smooth the inherited roughness of the incoming steel strip; The roughness formed by the imprinting of the work roll is represented by a model that, although divided into two parts, is influenced by strip parameters and leveling rolling process parameters. The specific model can be expressed as follows: In the formula These are the model coefficients; In step six above, the simultaneous model analysis and LM algorithm regression are performed: by combining the two parts of the model, the following GA strip roughness prediction model can be obtained: For multiple empirical coefficients in the coupled "hot-dip galvanizing-flattening" model ( and The regression analysis employs the Levenberg-Marquardt algorithm to address the resulting nonlinear least squares problem. The regression process iteratively adjusts these model coefficients to minimize the sum of squared residuals between the predicted values ​​of the "hot-dip galvanizing-smoothing" model and the observed roughness data. All undetermined coefficients in the model are defined as a parameter vector. Then, the loss function is defined as the sum of squared residuals, i.e., SSR, whose mathematical expression is: In the formula It is the actual measured roughness value of the i-th sample. At the current parameter estimate Below, the roughness value of the i-th sample predicted by the model, after defining the parameter vector and loss function, the initial parameter estimates of the LM algorithm are set. Initial step size adjustment factor =0.1 and the threshold for determining convergence The algorithm enters the iterative process. Through the iterative optimization of the LM algorithm, the coefficients in the "hot galvanizing-flattening" model can be effectively determined, thereby obtaining an accurate and reliable GA strip surface roughness prediction model. Tables 1 and 2 give the initial values ​​of the empirical coefficients of the initial parameter estimates and the regression values ​​after the iteration.

[0017] Table 1. Model coefficients for the hot-dip galvanizing process "hot-dip galvanizing-smoo Table 2. Model coefficients for the "hot-dip galvanizing-smoothing" process in the leveling step. Figure 4 The text demonstrates the model's performance in roughness prediction, including the relationship between predicted and actual values ​​in the test and training sets. It shows that the "hot-dip galvanizing-smoothing" model achieves high R-values ​​in both the test and training sets. 2 The values ​​were 0.8374 and 0.8386, respectively, with RMSE values ​​of 0.0725μm and 0.0737μm, indicating that the model has a certain degree of accuracy in practical applications and good generalization ability.

[0018] Figure 5 The model's prediction performance is shown for different steel grades (the most numerous in the test dataset) and the number of coils for each steel grade. It can be seen that the model performs excellently in most steel grades (DQ-IF, 1180DP, 590DP), with prediction accuracy exceeding 90% and reaching a maximum of 96.24%. For the 440DQ steel grade, which has fewer coils, although the prediction accuracy decreases, it still remains at 89.1%. This fully demonstrates the model's good robustness and effective handling of different data distributions, proving the effectiveness of the model.

[0019] Example 2: Please refer to Figures 1 to 7 A method for constructing a "hot-dip galvanizing-leveling" coupled model for predicting surface roughness of GA belts includes the following steps: Step 1, model data acquisition; Step 2, data standardization and dataset partitioning; Step 3, surface roughness prediction model construction; Step 4, surface roughness attenuation model construction for leveling work rolls; Step 5, surface roughness prediction model construction for leveling process; and Step 6, simultaneous model analysis and LM algorithm regression. In step one above, the parameters required for model construction are first collected. A total of 6551 industrial data points were collected, and each data point includes the surface roughness of the GA strip steel. The corresponding process parameters are as follows: Hot-dip galvanizing process parameters: alloying temperature Alloying speed and the iron content of the coating ; Leveling process parameters: Inlet thickness Yield strength Elongation at flatness The original roughness of the work roll Rolling kilometers of work rolls ; In step two above, all collected data features are standardized using z-scores, transforming the data into a distribution with a mean of 0 and a standard deviation of 1. The standardization formula is as follows: In the formula For standardized data points, For the data points before standardization, The mean of the dataset. To determine the standard deviation of the dataset, the standardized dataset is divided into a training set and a test set, using an 8:2 ratio. That is, 80% of the data is used for model training and 20% of the data is used for model testing. This division strategy helps to fully fit the model during training and objectively evaluate its performance on an independent test set. In step three above, by summarizing experience from the production site and through various types of correlation analysis, the alloying temperature was determined. Alloying speed and the iron content of the coating Surface roughness exhibits a strong correlation with surface roughness. To capture the necessary nonlinearity and feature interaction, a surface roughness prediction model for the hot-dip galvanizing stage is constructed. The following quadratic polynomial nonlinear model is established: in The surface roughness of the strip steel at the exit of the hot-dip galvanizing process. These are the model coefficients; In step four above, a surface roughness attenuation model for the leveling work rolls is constructed. During the leveling rolling process, the roughness of the work rolls is a key factor affecting the surface roughness of the strip. However, as the rolling process progresses, the surface roughness of the work rolls will attenuate. The current roughness of the work rolls can be expressed as a function of the original roughness of the work rolls and the rolling mileage. The specific model is as follows: in The surface roughness of the current working roll. These are the model coefficients; In step five above, a surface roughness prediction model for the leveling process is constructed: the surface roughness of the strip steel exiting the leveling process consists of two parts: the inheritance of the incoming material roughness and the imprint of the work roll roughness on the strip steel. The specific model is usually expressed as follows: In the formula To improve the surface roughness of the strip steel at the exit of the leveling process; To smooth the inherited roughness of the incoming steel strip; The roughness formed by the imprinting of the work roll is represented by a model that, although divided into two parts, is influenced by strip parameters and leveling rolling process parameters. The specific model can be expressed as follows: In the formula These are the model coefficients; In step six above, the simultaneous model analysis and LM algorithm regression are performed: by combining the two parts of the model, the following GA strip roughness prediction model can be obtained: For multiple empirical coefficients in the coupled "hot-dip galvanizing-flattening" model ( and The regression analysis employs the Levenberg-Marquardt algorithm to address the resulting nonlinear least squares problem. The regression process iteratively adjusts these model coefficients to minimize the sum of squared residuals between the predicted values ​​of the "hot-dip galvanizing-smoothing" model and the observed roughness data. All undetermined coefficients in the model are defined as a parameter vector. Then, the loss function is defined as the sum of squared residuals, i.e., SSR, whose mathematical expression is: In the formula It is the actual measured roughness value of the i-th sample. At the current parameter estimate Below, the roughness value of the i-th sample predicted by the model, after defining the parameter vector and loss function, the initial parameter estimates of the LM algorithm are set. Initial step size adjustment factor =0.1 and the threshold for determining convergence The algorithm enters the iterative process. Through the iterative optimization of the LM algorithm, the coefficients in the "hot galvanizing-flattening" model can be effectively determined, thereby obtaining an accurate and reliable GA strip surface roughness prediction model. Tables 3 and 4 give the initial values ​​of the empirical coefficients of the initial parameter estimates and the regression values ​​after the iteration.

[0020] Table 1. Model coefficients for the hot-dip galvanizing process "hot-dip galvanizing-smoo Table 2. Model coefficients for the "hot-dip galvanizing-smoothing" process in the leveling step. Figure 6 The model's performance in roughness prediction is demonstrated, including the relationship between predicted and actual values ​​in the test and training sets. It can be seen that the "hot-dip galvanizing-smoothing" model achieves high R-values ​​in both the test and training sets. 2 The values ​​were 0.8621 and 0.8745, respectively, with RMSE values ​​of 0.0751 μm and 0.0743 μm, indicating that the model has a certain degree of accuracy in practical applications and good generalization ability.

[0021] Figure 7 The model's prediction performance is shown for different steel grades (the most numerous in the test dataset) and the number of coils for each steel grade. It can be seen that the model performs excellently in most steel grades (DQ-IF, 1180DP, 590DP), with prediction accuracy exceeding 90% and reaching a maximum of 96.65%. For the 440DQ steel grade, which has fewer coils, although the prediction accuracy decreases, it still remains at 89.5%. This fully demonstrates the model's good robustness and effective handling of different data distributions, proving the effectiveness of the model.

[0022] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for constructing a "hot-dip galvanizing-leveling" coupled model for predicting surface roughness of GA strips, comprising: Step 1, model data acquisition; Step 2, data standardization and dataset partitioning; Step 3, surface roughness prediction model construction; Step 4, surface roughness attenuation model construction for leveling work rolls; Step 5, surface roughness prediction model construction for leveling process; Step 6, simultaneous model analysis and LM algorithm regression; characterized in that: In step one above, the parameters required for model construction are first collected. In step two above, all collected data features are standardized using z-score, transforming the data into a distribution with a mean of 0 and a standard deviation of 1. In step three above, a surface roughness prediction model for the hot-dip galvanizing stage is constructed: a nonlinear model in the form of a quadratic polynomial is established as follows; In step four above, a surface roughness attenuation model for the leveling work roll is constructed: the roughness of the current work roll can be expressed as a function of the original roughness of the work roll and its rolling mileage. In step five above, a surface roughness prediction model for the leveling process is constructed: the surface roughness of the strip steel exiting the leveling process consists of two parts: the inheritance of the incoming material roughness and the imprint of the work roll roughness on the strip steel. In step six above, the simultaneous model analysis and LM algorithm regression are performed: by combining the two models, the following GA strip roughness prediction model can be obtained: For multiple empirical coefficients in the coupled "hot-dip galvanizing-flattening" model ( and The regression of the model is performed using the Levenberg-Marquardt algorithm to solve the resulting nonlinear least squares problem. The regression process iteratively adjusts these model coefficients to minimize the sum of squared residuals between the predicted values ​​of the "hot-dip galvanizing-smoothing" model and the observed roughness data.

2. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step one, the required parameter is: alloying temperature. Alloying speed and the iron content of the coating Inlet thickness Yield strength Elongation at flatness The original roughness of the work roll Rolling kilometers of work rolls and the surface roughness of the finished GA strip steel .

3. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step two, the standardized formula is as follows: In the formula For standardized data points, For the data points before standardization, The mean of the dataset. Let $\frac{ ...

4. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step three, the nonlinear model in quadratic polynomial form is: in The surface roughness of the strip steel at the exit of the hot-dip galvanizing process. These are the model coefficients.

5. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step four, the specific model is represented as follows: in The surface roughness of the current working roll. These are the model coefficients.

6. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step five, the specific model is typically represented as follows: In the formula To improve the surface roughness of the strip steel at the exit of the leveling process; To smooth the inherited roughness of the incoming steel strip; The roughness is caused by the imprinting of the work roller.

7. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step five, although the model is divided into two parts, each part is affected by the strip parameters and the leveling and rolling process parameters. The specific model can be represented as follows: In the formula These are the model coefficients.

8. The method for constructing a "hot-dip coating-smoothing" coupled model for predicting surface roughness of GA strips according to claim 1, characterized in that: In step six, all the undetermined coefficients in the model are defined as a parameter vector. Then, the loss function is defined as the sum of squared residuals, i.e., SSR, whose mathematical expression is: In the formula It is the actual measured roughness value of the i-th sample. At the current parameter estimate Below, the roughness value of the i-th sample predicted by the model, after defining the parameter vector and loss function, the initial parameter estimates of the LM algorithm are set. Initial step size adjustment factor And the threshold for determining convergence The algorithm then enters the iterative process, and after the iteration is complete, it will output the optimal parameter estimates. Through iterative optimization of the LM algorithm, the coefficients in the "hot-dip galvanizing-smoothing" model can be effectively determined, thus obtaining an accurate and reliable GA strip surface roughness prediction model.

Images