A monitoring method for steel coil coating

Abstract

The present invention discloses a monitoring method for steel coil coating, which belongs to the technical field of steel coil galvanizing. The monitoring method for steel coil coating comprises the following steps: obtaining a historical monitoring set D, and using the historical monitoring set D to construct a pressure proximity model; obtaining a real-time blowing spacing X t , and based on the pressure proximity model, the real-time blowing distance X t And the thickness qualified range reverse to find the real-time blowing pressure P t ; With real-time blowing pressure P t The steel strip is blow-plated, and the leveling image of the blowing and plating position is photographed. The blowing and plating leveling image is processed to obtain the leveling information, and the leveling information and the leveling threshold are compared to obtain the monitoring result. The monitoring method proposed in the present invention effectively solves the problem of blowing and plating quality caused by the shaking of the steel coil during the galvanizing process of the steel coil by accurately controlling the blowing and plating pressure during the blowing and plating process, and combines the real-time monitoring and analysis of the leveling information, thereby improving the consistency and yield of the blowing and plating, and has a significant positive effect on improving the production efficiency and product quality of the galvanizing of the steel coil.

Description

Technical Field

[0001] The invention belongs to the technical field of steel coil galvanizing, in particular to a monitoring method for steel coil galvanizing. Background Art

[0002] Galvanizing of steel coils is a common metal surface treatment technology used to improve the corrosion resistance and service life of steel. During the galvanizing process, in order to control the coating thickness, an air knife system is usually used to blow away excess zinc liquid on the surface of the steel coil.

[0003] The invention patent with application number 202310347603.8 discloses a cutting edge knife for removing excess zinc liquid on the thickness edge. However, in actual production, due to factors such as roller non-parallelism or excessive roller spacing, the steel coil will inevitably shake when moving at high speed, which directly affects the purging effect of the cutting edge knife and the consistency of galvanizing quality. At present, in order to solve the problem of steel coil shaking, the cutting edge knife output is linearly controlled by the distance between the cutting edge knife and the steel strip, so as to achieve consistency in the purging effect.

[0004] After a long period of testing, it was found that the above-mentioned method of solving the problem of steel coil shaking can improve production quality to a certain extent, but the output of the edge cutting knife is in a scattered form, and its output effect is not in an ideal linear relationship with the steel coil spacing. This leads to limited improvement of the quality of steel strip blowing and plating by existing control methods, and a high defective rate of finished products. Therefore, a precise linkage method of distance monitoring and air knife output is needed, aiming to control the steel coil shaking and air knife output effect in real time, and achieve high-quality galvanizing through precise control. Summary of the invention

[0005] Purpose of the invention: To provide a method for monitoring the coating of steel coils to solve the above problems existing in the prior art.

[0006] Technical solution: A method for monitoring the coating of a steel coil comprises the following steps:

[0007] Step 1: Obtain the historical monitoring set D, and use the historical monitoring set D to build a pressure proximity model; where the historical monitoring set D = {X, P, H}, X is the blowing distance, P is the blowing pressure, and H is the quality inspection thickness;

[0008] Step 2: Get the real-time plating spacing X t , and based on the pressure proximity model, the real-time blowing distance X t And the thickness qualified range reverse to find the real-time blowing pressure P t ;

[0009] Step 3: Use the real-time blowing pressure P t The steel strip is blow-plated, and the leveling image of the blowing and plating position is taken. The leveling image is processed to obtain the leveling information, and the monitoring result is obtained by comparing the leveling information with the leveling threshold range.

[0010] Furthermore, the construction of the pressure proximity model includes the following steps:

[0011] Step 11: Normalize the plating spacing X and plating pressure P in the historical monitoring data set D to obtain the normalized spacing X g and normalized pressure P g , D g ={X g , P g , H} as the normalized monitoring set, and divide the normalized monitoring set into a training set and a validation set in proportion;

[0012] Step 12: Use the training set to train the pressure proximity model;

[0013] Step 13: Use the loss function to minimize and optimize the model parameters;

[0014] Step 14: Use the validation set to evaluate the accuracy of the pressure proximity model.

[0015] Furthermore, the training of the pressure proximity model using the training set includes the following steps:

[0016] Step 121: Using the formula Expressing the pressure proximity model; where w is the weight vector, b is the bias term, φ(X g ,P g ) is the feature mapping function;

[0017] Step 122: Using the formula Indirectly calculate the feature mapping function; where γ is the regularization parameter, X i is the plating spacing of the ith data point, P i Is with X i Corresponding blowing pressure;

[0018] Step 123: Input the training set into the pressure proximity model, and the pressure proximity model learns to obtain the weight vector w and the bias term b.

[0019] Furthermore, the loss function is , the model parameters include weight vector w, bias term b, regularization parameter γ and penalty parameter C; where, and is the slack variable, is a positive deviation, is a negative deviation.

[0020] Further, the blowing pressure P is found t The following steps are involved:

[0021] Step 21: Uniformly select N thickness values H within the qualified thickness range i ;

[0022] Step 22: Combine the real-time blowing plating spacing X t with the N thickness values H i respectively to form a search set U, U = {X t , H i};

[0023] Step 23: Input the search set U into the pressure proximity model and inversely solve to obtain N blowing plating values P i ;

[0024] Step 24: Take the P t-1 in min|P i - P i | or the H t-1 in min|H i - H i |, where the corresponding P i is used as the real-time blowing plating pressure P t ;

[0025] where i ∈ (1, N), P t-1 is the blowing plating pressure at the previous moment of the real-time blowing plating pressure P t , and H t-1 is the quality inspection thickness corresponding to P t-1 .

[0026] Furthermore, obtaining the leveling information corresponding to the blowing plating pressure P t includes the following steps:

[0027] Step 311: Grayscale and filter the leveling image corresponding to the blowing plating pressure P t to obtain a monitored leveling image;

[0028] Step 312: Use the Sobel edge detection algorithm to obtain the monitored ripple contour in the monitored leveling image;

[0029] Step 313: Use image moments and the Hough transform to determine the center point and ripple direction of the monitored ripple contour respectively, and take the center point and ripple direction of the monitored ripple contour as the leveling information.

[0030] Furthermore, obtaining the leveling threshold range includes the following steps:

[0031] Step 321: Screen the leveling images corresponding to the qualified quality inspection thickness H in the historical monitoring set to obtain a threshold image;

[0032] Step 322: Grayscale and filter the threshold image to obtain a preprocessed image;

[0033] Step 323: using the Sobel edge detection algorithm to obtain the threshold ripple profile in the preprocessed image;

[0034] Step 324: using image moment and Hough transform to respectively determine the center point and ripple direction of the threshold ripple profile;

[0035] Step 325: overlap and union all threshold ripple contours using the center point and ripple direction to obtain a union contour, and use the union contour as the leveling threshold range.

[0036] Furthermore, the comparing the leveling information and the leveling threshold to obtain the monitoring result includes the following steps:

[0037] Step 331: using image moment and Hough transform to respectively determine the center point and ripple direction of the union contour;

[0038] Step 332: Overlap the union contour with the monitoring ripple contour using the center point and the ripple direction;

[0039] Step 333: Determine whether the monitored corrugation profile is located inside the union profile. If so, the monitoring result is qualified blow plating, otherwise the monitoring result is unqualified blow plating.

[0040] Beneficial effects: The present invention uses historical monitoring data to construct a pressure proximity model, and can output the real-time blowing pressure in time according to the real-time monitoring of the blowing spacing of the steel coil, thereby ensuring the consistency of the blowing thickness. The blowing process is dynamically adjusted according to the change of the blowing spacing, avoiding the limitations of the traditional simple linear control of the cutting edge knife output and improving the blowing accuracy. Under high-precision blowing, the present invention accurately extracts the corrugated contour of the leveling image, providing an intuitive visual basis for blowing quality monitoring. Through real-time feature comparison, the blowing status can be fed back in time, which is conducive to timely adjustment of process parameters and reducing the defective rate of finished products.

[0041] In summary, the monitoring method proposed in the present invention effectively solves the problem of blowing plating quality caused by the shaking of steel coils during the galvanizing process by accurately controlling the blowing plating pressure during the blowing plating process and combining it with real-time monitoring and analysis of leveling information, thereby improving the consistency and yield of blowing plating, and has a significant positive effect on improving the production efficiency and product quality of galvanizing steel coils. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the flow of the monitoring method of the present invention. DETAILED DESCRIPTION

[0043] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features well known in the art are not described.

[0044] like Figure 1 As shown, a method for monitoring the coating of a steel coil comprises the following steps:

[0045] Step 1: Obtain the historical monitoring set D, and use the historical monitoring set D to build a pressure proximity model; where the historical monitoring set D = {X, P, H}, X is the blowing distance, P is the blowing pressure (air knife pressure), and H is the quality inspection thickness;

[0046] Step 2: Get the real-time plating spacing X t , and based on the pressure proximity model, the real-time blowing distance X t And the thickness qualified range reverse to find the real-time blowing pressure P t ;

[0047] Step 3: Use the real-time blowing pressure P t The steel strip is blow-plated, and the leveling image of the blowing and plating position is taken. The leveling image is processed to obtain the leveling information, and the monitoring result is obtained by comparing the leveling information with the leveling threshold range.

[0048] The monitoring method proposed in the present invention utilizes the pressure proximity model to accurately control the blowing pressure during the blowing process, and combines the real-time monitoring and analysis of leveling information to effectively solve the blowing quality problem caused by the shaking of the steel coil during the galvanizing process of the steel coil, thereby improving the consistency and yield of the blowing, and has a significant positive effect on improving the production efficiency and product quality of the galvanizing of the steel coil.

[0049] The construction of the pressure proximity model includes the following steps:

[0050] Step 11: Normalize the plating spacing X and plating pressure P in the historical monitoring data set D to obtain the normalized spacing X g and normalized pressure P g , D g ={X g , P g , H} as the normalized monitoring set, and divide the normalized monitoring set into a training set and a validation set according to the proportion. Among them, the normalization processing method is X g =(XX min ) / (X max -X min ), P g =(PP min ) / (P max -Pmin ), X max , X min , P max , P min are the maximum and minimum values ​​of X and P, respectively. The training set and validation set are allocated in a ratio of 7:3 or 8:2. By normalizing the plating spacing X and plating pressure P, the features that may have large numerical differences can be converted into data on a uniform scale, which helps to simplify the learning process of the model. When the feature scales are similar, the convergence speed of optimization algorithms such as gradient descent will be faster. The normalized feature value range is close, which helps to reduce the number of training iterations and shorten the model training time.

[0051] Step 12: Use the training set to train the pressure proximity model. Input the training set into the pressure proximity model. This model is a machine learning model that can predict the blowing pressure from the input blowing spacing and quality inspection thickness. During the model training phase, the model adjusts the internal weight vector w and bias term b by learning the data in the training set in order to achieve the best prediction effect. This is achieved through the back propagation algorithm, which updates the model parameters according to the gap between the predicted results and the actual results, so that it gradually approaches the optimal solution.

[0052] Step 13: Use the loss function to minimize and optimize the model parameters. In order to optimize the model parameters, a loss function is used to measure the gap between the model's predicted results and the actual results. The goal of training is to minimize this loss function so that the model can predict the blowing pressure as accurately as possible.

[0053] Step 14: Use the validation set to evaluate the accuracy of the pressure approximation model. After training, the performance of the model needs to be evaluated using the validation set. The validation set is a set of data that was not used in the model training and is used to check the generalization ability of the model, that is, the model's ability to predict new data. If the model performs well on the validation set, it means that the model has been properly trained and can predict the plating pressure well.

[0054] Training the pressure proximity model using the training set includes the following steps:

[0055] Step 121: Using the formula Expressing the pressure proximity model; where w is the weight vector, b is the bias term, φ(X g ,P g) is the feature mapping function. In the steel coil coating process, the relationship between the blowing pressure, blowing spacing and quality inspection thickness is very complex. It is not a simple linear or quadratic relationship, but contains nonlinear interactive relationships. The deep learning model can capture this complex relationship. The expression formula of the pressure proximity model does not directly give a specific mathematical formula, but describes the construction and use process of a machine learning model. In this model, instead of using traditional physical formulas, a data-driven approach is adopted, that is, historical monitoring data is used to train the model so that it can predict the blowing pressure under given blowing spacing and quality inspection thickness.

[0056] Step 122: Using the formula Indirect calculation of the feature mapping function; where γ is the regularization parameter, Xi is the blowing distance of the i-th data point, and Pi is the blowing pressure corresponding to Xi. Using the above formula to indirectly calculate the feature mapping function to simplify the operation is to convert the data in the original feature space to a higher-dimensional space, so that the data in the new space can become linearly separable, which can greatly reduce the calculation cost and improve the calculation efficiency.

[0057] Step 123: Input the training set into the pressure proximity model, and the pressure proximity model learns the weight vector w and the bias term b. Before starting training, the weight vector w and the bias term b are usually randomly initialized, and then continuously adjusted during the training process to minimize the gap between the predicted results and the actual results.

[0058] The loss function is , the model parameters include the weight vector w, the bias term b, the regularization parameter γ and the penalty parameter C; where, and is the slack variable, is a positive deviation, that is, the data point is above the regression line and exceeds the ε tubular area). is the negative deviation (i.e., the data point is below the regression line and outside the ε tubular region). In support vector regression, and These two slack variables play a key role and are introduced into the model to handle situations where the data points do not fully satisfy the strict linear or nonlinear regression bounds. Specifically, and Data points are allowed to deviate to a certain extent from the so-called "ε-tubular region", which is a boundary band around the prediction function with a width of 2ε. These variables are weighted in the objective function to minimize their sum, which helps to control the complexity of the model and the risk of overfitting. The weights are usually determined by a penalty coefficient C, which balances the error and complexity of the model. A larger value of C means that the model is more inclined to reduce the error, even if it means increasing the complexity of the model; a smaller value of C tends to simplify the model and tolerate more errors.

[0059] Search for the blowing pressure P t including the following steps:

[0060] Step 21: Uniformly select N thickness values H within the qualified thickness range i ;

[0061] Step 22: Combine the real-time blowing distance X t with the N thickness values H i respectively to form a search set U, U = {X t , H i};

[0062] Step 23: Input the search set U into the pressure proximity model and solve it inversely to obtain N blowing values P i ;

[0063] Step 24: Take the P t-1 in min|P i - P i | or the P t-1 corresponding to the H i in min|H i - H i | as the real-time blowing pressure P t ;

[0064] where i ∈ (1, N), P t-1 is the blowing pressure at the previous moment of the real-time blowing pressure P t , H t-1 is the quality inspection thickness corresponding to P t-1 . When the real-time blowing pressure P t = min|P t-1 - P i |, it is considered based on the stable change of the air knife output pressure, which is beneficial to the stable operation of the equipment. When the real-time blowing pressure P t = the P t-1 corresponding to the H i in min|H i - H i |, it is considered based on the stable change of the galvanized thickness, which is beneficial to the flatness of the galvanized steel strip.

[0065] Get the blowing pressure P t The corresponding leveling information includes the following steps:

[0066] Step 311: Adjust the plating pressure P t The corresponding leveling image is grayed and filtered to obtain the monitoring leveling image. Graying can remove color information and only retain the brightness level of the image, which helps to enhance the contrast of different areas in the image and make subsequent feature extraction more accurate. Filtering and denoising can eliminate random noise in the image, reduce unnecessary interference, and improve the accuracy of subsequent processing.

[0067] Step 312: Use the Sobel edge detection algorithm to obtain the monitoring ripple profile in the monitoring leveling image. The Sobel edge detection algorithm is a gradient operator widely used in image processing. It can effectively detect boundaries with large intensity changes in images, that is, the monitoring ripple profile in the monitoring leveling image. This step allows us to clearly identify the tiny ripples on the coating surface, providing a basis for further analysis. Image moments can help us calculate the center point of the monitoring ripple profile, which is an important part of the leveling information and can reflect the overall position and morphology of the coating surface. The Hough transform is used to identify the direction of the ripples, which is crucial for understanding the direction and effect of gas flow during the blow coating process. The combination of these two technologies not only provides a quantitative description of the leveling information, but also helps us diagnose and adjust the blow coating process to achieve the best leveling effect.

[0068] Step 313: Use image moment and Hough transform to determine the center point and corrugation direction of the monitored corrugation profile, and use the center point and corrugation direction of the monitored corrugation profile as leveling information. This leveling information acquisition method based on image processing technology is an important means to achieve automated monitoring and quality control in modern manufacturing, and has significant technological advancement significance.

[0069] Obtaining the leveling threshold range includes the following steps:

[0070] Step 321: Filter the leveling images corresponding to the locations where the thickness H meets the standard in the historical monitoring concentration quality inspection to obtain the threshold image;

[0071] Step 322: grayscale and filter the threshold image to obtain a preprocessed image;

[0072] Step 323: using the Sobel edge detection algorithm to obtain the threshold ripple profile in the preprocessed image;

[0073] Step 324: using image moment and Hough transform to respectively determine the center point and ripple direction of the threshold ripple profile;

[0074] Step 325: overlap and union all threshold ripple contours using the center point and ripple direction to obtain a union contour, and use the union contour as the leveling threshold range.

[0075] The method of obtaining the leveling threshold range is similar to the method of obtaining the leveling information. The difference is that the leveling threshold range is the merged result of the leveling image corresponding to the point where the thickness H meets the standard in the historical monitoring concentration quality inspection, which reflects the floating range of the corrugated profile that meets the galvanizing thickness quality inspection requirements.

[0076] Comparing the leveling information and the leveling threshold to obtain the monitoring results includes the following steps:

[0077] Step 331: using image moment and Hough transform to respectively determine the center point and ripple direction of the union contour;

[0078] Step 332: Overlap the union profile with the monitoring profile using the center point and the ripple direction; by overlapping the individual monitoring profiles with the union profile, it is possible to visually check whether each local ripple is consistent with the overall ripple distribution. This overlapping operation not only verifies the position and direction of individual ripples, but also ensures their consistency with the global ripple pattern, thereby improving the reliability of the monitoring results.

[0079] Step 333: Determine whether the monitoring corrugation profile is located inside the union profile. If so, the monitoring result is qualified blow plating, otherwise the monitoring result is unqualified blow plating. By determining whether the monitoring corrugation profile is completely located inside the union profile, it is possible to directly draw a conclusion on whether the blow plating is qualified. If all monitoring corrugation profiles are within the union profile, this means that the corrugations on the coating surface are evenly distributed, and there are no abnormal fluctuations beyond the normal range, so it can be determined as qualified blow plating. Otherwise, it may indicate defects or unevenness, and further investigation of the cause and process adjustments are required.

[0080] The rapid feedback of monitoring results helps to timely discover and solve problems in the blowing process, promote continuous process optimization, and ensure stable and reliable coating quality. In summary, through the above steps, we have built an efficient, accurate and automated blowing quality monitoring system, which is of great significance for improving production efficiency, reducing costs and ensuring the quality of the final product.

[0081] The preferred embodiments of the invention are described in detail above in conjunction with the accompanying drawings; however, the invention is not limited to the specific details in the above embodiments; within the technical concept of the invention, various equivalent transformations may be made to the technical solutions of the invention, and all of these equivalent transformations fall within the protection scope of the invention.

Claims

1. A method for monitoring steel coil coating, characterized in that: The following steps are involved: Step 1: Obtain the historical monitoring set D, and use the historical monitoring set D to build a pressure proximity model; where the historical monitoring set D = {X, P, H}, X is the blowing distance, P is the blowing pressure, and H is the quality inspection thickness; Step 2: Get the real-time plating spacing X t , and based on the pressure proximity model, the real-time blowing distance X t And the thickness qualified range reverse to find the real-time blowing pressure P t ; Step 3: Use the real-time blowing pressure P t Blow-coat the steel strip and take a leveling image of the blowing and plating position. Process the leveling image to obtain the leveling information. Compare the leveling information with the leveling threshold range to obtain the monitoring result. Wherein, in step 2, the blowing pressure P is found t The following steps are involved: Step 21: Evenly select N thickness values ​​H within the qualified thickness range i ; Step 22: Set the real-time plating spacing to X t Respectively with N thickness values ​​H i Combined into a search set U, U = {X t , H i }; Step 23: Input the search set U into the pressure proximity model and solve inversely to obtain N plating values ​​P i ; Step 24: Set min|P t-1 -P i |P in i or min|H t-1 -H i |Medium H i The corresponding P i As the real-time blowing pressure P t ; Where i∈(1,N), P t-1 is the real-time blowing pressure P t The blowing pressure at the previous moment, H t-1 P t-1 The corresponding quality inspection thickness; Get the blowing pressure P t The corresponding leveling information includes the following steps: Step 311: Adjust the plating pressure P t The corresponding leveling image is grayed and filtered to remove noise to obtain a monitoring leveling image; Step 312: using the Sobel edge detection algorithm to obtain the monitoring ripple contour in the monitoring leveling image; Step 313: using image moment and Hough transform to respectively determine the center point and the corrugation direction of the monitoring corrugation profile, and using the center point and the corrugation direction of the monitoring corrugation profile as leveling information; Obtaining the leveling threshold range includes the following steps: Step 321: Filter the leveling images corresponding to the locations where the thickness H meets the standard in the historical monitoring concentration quality inspection to obtain the threshold image; Step 322: grayscale and filter the threshold image to obtain a preprocessed image; Step 323: using the Sobel edge detection algorithm to obtain the threshold ripple profile in the preprocessed image; Step 324: using image moment and Hough transform to respectively determine the center point and ripple direction of the threshold ripple profile; Step 325: overlap and union all threshold ripple contours using the center point and the ripple direction to obtain a union contour, and use the union contour as the leveling threshold range; The comparison of the leveling information and the leveling threshold to obtain the monitoring result comprises the following steps: Step 331: using image moment and Hough transform to respectively determine the center point and ripple direction of the union contour; Step 332: Overlap the union contour with the monitoring ripple contour using the center point and the ripple direction; Step 333: Determine whether the monitored corrugation profile is located inside the union profile. If so, the monitoring result is qualified blow plating, otherwise the monitoring result is unqualified blow plating.

2. A method for monitoring steel coil coating according to claim 1, characterized in that: The construction of the pressure proximity model includes the following steps: Step 11: Normalize the plating spacing X and plating pressure P in the historical monitoring data set D to obtain the normalized spacing X g and normalized pressure P g , D g ={X g , P g , H} as the normalized monitoring set, and divide the normalized monitoring set into a training set and a validation set in proportion; Step 12: Use the training set to train the pressure proximity model; Step 13: Use the loss function to minimize and optimize the model parameters; Step 14: Use the validation set to evaluate the accuracy of the pressure proximity model.

3. A method for monitoring steel coil coating according to claim 2, characterized in that: The method of training the pressure proximity model using the training set comprises the following steps: Step 121: Using the formula Expressing the pressure proximity model; where w is the weight vector, b is the bias term, φ(X g ,P g ) is the feature mapping function; Step 122: Using the formula Indirectly calculate the feature mapping function; where γ is the regularization parameter, X i is the plating spacing of the ith data point, P i is with X i Corresponding blowing pressure; Step 123: Input the training set into the pressure proximity model, and the pressure proximity model learns to obtain the weight vector w and the bias term b.

4. A method for monitoring steel coil coating according to claim 3, characterized in that: The loss function is , the model parameters include weight vector w, bias term b, regularization parameter γ and penalty parameter C; where, and is the slack variable, is a positive deviation, is a negative deviation.

Images