A hot-dip strip steel surface quality control method and electronic device

Abstract

The present application relates to the technical field of strip production, and in particular to a hot-dip strip surface quality control method and electronic equipment. The method comprises: S1, establishing a data set required for model training; S2, constructing a condensation rate prediction model using the data set; S3, dynamically optimizing and controlling the process in real-time production based on the condensation rate prediction model. The dynamic optimization and control comprises: solving the bias compensation value of the condensation rate prediction model, and then obtaining a corrected condensation rate prediction model; based on the corrected condensation rate prediction model, optimizing the process parameters before the condensation of the hot-dip strip, and adjusting the process parameters based on the optimization results. The present application ensures the surface quality of the final hot-dip strip by controlling the position of the condensation line in the strip condensation process, thereby solving the problems of poor real-time performance, low control precision and hysteresis of manual control, and also being able to adapt to strip surface quality control of different requirements and processes.

Description

Technical Field

[0001] This invention relates to the field of strip steel production technology, specifically to a method and electronic device for controlling the surface quality of hot-dip galvanized strip steel. Background Technology

[0002] Hot-dip galvanized steel strip is a processing method in which a metallic or non-metallic coating is hot-dip galvanized onto the surface of steel strip. In the deep processing of steel strip, this typically refers to hot-dip galvanizing with zinc or other coatings to improve its corrosion resistance and service life. Galvanized steel strip is one of the most common hot-dip galvanized products. The production of galvanized steel strip mainly involves several stages: pickling, galvanizing, and cooling. This includes cleaning and drying after pickling, and air blowing after galvanizing. Each step directly affects the final quality of the galvanized steel strip; therefore, controlling the surface quality of galvanized steel strip is a complex and challenging technical problem.

[0003] To control the surface quality of galvanized steel strip, people usually conduct quality inspections on the final product after the galvanized steel strip is processed. When quality problems occur, the process parameters are adjusted in reverse. The parameter adjustment is usually based on experience judgment or continuous debugging and trial production. This control method not only has serious lag, but also the debugging is inaccurate, which affects the normal production.

[0004] Existing technology CN113136537A discloses a method for improving the surface quality of hot-dip galvanized strip steel, comprising: cleaning a hot-rolled substrate and then heating it to obtain an annealed steel sheet; the heating is carried out in a hydrogen atmosphere with an oxygen content ≤5ppm and a volume fraction of 18%–20%, and the pressure is controlled at 170–190Pa, the furnace temperature at 1280–1300℃, and the strip temperature of the hot-rolled substrate at 650–680℃; the annealed steel sheet is then hot-dip galvanized, cooled, and surface treated to obtain hot-dip galvanized strip steel with high surface quality. This method reduces the probability of surface quality defects in the strip steel and ensures the stability of product quality. However, the process parameters are not universal for galvanizing processes on different substrates or for galvanized strip steel with different requirements, so the method provided by the existing technology is only applicable to a small range of galvanized strip steel products. In addition, for other surface treatment processes of strip steel, such as hot-dip galvanizing, aluminum zinc silicon galvanizing, zinc aluminum galvanizing, etc., there is a cooling process after hot-dip galvanizing, which results in a serious lag in the control of surface treatment process parameters of strip steel.

[0005] Therefore, it is necessary to provide a method and electronic device for controlling the surface quality of hot-dip galvanized steel strip, which can quickly adjust important parameters in the hot-dip galvanizing process of steel strip, thereby improving the surface quality of hot-dip galvanized steel strip. Summary of the Invention

[0006] In order to solve the above-mentioned technical problems in the prior art, the present invention provides a method for controlling the surface quality of hot-dip galvanized steel strip and an electronic device.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] A method for controlling the surface quality of hot-dip galvanized steel strip includes:

[0009] S1. Establish the dataset needed for model training;

[0010] S2. Construct a condensation rate prediction model using the dataset;

[0011] S3. Dynamically optimize and control the process in real-time production based on the condensation rate prediction model;

[0012] Step S3 specifically includes:

[0013] S301. Obtain the n images of the hot-dip galvanized steel strip surface that have passed the test, and calculate the actual condensation rate corresponding to each image.

[0014] S302. Calculate n sets of strip steel process history data corresponding to n actual condensation rates.

[0015] S303. Based on the condensation rate prediction model in step S2, predict the n sets of process history data of the strip steel that have passed through, and obtain the average deviation between the predicted n sets of condensation rates and the actual condensation rates, which is used as the deviation compensation value of the condensation rate prediction model.

[0016] S304. Based on the deviation compensation values, the corrected condensation rate prediction model is obtained;

[0017] S305. Based on the modified condensation rate prediction model, the process parameters before the condensation of hot-dip galvanized steel strip are optimized, and the process parameters are adjusted based on the optimization results.

[0018] Further, the method for calculating the true condensation rate in step S301 includes: identifying the rectangular region where the strip is located, the width direction of the rectangular region being the width direction of the strip, and dividing the strip in the rectangular region image along the width direction of the strip into a left region with an area ratio of a, a middle region with an area ratio of 1-2a, and a right region with an area ratio of a.

[0019] The percentage of pixels in the condensed portion is calculated separately for the left and right regions, and denoted as R. left and R right ;but

[0020]

[0021] Where R is the actual condensation rate.

[0022] Furthermore, the range of values ​​for a is: 0 < a ≤ 0.5.

[0023] Furthermore, the method for calculating the deviation compensation value in step S303 is as follows:

[0024]

[0025] Where e is the real-time deviation, i.e., the deviation compensation value, and y i Let m(x) represent the true condensation rate of the i-th image. i ) indicates that the process parameter x is predicted using the condensation rate prediction model m(). i The predicted condensation rate is obtained.

[0026] Furthermore, the revised method for calculating the predicted condensation rate is as follows:

[0027] m new ()=m()+e

[0028] Where, m new () indicates the modified condensation rate prediction model.

[0029] Furthermore, step S305 specifically includes:

[0030] S3501, Obtain the k processes preceding the process where the hot-dip galvanized steel strip surface inspection camera is located: Process1, Process2, ..., Process... k Location;

[0031] S3502, Calculate the position of each process step. i The distance between the strip and the processes it has already passed through is calculated, and the actual process parameters of the processes that have already passed through are calculated when the strip passes through.

[0032] S3503. Based on the location of the camera, optimize the process of k steps using a near-to-far approach, including:

[0033] Obtain the process parameters of Process 1 at the current moment, and combine them with the actual process parameters experienced by the strip steel before Process 1. Use the condensation rate prediction model to predict the condensation rate of the strip steel before Process 1.

[0034] The process parameters corresponding to the process position Process1 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters corresponding to the process position Process1 are then denoted as p1.

[0035] Obtain the process parameters of Process2 at the current moment, and combine them with the actual process parameters of the strip steel before Process2 and p1. Use the condensation rate prediction model to predict the condensation rate of the strip steel before Process2.

[0036] The process parameters corresponding to the process position Process2 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters corresponding to the process position Process2 are then denoted as p2.

[0037] Repeat the optimization process until k processes are optimized, obtaining the optimization results p1, p2, ..., p for each process. k ;

[0038] The entire process for hot-dip galvanized steel strip was adjusted based on the optimization results.

[0039] Furthermore, step S1 specifically includes:

[0040] S101. Obtain real-time image data of hot-dip galvanized steel strip after condensation, calculate the true condensation rate, and use the true condensation rate as a label.

[0041] S102. Perform spatiotemporal transformation on the time-process data in the production process corresponding to each true condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each label after the transformation.

[0042] S103. Combine the process history information of each position of the strip with the actual condensation rate label to obtain the dataset required for model training.

[0043] Furthermore, the spatiotemporal transformation in step S102 includes: based on the process parameter information along the time axis corresponding to the processes before the hot-dip galvanized strip solidification, solving the production state experienced by the strip position corresponding to the current solidification rate label through integral calculation, wherein the integral calculation includes:

[0044] Obtain the distance L between the camera and the strip at the target process position, and calculate the inverse-time iterative integral of the strip movement speed curve:

[0045]

[0046] Among them, v i Let t represent the speed of the strip at time i. i This represents the duration of the strip at time i;

[0047] Until the integral value When the time equals L, the corresponding time i is the time when the strip passes through the target process position. The process parameter value corresponding to time i is obtained from the time-process data curve, which is the process parameter value of the target process position.

[0048] Furthermore, the condensation rate prediction model constructed in step S2 can be constructed using methods such as statistical analysis, machine learning, deep learning, and mechanistic models.

[0049] Furthermore, the surface quality control methods for hot-dip galvanized steel strips also include:

[0050] S4. Update the condensation rate prediction model based on the newly generated data; specifically including:

[0051] S401, Update the dataset;

[0052] S402. Based on the updated dataset, update the condensation rate prediction model.

[0053] Furthermore, updating the dataset includes: calculating the strip steel process parameter x corresponding to the actual condensation rate R obtained in real-time production; calculating the similarity between process parameter x and each existing data point in the dataset; and selecting the smallest similarity from the process parameter x and each existing data point in the dataset, denoted as d. i Set a similarity threshold d limit If the similarity d i Greater than the similarity threshold d limit Then, the process parameter x and the corresponding actual condensation rate R are added to the dataset; if the similarity d i Less than or equal to the similarity threshold d limit If so, then process parameter x is discarded.

[0054] Furthermore, similarity can be expressed using Euclidean distance, but Mahalanobis distance, cosine distance, etc., can also be used.

[0055] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described hot-dip galvanized steel strip surface quality control method.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The hot-dip galvanized strip steel surface quality control method provided by this invention establishes the dataset required for model training, constructs a condensation rate prediction model using the dataset, and dynamically optimizes and controls the process during real-time production based on the condensation rate prediction model. In turn, by controlling the position of the condensation line of the strip steel during the cooling process, the surface quality of the final hot-dip galvanized strip steel is guaranteed. This solves the problems of poor real-time performance, low control accuracy, and lag in manual control. Furthermore, it can adapt to the surface quality control of hot-dip galvanized strip steel with different needs and processes, and has wide adaptability. Attached Figure Description

[0058] Figure 1 This is an image of the strip steel taken by the camera at a certain moment in an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the equipment connection for the galvanizing and high-tower cooling sections in an embodiment of the present invention.

[0060] Figure 3 This is a schematic diagram illustrating the principle of inverse time integration in an embodiment of the present invention.

[0061] Figure 4 This is a diagram illustrating the condensation control effect in an embodiment of the present invention.

[0062] Explanation of reference numerals in the attached figures:

[0063] 1. Zinc pot; 2. Air knife; 3. First mobile fan; 4. Second mobile fan; 5. First high-tower fan; 6. Second high-tower fan; 7. Third high-tower fan; 8. Fourth high-tower fan; 9. Camera; 10. Steel strip. Detailed Implementation

[0064] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0065] It should be noted that, unless otherwise specifically stated, the relative arrangement and numerical expressions of the components and steps described in these embodiments should not be construed as limiting the scope of the invention.

[0066] The following description of exemplary embodiments is merely illustrative and is not intended to limit the invention or its application or use in any way. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but where applicable, such techniques, methods, and apparatus should be considered part of this specification.

[0067] This invention provides a method for controlling the surface quality of hot-dip galvanized steel strip, comprising:

[0068] S1. Establish the dataset required for model training; specifically including:

[0069] S101. Use a camera to acquire real-time image data of the hot-dip galvanized strip after condensation, calculate the true condensation rate, and use the true condensation rate as a label. The method for calculating the true condensation rate includes: identifying the rectangular area where the strip is located, with the width direction of the rectangular area being the width direction of the strip, and dividing the strip in the rectangular area image along the width direction of the strip into a left area with an area ratio of 'a', a middle area with an area ratio of 1-2a, and a right area with an area ratio of 'a'. The value of 'a' is in the range of 0 < a ≤ 0.5. The specific value of 'a' is determined based on actual debugging. The closer 'a' is to the air knife, the smaller its value; the farther 'a' is from the air knife, the larger its value.

[0070] The percentage of pixels in the condensed portion is calculated separately for the left and right regions, and denoted as R. left and R right ;but

[0071]

[0072] Where R is the actual condensation rate.

[0073] S102. Perform spatiotemporal transformation on the time-process data in the production process corresponding to each true condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each label after the transformation.

[0074] The spatiotemporal transformation includes: based on the process parameter information along the time axis corresponding to the process before the condensation of the hot-dip galvanized strip, the production state experienced by the strip position corresponding to the current condensation rate label is solved by integral calculation. The production state specifically includes the parameters corresponding to the current strip position when it passes through each cooling device (e.g., a fan). If the cooling device is a fan, the parameters are frequency, wind speed, air volume, etc. The process parameters of the air knife corresponding to the current strip position when it passes through the air knife are also included, such as air knife pressure, distance between the air knife and the strip, and distance of the air knife from the ground. The parameters of the current strip position when it passes through the zinc pot are also included, such as duration and zinc liquid temperature. Finally, the temperature of the current strip position at the zinc pot inlet pyrometer is also included.

[0075] The integration operation includes: obtaining the distance L between the camera and the strip at the target process position, and calculating the iterative integral of the strip's moving speed curve.

[0076]

[0077] Among them, v i Let t represent the speed of the strip at time i. i This represents the duration of the strip at time i;

[0078] Until the integral value When the time equals L, the corresponding time i is the time when the strip passes through the target process position. The process parameter value corresponding to time i is obtained from the time-process data curve, that is, the process parameter value of the target process position is obtained; then the correlation information between the strip position and the process data is obtained, that is, the process history information of the strip position.

[0079] S103. Combine the process history information of each position of the strip with the actual condensation rate label to obtain the dataset required for model training.

[0080] S2. Construct a condensation rate prediction model using the dataset; the construction of a condensation rate prediction model can employ methods such as statistical analysis, machine learning, deep learning, and mechanistic models; the goal of the condensation rate prediction model is to predict the condensation rate based on given process parameters.

[0081] S3. Dynamically optimize and control the process during real-time production based on a condensation rate prediction model; specifically including:

[0082] S301. Obtain the n images of the hot-dip galvanized steel strip surface that have passed the test, and calculate the actual condensation rate corresponding to each image; the method for calculating the actual condensation rate in this step is the same as the method for calculating the actual condensation rate in step S101.

[0083] S302. Calculate the process history data of the strip corresponding to the actual condensation rate; the calculation method is the same as step S102, that is, perform spatiotemporal transformation on the time-process data in the production process corresponding to each actual condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each actual condensation rate.

[0084] S303. Based on the process history data corresponding to the actual condensation rate in step S2, the condensation rate prediction model is used to predict the n sets of predicted condensation rate data of the strip that have passed through, and the deviation compensation value of the condensation rate prediction model is obtained; the calculation method of the deviation compensation value is as follows:

[0085]

[0086] Where e is the real-time deviation, i.e., the deviation compensation value, and y i Let m(x) represent the actual condensation rate of the i-th group. i ) indicates that the process parameter x is predicted using the condensation rate prediction model m(). i The predicted condensation rate is obtained.

[0087] S304. Based on the deviation compensation values, the corrected condensation rate prediction model is obtained; the description of the corrected condensation rate prediction model is as follows:

[0088] mnew ()=m()+e

[0089] Where, m new () indicates the modified condensation rate prediction model.

[0090] S305. Based on the modified condensation rate prediction model, optimize the process parameters of the processes preceding the process where the hot-dip galvanized steel strip surface inspection camera is located, and adjust the process parameters based on the optimization results. Specifically, this includes:

[0091] S3501, Obtain the k processes preceding the process where the hot-dip galvanized steel strip surface inspection camera is located: Process1, Process2, ..., Process... k The k processes are arranged in order of their distance from the camera, from closest to farthest.

[0092] S3502, Calculate the position of each process step. i The distance between the strip and the processes it has already passed through is calculated, and the actual process parameters of the processes that have already passed through are calculated when the strip passes through.

[0093] S3503. Optimize the process of k steps by prioritizing steps closest to the camera, i.e., first optimize Process1, which is closest to the camera, and finally optimize Process2, which is furthest from the camera. k Specifically, it includes:

[0094] Obtain the process parameters of Process 1 at the current moment, and combine them with the actual process parameters experienced by the strip steel before Process 1. Use the condensation rate prediction model to predict the condensation rate of the strip steel before Process 1.

[0095] The process parameters corresponding to the process position Process1 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters corresponding to the process position Process1 are then denoted as p1.

[0096] Obtain the process parameters of Process2 at the current moment, and combine them with the actual process parameters of the strip steel before Process2 and p1. Use the condensation rate prediction model to predict the condensation rate of the strip steel before Process2.

[0097] The process parameters corresponding to the process position Process2 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters corresponding to the process position Process2 are then denoted as p2.

[0098] Repeat the optimization process until k processes are optimized, obtaining the optimization results p1, p2, ..., p for each process. k ;

[0099] The entire process for hot-dip galvanized steel strip was adjusted based on the optimization results.

[0100] S4. Update the condensation rate prediction model based on the newly generated data; specifically including:

[0101] S401. Update the dataset; this includes: calculating the strip process parameter x corresponding to the actual condensation rate R obtained in real-time production; calculating the similarity between process parameter x and each existing data point in the dataset; and selecting the minimum similarity from the process parameter x and each existing data point in the dataset, denoted as d. i Set a similarity threshold d limit If the similarity d i Greater than the similarity threshold d limit Then, the process parameter x and the corresponding actual condensation rate R are added to the dataset; if the similarity d i Less than or equal to the similarity threshold d limit If so, then process parameter x is discarded.

[0102] The similarity is expressed using Euclidean distance, but Mahalanobis distance, cosine distance, etc., can also be used to represent the similarity.

[0103] S402. Based on the updated dataset, update the condensation rate prediction model.

[0104] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described hot-dip galvanized steel strip surface quality control method.

[0105] Example 1

[0106] This embodiment utilizes a surface quality control method for hot-dip galvanized steel strip provided by the present invention to control the quality of the steel strip. In this embodiment, the surface treatment of the steel strip is galvanizing. Of course, the present invention can also be applied to surface treatment processes such as aluminum-zinc-silicon galvanizing and zinc-aluminum galvanizing, which involve a hot-dip galvanizing followed by a cooling process. Some of the production steps of the steel strip in this embodiment are as follows: Figure 2As shown, the strip steel moves into a zinc bath for galvanizing. After galvanizing, it passes through an air knife, which uses air to blow away the galvanizing liquid from the strip surface, ensuring a uniform galvanized layer thickness. The strip steel then sequentially passes through a first moving fan, a second moving fan, a first high-tower fan, a second high-tower fan, a third high-tower fan, and a fourth high-tower fan to allow the galvanized coating to condense. A camera is positioned between the first and second high-tower fans. The specific quality control methods for the galvanized strip steel produced using this process include:

[0107] S1. Establish the dataset required for model training; specifically including:

[0108] S101. Utilize a camera to acquire real-time image data of the galvanized strip steel after condensation, calculate the true condensation rate, and use the true condensation rate as a label; the method for calculating the true condensation rate includes: identifying the rectangular area where the strip steel is located, such as... Figure 1 As shown, the width direction of the rectangular region is the width direction of the strip. The strip in the rectangular region image is divided along the width direction into a left region with an area ratio of 'a', a middle region with an area ratio of 1-2a, and a right region with an area ratio of 'a'. The value of 'a' ranges from 0 to 0.5. The specific value of 'a' is determined based on actual debugging; the closer 'a' is to the air knife, the smaller its value; the farther 'a' is from the air knife, the larger its value. For example... Figure 2 The camera position is shown, and the value of 'a' is 0.1.

[0109] The percentage of pixels in the condensed portion is calculated separately for the left and right regions, and denoted as R. left and R right ;but

[0110]

[0111] Where R is the actual condensation rate. According to Figure 1 The image of the galvanized steel strip shown is used to calculate R. left R is 0.32. right If the value is 0.31, then the actual condensation rate R is 0.315, or 31.5%.

[0112] S102. Perform spatiotemporal transformation on the time-process data in the production process corresponding to each true condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each label after the transformation.

[0113] The spatiotemporal transformation includes: based on the process parameter information along the time axis corresponding to the processes before the condensation of the galvanized strip, the production state experienced by the strip position corresponding to the current condensation rate label is solved by integral calculation. The production state specifically includes the parameters corresponding to the current strip position when it passes through each cooling device, namely the frequency of the first high tower fan, the frequency of the second moving fan, and the frequency of the first moving fan; as well as the process parameters of the air knife corresponding to the current strip position when it passes through the air knife, such as the air knife pressure, the distance between the air knife and the strip, and the distance of the air knife from the ground; the zinc liquid temperature when the current strip position passes through the zinc pot; and the temperature at the zinc pot inlet pyrometer when the current strip position passes through the zinc pot.

[0114] The integration operation includes: obtaining the distance L between the camera and the strip at the target process location (e.g., for the second moving fan, L is 10 meters); and calculating the iterative integral of the strip movement speed curve.

[0115]

[0116] Among them, v i Let t represent the speed of the strip at time i. i This represents the duration of the strip at time i;

[0117] Until the integral value The integral value up to L is: Figure 3 The shaded area represents the distance. At this time, time i corresponds to the moment the strip passes the second moving fan. The process parameter value corresponding to time i is obtained from the time-process data curve, i.e., the process parameter value for the target process position; thus, the correlation information between the strip position and the process data is obtained, i.e., the process history information of the strip position. In this embodiment, the frequency f of the second moving fan is obtained from the corresponding time-second moving fan frequency curve at this moment. mob2 It is 40Hz.

[0118] Using the same method, the frequency f of the first high-tower wind turbine was obtained. tower1 The frequency f of the first mobile fan is 21Hz. mob1 The Hz frequency is 27 Hz, the temperature of the molten zinc in the zinc pot (tempzn) is 445℃, and the temperature at the pyrometer at the zinc pot inlet (temp) is... strip The temperature is 490℃. Combining these process parameters and the condensation rate R = 0.315, a set of data can be obtained. Repeating the above process yields a dataset consisting of the condensation rates corresponding to the process parameters. This dataset can be denoted as Data = {width, thk, f}. tower1 f mob2 f mob1 temp zn temp strip ,R}, where width and thk represent the width and thickness of the strip, respectively.

[0119] S103. Combine the process history information of each position of the strip with the actual condensation rate label to obtain the dataset required for model training.

[0120] S2. Construct a condensation rate prediction model using the dataset; the construction of a condensation rate prediction model can employ methods such as statistical analysis, machine learning, deep learning, and mechanistic models; the goal of the condensation rate prediction model is to predict the condensation rate based on given process parameters.

[0121] S3. Dynamically optimize and control the process during real-time production based on a condensation rate prediction model; specifically including:

[0122] S301. Obtain n sets of real-time surface images of the galvanized strip after condensation, and calculate n sets of true condensation rates of the galvanized strip; the method for calculating the true condensation rate in this step is the same as the method for calculating the true condensation rate in step S101.

[0123] S302. Calculate the process history data of the strip corresponding to the actual condensation rate; the calculation method is the same as step S102, that is, perform spatiotemporal transformation on the time-process data in the production process corresponding to each actual condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each actual condensation rate.

[0124] S303. Based on the process history data corresponding to the actual condensation rate in step S2, a condensation rate prediction model is used to predict n sets of predicted condensation rate data for the strip that has passed through. For example, the condensation rate data within the last 3 minutes, predicted every 5 seconds, then n is 36; the deviation compensation value of the condensation rate prediction model is obtained; the calculation method of the deviation compensation value is as follows:

[0125]

[0126] Where e is the real-time deviation, i.e., the deviation compensation value, and y i Let m(x) represent the actual condensation rate of the i-th group. i ) indicates that the process parameter x is predicted using the condensation rate prediction model m(). i The predicted condensation rate is obtained.

[0127] S304. Based on the deviation compensation values, the corrected condensation rate prediction model is obtained; the corrected condensation rate prediction model is as follows:

[0128] m new ()=m()+e

[0129] Where, m new () indicates the modified condensation rate pre-model.

[0130] S305. Based on the modified condensation rate prediction model, optimize the process parameters before the process where the galvanized strip surface inspection camera is located, and adjust the process parameters based on the optimization results. Specifically, this includes:

[0131] S3501, the k processes before obtaining a real-time surface image of the galvanized strip: Process1, Process2, ..., Process... k Location; such as Figure 2 Given the camera positions shown, k is 3. Process1 represents the first tower fan, Process2 represents the second mobile fan, and Process3 represents the first mobile fan.

[0132] S3502, Calculate the position of each process step. i The preceding strip steel, and the process of this operation. i The previous process corresponds to the distance between equipment, thus obtaining the process steps. i The actual process parameters of the processes that the strip steel has already undergone;

[0133] S3503, Using the camera as a reference, optimize k processes from near to far, including:

[0134] Obtain the process parameters for Process 1 at the current moment. The current frequency of the first tower fan is 20Hz. Combine these parameters with the actual process parameters experienced by the strip steel before Process 1. The process parameters of the strip steel before Process 1 are x1 = {width, thk, f}. tower1 f mob2 f mob1 temp zn temp strip}={410, 2.2, 20, 42.5, 27.3, 452, 495}; The condensation rate R1 of the strip steel before Process 1, predicted by the condensation rate prediction model, is 0.07, or 7%.

[0135] The process parameters corresponding to Process1 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. For example, if the set target condensation rate is 0.03 (3%), the allowable condensation rate range is 0.029-0.031 (2.9%-3.1%). Therefore, the condensation rate predicted by the condensation rate prediction model after optimizing the process parameters needs to be within the range of 0.029-0.031. The optimized process parameters corresponding to Process1 are denoted as p1. The frequency of the first high-tower fan after optimization is 19.7Hz.

[0136] Obtain the process parameters for Process 2 at the current moment. The frequency of the second moving fan is 42.3Hz. Combine these parameters with the actual process parameters of the strip steel before Process 2 and p1 to obtain x2 = {width, thk, f}. tower1 f mob2 f mob1 temp zn temp strip}={410, 2.2, 19.7, 42.3, 27.5, 453, 497}; The condensation rate R2 of the strip steel before Process 2, predicted by the condensation rate prediction model, is 0.01, or 1%.

[0137] The process parameters corresponding to the process position Process2 are optimized, while the process parameters of Process1 are kept at p1. The optimized process parameters are used to predict the condensation rate through the condensation rate prediction model so that the condensation rate is within the set target condensation rate range. The optimized process parameters corresponding to the process position Process2 are then recorded as p2. The frequency of the second moving fan is 42.6Hz after optimization.

[0138] The optimization process is repeated until k processes are optimized, and the optimization results p1, p2, and p3 for each process are obtained. The final optimization results are p1 = 19.7 Hz, p2 = 42.6 Hz, and p3 = 27.2 Hz.

[0139] The entire process for galvanizing steel strip was adjusted based on the optimization results. Figure 4 The graph shows a comparison between the target condensation rate and the measured condensation rate over time. It can be seen from the graph that the optimized condensation rate fluctuates slightly around the target condensation rate, indicating that the condensation rate control method of the present invention has achieved good control effect.

[0140] S4. Update the condensation rate prediction model based on the newly generated data; specifically including:

[0141] S401. Update the dataset; this includes: calculating the strip process parameter x corresponding to the actual condensation rate R obtained in real-time production; calculating the similarity between process parameter x and each existing data point in the dataset; and selecting the minimum similarity from the process parameter x and each existing data point in the dataset, denoted as d. i Set a similarity threshold d limit If the similarity d i Greater than the similarity threshold d limit Then, the process parameter x and the corresponding actual condensation rate R are added to the dataset; if the similarity d i Less than or equal to the similarity threshold d limit If so, then process parameter x is discarded.

[0142] The similarity is expressed using Euclidean distance, but Mahalanobis distance, cosine distance, etc., can also be used to represent the similarity.

[0143] Assuming the dataset contains 1000 samples, determine the distance between the process parameters and each of the 1000 samples in the dataset, finding the minimum distance d. i Is it greater than the threshold d? limit If the conditions are met, the new sample x is added to the training database; otherwise, the sample is discarded.

[0144] S402. Based on the updated dataset, update the condensation rate prediction model.

[0145] The above specific embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for controlling the surface quality of hot-dip galvanized steel strip, characterized in that, include: S1. Establish the dataset needed for model training; S2. Construct a condensation rate prediction model using the dataset; S3. Dynamically optimize and control the process in real-time production based on the condensation rate prediction model; Step S3 specifically includes: S301. Obtain the n images of the hot-dip galvanized steel strip surface that have passed the test, and calculate the actual condensation rate corresponding to each image. S302. Calculate n sets of strip steel process history data corresponding to n actual condensation rates. S303. Based on the condensation rate prediction model in step S2, predict the n sets of process history data of the strip steel that have passed through, and obtain the average deviation between the predicted n sets of condensation rates and the actual condensation rates, which is used as the deviation compensation value of the condensation rate prediction model. S304. Based on the deviation compensation values, the corrected condensation rate prediction model is obtained; S305. Based on the modified condensation rate prediction model, the process parameters before the condensation of hot-dip galvanized steel strip are optimized, and the process parameters are adjusted based on the optimization results. The method for calculating the true condensation rate in step S301 includes: identifying the rectangular region where the strip is located, and dividing the strip in the rectangular region image along the strip width direction into a left region with an area ratio of a, a middle region with an area ratio of 1-2a, and a right region with an area ratio of a. The percentage of pixels in the condensed portion is calculated separately for the left and right regions, and denoted as follows: and ;but Where R is the actual condensation rate; the range of values ​​for a is: ; The method for calculating the deviation compensation value in step S303 is as follows: in, This represents the real-time deviation, i.e., the deviation compensation value. This represents the true condensation rate of the i-th image. This indicates the use of a condensation rate prediction model. () for process parameters x i The predicted condensation rate obtained; The revised condensation rate prediction model is as follows: in, This represents the corrected condensation rate prediction model; Step S305 specifically includes: S3501, Obtain the k processes preceding the process where the hot-dip galvanized steel strip surface inspection camera is located: Process1, Process2, ..., Process... k Location; S3502, Calculate the position of each process step. i The distance between the strip and the processes it has already passed through is calculated, and the actual process parameters of the processes that have already passed through are calculated when the strip passes through. S3503. Based on the location of the camera, optimize the process of k steps using a near-to-far approach, including: Obtain the process parameters of Process 1 at the current moment, and combine them with the actual process parameters experienced by the strip steel before Process 1. Use the condensation rate prediction model to predict the condensation rate of the strip steel before Process 1. The process parameters corresponding to Process1 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters for Process1 are then denoted as follows: ; Obtain the process parameters of process 2 at the current moment, and compare them with the actual process parameters of the strip steel before process 2. The combination of these methods involves using a condensation rate prediction model to predict the condensation rate of the strip steel before Process 2. The process parameters corresponding to Process2 are optimized so that the condensation rate predicted by the condensation rate prediction model is within the set target condensation rate range. The optimized process parameters for Process2 are then denoted as follows: ; Repeat the optimization process until k processes are optimized, and obtain the optimization result for each process. , ... ; The entire process for hot-dip galvanized steel strip was adjusted based on the optimization results.

2. The method for controlling the surface quality of hot-dip galvanized steel strip according to claim 1, characterized in that, Step S1 specifically includes: S101. Obtain real-time image data of hot-dip galvanized steel strip after condensation, calculate the true condensation rate, and use the true condensation rate as a label. S102. Perform spatiotemporal transformation on the time-process data in the production process corresponding to each true condensation rate, convert the time-process data into strip position-process data, and obtain the process history information of the strip position corresponding to each label after the transformation. S103. Combine the process history information of each position of the strip with the actual condensation rate label to obtain the dataset required for model training.

3. The method for controlling the surface quality of hot-dip galvanized steel strip according to claim 2, characterized in that, The spatiotemporal transformation in step S102 includes: based on the process parameter information along the time axis corresponding to the process before the hot-dip galvanized strip solidification, solving the production state experienced by the strip position corresponding to the current solidification rate label through integral calculation, wherein the integral calculation includes: Obtain the distance L between the camera and the strip at the target process position, and calculate the iterative integral of the strip movement speed curve: in, This represents the speed at which the strip moves at time i. This represents the duration of the strip at time i; Until the integral value Up to L, the corresponding time i is the time when the strip passes through the target process position. The process parameter value corresponding to time i is obtained from the time-process data curve, that is, the process parameter value of the target process position is obtained.

4. The method for controlling the surface quality of hot-dip galvanized steel strip according to claim 1, characterized in that, The surface quality control methods for hot-dip galvanized steel strips also include: S4. Update the condensation rate prediction model based on the newly generated data; specifically including: S401, Update the dataset; S402. Based on the updated dataset, update the condensation rate prediction model.

5. The method for controlling the surface quality of hot-dip galvanized steel strip according to claim 4, characterized in that, The updated dataset includes: calculating the strip process parameter x corresponding to the actual condensation rate R obtained in real-time production; calculating the similarity between process parameter x and each existing data point in the dataset; and selecting the smallest similarity from the process parameter x and each existing data point in the dataset as denoted as . d i Set a similarity threshold d limit If similarity d i Greater than the similarity threshold d limit Then, the process parameter x and the corresponding actual condensation rate R are added to the dataset; if the similarity d i Less than or equal to the similarity threshold d limit If so, then process parameter x is discarded.

6. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the program, it implements the steps of the hot-dip galvanized strip surface quality control method according to any one of claims 1-5.

Images