Plate surface cleaning control method and system based on segmented speed regulation

Abstract

The invention discloses a plate surface cleaning control method and system based on segmented speed regulation, and the method comprises the steps: S1, carrying out the trial washing of a plate at different conveying speeds, collecting an image, and building a data set D1; s2, segmenting the data set D1 to obtain a strip-shaped region image; identifying, marking and processing dark spots of the strip-shaped area image to form data sets D2 and D3; s3, establishing a detection and identification model of the dark spots on the surface of the plate, and training and deploying by using D2; s4, establishing an index prediction model of surface dark spots after the plate is cleaned, and training and deploying by using D3; and S5, acquiring an image of a to-be-cleaned plate, identifying to obtain dark spot region information, inputting the index prediction model of the cleaned surface dark spots, obtaining dark spot index prediction data, and calculating an ideal speed. Through combination of a staged speed regulation strategy and a deep learning model, efficient, low-consumption and high-consistency control of wide plate surface cleaning is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic control, and particularly relates to a control method and a control system for segmentally and real-time adjusting the conveying speed during the surface cleaning of wide plates in a high-pressure water jet surface cleaning device for plates. Background Art

[0002] The cleaning station is the core component of the high-pressure abrasive water jet surface cleaning device. When using the high-pressure abrasive water jet surface cleaning device to clean the surface of a metal strip, the conveying equipment drives the strip through the cleaning station. The jet nozzles in the cleaning station use a mixed liquid of high-pressure water and abrasive to impact the surface of the strip, removing the oxidized impurities on the strip surface to meet the subsequent processing requirements. The metal strip is relatively narrow, and the high-pressure abrasive water jet cleaning device only contains one cleaning station. All the jet nozzles are arranged in a single jet box. By setting an open-loop constant conveying speed or a closed-loop speed control, the conveyor belt conveys the strip through the cleaning station to complete the surface cleaning of the steel strip.

[0003] Different from the surface cleaning device corresponding to the relatively narrow strip, when cleaning the surface of a relatively wide metal plate, the scale of the water jet surface cleaning device will be significantly enlarged. There are multiple cleaning stations in the device, and a set of nozzles is arranged in each cleaning station, which is only responsible for cleaning a specific strip area along the conveying direction of the steel plate. The steel plate can only complete the whole-plate cleaning after passing through all the cleaning stations in sequence (the steel plate will be conveyed to the next cleaning station only after being cleaned by the previous cleaning station). At this time, if a single constant speed method is used to convey the steel plate through all the cleaning stations in sequence, in order to ensure the consistency of the cleaning effect of all strip areas in the steel plate, the conveying speed setting will tend to be conservative, resulting in too long operation time, not only increasing unnecessary electric energy consumption, but also increasing the burden on the cleaning equipment itself, and aggravating the potential failure risks of components such as the jet nozzles and high-pressure pipelines in the cleaning device. The patent with the publication number of CN118060249A adopts photographing the strip and performing real-time speed regulation on the strip cleaning according to the image processing result, but this method is not suitable for the speed control of wide plates. Compared with the strip, the wide plate is wider and shorter in length. When collecting images through the rear-mounted camera of the conveyor belt, the cleaning may have been completed, and real-time speed regulation control cannot be achieved. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to propose a phased speed control method to control the speed of the plate passing through each cleaning station by predicting the surface cleaning quality of the cleaned plate, aiming at the problems faced by the constant conveying speed control method during the cleaning of wide plates, such as energy consumption, potential equipment failures, and the inapplicability of the real-time speed regulation method for strip cleaning, so as to achieve the goal of balancing the cleaning quality and energy consumption during the surface cleaning of wide plates.

[0005] To solve the above technical problems, the technical solution proposed by the present invention is as follows: A method for controlling the surface cleaning of plates based on segmented speed regulation, comprising the following steps: S1. Use a cleaning device to wash several plates at different conveying speeds, collect the surface images of the plates before and after the trial wash, and establish an unstructured data set D1; S2. Divide the surface images of the plates before and after the trial wash in the unstructured data set D1 into several parts along the width direction of the plates according to the number of cleaning positions in the cleaning device, to obtain several strip-shaped area images of the plates before cleaning and strip-shaped area images of the plates after cleaning; identify, label and process the dark spots in the several strip-shaped area images of the plates before cleaning and strip-shaped area images of the plates after cleaning, to form an image structured data set D2 and a numerical structured data set D3; S3. Establish a detection and recognition model for dark spots on the plate surface, and use the image structured data set D2 to train and deploy the detection and recognition model for dark spots on the plate surface; S4. Establish an index prediction model for dark spots on the surface of the plate after cleaning, and use the numerical structured data set D3 to train and deploy the index prediction model for dark spots on the surface of the plate after cleaning; S5. Collect the pre-cleaning image of the plate to be cleaned at a set conveying speed, input the pre-cleaning image into the detection and recognition model for dark spots on the plate surface, identify the dark spot area information in the pre-cleaning image, input the dark spot area information into the index prediction model for dark spots on the surface of the plate after cleaning, obtain the predicted index data of the dark spots after cleaning, calculate the ideal speed of the plate passing through each cleaning position according to the predicted index data of the dark spots after cleaning, and control the plate to be cleaned for cleaning at the ideal speed.

[0006] For the cleaning method of wide plates with multiple cleaning positions of the present invention, the cleaning effect images of the plates are segmented and recognized according to the distribution mode of the cleaning positions in the cleaning device, and the corresponding dark spot detection and recognition model and dark spot index prediction model are constructed. The above models are trained through the effect images and data of several trial washes. Through the collected images of the plates to be cleaned, the ideal speed of each cleaning position can be directly calculated and output according to the detection, recognition and prediction results, so as to perform segmented control on the conveying speed of the plates. On the one hand, it can accurately ensure that each cleaning position can exert the best effect and obtain the desired cleaning effect. On the other hand, it also avoids the conservative and unified low-speed setting, reducing the energy consumption of the equipment.

[0007] As a further optimization of the above technical solution, in S1, when washing several plates at different conveying speeds, set the conveying speed range of the plates and the speed increase step , starting from Start cleaning n sheets of plates, then increase the speed , and then clean n sheets of plates until the cleaning of n sheets of plates at the speed is completed.

[0008] As a further optimization of the above technical solution, in S2, the number of segmented surface images of the plates before and after trial cleaning corresponds to the number of cleaning positions in the cleaning device.

[0009] As a further optimization of the above technical solution, in S2, the image structured dataset D2 includes the strip region images of the plates before cleaning with dark spots marked and the strip region images after cleaning with dark spots marked; the numerical structured dataset D3 includes the average dark spot area, the maximum dark spot area of the strip region images of the plates before cleaning, the plate conveying speed, and the average dark spot area, the maximum dark spot area of the strip region images of the plates after cleaning.

[0010] As a further optimization of the above technical solution, in S3, a detection and recognition model for dark spots on the plate surface is established based on the single-stage object detection network YOLOv8n. The image structured dataset D2 is divided into a training set, a validation set, and a test set. The training set is used to train YOLOv8n. During the training process, the validation set is used to evaluate the model's metrics, and the hyperparameters are adjusted as needed (the hyperparameters such as the input resolution (imgsz), data augmentation (hsv), learning rate (lr0), loss weight (cls), etc. can be adjusted according to the performance of the model evaluated by the validation set, such as loss curves, mAP, etc.). When the performance of the detection and recognition model for dark spots on the plate surface on the validation set meets the preset standard, an independent test set is used for a one-time test. When the test results meet the requirements, the training of the model is ended and a deployable detection and recognition model for dark spots on the plate surface is generated.

[0011] As a further optimization of the above technical solution, during the training of YOLOv8n, a composite loss function is used for model training. The expression of the composite loss function is: ; In the formula , and are weight coefficients, adjusted by experience, is the confidence loss, is the oxide dark spot recognition and localization loss, is the oxide dark spot recognition and annotation box envelope area loss; among them, the oxide dark spot recognition annotation box refers to the dark spot area reference box generated manually or by tools during the training process to guide the model training and learning; the oxide dark spot recognition prediction box refers to the dark spot recognition box output by YOLOv8n, including the predicted position and confidence; Among them, ; Wherein, is the intersection over union (IoU) of the oxide dark spot recognition prediction box and the oxide dark spot true annotation box; the oxide dark spot true annotation box refers to the standard box of the dark spot in the dataset, which is used for IoU and area loss calculation; ; Wherein, is the true value, is the predicted probability, N is the number of samples; ; Wherein, represents the area of the oxide dark spot recognition prediction box, represents the area of the oxide dark spot annotation box. The model is trained using a composite loss function of localization error IoU, envelope area error, and confidence error, which can improve the localization accuracy and enhance the robustness.

[0012] As a further optimization of the above technical solution, when using the validation set to evaluate the model metrics, the metrics include the accuracy, recall, and F1-score of YOLOv8n; the hyperparameters include the initial learning rate, learning rate decay factor, and the number of training samples per batch.

[0013] As a further optimization of the above technical solution, in S4, an index prediction model for the surface dark spots after sheet cleaning is established based on the MLP multi-layer network model. The numerical structured dataset D3 is divided into a training set, a validation set, and a test set. The training set is used to train the MLP multi-layer network model. After the metric evaluation of the validation set and the generalization ability test of the test set, the training of the model is ended and a deployable index prediction model for the surface dark spots after cleaning is generated.

[0014] As a further optimization of the above technical solution, when using the training set to train the MLP multi-layer network model, the feature vector of the strip area before sheet cleaning is used as the input value, the feature vector of the strip area after sheet cleaning is used as the output value, multi-layer fully connected neurons are used as the hidden layer, the ReLU function is used as the activation function, and the mean squared error is used as the loss function , the gradient is calculated through the backpropagation algorithm, and the Adam optimizer is used for parameter update. Using the mean squared error (MSE) as the loss function can minimize the error between the predicted value and the true value, and using the Adam optimizer for parameter update can ensure the rapid convergence of the model.

[0015] As a further optimization of the above technical solution, the feature vector ; Wherein, is the average area of the dark spots in the strip area before sheet cleaning, is the maximum value of the dark spot area in the strip area before sheet cleaning, is the sheet conveying speed; the eigenvector ; where, is the average value of the dark spot area in the strip area after sheet cleaning, is the maximum value of the dark spot area in the strip area after sheet cleaning.

[0016] As a further preference of the above technical solution, the loss function ; where, , are the predicted value of the average value of the dark spot area and the predicted value of the maximum value of the dark spot area in the strip area after sheet cleaning.

[0017] As a further preference of the above technical solution, in S5, after collecting the pre-cleaning image of the sheet to be cleaned at the set conveying speed, the pre-cleaning image is segmented into several rectangular strips according to the method of S2, and then each of the rectangular strips is input into the detection and recognition model of the dark spots on the sheet surface to identify the dark spot area information in the pre-cleaning image, and the dark spot area information includes the average value of the dark spot area, the maximum value of the dark spot area and the conveying speed of the sheet.

[0018] As a further preference of the above technical solution, the dark spot area information is input into the index prediction model of the dark spots on the post-cleaning surface to obtain the predicted value of the average value of the dark spot area and the predicted value of the maximum value of the dark spot area, and the ideal speed of the sheet passing through each cleaning position is calculated by combining the conveying speed of the sheet and the desired cleaning effect.

[0019] As a further preference of the above technical solution, the ideal speed is calculated as follows: ; where, is the conveying speed of the sheet, is the predicted value of the average value of the dark spot area, is the predicted value of the maximum value of the dark spot area, is the expected value of the average value of the dark spot area, is the expected value of the maximum value of the dark spot area, which is determined by operation experience and actual requirements, and are weight coefficients, which are adjusted by experience (by analyzing the image data, statistically calculating the change ranges of the predicted value of the average value of the dark spot area and the predicted value of the maximum value of the dark spot area, and combining the speed adjustment formula, first determining that the adjusted speed is within the reasonable process range, and then respectively considering the contribution rates of the average value of the dark spot area and the maximum value of the dark spot area at the same time).

[0020] Based on the same inventive concept, the present invention also provides a sheet surface cleaning control system based on segmented speed regulation, which includes a vision subsystem, a sheet surface dark spot detection and recognition subsystem, a sheet surface dark spot index prediction subsystem, a speed automatic control subsystem, and a computing and processing device; The vision subsystem includes a front camera assembly arranged in front of the first cleaning position of the cleaning device and a rear camera assembly arranged behind the last cleaning position of the cleaning device, and is used to acquire the surface image of the sheet; The sheet surface dark spot detection and recognition subsystem is used to generate and deploy a detection and recognition model for dark spots on the sheet surface. The model is used to identify the surface image of the sheet acquired by the vision subsystem and output the dark spot area information in the surface image, including the average dark spot area and the maximum dark spot area; The sheet surface dark spot index prediction subsystem is used to generate and deploy an index prediction model for dark spots on the sheet surface after cleaning. The model is used to process the acquired dark spot area information. The inputs are the average dark spot area, the maximum dark spot area, and the sheet conveying speed, and the outputs are the predicted value of the average dark spot area of the sheet after cleaning and the predicted value of the maximum dark spot area of the sheet after cleaning; The speed automatic control subsystem deploys a detection and recognition model for dark spots on the sheet surface and an index prediction model for dark spots on the sheet surface after cleaning, and is used to calculate the ideal speed of the sheet passing through each cleaning position and control the conveying speed of the sheet in the cleaning device; The computing and processing device is composed of a central processor, a memory, a graphics card, and a storage, and is used to implement the storage and processing of the data set, the training and deployment of the sheet surface dark spot detection and recognition subsystem, the training and deployment of the sheet surface dark spot index prediction subsystem, and the related computing and processing functions of the speed automatic control subsystem.

[0021] As a further preference of the above technical solution, it further includes a sheet in-place detection device. The sheet in-place detection device includes a plurality of sensors arranged at the entrance of each cleaning position of the cleaning device, and is used to judge the cleaning position where the sheet is located and send a signal to the speed automatic control subsystem to assist the speed automatic control subsystem in adjusting and controlling the conveying speed of the sheet.

[0022] Compared with the prior art, the advantages of the present invention are as follows: By combining the segmented speed regulation strategy with the deep learning model, the present invention realizes the efficient, low-consumption, and high-consistency control of the wide sheet surface cleaning. By using the dark spot index and conveying speed of the sheet surface before cleaning to predict the dark spot index of the sheet surface after passing through the current cleaning position, as the basis for calculation and control decisions, the conveying speed of each position is dynamically adjusted, avoiding the energy consumption waste caused by the traditional constant low-speed mode, significantly reducing the equipment operation energy consumption and reducing unnecessary power consumption on the premise of ensuring the cleaning quality compliance. Description of the Drawings

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0024] Figure 1 It is a flowchart of the control method for cleaning the surface of a sheet based on segmented speed regulation in Embodiment 1; Figure 2 It is a schematic structural diagram of the control system for cleaning the surface of a sheet based on segmented speed regulation in Embodiment 2; Figure 3 It is a schematic diagram of the setting of the control system for cleaning the surface of a sheet based on segmented speed regulation in Embodiment 2 on the cleaning device; Figure 4 It is a schematic diagram of the segmentation of the steel plate image in the control method for cleaning the surface of a sheet based on segmented speed regulation in Embodiment 1. Detailed implementation manners

[0025] To facilitate the understanding of the present invention, the following will describe the present invention more comprehensively and in detail in conjunction with the accompanying drawings of the specification and preferred embodiments. However, the protection scope of the present invention is not limited to the following specific embodiments.

[0026] Unless otherwise defined, all professional terms used hereinafter have the same meaning as commonly understood by those skilled in the art. The professional terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the protection scope of the present invention.

[0027] Unless otherwise specifically stated, various instruments and equipment used in the present invention can be obtained through the market or can be obtained by existing methods.

[0028] Embodiment 1: The control method for cleaning the surface of a sheet based on segmented speed regulation in this embodiment is used to clean a steel plate. As Figure 1 shown, it includes the following steps: S1. Conduct a trial wash of the steel plate at different conveying speeds, collect the surface images of the steel plate before and after cleaning, and establish an unstructured data set D1. The specific process is as follows: According to the operation experience of the cleaning device, set the range of the steel plate conveying speed and the speed increase step , start cleaning the wide steel plate from the lower limit value of the speed . A total of n steel plates are cleaned, and then the speed is increased by , and another n steel plates are cleaned until the cleaning is completed at the speed Cleaning of n steel plates. During the trial cleaning process of each steel plate, the speed passing through each cleaning station remains unchanged. During this process, a high-resolution imaging device is used to photograph the steel plate before and after cleaning, and an unstructured dataset D1 of steel plate surface cleaning is established. Each data record in this dataset consists of three elements: the image of the steel plate before cleaning, the image of the steel plate after cleaning, and the conveying speed corresponding to the steel plate. Specifically in this embodiment, the range of the steel plate conveying speed determined by experience is 6 m / min to 18 m / min, and the set speed increment step is 0.2 m / min. Therefore, the speed is divided into 60 gears in total. Starting from a speed of 6 m / min, 10 steel plates are cleaned. Similarly, for each increase in the speed gear, 10 steel plates are cleaned. During the cleaning process of each steel plate, the speed gear passing through each cleaning station remains unchanged, and the imaging device is used to photograph the surface of the steel plate before and after cleaning. A total of 1200 steel plate images are obtained to form dataset D1. Each record in this dataset includes the image of the steel plate before cleaning, the image of the steel plate after cleaning, and the corresponding conveying speed; S2. Divide the aforementioned unstructured dataset D1 into structured datasets corresponding to the steel plate surface oxide dark spot detection and recognition model and the steel plate surface oxide dark spot index prediction model after cleaning. The specific process is as follows: The images of the steel plate before cleaning and the images of the steel plate after cleaning in each steel plate in dataset D1 are divided into equally spaced strip-shaped regions along the width direction of the steel plate (which is also the conveying direction), becoming several steel plate strip-shaped region images. The number of divisions is equal to the number of cleaning stations in the surface cleaning device. Mark the positions of the oxide dark spots in the steel plate strip-shaped region images, and calculate the mean and maximum values of the dark spot area in each steel plate strip-shaped region image to form structured datasets D2 and D3. Among them, D2 is an image dataset, consisting of the steel plate strip-shaped region images before cleaning and the steel plate strip-shaped region images after cleaning with the oxide dark spots marked. D3 is a numerical dataset, consisting of five data: the mean value of the dark spot area before cleaning of the steel plate strip-shaped region, the maximum value of the dark spot area, the steel plate conveying speed, the mean value of the dark spot area of the steel plate strip-shaped region after cleaning, and the maximum value of the dark spot area of the steel plate strip-shaped region after cleaning. Specifically in this embodiment, according to Figure 4 The shown segmentation method equally divides the unstructured dataset D1 along the width direction of the steel plate into 4 steel plate strip-shaped region images (corresponding to the number and position of the cleaning stations in the cleaning device), uses the labeling tool LabelImg to label the positions of the oxide dark spots in each steel plate strip-shaped region image, and calculates the average value and maximum value of the dark spot area in each steel plate strip-shaped region image to form structured datasets D2 and D3. The D2 and D3 datasets each contain 4800 records; S3. Establish a detection and recognition model for dark spots on the surface of the plate, and use the structured data set D2 to train and deploy the detection and recognition model for dark spots on the surface of the plate. The specific process is: based on the single-stage target detection network YOLOv8n, establish a detection and recognition model for dark spots on the surface of the plate. In this embodiment, D2 is divided into a training set, a validation set, and a test set in a ratio of 7:2:1. YOLOv8n is trained using the training set, and the model is trained using a composite loss function of positioning error IoU, envelope area error, and confidence error. To optimize network parameters, the Adam optimizer is used to reduce losses. During the training process, the model is evaluated using the validation set, and the hyperparameters are adjusted as needed. When the performance of the model on the validation set meets the preset standards, an independent test set is used for a one-time test to evaluate the generalization ability of the model. If the test results meet the requirements, the model training is terminated and a deployable detection and recognition model is generated. In the Python environment, using the functions provided by the ultralytics library, you can complete the processes of YOLOv8n model loading, configuring training parameters, model training, and verifying model performance. If the test reaches the predetermined indicators, the detection and recognition model for dark spots on the surface of the plate established using YOLOv8n will be exported as a deployable model in ONNX format.

[0029] In this step, the composite loss function expression is: The composite loss function expression is: ; In the formula , and is the weight coefficient, is the confidence loss, It is the oxide dark spot identification and positioning loss, It is the area loss of the oxide dark spot identification annotation box envelope; in, ; In the formula, It is the intersection-over-union ratio of the oxide dark spot recognition prediction box and the oxide dark spot real annotation box; ; In the formula, is the true value, is the predicted probability, N is the sample size; ; In the formula, represents the area of ​​the oxide dark spot recognition prediction box, Represents the area of ​​the oxide dark spot labeling box.

[0030] In this step, when evaluating the model using the validation set, the metrics include the accuracy, recall, and F1 score of YOLOv8n. The hyperparameters such as the initial learning rate, learning rate decay factor, and the number of training samples per batch are optimized and adjusted according to the requirements of the metrics.

[0031] In this step, if the results on the test set do not meet the requirements, the speed increase step size in S1 is further reduced. Or increase the cleaning quantity n at each speed. Meanwhile, methods such as image enhancement technology are used to further expand the quantity of dataset D2. After that, this step is performed again.

[0032] S4. Establish an index prediction model for the dark spots on the surface of the sheet after cleaning. Use the structured dataset D3 to train and deploy the index prediction model for the dark spots on the surface of the sheet after cleaning. The specific process is as follows: Construct an MLP multi-layer network model to predict the mean value and maximum value of the oxide dark spot area on the steel sheet after cleaning. In this embodiment, D3 is divided into a training set, a validation set, and a test set according to the ratio of 7:2:1. The training set is used to train the MLP. The input of this model is the feature vector of the steel sheet bar area before cleaning. The output is the feature vector of the steel sheet bar area after cleaning. The hidden layer is a multi-layer fully connected neuron, and the ReLU function is used as the activation function. The mean squared error (MSE) is used as the loss function. Minimize the error between the predicted value and the true value. Calculate the gradient through the backpropagation algorithm, update the weights and biases, and use the Adam optimizer for parameter update to ensure the rapid convergence of the model. After the metric evaluation of the validation set and the generalization ability test on the test set, the training of the model is ended and a deployable prediction model is generated. The definition of the MLP network, the definition of the loss function and the optimizer, the execution of training and model evaluation can be completed through the functions provided by the PyTorch framework. If the test reaches the predetermined metrics, the index prediction model for the dark spots on the surface of the sheet after cleaning established using the MLP is exported as a deployable model in ONNX format.

[0033] In this step, the feature vector ; where is the mean value of the dark spot area in the bar area of the sheet before cleaning, is the maximum value of the dark spot area in the bar area of the sheet before cleaning, is the sheet conveying speed; the feature vector ; where is the mean value of the dark spot area in the bar area of the sheet after cleaning, is the maximum value of the dark spot area in the bar area of the sheet after cleaning.

[0034] In this step, the loss function ; where , are the predicted values of the average area and the maximum area of the dark spots in the strip area after the sheet is cleaned.

[0035] S5. In the implementation stage of segmented speed control, the pre-cleaning image of the steel sheet to be cleaned is collected at the set conveying speed, and the pre-cleaning image is input into the detection and recognition model of the dark spots on the sheet surface. The information of the dark spot area in the pre-cleaning image is recognized, and the dark spot area information is input into the index prediction model of the dark spots on the post-cleaning surface to obtain the predicted data of the dark spot index after cleaning. According to the predicted data of the dark spot index after cleaning, the ideal speed of the sheet passing through each cleaning position is calculated, and the current conveying speed is adjusted to the ideal speed to pass through the current cleaning position. Specifically: at the preset steel sheet conveying speed the steel sheet is conveyed to the front of the first cleaning position, and the pre-cleaning image of the steel sheet is obtained. The image is equally divided into several rectangular strips at equal intervals along the conveying direction of the steel sheet (the splitting method in this embodiment is the same as Figure 4 shown, split into 4 rectangular strips), and each rectangular strip image is input into the detection and recognition model of the dark spots on the sheet surface deployed in S3 to detect the oxide dark spot area in the strip image, and at the same time calculate the average area of the oxide dark spots and the maximum area of the dark spots . Combining the steel sheet conveying speed , the input feature variables of the index prediction model of the dark spots on the post-cleaning surface deployed in S4 are constructed . Through the output of the model, the predicted value of the average area of the dark spots and the predicted value of the maximum area of the dark spots in each rectangular strip after the steel sheet is cleaned are obtained . According to the predicted values and as the basis for control decision-making, the actual speed required for the steel sheet to pass through each cleaning position is obtained through the following formula:

[0036] In the formula, is the preset conveying speed of the sheet, is the predicted value of the average area of the dark spots, is the predicted value of the maximum area of the dark spots, is the expected value of the average area of the dark spots, is the expected value of the maximum area of the dark spots, which is determined by operating experience and actual requirements, and are the weight coefficients, which are adjusted by experience.

[0037] In this embodiment, if the preset conveying speed of a certain sheet , before reaching the 2nd cleaning position, according to the predicted value , the predicted value , expected value , expected value , , , then the ideal speed of passing through the No. 2 cleaning position , indicating that the speed needs to be reduced to improve the cleaning quality. After analyzing the images collected by the rear camera module, after the speed is reduced and the steel plate passes through the No. 2 cleaning position, the actual value of the average area of the dark spots after cleaning , the actual value of the maximum area of the dark spots after cleaning , both have achieved approximation to the expected value, improving the cleaning quality and enhancing the cleaning consistency; For another example, the conveying speed of a certain steel plate , before reaching the No. 2 cleaning position, the predicted value , the predicted value , expected value , expected value , , , then the ideal speed , indicating that the speed can be increased to reduce energy consumption. After analyzing the images collected by the rear camera module, after the speed is increased and the steel plate passes through the No. 2 cleaning position, the actual value of the average area of the dark spots after cleaning , the actual value of the maximum area of the dark spots after cleaning , while ensuring the cleaning quality, reducing the cleaning time in the No. 2 position and lowering the energy consumption of the equipment.

[0038] Embodiment 2: Such as Figure 2 and Figure 3 shown ( Figure 3 in, 1 is the front camera module, 2 is the rear camera module at the cleaning outlet end, 3 is the sheet material in-place detection device, 4 is the calculation and processing device, 5 is the conveying roller, 6 is the steel plate before cleaning, 7 is the steel plate after cleaning, 8 is the cleaning position in the cleaning device), the sheet material surface cleaning control system based on segmented speed regulation of this embodiment includes a vision subsystem, a sheet material surface dark spot detection and recognition subsystem, a sheet material surface dark spot index prediction subsystem, a speed automatic control subsystem, a sheet material in-place detection device and a calculation and processing device; Among them, the vision subsystem includes a front camera module arranged in front of the first cleaning position of the cleaning device and a rear camera module arranged behind the last cleaning position of the cleaning device, which is used to obtain the surface image of the sheet material, and the obtained image is used for the sheet material surface dark spot detection and recognition subsystem and the sheet material surface dark spot index prediction subsystem.

[0039] The dark spot detection and recognition subsystem for the sheet metal surface is used to generate and deploy a detection and recognition model for dark spots on the sheet metal surface. The model is used to recognize the surface image of the sheet metal obtained by the vision subsystem and output the dark spot area information in the surface image, including the average dark spot area and the maximum dark spot area. During the implementation stage of segmented speed control, this subsystem is in an offline state, but still accepts the surface images of the steel plates taken by the vision subsystem to expand the dataset for subsequent optimization of the network.

[0040] The dark spot index prediction subsystem for the sheet metal surface is used to generate and deploy an index prediction model for dark spots on the sheet metal surface after cleaning. The model is used to process the obtained dark spot area information. The inputs are the average dark spot area, the maximum dark spot area, and the sheet metal conveying speed, and the outputs are the predicted values of the average dark spot area and the predicted values of the maximum dark spot area of the sheet metal after cleaning. During the implementation stage of segmented speed control, this subsystem is in an offline state, but still accepts the input of the feature data from the dark spot detection and recognition subsystem for the sheet metal surface to expand the dataset for subsequent optimization of the network.

[0041] Sheet metal in-place detection device: It includes several sensors set at the entrance of each cleaning position of the cleaning device. When the sensors detect that the steel plate to be cleaned being conveyed arrives before the entrance of the cleaning position, corresponding in-place signals are sent. The speed automatic control subsystem can know which cleaning position the steel plate has been conveyed to according to the corresponding signals.

[0042] Speed automatic control subsystem: It deploys a detection and recognition model for dark spots on the sheet metal surface and an index prediction model for dark spots on the sheet metal surface after cleaning in ONNX format. Through the predicted values of the average dark spot area and the predicted values of the maximum dark spot area at the current conveying speed output by the dark spot index prediction model for the sheet metal surface, the ideal speed for the steel plate to pass through each cleaning position is calculated, and then the conveying speed of the steel plate in the cleaning device is controlled according to the signals from the sheet metal in-place detection device and the ideal speed.

[0043] Computing and processing device: It consists of a central processing unit CPU, a memory SDRAM, a graphics card GPU, and a storage. The storage and processing of the dataset, the training and deployment of the dark spot detection and recognition subsystem for the sheet metal surface, the training and deployment of the dark spot index prediction subsystem for the sheet metal surface, and the relevant computing and processing of the speed automatic control subsystem are all implemented in the computing and processing device.

Claims

1. A control method for cleaning the surface of a plate based on segmented speed regulation, characterized in that, It includes the following steps: S1. Use a cleaning device to wash several plates at different conveying speeds, collect the surface images of the plates before and after the trial wash, and establish an unstructured data set D1; S2. Divide the surface images of the plates before and after the trial wash in the unstructured data set D1 into several parts along the width direction of the plates according to the number of cleaning positions in the cleaning device, to obtain several strip region images of the plates before cleaning and strip region images of the plates after cleaning; Identify, label and process the dark spots in the several strip region images of the plates before cleaning and strip region images of the plates after cleaning, to form an image structured data set D2 and a numerical structured data set D3; S3. Establish a detection and recognition model for dark spots on the plate surface, and use the image structured data set D2 to train and deploy the detection and recognition model for dark spots on the plate surface; S4. Establish an index prediction model for dark spots on the plate surface after cleaning, and use the numerical structured data set D3 to train and deploy the index prediction model for dark spots on the plate surface after cleaning; S5. Collect the pre-cleaning image of the plate to be cleaned at a set conveying speed, input the pre-cleaning image into the detection and recognition model for dark spots on the plate surface, obtain the dark spot area information in the pre-cleaning image, input the dark spot area information into the index prediction model for dark spots on the plate surface after cleaning, obtain the predicted dark spot index data after cleaning, calculate the ideal speed of the plate passing through each cleaning position according to the predicted dark spot index data after cleaning, and control the plate to be cleaned for cleaning at the ideal speed.

2. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 1, wherein In S1, when trial-washing a number of plates at different conveying speeds, set the conveying speed range of the plates and the speed increment step , start cleaning n plates from , then increase the speed by , and clean n plates again until the cleaning of n plates at the speed is completed.

3. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 1, characterized in that, In S2, the number of divisions of the surface images of the plates before and after the trial wash corresponds to the number of cleaning positions in the cleaning device.

4. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to any one of claims 1-3, characterized in that, In S2, the image structured data set D2 includes the strip region images of the plates before cleaning with dark spots marked and the strip region images of the plates after cleaning with dark spots marked; The numerical structured data set D3 includes the average dark spot area, the maximum dark spot area of the strip region images of the plates before cleaning, the plate conveying speed, and the average dark spot area, the maximum dark spot area of the strip region images of the plates after cleaning.

5. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 4, characterized in that, In S3, establish a detection and recognition model for dark spots on the plate surface based on the single-stage object detection network YOLOv8n, divide the image structured data set D2 into a training set, a validation set and a test set, use the training set to train YOLOv8n, during the training process, use the validation set to evaluate the model's indicators, and adjust the hyperparameters as needed. When the performance of the detection and recognition model for dark spots on the plate surface on the validation set meets the preset standard, use an independent test set for a one-time test. When the test result meets the requirements, end the model training and generate a deployable detection and recognition model for dark spots on the plate surface.

6. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 5, characterized in that, During the training of YOLOv8n, a composite loss function is used for model training, and the expression of the composite loss function is: ； where , and are weight coefficients, is the confidence loss, is the oxide dark spot recognition and localization loss, is the oxide dark spot recognition annotation box envelope area loss; Where, ; wherein, is the intersection over union of the oxide dark spot recognition prediction box and the oxide dark spot true annotation box; ; where, is the true value, is the predicted probability, N is the number of samples; ; wherein, represents the area of the oxide dark spot recognition prediction box, represents the area of the oxide dark spot annotation box; Among them, the oxide dark spot recognition annotation box refers to the reference box of the dark spot area generated manually or by tools during the training process, guiding the model training and learning; the oxide dark spot recognition prediction box refers to the dark spot recognition box output by YOLOv8n, including the predicted position and confidence; the oxide dark spot true annotation box is the standard box of the dark spot in the dataset, used for IoU and area loss calculation.

7. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 5, wherein, When using the validation set to evaluate the model's metrics, the metrics include the accuracy, recall, and F1 score of YOLOv8n; the hyperparameters include the initial learning rate, learning rate decay factor, and the number of training samples per batch.

8. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 4, wherein, In S4, an index prediction model for the surface dark spots of the sheet after cleaning is established based on the MLP multi-layer network model. The numerical structured dataset D3 is divided into a training set, a validation set, and a test set. The training set is used to train the MLP multi-layer network model. After the metric evaluation of the validation set and the generalization ability test of the test set, the training of the model is ended and a deployable index prediction model for the surface dark spots after cleaning is generated.

9. The control method for cleaning the surface of a sheet based on segmented speed regulation according to claim 8, characterized in that When training the MLP multi-layer network model using the training set, the feature vector of the strip area before sheet cleaning is used as the input value, and the feature vector of the strip area after sheet cleaning is used as the output value. A multi-layer fully connected neuron is used as the hidden layer, the ReLU function is used as the activation function, and the mean squared error is used as the loss function , and the gradient is calculated through the backpropagation algorithm, and the Adam optimizer is used for parameter update.

10. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 9, wherein, The feature vector ; where is the average area of the dark spots in the strip area before the sheet metal is cleaned, is the maximum area of the dark spots in the strip area before the sheet metal is cleaned, is the conveying speed of the sheet metal; the feature vector ; where is the average area of the dark spots in the strip area after the sheet metal is cleaned, is the maximum area of the dark spots in the strip area after the sheet metal is cleaned.

11. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 9, characterized in that, The loss function has the expression of ; in the formula, , are the predicted values of the average area and the maximum area of the dark spots in the strip area after the sheet cleaning.

12. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 4, characterized in that, In S5, after collecting the pre-cleaning image of the sheet to be cleaned at the set conveying speed, the pre-cleaning image is segmented into several rectangular strips according to the method of S2, and then each of the rectangular strips is input into the detection and recognition model of the dark spots on the sheet surface to identify the dark spot area information in the pre-cleaning image. The dark spot area information includes the average dark spot area, the maximum dark spot area, and the conveying speed of the sheet.

13. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 4, wherein The dark spot area information is input into the index prediction model of the surface dark spots after cleaning to obtain the predicted value of the average dark spot area and the predicted value of the maximum dark spot area, and the ideal speed of the sheet passing through each cleaning position is calculated by combining the conveying speed of the sheet and the desired cleaning effect.

14. The method for controlling the surface cleaning of a sheet based on segmented speed regulation according to claim 13, characterized in that, The ideal speed is calculated as follows: ; where, is the conveying speed of the sheet, is the predicted value of the average area of the dark spots, is the predicted value of the maximum area of the dark spots, is the expected value of the average area of the dark spots, is the expected value of the maximum area of the dark spots, and are the weight coefficients.

15. A sheet surface cleaning control system based on segmented speed regulation, characterized in that, A sheet surface cleaning control method based on segmented speed regulation for implementing any one of claims 1-14, including a vision subsystem, a sheet surface dark spot detection and recognition subsystem, a sheet surface dark spot index prediction subsystem, a speed automatic control subsystem, and a computing and processing device; The vision subsystem includes a front camera assembly arranged in front of the first cleaning position of the cleaning device and a rear camera assembly arranged behind the last cleaning position of the cleaning device, for acquiring the surface image of the sheet; The sheet surface dark spot detection and recognition subsystem is used to generate and deploy a detection and recognition model for the dark spots on the sheet surface. The detection and recognition model for the dark spots on the sheet surface is used to identify the surface image of the sheet acquired by the vision subsystem and output the dark spot area information in the surface image, including the average dark spot area and the maximum dark spot area; The sheet surface dark spot index prediction subsystem is used to generate and deploy an index prediction model for the surface dark spots after cleaning the sheet. The index prediction model for the surface dark spots after cleaning the sheet is used to process the acquired dark spot area information and output the predicted value of the average dark spot area after cleaning the sheet and the predicted value of the maximum dark spot area after cleaning the sheet; The speed automatic control subsystem deploys a detection and recognition model for the dark spots on the sheet surface and an index prediction model for the surface dark spots after cleaning the sheet, and is used to calculate the ideal speed of the sheet passing through each cleaning position and control the conveying speed of the sheet in the cleaning device; The computing and processing device consists of a central processing unit, a memory, a graphics card, and a storage, and is used to implement the storage and processing of data sets, the training and deployment of the dark spot detection and recognition subsystem on the sheet surface, the training and deployment of the dark spot index prediction subsystem on the sheet surface, and the related computing and processing functions of the automatic speed control subsystem.

16. The surface cleaning control system for sheet materials based on segmented speed regulation according to claim 15, characterized in that, It further includes a sheet in-place detection device, and the sheet in-place detection device includes a number of sensors arranged at the entrance of each cleaning position of the cleaning device, which is used to judge the cleaning position where the sheet is located and send a signal to the automatic speed control subsystem to assist the automatic speed control subsystem in adjusting and controlling the conveying speed of the sheet.

Images