Strip steel pickling prediction control system based on neural network model
Through the neural network model of strip pickling prediction control system, the concentration and speed of the pickling solution are adjusted in real time, which solves the problem of factor matching in the pickling process, improves the pickling effect and hydrochloric acid consumption efficiency, and reduces the discharge of acidic wastewater.
Patent Information
- Application Number
- CN202411200173.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-06
AI Technical Summary
When treating the iron oxide sheet on the strip steel surface, it is difficult to accurately match the strip steel material factors and acid liquid factors, resulting in large differences in the pickling effect, which may cause over-washing or incomplete washing, and excessive hydrochloric acid consumption, which will affect the Cost and environment.
Adopting a strip steel pickling prediction control system based on neural network model, by detecting the surface quality and acid properties of the strip steel, predicting and adjusting the concentration and pickling speed of the pickling liquid in real time, and using deep learning to optimize the pickling process parameters to realize pickling Minimum consumption and maximum efficiency of liquid.
The surface quality of the strip steel pickling is improved, the consumption of pickling liquid is reduced, the discharge of acidic wastewater is reduced, the concentration and speed of the pickling liquid is achieved, and the process efficiency and environmental protection performance are improved.
Smart Images

Figure CN119087931A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of steel rolling, and in particular relates to a strip pickling prediction control system based on a neural network model. Background Art
[0002] At present, the cold rolling section of steel enterprises generally adopts shallow turbulent pickling process to pickle hot-rolled strip to remove the iron oxide scale in the strip. There are many factors that affect the quality of the pickling process, which are mainly divided into material factors and process factors according to the source. Material factors mainly include: strip thickness, strip temperature, strip surface iron oxide scale thickness and strip surface iron oxide scale composition; process factors mainly include: acid type, acid concentration, acid temperature, Fe ion concentration in the acid and other turbulent factors that form the acid.
[0003] The pickling process of the pickling unit changes with the material factors of the incoming materials and the acid factors in the process, and the pickling effect varies greatly, resulting in a large difference in the removal effect of the pickling process on the surface of the strip steel oxide scale. If the pickling speed does not match the material factors of the strip steel and the acid factors of the pickling process, there will be two possibilities: one is over-washing, which will produce a large number of surface defects and also lead to excessive consumption of acid; the other is incomplete pickling, which cannot achieve the effect of removing the surface steel oxide scale.
[0004] In the current pickling process design, the pickling speed is generally set by combining the pickling target value with manual experience, while ignoring the influence of strip material factors and pickling liquid process factors on the pickling effect. With the increasing difference in the structure of the hot-rolled surface oxide layer, it is impossible to achieve a good match between the pickling speed and the hot-rolled strip and the pickling liquid by relying solely on the traditional pickling speed setting method.
[0005] The hydrochloric acid pickling process is a commonly used and efficient pickling process. However, how to minimize the consumption of hydrochloric acid while ensuring the quality of strip pickling has always been a hot topic in the field of cold rolling pickling. Minimizing the consumption of hydrochloric acid can save the pickling cost of enterprises on the one hand, and reduce the production of acidic wastewater on the other hand. Among them, some possible problems and solutions include:
[0006] 1. The strip properties do not match the concentration of pickling hydrochloric acid. The initial concentration of pickling hydrochloric acid is controllable. In the traditional continuous strip pickling system, the hydrochloric acid concentration is always a certain value, while the surface quality of strip produced in different batches varies, such as the strip surface temperature, strip surface composition, and strip thickness, etc. These factors require a suitable hydrochloric acid concentration and strip pickling speed to match them. If these factors are ignored, it may cause insufficient or over-pickling of the strip, thereby affecting the surface quality of the strip and causing a decrease in the utilization of hydrochloric acid during the pickling process.
[0007] 2. Inaccurate control of pickling time: At present, the pickling speed of the strip in the traditional pickling system is set according to experience, but it is not accurately correlated with various factors such as the system's acid properties and the strip surface properties, resulting in the pickling time being too short or too long, resulting in insufficient or over-pickling of the strip.
[0008] 3. Previous excellent production data or production experience has not been fully utilized: The pickling section of the enterprise generates a large amount of pickling production data every year, and these data contain relatively excellent process parameters in the pickling process. How to repeat the excellent production process parameters of the enterprise or predict the quality of the pickled steel plate according to certain pickling conditions is the future development direction of the strip pickling system.
[0009] By systematically analyzing and improving these aspects, the efficiency of the strip pickling process can be improved, thereby improving the pickling quality of the strip surface on the one hand, and saving the amount of hydrochloric acid pickling solution to the greatest extent on the other hand, reducing the generation of acidic wastewater in the enterprise. In terms of the prediction and control of pickling solution concentration, some directions worth studying include:
[0010] 1. Intelligent prediction model: Develop models based on machine learning or artificial intelligence to predict the optimal hydrochloric acid concentration and pickling speed by analyzing the property data of hydrochloric acid pickling solution and the surface quality data of strip steel. These models can improve the accuracy and efficiency of pickling.
[0011] 2. Feedback control system: Design and implement a feedback control system to adjust the concentration of hydrochloric acid pickling solution based on real-time strip surface properties and strip surface quality feedback after pickling. This system can achieve automated acid concentration control, improve pickling efficiency and reduce pickling solution waste.
[0012] 3. Optimization algorithm: Use optimization algorithms, such as genetic algorithm, particle swarm optimization, etc., to find the optimal pickling solution concentration and strip pickling speed to maximize the strip pickling effect and minimize the pickling solution consumption.
[0013] 4. Sensor-based monitoring technology: Research and develop sensor-based monitoring technology to monitor the properties of hydrochloric acid pickling solution and the pickling reaction process in real time, providing reliable data support for the precise control of hydrochloric acid pickling concentration and strip pickling speed.
[0014] Therefore, it is urgent to develop a better method to predict and control the acid pickling concentration and strip pickling speed, and automatically adjust the pickling solution concentration and strip pickling speed according to the prediction results. On the one hand, it improves the processing efficiency of the strip pickling process, and on the other hand, it reduces the consumption of pickling solution, thereby further reducing the acid wastewater discharge of the enterprise. Summary of the invention
[0015] The purpose of the present invention is to provide a prediction control system for pickling of steel strips based on a neural network model, which provides a prediction model for the concentration of pickling solution and the pickling speed of pickling of steel strips, and adjusts the concentration of pickling solution and the pickling speed of steel strips according to the material factors of steel strips and the real-time change state of the pickling solution of the pickling system, so as to avoid over-pickling or under-pickling of steel strips, thereby causing surface defects of steel strips, and at the same time, the consumption of pickling solution can be reduced to the greatest extent. The prediction model in the present invention deeply learns the relationship between the historical pickling solution concentration and pickling speed of the pickling system and the thickness of the iron oxide scale, the composition of the iron oxide scale, the thickness of the steel strip, the temperature of the steel strip, the concentration of the acid solution, the temperature of the acid solution, and the Fe ion concentration in the acid solution, optimizes the neural network model, quantitatively predicts the concentration of pickling solution and the pickling speed of steel strips, and adjusts the concentration of pickling solution and the pickling speed in real time to eliminate the disadvantages caused by the untimely adjustment of the concentration of pickling solution and the pickling speed, thereby improving the pickling process effect and reducing the consumption of pickling agents.
[0016] The technical solution adopted by the present invention is as follows:
[0017] According to a first aspect of the present invention, a strip pickling prediction control system based on a neural network model is provided, which is a system for predicting and adjusting the pickling solution concentration and pickling speed of the strip pickling, comprising: a strip surface quality detection unit, an acid solution property online detection unit, a neural network model unit, a pickling solution concentration control unit, a pickling speed control unit and a strip surface quality detection unit after pickling;
[0018] The strip surface quality detection unit and the acid liquid property online detection unit are used to detect the data of the strip and the pickling solution respectively; the strip surface quality detection unit after pickling is used to detect the surface quality of the strip after pickling; the neural network model unit is used to predict the pickling solution concentration and the pickling speed according to the detection data of the strip surface quality detection unit and the acid liquid property online detection unit, so as to obtain the corresponding pickling solution concentration and pickling speed, and dynamically adjust the pickling speed in real time according to the strip surface quality data fed back by the strip surface quality detection unit after pickling; the pickling solution concentration control unit includes a pickling solution dosing pump, which is used to control the amount of pickling solution added according to the pickling solution concentration given by the neural network model unit; the pickling speed control unit includes a strip feeder, which is used to control the feeding speed of the strip according to the pickling speed given by the neural network model unit.
[0019] Optionally, the strip surface quality detection unit is used to detect the strip thickness, strip temperature, strip surface oxide scale thickness and strip surface oxide scale composition; the acid liquid property online detection unit is used to detect the acid liquid concentration, acid liquid temperature and Fe ion concentration in the acid liquid;
[0020] The neural network model unit has a built-in neural network model for predicting the pickling solution concentration and pickling speed of strip pickling. The neural network model takes the strip thickness, strip temperature, strip surface oxide scale thickness, strip surface oxide scale composition, acid solution concentration, acid solution temperature and Fe ion concentration in the acid solution as input, and takes the corresponding pickling solution concentration and pickling speed as output.
[0021] Optionally, the strip surface quality detection unit includes an online strip thickness detector, an online strip surface temperature detector, an online strip surface oxide scale thickness detector and an online strip surface composition detector, which are used to detect the strip thickness, strip temperature, strip surface oxide scale thickness and strip surface oxide scale composition respectively;
[0022] The acid liquid property online detection unit comprises an acid concentration online detector, an acid liquid temperature detector and an acid Fe concentration online detector, which are respectively used to detect the concentration of the acid liquid, the temperature of the acid liquid and the Fe ion concentration in the acid liquid.
[0023] Optionally, the neural network model includes a BP neural network model, a RBF neural network model and a RNN neural network model, and each neural network model is trained and predicted respectively, and the prediction results of each model are compared to select the optimal neural network model.
[0024] Optionally, the system also includes a pickling system operation data storage unit, which stores data on optimal pickling parameters of the pickling workshop in previous years, including strip thickness, strip temperature, strip surface oxide scale thickness, strip surface oxide scale composition, acid solution concentration, acid solution temperature, Fe ion concentration in the acid solution, and corresponding pickling solution concentration and pickling speed.
[0025] Optionally, the strip surface quality detection unit after pickling includes a strip surface imaging detector, which is arranged at a strip pickling quality inspection station after the pickling, rinsing and drying of the strip. It quantifies the surface quality of the strip after pickling by analyzing the surface imaging of the strip after pickling, and feeds back to the neural network model unit, so that the neural network model unit can dynamically adjust the pickling speed in real time.
[0026] Optional, dynamic adjustment of pickling speed, including:
[0027] Determine the feedback pickling speed based on the surface quality data of the strip steel, that is, if the strip steel is not pickled enough, reduce the pickling speed; if the strip steel is over pickled, increase the pickling speed;
[0028] Determine whether the difference between the feedback pickling speed and the pickling speed predicted by the neural network model unit is within a preset range; if so, control the feeding speed of the strip with the feedback pickling speed; otherwise, control the feeding speed of the strip with the predicted pickling speed.
[0029] Optionally, the neural network model unit further includes a display screen; the display screen is used to display data of the steel strip and the pickling solution in real time, including the thickness of the steel strip, the temperature of the steel strip, the thickness of the iron oxide scale on the surface of the steel strip, the composition of the iron oxide scale on the surface of the steel strip, the concentration of the acid solution, the temperature of the acid solution, and the Fe ion concentration in the acid solution;
[0030] The display screen is also used to display the corresponding pickling solution concentration and pickling speed in real time.
[0031] Optionally, the acid liquid property online detection unit further includes an alarm; when the data of the pickling liquid is abnormal, the alarm sounds an alarm.
[0032] According to a second aspect of the present invention, there is provided a strip pickling system, the strip pickling system comprising any one of the above-mentioned strip pickling prediction control systems based on a neural network model, a turbulent pickling shallow tank, an acid spray head and a strip advancing roller;
[0033] The turbulent pickling shallow trough is used to place the pickling liquid, the acid spray head is used to spray the pickling liquid onto the strip, and the strip advancing roller is used to make the strip move forward smoothly and continuously in the pickling line.
[0034] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:
[0035] The present invention predicts the pickling solution concentration and the pickling speed of the strip steel pickling process in real time through a neural network model according to strip steel material factors and acid factors in the process, including strip steel thickness, strip steel temperature, strip steel surface iron oxide scale thickness, strip steel surface iron oxide scale composition, acid solution concentration, acid solution temperature and Fe ion concentration in the acid solution, thereby achieving minimum pickling solution consumption and maximum pickling efficiency in the strip steel pickling process, reducing enterprise pickling solution consumption, and improving strip steel pickling surface quality, and has relatively broad practical significance and application prospects.
[0036] In addition, the present invention dynamically adjusts the pickling speed in real time according to the feedback data of the strip surface quality: first, the feedback pickling speed is determined according to the feedback data of the strip surface quality, that is, if the strip is not pickled enough, the pickling speed is reduced; if the strip is over-pickled, the pickling speed is increased to implement feedback control and further improve the pickling surface quality of the strip; however, the quality of the feedback is generally poor, so the present invention determines whether the difference between the feedback pickling speed and the pickling speed predicted by the neural network model unit is within a preset range; if so, the feeding speed of the strip is controlled by the feedback pickling speed; otherwise, the feeding speed of the strip is controlled by the predicted pickling speed, so as to keep the pickling quality optimal. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a schematic diagram of the strip pickling prediction control system based on the neural network model of the present invention. DETAILED DESCRIPTION
[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0039] The present invention provides a prediction and control system for pickling solution concentration and pickling speed of steel strip based on a neural network model, which accurately predicts the pickling solution concentration and pickling speed of steel strip according to the influencing factors of the steel rolling pickling process, mainly including the factors of the steel strip itself and the pickling solution factors, thereby improving the pickling quality and reducing the consumption of the pickling solution. In particular, it refers to a prediction and control system for the pickling solution concentration and pickling speed in the steel strip pickling process using hydrochloric acid as the pickling solution.
[0040] Embodiment 1
[0041] like Figure 1As shown, the embodiment of the present application provides a prediction control system for pickling solution concentration and pickling speed of steel strip based on a neural network model, which is a system for predicting and adjusting the pickling solution concentration and pickling speed of steel strip, and includes: a steel strip surface quality detection unit, an acid liquid property online detection unit, a neural network model unit, a pickling solution concentration control unit, a pickling speed control unit, and a strip surface quality detection unit after pickling. The neural network model unit is used to simulate and calculate the pickling solution concentration and pickling speed according to the detection data of the feedforward detection units such as the steel strip surface quality detection unit and the acid liquid property online detection unit to obtain the corresponding pickling solution concentration and pickling speed, and to dynamically adjust the pickling speed in real time according to the strip surface quality data fed back by the strip surface quality detection unit after pickling; the pickling solution concentration control unit includes a pickling solution dosing pump, which is used to control the dosing amount of the pickling solution according to the pickling solution concentration given by the neural network model unit; the pickling speed control unit includes a strip feeder, which is used to control the feeding speed of the strip according to the pickling speed given by the neural network model unit.
[0042] In this embodiment, the neural network model unit can select the BP neural network model and the RBF neural network model, and respectively use the data of the actual pickling parameters collected from the pickling workshop in previous years for training to obtain the optimal model to control the pickling solution concentration and pickling speed. Among them, each model uses 7 parameters such as strip thickness, strip temperature, strip oxide scale thickness, oxide scale composition, concentration of acid in the pickling tank, temperature of the acid, and Fe ion concentration in the acid as input parameters, and takes the strip pickling speed and the concentration of hydrochloric acid pickling solution as output, and respectively performs training and prediction to determine the optimal parameters of each network model, compares the prediction results of each model, and selects the optimal model according to the prediction results to realize the simulation calculation of strip pickling solution concentration and pickling speed.
[0043] In this embodiment, the online detection unit for the properties of the pickling liquid includes: an online detector for acid concentration, an acid liquid temperature detector, and an online detector for Fe concentration in the acid; the strip surface quality detection unit includes: an online detector for strip thickness, an online detector for strip surface temperature, an online detector for strip surface oxide scale thickness, and an online detector for strip surface composition.
[0044] The feedforward hydrochloric acid pickling liquid property detection unit and the strip surface property online detection unit obtain data such as the acid concentration of the pickling liquid, the pickling liquid temperature, the Fe ion concentration in the pickling liquid, the strip thickness, the strip surface temperature, the strip surface oxide scale thickness, and the strip surface composition, and then feed back to the neural network model unit so that the neural network model unit can simulate and calculate the hydrochloric acid concentration and the strip pickling speed to obtain the corresponding parameters.
[0045] In this embodiment, the surface quality detection unit for the steel strip after pickling, which feeds back the surface quality of the steel strip after pickling, includes a steel strip surface imaging detector, which is arranged at the steel strip pickling quality inspection station after the pickling, rinsing and drying of the steel strip. The unit analyzes the surface imaging of the steel strip after pickling to quantify the surface quality of the steel strip after pickling, mainly the degree of over-pickling or under-pickling of the steel strip, and feeds back to the neural network model unit so that the neural network model unit can dynamically adjust the pickling speed and the hydrochloric acid concentration of the pickling solution in real time.
[0046] For example, if it is found that the strip is not pickled enough, the pickling speed is appropriately reduced. If the strip is over-pickled, the pickling speed is increased as the feedback pickling speed to control the pickling speed of the strip. The amount of pickling speed reduction and increase each time can be set according to the actual situation. However, since the quality of post-feedback is generally poor and has hysteresis, if the adjusted pickling speed is used directly, a large error may occur. Therefore, the present application will then determine whether the difference between the feedback pickling speed and the pickling speed predicted by the neural network model unit is within a preset range; if so, the feeding speed of the strip is controlled by the feedback pickling speed; otherwise, the feeding speed of the strip is controlled by the predicted pickling speed to ensure the pickling quality.
[0047] In this embodiment, the neural network model unit includes a data processing unit and a system control unit. The data processing unit is used to calculate the hydrochloric acid concentration of the pickling solution and the pickling speed of the strip using the neural network model. The system control unit is used to control the pickling solution concentration control unit and the pickling speed control unit according to the hydrochloric acid concentration of the pickling solution and the pickling speed of the strip calculated by the data processing unit, thereby controlling the pickling solution hydrochloric acid dosing pump and adjusting the pickling speed of the strip.
[0048] The system control unit is provided with a display screen for real-time display of the strip surface property data and the pickling liquid property data. The strip surface property data include strip thickness, strip surface temperature, strip surface oxide scale thickness, strip surface composition and other data. The pickling liquid property data include the acid concentration of the pickling liquid, the Fe concentration in the pickling liquid, and the pickling liquid temperature.
[0049] In addition, the display screen can also display the pickling solution concentration and pickling speed predicted by the neural network model unit. The staff can roughly judge whether the prediction result is correct based on the data displayed on the display screen, and then make timely adjustments. For example, when the thickness of the iron oxide scale on the surface of the strip is thicker, it generally corresponds to a higher pickling speed.
[0050] The system control unit can also be provided with an alarm to sound an alarm when the hydrochloric acid liquid has abnormal properties. For example, if the acid concentration of the pickling liquid is too low and exceeds the limit, the alarm will sound an alarm.
[0051] In this embodiment, the strip surface property detection unit, the acid liquid property online detection unit and the neural network model unit transmit input and output information through network signals. The neural network model unit is connected to a metering dosing pump for replenishing hydrochloric acid pickling solution and a strip feeder for conveying the strip to respectively control the pickling solution concentration and the pickling speed during pickling of the strip.
[0052] In summary, the present invention integrates big data models, and uses the operating data of the enterprise's pickling section in previous years as the decision-making basis for the later pickling hydrochloric acid consumption and the pickling speed of the strip. Through data quantification, the past operating experience and actual production are combined to realize the prediction model of hydrochloric acid concentration and strip pickling speed in the pickling process of steel cold rolling enterprises based on machine deep learning, so as to obtain accurate pickling process parameters and adjust the amount of hydrochloric acid in real time to eliminate the drawbacks of the traditional control process and improve the pickling effect.
[0053] Embodiment 2
[0054] like Figure 1 As shown, the cold-rolled strip pickling system is provided with a turbulent pickling shallow tank, a strip push roller, an acid spray head, a pickling solution hydrochloric acid property detection unit, a strip surface quality detection unit, etc. The pickling solution hydrochloric acid property detection unit includes online detection of hydrochloric acid concentration, online detection of hydrochloric acid temperature, and online detection of Fe content in hydrochloric acid. The strip surface quality detection unit includes a strip surface temperature detector, a strip thickness detector, a strip surface composition detection unit, and a strip surface quality detection unit after rinsing and drying. The hydrochloric acid concentration online detection is used to detect the hydrochloric acid concentration of the pickling solution in the pickling shallow tank, the hydrochloric acid temperature online detector is used to detect the temperature of the pickling solution in the pickling shallow tank online, the Fe content in hydrochloric acid online detector is used to detect the Fe ion content of the pickling solution in the pickling shallow tank online, the strip surface temperature detector is used to detect the surface temperature of the strip before pickling, the strip thickness detector is used to detect the thickness of the strip before pickling, and the strip surface composition detector is used to detect the surface composition of the strip before pickling, mainly to detect the type of iron oxide on the strip surface. The strip surface quality detector after rinsing and drying mainly refers to the detection of the surface quality of the strip after pickling, rinsing and drying. It is mainly through the online surface quality detector after steel pickling, that is, the strip surface imaging detector, which is set at the strip pickling quality inspection station after the strip pickling, rinsing and drying. The unit quantifies the surface quality of the strip after pickling by analyzing the surface imaging of the strip after pickling.
[0055] The detection data in the above-mentioned detection units are fed back to the neural network model unit so that the neural network model unit can simulate the calculation of the hydrochloric acid concentration of the pickling solution and the pickling speed of the strip to obtain the corresponding hydrochloric acid concentration of the pickling solution and the pickling speed.
[0056] The neural network model unit can select neural network models such as BP neural network, RBF neural network, RNN neural network model, etc., and use the collected good data for training to determine the optimal hyperparameters of the network, compare the prediction results of each model, and select the optimal model according to the prediction results to achieve the adjustment of the hydrochloric acid concentration of the cold rolling pickling solution and the pickling speed of the strip.
[0057] Specifically, the actual operation data of the system is collected, including the surface composition of the strip, the thickness of the strip, the temperature of the strip, the hydrochloric acid concentration of the pickling solution, the temperature of the pickling solution, and the Fe ion concentration of the pickling solution. The actual operation data collected is divided into training samples and test samples, and the training samples are used to train neural network models such as BP neural network, RBF neural network, RNN neural network model, etc., where each model takes 7 parameters such as strip thickness, strip temperature, strip oxide scale thickness, oxide scale composition, acid concentration in the pickling tank, acid temperature, and Fe ion concentration in the acid as input parameters, and takes pickling solution hydrochloric acid concentration and strip pickling speed as output. Training and prediction are carried out respectively to determine the optimal hyperparameters of the network of each model, compare the prediction results of each model, and select the optimal model according to the prediction results to realize the simulation calculation of the dosage and pickling speed of cold rolling pickling. Through analysis, it is found that the BP neural network model has better prediction ability, and its simulation prediction accuracy is as high as 97.31%.
[0058] Embodiment 3
[0059] The embodiment of the present application provides a strip pickling system, which is provided with a turbulent pickling shallow tank, an acid spray head, a strip steel push roller, and a strip steel pickling liquid concentration and pickling speed prediction control system based on a neural network model in the above scheme to complete the strip steel surface pickling process. Among them, the turbulent pickling shallow tank is used to place the pickling liquid, the acid spray head is used to spray the pickling liquid to the strip steel, and the strip steel push roller is used to make the strip steel move forward steadily and continuously in the pickling line.
[0060] In summary, the present invention proposes a prediction and control system for pickling solution concentration and pickling speed of steel strip based on a neural network model. The system is used to predict and adjust the pickling solution concentration and pickling speed of steel strip pickling. According to the material factors of steel strip and the acid factors in the process, including the thickness of steel strip, the temperature of steel strip, the thickness of iron oxide scale on the surface of steel strip, the composition of iron oxide scale on the surface of steel strip, the concentration of acid solution, the temperature of acid solution and the Fe ion concentration in the acid solution, the pickling solution concentration and the pickling speed of steel strip in the pickling process are predicted in real time through the neural network model, so as to achieve the minimum consumption of pickling solution and the maximization of pickling efficiency in the pickling process of steel strip, reduce the enterprise's pickling solution consumption, and improve the surface quality of steel strip pickling. The system has a relatively broad practical significance and application prospect.
[0061] In addition, the present invention dynamically adjusts the pickling speed in real time according to the feedback data of the strip surface quality: first, the feedback pickling speed is determined according to the feedback data of the strip surface quality: if the strip surface quality detection unit detects that the strip is insufficiently pickled after pickling, the pickling speed is appropriately reduced; if the strip surface quality detection unit detects that the strip is over-pickled after pickling, the pickling speed is appropriately increased to implement feedback control and further improve the pickling surface quality of the strip; however, the quality of the feedback is generally poor, so the present invention determines whether the difference between the feedback pickling speed and the pickling speed predicted by the neural network model unit is within a preset range; if so, the feeding speed of the strip is controlled by the feedback pickling speed; otherwise, the feeding speed of the strip is controlled by the predicted pickling speed, so as to keep the pickling quality optimal.
[0062] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0063] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.
[0064] It will be easily understood by those skilled in the art that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A predictive control system for pickling of steel strip based on a neural network model, characterized in that: The system is used to predict and adjust the pickling solution concentration and pickling speed of strip steel pickling, and includes: a strip steel surface quality detection unit, an acid solution property online detection unit, a neural network model unit, a pickling solution concentration control unit, a pickling speed control unit and a strip steel surface quality detection unit after pickling; The strip surface quality detection unit and the acid liquid property online detection unit are used to detect the data of the strip and the pickling solution respectively; the strip surface quality detection unit after pickling is used to detect the surface quality of the strip after pickling; the neural network model unit is used to predict the pickling solution concentration and the pickling speed according to the detection data of the strip surface quality detection unit and the acid liquid property online detection unit, so as to obtain the corresponding pickling solution concentration and pickling speed, and dynamically adjust the pickling speed in real time according to the strip surface quality data fed back by the strip surface quality detection unit after pickling; the pickling solution concentration control unit includes a pickling solution dosing pump, which is used to control the amount of pickling solution added according to the pickling solution concentration given by the neural network model unit; the pickling speed control unit includes a strip feeder, which is used to control the feeding speed of the strip according to the pickling speed given by the neural network model unit.
2. The strip pickling prediction control system based on the neural network model according to claim 1 is characterized in that: The strip surface quality detection unit is used to detect the strip thickness, strip temperature, strip surface oxide scale thickness and strip surface oxide scale composition; the acid liquid property online detection unit is used to detect the acid liquid concentration, acid liquid temperature and Fe ion concentration in the acid liquid; The neural network model unit has a built-in neural network model for predicting the pickling solution concentration and pickling speed of strip pickling. The neural network model takes the strip thickness, strip temperature, strip surface oxide scale thickness, strip surface oxide scale composition, acid solution concentration, acid solution temperature and Fe ion concentration in the acid solution as input, and takes the corresponding pickling solution concentration and pickling speed as output.
3. The strip pickling prediction control system based on the neural network model according to claim 2 is characterized in that: The strip surface quality detection unit includes an online strip thickness detector, an online strip surface temperature detector, an online strip surface oxide scale thickness detector and an online strip surface composition detector, which are used to detect the strip thickness, strip temperature, strip surface oxide scale thickness and strip surface oxide scale composition respectively; The acid liquid property online detection unit comprises an acid concentration online detector, an acid liquid temperature detector and an acid Fe concentration online detector, which are respectively used to detect the concentration of the acid liquid, the temperature of the acid liquid and the Fe ion concentration in the acid liquid.
4. The strip pickling prediction control system based on the neural network model according to claim 2 is characterized in that: The neural network models include a BP neural network model, a RBF neural network model and a RNN neural network model. Each neural network model is trained and predicted respectively, and the prediction results of each model are compared to select the optimal neural network model.
5. The strip pickling prediction control system based on the neural network model according to claim 2 is characterized in that: The system also includes a pickling system operation data storage unit, which stores data on the optimal pickling parameters of the pickling workshop in previous years, including strip thickness, strip temperature, strip surface oxide scale thickness, strip surface oxide scale composition, acid solution concentration, acid solution temperature, Fe ion concentration in the acid solution, and corresponding pickling solution concentration and pickling speed.
6. The strip pickling prediction control system based on the neural network model according to claim 1 is characterized in that: The strip surface quality detection unit after pickling includes a strip surface imaging detector, which is arranged at a strip pickling quality inspection station after the pickling, rinsing and drying of the strip. The surface quality of the strip after pickling is quantified by analyzing the surface imaging of the strip after pickling, and the surface quality is fed back to the neural network model unit so that the neural network model unit can dynamically adjust the pickling speed in real time.
7. The strip pickling prediction control system based on the neural network model according to claim 1 is characterized in that: Dynamic adjustment of pickling speed, including: According to the feedback of strip surface quality data, the feedback pickling speed is determined, that is, if the strip is not pickled enough, the pickling speed is reduced, and if the strip is over pickled, the pickling speed is increased; Determine whether the difference between the feedback pickling speed and the pickling speed predicted by the neural network model unit is within a preset range; if so, control the feeding speed of the strip with the feedback pickling speed; otherwise, control the feeding speed of the strip with the predicted pickling speed.
8. The strip pickling prediction control system based on the neural network model according to claim 1 is characterized in that: The neural network model unit also includes a display screen; the display screen is used to display data of the strip and the pickling solution in real time, including the thickness of the strip, the temperature of the strip, the thickness of the iron oxide scale on the surface of the strip, the composition of the iron oxide scale on the surface of the strip, the concentration of the acid solution, the temperature of the acid solution and the Fe ion concentration in the acid solution; The display screen is also used to display the corresponding pickling solution concentration and pickling speed in real time.
9. The strip pickling prediction control system based on the neural network model according to claim 1 is characterized in that: The acid liquid property online detection unit also includes an alarm; when the data of the pickling liquid is abnormal, the alarm sounds an alarm.
10. A strip steel pickling system, characterized in that: The strip steel pickling system comprises the strip steel pickling prediction control system based on the neural network model as claimed in any one of claims 1 to 9, a turbulent pickling shallow tank, an acid liquid spray head and a strip steel advancing roller; The turbulent pickling shallow tank is used to place the pickling liquid, the acid spray head is used to spray the pickling liquid onto the strip, and the strip advancing roller is used to make the strip move forward smoothly and continuously in the pickling line.
Citation Information
Cited By
Intelligent risk supervision and early warning system for steel strip production
CN121258218A
Intelligent control system for pickling steam based on edge computing and lstm neural network
CN123010430A
