Method and system for optimally setting process parameters of three-roller roller coater

Abstract

The invention discloses a method and a system for optimally setting process parameters of a three-roller roller coater, and belongs to the technical field of cold-rolled strip steel processing lines. According to the method, production data are collected in real time, an artificial neural network model is used for predicting the coating film thickness, key process variables are determined through correlation analysis, finally, multi-target optimization is carried out with the target film thickness and energy medium consumption minimization as the target, and process parameters are automatically set. The system adopts a closed-loop architecture of a data acquisition and execution layer, a control and communication layer and an optimization decision layer, and integrates a roller diameter online detection module, a film thickness prediction module, a correlation analysis module and an optimization setting module. The problems of unstable coating quality and high cost caused by traditional manual adjustment are effectively solved, and high-precision, high-stability and low-cost control over the coating thickness is achieved.

Description

Technical Field

[0001] This invention relates to the field of cold-rolled strip steel processing line technology, and in particular to a method and system for intelligent optimization setting of process parameters of a three-roller coater for surface coating of cold-rolled strip steel. Background Technology

[0002] Roll coating is an effective method for processing steel plate surfaces, altering the chemical composition, microstructure, and stress state of steel products to achieve desired surface properties. It is a highly efficient and high-quality method. The thickness and stability of the coating determine the product's corrosion resistance, insulation, and other properties.

[0003] Currently, the strip coating process parameters of silicon steel, color coating, and galvanizing units are mainly adjusted manually based on experience, which easily leads to unstable coating quality and waste of resources. Moreover, in order to ensure product quality, positive tolerances are generally used for coating thickness control, which increases production costs. Summary of the Invention

[0004] In view of the technical defects and drawbacks existing in the prior art, the present invention provides a method and system for optimizing the setting of process parameters of a three-roller coating machine to overcome the above problems or at least partially solve the above problems. The specific solution is as follows:

[0005] As a first aspect of the present invention, a method for optimizing and setting process parameters of a three-roller coating machine is provided, characterized by comprising the following steps:

[0006] The process parameters and coating thickness data of the three-roller coating machine are collected in real time using industrial cameras, field sensors and online film thickness measuring instruments.

[0007] Based on the collected data, the coating thickness is predicted using an artificial neural network model, and key process variables with a correlation greater than a preset threshold are identified through correlation analysis.

[0008] With the goal of minimizing the target coating thickness and energy consumption, the key process variables are combined and optimized to generate optimized process parameter instructions, which are then sent to the three-roller coating machine for execution via the unit's PLC.

[0009] In some embodiments, the real-time acquired process parameter data includes: strip thickness, strip speed, cumulative strip coating length, strip tension, coating roll diameter, coating liquid composition, concentration and viscosity, speed ratio and gap between metering roll and liquid taking roll, pressure between liquid taking roll and coating roll, liquid taking roll speed ratio, direction and speed ratio of coating roll (i.e., rotation direction of coating roll and speed ratio of coating roll to other rolls), and pressure between coating roll and strip; the coating film thickness data is obtained by online detection using an online film thickness detector; the coating roll diameter is obtained by capturing real-time images of the coating roll with an industrial camera and detecting them online using image processing algorithms.

[0010] In some embodiments, before using an artificial neural network model to predict the coating thickness, a sample generation and preprocessing step is included: collecting historical process parameter data and corresponding coating thickness data to form an initial sample set; using at least one of the threshold method, angular distance method and robust regression method to remove abnormal data in the initial sample set, generating a three-roller coating sample set for model training.

[0011] In some embodiments, the artificial neural network model has a three-layer structure, including an input layer, a hidden layer, and an output layer; the input parameters of the input layer are the process parameter data, and the output parameters of the output layer are the predicted coating thickness; the artificial neural network model uses the Sigmoid function as the transformation function and is trained using the BP algorithm; during training, a portion of the three-roller coating sample set is used as training data, and another portion is used as validation data, until the model error is controlled within the allowable range.

[0012] In some embodiments, the correlation analysis calculates the correlation coefficient between each process parameter and the coating thickness using the following formula:

[0013]

[0014] Where x is the independent variable of the process parameter, and y is the dependent variable of the coating thickness. The mean of the independent variable. The mean of the dependent variable is given; the correlation coefficient R(t) is positive if positive and negative if negative; process parameters with an absolute value of the correlation coefficient greater than a preset threshold are selected as the key process variables.

[0015] In some embodiments, the key process variables include: the speed ratio and gap between the metering roller and the liquid taking roller, the pressure between the liquid taking roller and the coating roller, the speed ratio of the liquid taking roller, the direction and speed ratio of the coating roller, and the pressure between the coating roller and the strip.

[0016] In some embodiments, the combination optimization of key process variables specifically includes: using the trained artificial neural network model to calculate the predicted coating thickness under multiple combinations of key process variables; and selecting the combination with the minimum energy medium consumption from all combinations of predicted coating thicknesses that meet the target value as the optimal combination of process parameters.

[0017] As a second aspect of the present invention, a three-roller coating machine process parameter optimization setting system for implementing the method described in any of the above claims is also provided, comprising:

[0018] The data acquisition and execution layer includes field sensors, an online film thickness detector, an industrial camera, and a three-roller coating machine. The field sensors and the online film thickness detector are used to acquire unit operating parameters and coating film thickness data in real time. The industrial camera is used to acquire real-time images of the coating rolls. The three-roller coating machine is used to execute optimized process parameter instructions.

[0019] The control and communication layer includes a unit PLC and an industrial Ethernet. The unit PLC connects to and controls the devices in the data acquisition and execution layer, and is used to collect various real-time data from the data acquisition and execution layer and issue control commands. The industrial Ethernet provides a communication backbone for the system.

[0020] The optimization decision layer is a roller coating model computer, which communicates with the unit PLC and the industrial camera via the industrial Ethernet to receive the real-time data and images, and run the roller coating model software to calculate optimized process parameter instructions based on the real-time data, images and target film thickness, and then send them to the three-roller coating machine via the unit PLC.

[0021] In some embodiments, the roll coating model software running on the roll coating model computer includes four functional modules that work in sequence:

[0022] The online detection module for coating roller diameter receives real-time images captured by the industrial camera and identifies and outputs real-time roller diameter data of the coating roller online through image processing algorithms.

[0023] The coating thickness prediction module receives various real-time process parameters collected by the unit PLC and the real-time roller diameter data, and calls a pre-trained artificial neural network model to output the predicted coating thickness.

[0024] The correlation analysis module calculates the correlation coefficient between each process parameter and the coating thickness based on historical data sequences, and selects the set of key process variables based on the absolute value of the coefficient.

[0025] The roller coating process optimization setting module performs combination optimization within the parameter space of the key process variable set based on the current real-time status of the unit and the target film thickness. It then uses the coating film thickness prediction module for evaluation to select the process parameter combination that meets the target film thickness and minimizes energy consumption, which is output as the optimized process parameter command.

[0026] In some embodiments, the artificial neural network model used by the coating thickness prediction module is a three-layer network structure. The input parameters received by the input layer are determined according to the unit configuration and include at least the strip thickness, strip speed, strip tension, coating roll diameter, coating liquid characteristic parameters, and the speed ratio, gap, and pressure parameters between the metering roll, liquid taking roll, and coating roll in the three-roll coater, which are collected by the data acquisition and execution layer. The output layer outputs a single predicted coating thickness value. The artificial neural network model is trained with historical production data and has nonlinear mapping capability to achieve the prediction of coating thickness.

[0027] The present invention has the following beneficial effects:

[0028] This invention overcomes the problems of large coating thickness fluctuations, unstable quality, and high costs caused by traditional methods that rely on manual experience for adjustment by establishing a closed-loop control process of "data acquisition - model prediction - optimization setting - automatic execution". This method can dynamically adapt to changes in production conditions, proactively optimizing energy consumption while ensuring target film thickness accuracy, significantly improving production efficiency and product consistency, and reducing production costs. Attached Figure Description

[0029] Figure 1 A flowchart illustrating a method for optimizing process parameters of a three-roller coating machine, provided in an embodiment of the present invention;

[0030] Figure 2 A schematic diagram of an artificial neural network model provided in an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of the composition of a process parameter optimization setting system for a three-roller coating machine provided in an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0033] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0034] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0035] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0036] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0037] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0038] To address at least one of the technical problems existing in the aforementioned related technologies, the present invention provides a method for optimizing and setting process parameters of a three-roller coating machine. Figure 1 A flowchart illustrating a method for optimizing process parameters of a three-roller coating machine, provided in an embodiment of the present invention, includes the following steps:

[0039] S1. Real-time acquisition of process parameter data and coating thickness data of the three-roller coating machine through industrial cameras, field sensors and online film thickness measuring instruments;

[0040] S2. Based on the collected data, the coating thickness is predicted using an artificial neural network model, and key process variables with a correlation greater than a preset threshold are identified through correlation analysis.

[0041] S3. With the goal of minimizing the target coating thickness and energy consumption, the key process variables are combined and optimized to generate optimized process parameter instructions, which are then sent to the three-roller coating machine for execution via the unit PLC.

[0042] This invention overcomes the problems of large coating thickness fluctuations, unstable quality, and high costs caused by traditional methods that rely on manual experience for adjustment by establishing a closed-loop control process of "data acquisition - model prediction - optimization setting - automatic execution". This method can dynamically adapt to changes in production conditions, proactively optimizing energy consumption while ensuring target film thickness accuracy, significantly improving production efficiency and product consistency, and reducing production costs.

[0043] In the high-speed roll coating process of a silicon steel production line in a cold rolling mill, operators need to set the target coating thickness in the control system. In the traditional mode, operators need to manually adjust more than ten parameters, such as the speed ratio between the metering roll and the coating roll, which is time-consuming and the results are difficult to guarantee.

[0044] Taking the requirement of coating a new batch of silicon steel strips with an insulating coating of a specific thickness, with a target film thickness of 5μm, as an example, the solution of this invention is as follows:

[0045] The system automatically collects all relevant process parameters, such as current strip speed, tension, coating roll diameter, coating liquid viscosity, and current real-time film thickness data, through on-site sensors, industrial cameras, and film thickness measuring instruments.

[0046] The coating thickness prediction artificial neural network model in the roll coating model computer predicted a film thickness of 4.8 μm under the current parameters based on real-time data, which did not meet the target. Subsequently, the correlation analysis module identified "coating roll and strip pressure" and "liquid take-up roll speed ratio" as the most critical adjustment variables. The optimization setting module then calculated multiple combinations of these two variables and quickly selected the set of parameters that could stabilize the predicted film thickness at 5.1 μm and minimize energy consumption.

[0047] The optimized parameter instructions are automatically sent to the three-roller coater via the unit's PLC. Throughout the process, no manual intervention is required; the system completes adjustments within seconds, and the final measured film thickness stabilizes within the range of 5.0μm±0.1μm, fully meeting quality requirements. Compared to manual settings, this process saves approximately 8% of coating liquid consumption.

[0048] In some embodiments, the real-time acquired process parameter data includes: strip thickness, strip speed, cumulative strip coating length, strip tension, coating roll diameter, coating liquid composition, concentration and viscosity, speed ratio and gap between metering roll and liquid taking roll, pressure between liquid taking roll and coating roll, liquid taking roll speed ratio, coating roll direction and speed ratio, and pressure between coating roll and strip; the coating film thickness data is obtained through online detection by an online film thickness detector; and the coating roll diameter is obtained by capturing real-time images of the coating roll with an industrial camera and detecting them online using image processing algorithms.

[0049] The above embodiments solve the problem of inaccurate model prediction caused by roller wear by clearly defining the specific types of parameters to be collected and innovatively introducing image recognition technology to detect the diameter of the coating roller online. This ensures the authenticity and integrity of the data source on which subsequent model prediction and optimization depend, lays the foundation for the precise control of the entire system, and avoids systematic deviations caused by missing or inaccurate key parameters.

[0050] In some embodiments, before using an artificial neural network model to predict the coating thickness, a sample generation and preprocessing step is included: collecting historical process parameter data and corresponding coating thickness data to form an initial sample set; using at least one of the threshold method, angular distance method and robust regression method to remove abnormal data in the initial sample set, generating a reliable three-roll coating sample set for model training.

[0051] The above embodiments introduce various data cleaning methods (threshold method, angle distance method, robust regression method) to preprocess the original sample data, which can effectively remove abnormal data points caused by factors such as instantaneous sensor failure and abnormal production conditions, prevent "dirty data" from misleading the training of artificial intelligence models, and ensure the accuracy, stability and credibility of the final prediction and optimization results.

[0052] In the initial system setup phase, a large amount of historical production data needs to be collected to train the coating thickness prediction model. For example, nearly three months' worth of production data records need to be exported from the historical database to train the initial artificial neural network model. After adopting the solution of this invention:

[0053] After receiving the initial 10,000 samples, the coating thickness prediction module does not use them directly for training. Instead, it performs data preprocessing: First, a threshold method is used to remove records with film thickness values ​​that significantly exceed the equipment's process capability range (e.g., negative numbers or maximum values). Next, an angular distance method is used to identify and remove "outliers" that differ significantly from other data points in the multidimensional parameter space. These outliers may correspond to temporary equipment failures or special debugging conditions. Finally, a robust regression method is used for secondary screening to further reduce the impact of a few extreme outliers on the overall data distribution.

[0054] After the above steps, a highly reliable "three-roller coating sample set" containing approximately 9,800 data sets was finally generated. The neural network model trained with this "clean" dataset showed significantly better prediction accuracy and generalization ability than the model trained directly with the original data. In subsequent online applications, the model demonstrated extremely strong stability and could provide reliable predictions even when production conditions fluctuated slightly, providing a credible basis for optimization settings.

[0055] In some embodiments, the artificial neural network model has a three-layer structure, including an input layer, a hidden layer, and an output layer; the input parameters of the input layer are the process parameter data, and the output parameters of the output layer are the predicted coating thickness; the artificial neural network model uses the Sigmoid function as the transformation function and is trained using the BP algorithm; during training, 80% of the three-roller coating sample set is used as training data and 20% is used as validation data until the model error is controlled within the allowable range.

[0056] Artificial neural network models such as Figure 2 As shown

[0057] By clearly defining the three-layer structure, activation function (Sigmoid), and training algorithm (BP algorithm) of the artificial neural network, and specifying a scientific training and validation data split ratio (80% / 20%), this method ensures that the model can effectively learn complex nonlinear relationships from historical data. This approach enables the model to have strong approximation and generalization capabilities, and can accurately predict coating thickness, providing a reliable decision-making basis for subsequent process parameter optimization. It avoids the shortcomings of traditional mathematical models in handling complex process nonlinear problems.

[0058] For example, during the initial commissioning phase of a newly built roll coating production line in a galvanizing unit, it is necessary to establish a model that can accurately predict the zinc layer thickness. The production line has a complex process, and the coating thickness is affected nonlinearly by more than ten parameters such as strip speed, roller pressure, and zinc liquid temperature. Traditional empirical formulas have large prediction errors.

[0059] After adopting the solution of this invention: the technicians designed as follows Figure 2 The three-layer artificial neural network model shown is illustrated. The number of nodes in the input layer corresponds to the 12 selected key process parameters, and the output layer has one node (predicting film thickness). The number of nodes in the hidden layer was determined experimentally.

[0060] During the model training and validation phases, the system divides the 5000 sets of valid historical sample data collected into two parts: 80% (4000 sets) for training and 20% (1000 sets) for validation. During training, the Sigmoid function is used as the activation function for neurons, and the backpropagation algorithm is employed for error backpropagation to continuously adjust the network weights.

[0061] After multiple iterations of training, the model's prediction error on the validation set was successfully controlled within the allowable range (e.g., ±3%). Once deployed, the trained model demonstrated highly accurate predictions of zinc layer thickness for new production orders, with a correlation coefficient exceeding 0.98 between the predicted and actual measured values. This high-precision predictive capability laid a solid foundation for further precise optimization of process parameters, enabling the production line to achieve high-quality thickness control from the initial commissioning stage.

[0062] In some embodiments, the correlation analysis calculates the correlation coefficient between each process parameter and the coating thickness using the following formula:

[0063]

[0064] Where, x i Let y be the independent variable value of the i-th sample, that is, the specific value of a certain process parameter (such as strip speed, coating roll pressure, etc.) in the i-th sampling or historical data record. i It is the dependent variable value of the i-th sample, that is, the value related to x. i Corresponding to the actual coating thickness value obtained under the same production conditions, The mean of the independent variable x across all samples It is the average value of the dependent variable y (coating thickness) in all samples; the correlation coefficient R(t) is positive to indicate positive correlation and negative to indicate negative correlation; process parameters with an absolute value of correlation coefficient greater than a preset threshold are selected as the key process variables.

[0065] In the above embodiments, by introducing explicit mathematical formulas to calculate the correlation coefficient (R(t)) between each process parameter and the coating thickness, the traditional factor importance analysis that relies on subjective experience judgment is transformed into a data-driven scientific analysis. This method can accurately identify parameters that have a significant positive or negative impact on the film thickness, effectively avoid wasting resources on secondary variables during the optimization process, thereby "grasping the main contradiction", greatly improving optimization efficiency and pertinence, and ensuring that the optimization algorithm searches within the most important parameter space and quickly converges to the optimal solution.

[0066] For example, calculations revealed a strong positive correlation coefficient of +0.85 for the "coating roller and strip pressure," a strong negative correlation coefficient of -0.72 for the "gap between the metering roller and the liquid collection roller," and a weak positive correlation coefficient of only +0.15 for the "strip tension." The analysis clearly shows that the "coating roller and strip pressure" and the "gap between the metering roller and the liquid collection roller" are key variables controlling film thickness. Therefore, in subsequent optimization settings, the system focuses primarily on optimizing the combination of these two parameters, rather than wasting effort on parameters with negligible impact, such as "strip tension." This makes the optimization process faster and more efficient, significantly reducing setup time when changing production specifications.

[0067] In some embodiments, the key process variables include: the speed ratio and gap between the metering roller and the liquid taking roller, the pressure between the liquid taking roller and the coating roller, the speed ratio of the liquid taking roller, the direction and speed ratio of the coating roller, and the pressure between the coating roller and the strip.

[0068] In the above embodiments, by specifically specifying the set of key process variables (speed ratio and gap between the metering roller and the liquid-taking roller, pressure between the liquid-taking roller and the coating roller, speed ratio of the liquid-taking roller, direction and speed ratio of the coating roller, and pressure between the coating roller and the strip), a clear and effective search space is defined for the optimization algorithm. This not only significantly reduces the dimensionality and computational complexity of the optimization problem and improves the feasibility of real-time optimization, but also ensures that any optimization adjustment directly affects the link most sensitive to the film thickness, making the control behavior more precise and efficient.

[0069] For example, a silicon steel production line needs to frequently switch between different insulation coating thickness specifications. After each switch, the film thickness needs to be quickly adjusted to the target value. Traditionally, engineers need to adjust multiple parameters at the same time, which is cumbersome and prone to mutual interference, resulting in excessive adjustment time and a large number of defective products.

[0070] After adopting the solution of this invention: before optimization, the optimization variables are clearly locked on five key adjustable parameters: the speed ratio and gap between the metering roller and the liquid taking roller (A), the pressure between the liquid taking roller and the coating roller (B), the speed ratio of the liquid taking roller (C), the direction and speed ratio of the coating roller (D), and the pressure between the coating roller and the strip (E). When it is necessary to switch from a 3μm film thickness to a 5μm film thickness, the roller coating process optimization setting module only needs to perform combined simulation and search within the parameter space composed of the five variables A, B, C, D, and E using a trained neural network model.

[0071] Because of its highly focused variable set, the system can calculate multiple parameter combinations that achieve a 5μm film thickness within seconds and select the one with the lowest energy consumption. Engineers only need to confirm the settings, and the equipment will automatically complete the adjustment. Compared to the previous method of manually trying and adjusting more than ten parameters repeatedly, the system now directly optimizes the core variables, reducing switchover time by more than 70%, significantly improving the first-pass yield, and effectively reducing material and energy waste.

[0072] In some embodiments, the combination optimization of key process variables specifically includes: using the trained artificial neural network model to calculate the predicted coating thickness under multiple combinations of key process variables; and selecting the combination with the minimum energy medium consumption from all combinations of predicted coating thicknesses that meet the target value as the optimal combination of process parameters.

[0073] The above embodiments explicitly define "meeting the target film thickness" as a quality constraint and "minimizing energy medium consumption" as an economic optimization objective, utilizing a neural network model for rapid simulation and selection. This method surpasses traditional single-objective control that merely pursues quality compliance or blindly reduces usage. It intelligently identifies the most economical and energy-efficient combination of process parameters while ensuring high product quality, thereby directly reducing the consumption of coating liquid and electrical energy, achieving cost reduction and efficiency improvement, and enhancing the level of lean production.

[0074] For example, a color-coated steel sheet manufacturer receives a large order from a customer with strict requirements for coating color and thickness consistency. At the same time, factory management wants to minimize production costs for this order, such as applying a blue coating with a target thickness of 25μm to a batch of substrates. It is known that different roller speed ratios and pressure combinations affect the transfer efficiency of the coating liquid, thus affecting consumption and energy media usage.

[0075] After adopting the solution of this invention: the roller coating process optimization setting module generates hundreds of different parameter combinations based on the current production status and within a reasonable range of key process variables (such as the pressure and speed ratio of each roller). For each set of parameters, the system calls the coating thickness prediction artificial neural network model to quickly calculate the corresponding predicted coating thickness and the estimated energy medium consumption (such as the power loss of pumps and motors).

[0076] For example, the system first selects combinations from all available options that predict film thickness between 24.5 μm and 25.5 μm (i.e., meet the target). Then, among these qualified combinations, the system compares their estimated energy medium consumption and finally selects the set of parameters with the lowest energy medium consumption value. The system automatically sends this optimal set of parameters to the generator unit.

[0077] See Figure 3As shown, based on the same inventive concept, the present invention also provides a three-roller coating machine process parameter optimization setting system for implementing any of the above-described methods, comprising:

[0078] The data acquisition and execution layer includes field sensors, an online film thickness measuring instrument, an industrial camera, and a three-roll coating machine. The field sensors and the online film thickness measuring instrument are used to acquire the unit's operating parameters (i.e., process parameter data) and coating film thickness data in real time. The industrial camera is used to acquire real-time images of the coating rolls, and the three-roll coating machine is used to execute optimized process parameter instructions.

[0079] The control and communication layer includes a unit PLC and an industrial Ethernet. The unit PLC connects to and controls the devices in the data acquisition and execution layer, and is used to collect various real-time data from the data acquisition and execution layer and issue control commands. The industrial Ethernet provides a communication backbone for the system.

[0080] The optimization decision layer is a roller coating model computer, which communicates with the unit PLC and the industrial camera via the industrial Ethernet to receive the real-time data and images, and run the roller coating model software to calculate optimized process parameter instructions based on the real-time data, images and target film thickness, and then send them to the three-roller coating machine via the unit PLC.

[0081] The above embodiments divide the system into a "data acquisition and execution layer", a "control and communication layer" and an "optimization decision layer", clarifying the roles of devices in each layer and the interaction relationships between layers. This architecture ensures that the entire process from raw data acquisition to intelligent decision-making and final execution is efficient, reliable and automated.

[0082] Take, for example, a modern steel plant that plans to upgrade its old semi-automatic roller coating unit to an intelligent system. The original system operated independently, data was recorded manually, and parameter adjustments depended on on-site operation by engineers, resulting in slow response and poor consistency.

[0083] After adopting the solution of this invention: During upgrades, the construction is strictly carried out according to the above control architecture. Data acquisition and execution layer: New high-precision sensors, film thickness gauges, and industrial cameras are installed and connected to the three-roll coating machine. Control and communication layer: Utilizing the existing unit PLC and the newly added industrial Ethernet switch, high-speed and stable transmission of all data and control commands is ensured. Optimization decision layer: A high-performance roll coating model computer is added, communicating with the PLC and camera via Ethernet. During production, the bottom-level sensors and cameras collect data in real time, which is then aggregated to the upper-level model computer via the PLC and network. The software within the model computer performs calculations and optimizations, and after generating commands, they are immediately sent to the PLC through the same path. The PLC then controls the roll coating machine to execute adjustments.

[0084] This architecture makes the entire production line an organic whole. For example, when a change in the viscosity of the coating liquid is detected, the system can automatically complete the entire process from detection, prediction, optimization to adjustment within seconds, without human intervention. The data flow at all three levels can be clearly seen on the large screen in the workshop control room, achieving transparent production management and greatly improving production efficiency and stability.

[0085] In some embodiments, the roll coating model software running on the roll coating model computer includes four functional modules that work in sequence:

[0086] The online detection module for coating roller diameter receives real-time images captured by the industrial camera and identifies and outputs real-time roller diameter data of the coating roller online through image processing algorithms.

[0087] The coating thickness prediction module receives various real-time process parameters collected by the unit PLC and the real-time roller diameter data, and calls a pre-trained artificial neural network model to output the predicted coating thickness.

[0088] The correlation analysis module calculates the correlation coefficient between each process parameter and the coating thickness based on historical data sequences, and selects the set of key process variables based on the absolute value of the coefficient.

[0089] The roller coating process optimization setting module performs combination optimization within the parameter space of the key process variable set based on the current real-time status of the unit and the target film thickness. It then uses the coating film thickness prediction module for evaluation to select the process parameter combination that meets the target film thickness and minimizes energy consumption, which is output as the optimized process parameter command.

[0090] The above embodiment decomposes the complex optimization task into four sequentially executed modules: "roller diameter detection → film thickness prediction → variable analysis → parameter optimization". Each module has a single responsibility and a clear output, forming a pipeline-like information processing flow. This design not only ensures that the software system has a clear structure, is easy to develop and maintain, but more importantly, it ensures that the optimization decision is based on accurate basic data (such as real-time roll diameter), scientific prediction, and key factor analysis. This makes the final optimization setting instructions both accurate and reliable, greatly improving the intelligence level and decision quality of the entire system.

[0091] For example, during normal operation of the unit, the coating roller may experience a slight reduction in diameter due to normal wear. The system needs to automatically compensate for the impact of this change on the coating thickness. If the change in roller diameter is not considered and the original parameters are continued to be used, the coating thickness will gradually deviate from the target value.

[0092] After adopting the solution of this invention, the modules cooperate sequentially:

[0093] The online detection module for coating roller diameter is activated first. It processes the image transmitted from the industrial camera, identifies that the current roller diameter has worn down from 300.0 mm to 299.5 mm, and outputs this new data.

[0094] The coating thickness prediction module then receives this new roller diameter data and, in conjunction with other real-time process parameters, uses a neural network model to predict that the current film thickness will be 3% thinner than the target value.

[0095] The correlation analysis module (based on historical data) has already determined that "coating roller pressure and strip pressure" and "liquid extraction roller speed ratio" are the two variables most sensitive to film thickness. The optimization module will primarily focus on optimizing these two variables.

[0096] The roller coating process optimization setting module calls the prediction model to calculate various pressure and speed ratio combinations, and quickly finds a new set of parameters that can pull the predicted film thickness back to the target value while minimizing energy consumption.

[0097] The four modules of this invention work automatically in sequence like an intelligent production line. Before the operator even notices the thickness deviation, the system has already automatically issued a fine-tuning command, perfectly offsetting the effects of roller wear. This achieves truly fully automatic and proactive control, ensuring continuous production and extremely stable product quality.

[0098] In some embodiments, the artificial neural network model used by the coating thickness prediction module is a three-layer network structure. The input parameters received by the input layer are determined according to the unit configuration and include at least the strip thickness, strip speed, strip tension, coating roll diameter, coating liquid characteristic parameters, and the speed ratio, gap, and pressure parameters between the metering roll, liquid taking roll, and coating roll in the three-roll coater, which are collected by the data acquisition and execution layer. The output layer outputs a single predicted coating thickness value. The artificial neural network model is trained with historical production data and has nonlinear mapping capabilities to achieve accurate prediction of coating thickness.

[0099] The above embodiments, by clearly defining the source (data acquisition and execution layer) and specific types (including strip characteristics, coating liquid characteristics, roll parameters, etc.) of the input parameters of the artificial neural network model, ensure that the model can comprehensively reflect the key influencing factors in the production process. At the same time, the model can be trained through historical data and has nonlinear mapping capabilities. This makes the model not a fixed formula, but an intelligent agent that can continuously learn and self-optimize from massive amounts of data from a specific production line.

[0100] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A method for optimizing process parameters of a three-roller coating machine, characterized in that, Includes the following steps: The process parameters and coating thickness data of the three-roller coating machine are collected in real time using industrial cameras, field sensors and online film thickness measuring instruments. Based on the collected data, the coating thickness is predicted using an artificial neural network model, and key process variables with a correlation greater than a preset threshold are identified through correlation analysis. With the goal of minimizing the target coating thickness and energy consumption, the key process variables are combined and optimized to generate optimized process parameter instructions, which are then sent to the three-roller coating machine for execution via the unit's PLC.

2. The method according to claim 1, characterized in that, The real-time acquired process parameter data includes: strip thickness, strip speed, cumulative strip coating length, strip tension, coating roll diameter, coating liquid composition, concentration and viscosity, speed ratio and gap between metering roll and liquid taking roll, pressure between liquid taking roll and coating roll, liquid taking roll speed ratio, coating roll direction and speed ratio, and pressure between coating roll and strip. The coating film thickness data is obtained through online detection by an online film thickness measuring instrument. The coating roll diameter is obtained by capturing real-time images of the coating roll with an industrial camera and detecting them online using image processing algorithms.

3. The method according to claim 1, characterized in that, Before using an artificial neural network model to predict coating thickness, the process includes sample generation and preprocessing steps: collecting historical process parameter data and corresponding coating thickness data to form an initial sample set; using at least one of the threshold method, angle distance method and robust regression method to remove abnormal data from the initial sample set, generating a three-roll coating sample set for model training.

4. The method according to claim 1, characterized in that, The artificial neural network model has a three-layer structure, including an input layer, a hidden layer, and an output layer. The input parameters of the input layer are the process parameter data, and the output parameters of the output layer are the predicted coating thickness. The artificial neural network model uses the Sigmoid function as the transformation function and is trained using the BP algorithm. During training, a portion of the three-roller coating sample set is used as training data, and the other portion is used as validation data.

5. The method according to claim 1, characterized in that, The correlation analysis calculates the correlation coefficient between each process parameter and the coating thickness using the following formula: Where x is the independent variable of the process parameter, and y is the dependent variable of the coating thickness. The mean of the independent variable. The mean of the dependent variable is given; the correlation coefficient R(t) is positive if positive and negative if negative; process parameters with an absolute value of the correlation coefficient greater than a preset threshold are selected as the key process variables.

6. The method according to claim 1, characterized in that, The key process variables include: the speed ratio and gap between the metering roller and the liquid taking roller, the pressure between the liquid taking roller and the coating roller, the speed ratio of the liquid taking roller, the direction and speed ratio of the coating roller, and the pressure between the coating roller and the strip.

7. The method according to claim 1, characterized in that, The optimization of key process variables specifically includes: using the trained artificial neural network model to calculate the predicted coating thickness under multiple combinations of key process variables; and selecting the combination with the minimum energy medium consumption from all combinations of predicted coating thicknesses that meet the target value as the optimal process parameter combination.

8. A system for optimizing and setting process parameters of a three-roller coating machine for implementing the method according to any one of claims 1-7, characterized in that, include: The data acquisition and execution layer includes field sensors, an online film thickness detector, an industrial camera, and a three-roller coating machine. The field sensors and the online film thickness detector are used to acquire unit operating parameters and coating film thickness data in real time. The industrial camera is used to acquire real-time images of the coating rolls. The three-roller coating machine is used to execute optimized process parameter instructions. The control and communication layer includes a unit PLC and an industrial Ethernet. The unit PLC connects to and controls the devices in the data acquisition and execution layer, and is used to collect various real-time data from the data acquisition and execution layer and issue control commands. The industrial Ethernet provides a communication backbone for the system. The optimization decision layer is a roller coating model computer, which communicates with the unit PLC and the industrial camera via the industrial Ethernet to receive the real-time data and images, and run the roller coating model software to calculate optimized process parameter instructions based on the real-time data, images and target film thickness, and then send them to the three-roller coating machine via the unit PLC.

9. The system according to claim 8, characterized in that, The roll coating model software running on the roll coating model computer includes four functional modules that work in sequence: The online detection module for coating roller diameter receives real-time images captured by the industrial camera and identifies and outputs real-time roller diameter data of the coating roller online through image processing algorithms. The coating thickness prediction module receives various real-time process parameters collected by the unit PLC and the real-time roller diameter data, and calls a pre-trained artificial neural network model to output the predicted coating thickness. The correlation analysis module calculates the correlation coefficient between each process parameter and the coating thickness based on historical data sequences, and selects the set of key process variables based on the absolute value of the coefficient. The roller coating process optimization setting module performs combination optimization within the parameter space of the key process variable set based on the current real-time status of the unit and the target film thickness. It then uses the coating film thickness prediction module for evaluation to select the process parameter combination that meets the target film thickness and minimizes energy consumption, which is output as the optimized process parameter command.

10. The system according to claim 9, characterized in that, The artificial neural network model used in the coating thickness prediction module has a three-layer network structure. The input parameters received by the input layer are determined according to the unit configuration and include at least the strip thickness, strip speed, strip tension, coating roll diameter, coating liquid characteristic parameters, and the speed ratio, gap, and pressure parameters between the metering roll, liquid taking roll, and coating roll in the three-roll coater, which are collected by the data acquisition and execution layer. The output layer outputs a single predicted coating thickness value. The artificial neural network model is trained with historical production data and has nonlinear mapping capabilities to achieve the prediction of coating thickness.

Images