A 3PE anticorrosion layer production process parameter control method based on an intelligent algorithm

Abstract

The present application belongs to the technical field of intelligent control, and particularly relates to a 3PE anticorrosion layer production process parameter control method based on intelligent algorithm. The method first collects historical data and real-time production data, and obtains a standardized data set through outlier rejection, missing value filling and data normalization processing. Then, an improved multi-objective optimization algorithm is used for training, taking yield rate and coating uniformity as the target, and outputting the first predicted process parameter containing the optimal parameter interval. Then, a process parameter dynamic graph model is constructed, and the node correlation entropy is calculated and the parameter adjustment is triggered according to the deviation degree of real-time parameters and predicted parameters, historical correlation coefficient between parameters and expert experience, and the optimized production parameters are obtained through multiple iterations. The present application can accurately determine and dynamically adjust the parameters, effectively handle the complex correlation between data, and improve the stability of 3PE anticorrosion layer production and product quality.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent control, and particularly relates to a 3PE anticorrosion layer production process parameter control method based on an intelligent algorithm. BACKGROUND

[0002] 3PE anticorrosion layer is a widely used anticorrosion structure for buried steel pipelines, which is composed of three layers of epoxy powder, adhesive and polyethylene. The epoxy powder layer provides good adhesion and corrosion resistance, the adhesive layer serves to connect the epoxy powder and polyethylene, enhancing the interlayer bonding force, and the polyethylene layer has excellent mechanical properties, weather resistance and corrosion resistance, which can effectively protect the pipeline from erosion by external factors such as soil, moisture and chemicals, prolonging the service life of the pipeline.

[0003] In the production process of 3PE anticorrosion layer, production process parameters have a crucial influence on product quality. However, the existing production process parameter control method has certain limitations. Traditional methods are difficult to comprehensively consider numerous complex process parameters and their mutual relationships, and cannot accurately determine the optimal production parameters, resulting in unstable product yield and coating uniformity. At the same time, in the face of parameter fluctuations in real-time production process, there is a lack of effective dynamic adjustment mechanism, which cannot optimize the parameters in time, affecting production efficiency and product quality. SUMMARY

[0004] The application proposes a 3PE anticorrosion layer production process parameter control method based on an intelligent algorithm to solve the technical problems in the above background technology.

[0005] To achieve the above purpose, the technical solution adopted by the application comprises the following steps:

[0006] S1, collecting current real-time production data of historical data in 3PE anticorrosion layer production;

[0007] S2, preprocessing the collected historical data, the preprocessing including outlier rejection, missing value filling and data normalization processing of the historical data, to obtain a standardized data set;

[0008] S3, training the standardized data set by using an improved multi-objective optimization algorithm, taking yield and coating uniformity as optimization objectives, and outputting first predicted process parameters containing parameter optimal interval;

[0009] S4. Construct a dynamic graph model of process parameters, encode the deviation between real-time production parameters and the first predicted process parameters as node features, label edge weights with historical correlation coefficients between parameters and expert experience, calculate node association entropy, and trigger parameter adjustment based on the deviation of core nodes in the graph, and spread adjustment instructions to neighboring nodes according to edge weights. Finally, after multiple rounds of adjustment iterations, the optimized production parameters are output.

[0010] The specific steps for obtaining the optimized production parameters in step S4 are as follows:

[0011] S41. First, collect production parameters in real time. Calculate the deviation from the first predicted process parameter. The deviation is calculated as follows: in It is to take the midpoint value of the first predicted process parameter interval;

[0012] S42. Select the top six nodes by association entropy and denote them as core nodes. When the deviation d of the core nodes... i When ≥0.15, adjust according to equipment accuracy. The convergence to the target interval yields the first-level adjustment parameter, calculated as follows: in This is represented as the first-level adjustment parameter, and the remaining adjusted parameters are obtained by spreading the adjustment instructions according to the edge weights;

[0013] S43. Perform constraint verification on the adjusted parameters. If the parameter value is within the physical range, use it directly. If it exceeds the limit, take the nearest boundary value to form the secondary adjustment parameter.

[0014] S44. After one round of adjustment, recalculate the core node deviation and perform iterative calculations until the deviation of all core nodes is equal to d. i Output all final optimized production parameters if the value is ≤0.1.

[0015] Preferably, the historical and production data in step S1 include preheating temperature, medium-frequency heating temperature, post-heating temperature, traction speed, coating thickness, adhesive layer ratio, epoxy powder coating particle size, epoxy powder coating curing temperature, epoxy powder coating speed, adhesive extrusion temperature, adhesive extrusion speed, adhesive extrusion temperature, polyethylene extrusion temperature, steel pipe speed, cooling time, and cooling temperature.

[0016] As a preferred embodiment, the specific implementation method of data preprocessing in step S1 includes outlier detection and missing value imputation;

[0017] The implementation method of the outlier detection is to construct an anomaly detection graph, for each parameter node, extract the directly associated nodes with edge weight ≥ 0.6 in the graph to form a local associated cluster; calculate the Mahalanobis distance of the real-time data and the historical mean in each associated cluster, set a threshold, and if the calculated Mahalanobis distance threshold is greater than the threshold, it is determined that the parameter is abnormal; all node data in the associated cluster determined to be parameter abnormal are excluded;

[0018] The missing value filling is to fill the data position after excluding the abnormal values, and the bidirectional strong associated nodes are screened according to the edge weight in ascending order. The bidirectional strong associated nodes are obtained by starting from the missing value parameter, first screening the parameters having direct causal effect on the missing value parameter, taking the top three in descending order of edge weight, then screening the parameters affected by the direct feedback of the missing value parameter, taking the top three in descending order of edge weight, forming a set of six strong associated nodes including the cause and effect and feedback, and finally detecting whether the filling value obtained by interpolation meets the process constraints. If the filling value is within a reasonable range, it is directly used, if it exceeds the upper limit, the maximum value boundary is taken, and if it is lower than the lower limit, the minimum value boundary is taken.

[0019] As preferred, the step S3 adopts a modified multi-objective optimization algorithm to train the standardized data set, taking the yield and coating uniformity as optimization objectives, and the specific implementation of outputting the first predicted process parameter containing the optimal parameter interval is:

[0020] S31, first construct a deep solution generation network, input the historical standardized parameters, and output the initial candidate solution, so that the initial population distribution preferentially covers the historical efficient area, and the calculation method of the initial candidate solution is: init = clip(σ(W2σ(W1X std +b1)+b2)⊙(x max -x min )+x min ,x min ,x max ), wherein W2, W1 represent weight matrices, b1, b2 represent bias vectors, σ represents an activation function, x max ,x min are the maximum and minimum values of the physical interval of the process parameters, respectively, ⊙ represents element-wise multiplication, and clip is an interval clipping function;

[0021] S32, set an adaptive weight network, input the target statistical characteristics of the current population, and output the dynamic weight ω(t), and the calculation method of the dynamic weight output is: wherein Ω represents an adaptive weight matrix, σ Y ,σ U are the product yield standard deviation and the coating uniformity standard deviation, respectively, and σ YU represents the correlation of the two objectives.

[0022] S33, constructing a target function according to the dynamic weight, the target function being specifically: f(x) = ω Y ·(1-Y(x))+ω Y ·U(x)+λ·D HD (x), wherein Y(x) is a yield prediction value of the parameter combination x, U(x) represents a coating uniformity prediction value, D HD (x) represents a diversity measure based on Hausdorff distance, representing the spatial distribution difference between the solution x and other solutions in the population, and λ represents a diversity adjustment coefficient;

[0023] S34, then generating a non-traditional parameter combination x gen based on the target value (Y, U), which is specifically realized by using a generative adversarial network, wherein a generator G inputs noise z and a conditional target output x gen =G(z,Y * ,U * ), wherein Y * , U * represent the expected target yield and target coating uniformity; a discriminator D distinguishes between real solutions and generated solutions, and the optimization target is: wherein P real represents a real data distribution, P (z) represents a probability distribution of noise z,

[0024] S35, finally determining a continuous interval of the first 80% of the density based on the Pareto frontier solution by kernel density estimation, eliminating sub-intervals that violate process hard constraints, verifying by 1000 times Monte Carlo simulation that intervals with a target fluctuation of ≤5% are retained, and outputting the first predicted process parameter containing the optimal parameter interval.

[0025] As a preferred, for the generated solution in the step S34, every 10 generations will replace the solution with the smallest crowding distance in the population with a 30% proportion of the generated solution, so as to make the algorithm explore the low probability parameter region.

[0026] As a preferred, the training loss function of the deep solution generation network in the step S31 is: wherein MSE is a mean square error, measuring the difference between solutions, this term represents the repulsive force between the generated solution and the existing solution, avoiding the occurrence of repeated solutions.

[0027] As preferred, the edge weight is marked by the historical correlation coefficient between parameters and expert experience in the step S4, and the specific implementation manner of calculating the correlation entropy is that, first, the Pearson correlation coefficient p between parameters is calculated, and the prior knowledge of the edge by the process expert is introduced to obtain an expert correction coefficient d, the final edge weight is obtained by averaging the two, the correlation entropy is calculated by calculating the weight, and the calculation manner of the correlation entropy is: Wherein, N(i) is a neighbor node set, and β ij The edge weight between node i and node j.

[0028] Compared with the prior art, the application has the advantages and positive effects that:

[0029] 1. The abnormal value is identified by constructing an anomaly detection graph, the abnormal correlation cluster data is removed by calculating the Mahalanobis distance, the accuracy of the data is ensured, the missing values are filled by screening the bidirectional strong correlation nodes according to the graph edge weight, the rationality is ensured by process constraint verification, the data strong coupling correlation is effectively processed, and reliable data basis is provided for subsequent analysis and optimization.

[0030] 2. The deep solution generation network is constructed, the initial candidate solution covers the historical efficient area, and the quality of the optimization starting point is improved; the adaptive weight network outputs dynamic weight, the weight is adjusted according to the target statistical characteristics, the yield and the coating uniformity are balanced; the non-traditional parameter combination is generated by combining the adversarial network, the low-probability parameter area is explored, and the local optimum is avoided; the precise parameter optimal interval is determined based on the Pareto front solution screening, the kernel density estimation, the removal of the illegal interval and the Monte Carlo simulation verification.

[0031] 3. The deviation degree of real-time parameters and predicted parameters is set as a node feature, the correlation entropy is calculated by combining the historical correlation coefficient and the edge weight marked by expert experience; the deviation degree of the core node is triggered to adjust, the instruction is diffused to the neighbor nodes according to the edge weight, the production parameters are optimized by multiple iterations, the parameter fluctuation in real-time production is responded in time, and the complex parameter coupling relationship is adapted. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical scheme of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0033] Figure 1 It is a structure flow schematic diagram of a 3PE anti-corrosion layer production process parameter control method based on intelligent algorithm. DETAILED DESCRIPTION

[0034] In order to enable more clearly understanding the above-mentioned purposes, features and advantages of the present application, the present application is further explained below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0035] In the following description, a lot of specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can be practiced in other manners different from those described herein, therefore, the present application is not limited to the specific embodiments disclosed in the following description.

[0036] Embodiment, in order to solve the problem that the optimal production parameter cannot be accurately determined in the 3PE anticorrosion layer production line, the present application provides a 3PE anticorrosion layer production process parameter control method based on intelligent algorithm, through data preprocessing, multi-objective optimization algorithm and dynamic graph model, the accurate control and dynamic adjustment of the parameter are realized, the production stability and product quality are improved, and the overall process is as shown in Figure 1

[0037] Firstly, the current real-time production data of the historical data in the 3PE anticorrosion layer production is collected; the historical data and the production data include preheating temperature, intermediate frequency heating temperature, post-heating temperature, traction speed, coating thickness, adhesive layer ratio, epoxy powder spraying powder particle size, epoxy powder spraying powder curing temperature, epoxy powder spraying speed, adhesive extrusion temperature, adhesive extrusion speed, adhesive extrusion temperature, polyethylene extrusion temperature, steel pipe rotating speed and cooling time and cooling temperature. There is a correlation and mutual influence between various data, and there is a strong coupling correlation between the data.

[0038] ​Considering that there is strong coupling between data, such as the influence of preheating temperature on the melting state of epoxy powder, which together with the intermediate frequency heating temperature determines the adhesion of the coating; the coupling of traction speed and polyethylene extrusion temperature affects the uniformity of the coating thickness, and the coupling of the adhesive extrusion speed and the adhesive layer ratio determines the interlayer bonding force. Parameter fluctuations are transmitted through a physical and chemical reaction chain, and a single parameter anomaly may trigger multiple index degradation, making it difficult for traditional methods to quantify complex correlations, and requiring the construction of a parameter network through a graph model to mine causal and feedback relationships and provide correlation basis for intelligent control. The specific implementation method of the data preprocessing of the application includes outlier detection and missing value filling; the implementation method of the outlier detection is to construct an anomaly detection graph, for each parameter node, extract the directly associated nodes in the graph with an edge weight greater than or equal to 0.6 to form a local association cluster; calculate the Mahalanobis distance between the real-time data and the historical mean in each association cluster, set a threshold, and if the calculated Mahalanobis distance threshold is greater than the threshold, it is determined that the parameter is abnormal; all node data in the association cluster determined to be parameter abnormal are excluded; the missing value filling is to fill the data position after excluding the abnormal values, and the missing position is screened according to the order of the edge weight of the bidirectional strong association nodes, the bidirectional strong association nodes are obtained according to the weight of the structure edge in the graph, starting from the missing value parameter, first selecting the parameters that have a direct causal effect on it, taking the top three in descending order of edge weight, then selecting the parameters that are directly affected by it, taking the top three in descending order of edge weight, forming a set of six strong association nodes containing causality and feedback, and finally detecting whether the interpolated filling value meets the process constraints, if the filling value is within a reasonable range, it is directly used, if it exceeds the upper limit, the maximum value boundary is taken, and if it is lower than the lower limit, the minimum value boundary is taken.

[0039] After processing the data, considering that traditional single-objective optimization cannot balance yield and coating uniformity, and the initial solution distribution is random, leading to low optimization efficiency and easy falling into local optimum. The improved multi-objective optimization algorithm is used to train the standardized data set, taking yield and coating uniformity as optimization objectives, and outputting the first predicted process parameter containing the optimal parameter interval. Specifically, first, a deep solution generation network is constructed, and the training loss function of the deep solution generation network is: Where MSE is the mean square error, which measures the difference between the solutions, This term represents the repulsive force between the generated solution and the existing solution, avoiding duplicate solutions. Then input the historical standardized parameters, output the initial candidate solution, and make the initial population distribution preferentially cover the historical high-efficiency area, the calculation method of the initial candidate solution is: init = clip(σ(W2σ(W1X std +b1)+b2)⊙(x max -x min )+x min ,x min ,x max), wherein W2, W1 represent weight matrix, b1, b2 represent bias vector, σ represents activation function, x max ,x min are maximum and minimum values of the physical interval of the process parameters respectively, ⊙ represents element-by-element multiplication, and clip is an interval clipping function; an adaptive weight network is set, target statistical characteristics of the current population are input, and a dynamic weight ω(t) is output, and the calculation manner of the dynamic weight output is: wherein Ω represents an adaptive weight matrix, σ Y ,σ U are product yield standard deviation and coating uniformity standard deviation respectively, σ YU represents the correlation of the two objectives; a target function is constructed according to the dynamic weight, and the target function is specifically: f(x) = ω Y ·(1-Y(x))+ω Y ·U(x)+λ·D HD (x), wherein Y(x) is a yield prediction value of the parameter combination x, U(x) represents a coating uniformity prediction value, D HD (x) represents a diversity measure based on Hausdorff distance, representing the spatial distribution difference between the solution x and other solutions in the population, and λ represents a diversity adjustment coefficient; then, a non-traditional parameter combination x gen is generated under the condition of the target values (Y, U), and the specific implementation is to use a generative adversarial network, a generator G inputs noise z and conditional target to output x gen =G(z,Y * ,U * ), wherein Y * ,U * represent the expected target yield and target coating uniformity; a discriminator D distinguishes between real solutions and generated solutions, wherein every 10 generations, the generated solutions are used to replace the solution with the smallest crowding distance in the population at a proportion of 30%, so that the algorithm explores the low-probability parameter region. The optimization target is: wherein P real represents a real data distribution, P (z) represents a probability distribution of noise z, and finally, based on the Pareto front solution, the continuous interval of the first 80% of the density is determined through kernel density estimation, the sub-interval violating the process hard constraint is removed, the interval with target fluctuation ≤5% is reserved through 1000 times of Monte Carlo simulation verification, and the first predicted process parameter containing the optimal parameter interval is output

[0040] After obtaining the predicted parameters, the deviation of the real-time production parameters from the first predicted process parameters is encoded as a node feature to construct a process parameter dynamic graph model, the historical correlation coefficient between parameters is combined with expert experience to label the edge weight, the node correlation entropy is calculated, the parameter adjustment is triggered based on the deviation of the core node in the graph, the adjustment instruction is diffused to the neighbor nodes according to the edge weight, and finally the optimized production parameters are output after multiple rounds of adjustment iteration. Specifically, first, the production parameters are collected in real time The deviation of the first predicted process parameters is calculated, and the deviation calculation method is as follows: Wherein is the midpoint value in the interval of the first predicted process parameters; the nodes ranked in the top six of the correlation entropy are recorded as core nodes, and when the deviation d of the core node is greater than or equal to 0.15, the target interval converges according to the equipment adjustment accuracy i The first-order adjustment parameter is obtained, and the calculation method is as follows: Wherein is the first-order adjustment parameter, and the remaining adjusted parameters are obtained by diffusing the adjustment instruction according to the edge weight; the adjusted parameters are subjected to constraint verification, if the parameter value is within the physical interval, it is directly used, if it is out of range, the nearest boundary value is taken to form the second-order adjustment parameter; the deviation of the core node is recalculated after one round of adjustment, and iterative operation is performed until the deviation d of all core nodes is less than or equal to 0.1, and all final optimized production parameters are output. i

[0041] The above is only a preferred embodiment of the present application, and is not intended to limit the other forms of the present application, any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification of the above embodiments without departing from the technical solution content of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.​​

Claims

1. A method for controlling process parameters in the production of 3PE anti-corrosion coatings based on intelligent algorithms, characterized in that, Includes the following steps: S1. Collect historical data and current real-time production data from the production of 3PE anti-corrosion coating; S2. Preprocess the collected historical data, including outlier removal, missing value filling, and data normalization, to obtain a standardized dataset. S3. The standardized dataset is trained using an improved multi-objective optimization algorithm, with yield rate and coating uniformity as optimization objectives, and the first predicted process parameters containing the optimal parameter range are output. S4. Construct a dynamic graph model of process parameters, encode the deviation between real-time production parameters and the first predicted process parameters as node features, label edge weights with historical correlation coefficients between parameters and expert experience, calculate node association entropy, and trigger parameter adjustment based on the deviation of core nodes in the graph, and spread adjustment instructions to neighboring nodes according to edge weights. Finally, after multiple rounds of adjustment iterations, the optimized production parameters are output. The specific steps for obtaining the optimized production parameters in step S4 are as follows: S41. First, collect production parameters in real time. Calculate the deviation from the first predicted process parameter. The deviation is calculated as follows: in It is to take the midpoint value of the first predicted process parameter interval; S42. Select the top six nodes by association entropy and denote them as core nodes. When the deviation d of the core nodes... i When ≥0.15, adjust according to equipment accuracy. The convergence to the target interval yields the first-level adjustment parameter, calculated as follows: in This is represented as the first-level adjustment parameter, and the remaining adjusted parameters are obtained by spreading the adjustment instructions according to the edge weights; S43. Perform constraint verification on the adjusted parameters. If the parameter value is within the physical range, use it directly. If it exceeds the limit, take the nearest boundary value to form the secondary adjustment parameter. S44. After one round of adjustment, recalculate the core node deviation and perform iterative calculations until the deviation of all core nodes is equal to d. i Output all final optimized production parameters if the value is ≤0.

1.

2. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 1, characterized in that, The historical and production data in step S1 include preheating temperature, medium-frequency heating temperature, post-heating temperature, traction speed, coating thickness, adhesive layer ratio, epoxy powder coating particle size, epoxy powder coating curing temperature, epoxy powder coating speed, adhesive extrusion temperature, adhesive extrusion speed, adhesive extrusion temperature, polyethylene extrusion temperature, steel pipe speed, cooling time, and cooling temperature.

3. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 1, characterized in that, The specific implementation method of data preprocessing in step S1 includes outlier detection and missing value imputation. The outlier detection method involves constructing an anomaly detection graph. For each parameter node, directly related nodes with edge weights ≥ 0.6 are extracted to form local association clusters. Within each association cluster, the Mahalanobis distance between the real-time data and the historical mean is calculated, and a threshold is set. If the calculated Mahalanobis distance threshold is greater than the threshold, the parameter is considered an anomaly. All node data within the association clusters identified as having parameter anomalies are removed. The missing value imputation involves filling the data positions after outliers have been removed. For missing positions, bidirectional strongly correlated nodes are selected in ascending order of graph edge weights. These bidirectional strongly correlated nodes are determined based on the weights of the structural edges in the graph. Starting with the missing value parameters, the first step is to select parameters that have a direct causal impact, and then select the top three in descending order of edge weight. Next, the parameters that are directly affected by the feedback are selected, and the top three in descending order of edge weight are also selected. This forms a set of six strongly correlated nodes, including both causal and feedback nodes. Finally, the imputed values ​​obtained by interpolation are checked to see if they meet the process constraints. If the imputed values ​​are within a reasonable range, they are used directly. If they exceed the upper limit, the maximum value boundary is used. If they are below the lower limit, the minimum value boundary is used.

4. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 1, characterized in that, Step S3 uses an improved multi-objective optimization algorithm to train the standardized dataset, with yield rate and coating uniformity as optimization objectives, and outputs the first predicted process parameters containing the optimal parameter range. The specific implementation of this method is as follows: S31. First, construct a deep solution generation network, input the standardized parameters of the history, and output initial candidate solutions, so that the initial distribution of the population preferentially covers the historically efficient region. The calculation method of the initial candidate solutions is: x init =clip(σ(W2σ(W1X)) std +b1)+b2)⊙(x max -x min )+x min ,x min ,x max ), where W2 and W1 represent weight matrices, b1 and b2 represent bias vectors, σ represents the activation function, and x max ,x min These represent the maximum and minimum values ​​of the physical range of the process parameters, respectively. ⊙ represents element-wise multiplication, and clip is the range clipping function. S32. Set up an adaptive weighted network, input the target statistical features of the current population, and output dynamic weights ω(t). The calculation method for the dynamic weight output is as follows: Where Ω represents the adaptive weight matrix, σ Y ,σ U These are the standard deviations of product quality and coating uniformity, respectively, σ YU Represents the correlation between two objectives; S33. Construct an objective function based on dynamic weights. The specific objective function is: f(x) = ω Y ·(1-Y(x))+ω Y ·U(x)+λ·D HD (x), where Y(x) is the predicted yield value of parameter combination x, U(x) represents the predicted coating uniformity value, and D HD (x) represents a diversity measure based on Hausdorff distance, characterizing the spatial distribution difference between solution x and other solutions in the population, and λ represents the diversity adjustment coefficient; S34. Then, using the target value (Y,U) as a condition, generate a non-traditional parameter combination x. gen Specifically, this is achieved by using an adversarial network, where the generator G takes noise z as input and a conditional target output x as output. gen =G(z,Y) * U * ), where Y * U * The target yield and target coating uniformity are represented by the desired values; the discriminator D distinguishes between the true solution and the generated solution, and the optimization objective is: Among them, P real P represents the true data distribution. (z) This represents the probability distribution of the noise z. S35. Finally, based on the Pareto front solution, the top 80% of the continuous intervals in terms of density are determined through kernel density estimation. Subintervals that violate the hard process constraints are eliminated. After verification through 1000 Monte Carlo simulations, the intervals with target fluctuations ≤5% are retained, and the first predicted process parameters containing the optimal parameter intervals are output.

5. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 4, characterized in that, For the generated solutions in step S34, every 10 generations the generated solutions replace the solutions with the smallest crowding distance in the population at a ratio of 30%, allowing the algorithm to explore regions with low probability parameters.

6. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 4, characterized in that, The training loss function of the deep solution generation network in step S31 is: MSE, or mean squared error, measures the difference in solutions. This term represents the repulsive force between the generated solution and the existing solution, thus avoiding duplicate solutions.

7. The method for controlling the production process parameters of 3PE anti-corrosion coating based on intelligent algorithms according to claim 1, characterized in that, In step S4, the specific implementation of calculating the association entropy by combining the historical correlation coefficient between parameters with expert experience to label edge weights is as follows: First, the Pearson correlation coefficient ρ between parameters is calculated, and the prior knowledge of process experts regarding the edges is introduced to obtain the expert correction coefficient δ. The two are averaged to obtain the final edge weights, and the association entropy is calculated based on the weights. The method for calculating the association entropy is as follows: Where N(i) is the set of neighboring nodes, β ij This represents the edge weight from node i to node j.

Images