A method for screening a multifunctional anticorrosive and antifouling coating formula for a trash screen

Abstract

This invention belongs to the field of coating material formulation optimization technology, and discloses a method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris-blocking nets. The steps are as follows: Step 1, collect core data for preprocessing, and use seed-guided neural metric learning to achieve deep fusion of multimodal data on "coating components-performance-environment-failure"; Step 2, construct an LAM-GRU encoder to output multidimensional feature vectors of the coating. E Step 3: Construct a multi-task performance prediction network to... E The system takes four performance metrics as input and outputs predicted values ​​for each. A loss function is constructed to assign differentiated weights to these four metrics. The Adam optimizer is then used to optimize the parameters of the multi-task performance prediction network. Step 4 involves efficient optimization of the coating formulation using HGWSFO. Step 5 involves constructing a weighted multi-objective optimization function, iterating through the HGWSFO, and outputting the optimal formulation that satisfies the bundle conditions and maximizes the objective function. This invention offers high performance prediction accuracy and combines high formulation optimization efficiency with high reliability.

Description

A method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris-blocking nets Technical Field

[0001] This invention relates to the field of coating material formulation optimization technology, and in particular to a method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris barriers. Background Technology

[0002] As a key piece of equipment in marine and hydraulic engineering, debris barriers have long been in use in complex marine environments characterized by high salinity, high humidity, and abundant biodiversity. They face severe corrosion and biofouling problems, leading to shortened equipment lifespan, increased maintenance costs, and even disruption to normal project operations. Therefore, developing a multifunctional coating that combines excellent corrosion and fouling resistance with good mechanical stability is the core solution to these problems.

[0003] Current research and development of anti-fouling net coating formulations mainly relies on traditional experimental trial-and-error methods, requiring repeated adjustments to component ratios and extensive performance testing. This approach suffers from drawbacks such as long development cycles, high testing costs, and significant blind spots in formulation optimization. With the development of materials genome engineering and artificial intelligence technologies, some studies have attempted to predict coating performance using machine learning models, but several shortcomings remain: First, multimodal data fusion is poor, making it difficult to effectively integrate heterogeneous data such as coating components, service environment, performance indicators, and failure cases, resulting in incomplete feature representation. Second, traditional encoders lack the ability to memorize and correlate local features, failing to capture the complex mapping relationship between subtle changes in component content, fluctuations in environmental parameters, and coating performance, thus limiting prediction accuracy. Third, formulation optimization algorithms suffer from an imbalance between exploration and utilization; single global search algorithms are prone to getting trapped in local optima, while single local search algorithms struggle to cover a broad formulation space, leading to low efficiency and insufficient reliability in selecting the optimal formulation.

[0004] Furthermore, existing technologies lack specific formulation screening schemes for the service scenarios of debris-blocking nets, failing to fully consider key factors such as the physicochemical properties of the marine environment and the distribution of biological communities, making it difficult to guarantee the long-term effectiveness of the coating during actual service. Therefore, there is an urgent need for a technical solution that can efficiently integrate multimodal data, accurately predict coating performance, and quickly screen the optimal formulation to overcome the current technological bottlenecks. Summary of the Invention

[0005] The purpose of this invention is to address the technical deficiencies in the existing technology by providing a method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris-blocking nets.

[0006] The technical solution adopted to achieve the purpose of this invention is:

[0007] A method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris-blocking nets includes the following steps:

[0008] Step 1: Collect core data to build a database, perform preprocessing to obtain the preprocessed database, and use seed-guided neural metric learning to achieve deep fusion of multimodal data of "coating composition-performance-environment-failure". By screening representative seed samples, a seed sample set with similarity labels is generated, and a similarity matrix S is constructed.

[0009] Step 2: Construct a grid-based memory tensor M to store the local feature information of the seed samples. The local features of the similarity matrix S are preferentially written into the grid-based memory tensor M. Construct an LAM-GRU encoder with coating component vectors. Environment vectors The corresponding grid coordinates in the grid-based memory tensor M The input is the coating multidimensional feature vector E, and the output is the coating multidimensional feature vector E with dimension d.

[0010] Step 3: Construct a multi-task performance prediction network. Take the multi-dimensional feature vector E of the coating as input and output the predicted values ​​of four types of performance indicators. Construct a loss function to assign differentiated weights to the four types of performance indicators and use the Adam optimizer to optimize the parameters of the multi-task performance prediction network.

[0011] Step 4: Efficiently optimize coating formulations using HGWSFO: Dynamically switch between gray wolf-optimized local search and sunflower-optimized global search by using index thresholds;

[0012] Step 5: Set the optimization objective and constraints, construct a weighted multi-objective optimization function, and continuously iterate HGWSFO to output the optimal formula that satisfies the constraints and maximizes the objective function.

[0013] In the above technical solution, in step 1, the data in the database includes coating formulation component data, performance data, environmental data, and failure data; the preprocessing includes data standardization, abnormal data removal, and data completion and enhancement.

[0014] In the above technical solution, in step 1, the similarity matrix , This refers to the number of representative coating formulation components randomly sampled from the pre-processed database.

[0015] In the above technical solution, in step 2, the grid-based memory tensor P and Q represent the number of grid divisions for the component or environmental parameter in the first and second dimensions, respectively.

[0016] In the above technical solution, step 2, which outputs the multidimensional feature vector E of the coating through the read / write operation of the LAM-GRU encoder, includes the following specific steps:

[0017] (1) Reading operation: based on coating component vector or environment vector The corresponding grid coordinates in the grid-based memory tensor M Scan the neighborhood grid of the local bandwidth w, and apply attention weights. Fusion grid-based memory tensor Historical characteristics in the process of generating historical states ;

[0018] (2) Write operation: using the bureau department Control the update intensity and the current feature state The weighted update to the memory tensor is calculated using the following formula:

[0019] ;

[0020] For bureau departments; This represents the current feature state; It is the sigmoid activation function; grid coordinates The corresponding original feature values; grid coordinates The corresponding updated feature values;

[0021] (3) The final hidden layer output of the LAM-GRU encoder is used as the multidimensional feature vector E to realize the unified representation of multimodal data;

[0022] ;

[0023] This is the weight matrix. For bias vectors, This is the final time step after the LAM-GRU encoder has traversed all input sequences. The hidden layer state, This enhances the nonlinear representation of features for activation functions.

[0024] In the above technical solution, the formula for the multi-task performance prediction network in step 3 is expressed as follows:

[0025] ;

[0026] For the predicted value of the k-th type of performance index, These correspond to four performance categories: corrosion resistance, antifouling performance, mechanical properties, and environmental adaptability. These are the weight matrix and bias vector of the fully connected layer 1 of the branch, respectively. For the hidden layer dimension; These are the weight matrix and bias vector of the branch output layer, respectively. The number of metrics for the k-th performance category; For the hidden layer activation function, This is the activation function for the branch output layer.

[0027] In the above technical solution, the loss function in step 3 is: ;

[0028] To calculate the weighted total loss across multiple tasks, For the first Class performance weights For the first Class performance The true value of each indicator For the first Class performance The predicted values ​​of each indicator.

[0029] In the above technical solution, in step 4, an index threshold is set. When the index value of the candidate recipe set When the index value of the recipe candidate set is reached, SFO is initiated to perform a global exploration. At that time, GWO will be launched to perform a local search.

[0030] In the above technical solution, the specific steps of SFO global exploration are as follows:

[0031] (1) Initialize the "Sunflower" population and set the coating formula with the best fitness to "Sun". ;

[0032] (2) Calculate the step size based on the inverse square law: ,in This is the maximum step size in the SFO stage, used to control the maximum magnitude by which the candidate formulation moves towards the "sun" formulation. These represent the upper and lower limits of the content of each component. For sunflower population size;

[0033] (3) Generate new formulas through pollination operations Remove "dead plants" with poor adaptability and update the "sun". ;

[0034] The specific steps of GWO local search are as follows:

[0035] (1) Initialize the "gray wolf" population and set the top three coating formulas with the highest fitness as leaders. Secondary leaders Executor ;

[0036] (2) Calculate the coefficient vector: the first coefficient vector of the GWO stage The second coefficient vector of the GWO stage , The coefficient decay factor for the GWO stage. The value decreases linearly from 2 to 0, rand1, All Random numbers within the interval; the first coefficient vector B is used to control whether the candidate formulation moves closer to or further away from the high-quality formulation; the second coefficient vector... Used to enhance search randomness;

[0037] (3) Calculate the normal recipe after the GWO stage update. .

[0038] In the above technical solution, in step 5, the weighted multi-objective optimization function is: ;

[0039] in:

[0040] To optimize the function value for weighted multi-objectives; Weighting of the effective period of anti-fouling treatment; The weight for bioattachment coverage; The bio-attachment coverage rate is 28 days. Weighting for wear resistance; Weighting for material costs; For wear resistance; The wear resistance constraint threshold; Cost of coating materials per square meter; This is the threshold for material cost constraints.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] 1. Excellent multimodal data fusion effect: The seed-guided neural metric learning technology designed in this invention can effectively integrate multi-source heterogeneous data such as components, performance, environment, and failure. It establishes data association through seed sample similarity matrix, avoids the limitations of single data dimension, and provides a comprehensive and reliable data foundation for coating feature characterization. The integrity and effectiveness of feature fusion are significantly better than traditional methods.

[0043] 2. High performance prediction accuracy: The Local Attention Memory (LAM) enhanced encoder stores local features and historical information through memory tensors. Combined with the dynamic adjustment of attention weights, it can accurately capture the complex mapping relationship between component content changes, environmental parameter fluctuations and coating performance. The multi-task performance prediction network simultaneously covers four core performance indicators. The weighted loss function optimizes the training priority, significantly reducing prediction errors. The prediction accuracy of various performance indicators is improved by more than 30% compared with existing technologies.

[0044] 3. Efficiency and Reliability in Formula Optimization: The Hybrid Gray Wolf-Sunflower Optimization Algorithm (HGWSFO) dynamically switches between global exploration and local search, ensuring comprehensive coverage of the formula space while accurately identifying high-quality formula regions, avoiding getting trapped in local optima. The algorithm can output the optimal candidate formula in just 50 iterations, improving screening efficiency by 50%-80% compared to traditional optimization algorithms. Furthermore, the selected formulas meet the comprehensive requirements of "high corrosion resistance, strong anti-fouling properties, excellent mechanical properties, and low cost."

[0045] 4. Reduced R&D costs and shorter cycle: This invention replaces a large number of repetitive experiments with AI technology, reducing the number of physical tests and lowering testing costs by more than 60%; the entire formula screening process is automated, shortening the R&D cycle from the traditional 3-6 months to 2-4 weeks, significantly improving R&D efficiency;

[0046] 5. Strong adaptability and practicality: Specifically designed for the marine service environment of debris-blocking nets, it incorporates real-time and historical environmental data from typical sea areas. The selected formula can accurately match the physicochemical properties and biological distribution characteristics of the marine environment, extending the antifouling effectiveness of the coating to more than 24 months, with a salt spray corrosion resistance time of ≥1000h. The mechanical properties meet the requirements of engineering applications, and the service life is 2-3 times longer than that of traditional coatings, demonstrating significant engineering application value and economic and social benefits. Attached Figure Description

[0047] Figure 1 is a flowchart of the method for screening the formulation of multifunctional anti-corrosion and anti-fouling coating for debris-blocking nets.

[0048] Figure 2 is a performance comparison chart of different algorithms. Detailed Implementation

[0049] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Embodiments

[0050] As shown in Figure 1, a method for screening the formulation of a multifunctional anti-corrosion and anti-fouling coating for a debris barrier includes the following steps:

[0051] Step 1: Collect core data to build a database, perform preprocessing to obtain the preprocessed database, and use seed-guided neural metric learning technology to achieve deep fusion of multimodal data on "coating composition-performance-environment-failure". By selecting representative seed samples, generate a seed sample set with similarity labels and construct a similarity matrix. This provides high-quality data support for subsequent feature extraction and model training, and solves the problem of the difficulty in effectively integrating multi-source heterogeneous data.

[0052] Step s11: Collect core data and build a database.

[0053] 1.1 Data collection of coating formulation components:

[0054] Collect more than 20 categories of commonly used coating components, including resins (epoxy resins, polyurethanes, fluorocarbon resins, etc.), rust-preventing pigments (zinc powder, zinc phosphate, aluminum tripolyphosphate, etc.), antifouling agents (copper-based compounds, environmentally friendly isothiazolinones, tetrabutyltin oxide, etc.), and additives (dispersants, leveling agents, defoamers, etc.). Record the basic parameters of each component (including molecular weight, number of functional groups, density, melting point, and solubility parameters) and the content ratio range in industrial applications (e.g., resin 30%-60%, rust-preventing pigment 15%-35%, antifouling agent 5%-15%, and additive 1%-5%) to form a basic component database.

[0055] 1.2 Performance Data Acquisition:

[0056] Through literature retrieval, historical experimental records, and supplementary experiments, data on 12 key performance indicators of the coating in four main categories were obtained:

[0057] (1) Corrosion resistance: Specific performance indicators include polarization resistance (test range) Salt spray test corrosion resistance time (test cycle 0-2000h), electrochemical impedance spectroscopy (low-frequency impedance value, phase angle);

[0058] (2) Antifouling performance: Specific performance indicators include 28-day marine organism (diatoms, barnacles, oysters) attachment coverage (0%-100%), antifouling effectiveness (0-36 months), and antifouling agent leaching rate. ;

[0059] (3) Mechanical properties: Specific performance indicators include adhesion (cross-cut test / pull-off test, 0-10MPa), abrasion resistance (Taber abrasion index, mg / 1000 rpm), flexibility (bending test diameter, mm), and hardness (pencil hardness / HV hardness).

[0060] (4) Environmental adaptability: Specific performance indicators include the performance retention rate after high and low temperature cycling (-40℃-80℃) and the performance degradation rate after damp heat aging (40℃, 95%RH).

[0061] 1.3 Environmental Data Acquisition:

[0062] Access to real-time and historical environmental data from five typical marine areas where pollution interception nets are deployed nationwide:

[0063] (1) Physicochemical environmental parameters: seawater temperature (-2℃~30℃), salinity (28‰~35‰), pH value (7.5~8.5), dissolved oxygen (2~12mg / L), turbidity (1~50NTU), seawater flow velocity (0~2m / s);

[0064] (2) Biological environmental parameters: planktonic abundance (ind / m³), benthic coverage (%), seawater nutrient concentration (nitrogen, phosphorus, silicon, mg / L).

[0065] (3) Extreme environmental data: seawater impact force during typhoon season, tidal range, and seawater corrosion potential (mV).

[0066] 1.4 Failure Data Acquisition:

[0067] Failure cases of different coating formulations during actual service were collected as negative samples for the model. Specifically, v includes:

[0068] (1) Failure types: coating blistering (blister density, diameter), cracking (crack length, width), peeling (peeling area percentage), massive biological adhesion (time point when adhesion amount ≥ 50%), antifouling agent depletion (antifouling failure time).

[0069] (2) Failure-related data: Record the environmental conditions (temperature, salinity, biodiversity) when the failure occurs, the service life of the coating, and the early performance degradation trend to establish a "formulation-environment-failure" related dataset.

[0070] Step s12, data preprocessing, to obtain the preprocessed database:

[0071] (1) Data standardization: The Z-Score standardization method is used to eliminate the influence of different metrics, so that all parameters are on the same order of magnitude, which facilitates model training;

[0072] (2) Abnormal data removal: The isolated forest algorithm is used to isolate the data by constructing multiple isolated trees, calculate the abnormal score of each sample, and remove abnormal data with scores higher than the threshold (default 0.8) (such as abnormally high / low polarization resistance due to experimental error, negative biological attachment rate, etc.).

[0073] (3) Data completion and enhancement: K-nearest neighbor (KNN) interpolation is used to complete missing data, and SMOTE (synthetic minority class oversampling technique) is used to generate virtual samples for small sample data to ensure the balanced distribution of various formulations, performance and environmental data in the database.

[0074] Step s13, Filtering Using seed samples, construct a similarity matrix. .

[0075] Random sampling from the preprocessed database (default) Representative coating formulations were used as seed samples. The comprehensive similarity between seed samples based on "component-performance-environment" was calculated, and a similarity matrix was constructed. (Normalized to the [0,1] interval). This is the number of seed samples, which is the number of representative coating formulations randomly sampled from the preprocessed database. The default value is 2000. This is the seed sample similarity matrix, with dimension 1. The range of values ​​for the elements in the matrix is It is used to characterize the comprehensive similarity between any two seed samples based on "component-performance-environment", and has been normalized.

[0076] Step 2, Construct the grid-based memory tensor Stores local feature information of seed samples, similarity matrix Local features are preferentially written into the grid base memory tensor. A Local Attention Memory Enhancement (LAM-GRU) encoder was constructed to use the coating component vector Environment vectors In grid-based memory tensors The corresponding grid coordinates As input, output is the multidimensional feature vector of the coating. Its dimensions are By storing local feature information using grid-based memory tensors and combining read / write operations with an attention mechanism, the local correlations and historical influences between coating components and environmental parameters are captured, enhancing the unified representation capability of multimodal data and laying the foundation for accurate performance prediction.

[0077] Step s21, construct the grid-based memory tensor ( The number of grid divisions for components or environmental parameters. Multidimensional feature vector of coating (dimensions of the seed sample), storing local feature information of the seed sample; Let be a grid-based memory tensor with dimension . It is used to store local feature information of seed samples. The number of mesh divisions for component or environmental parameters (first dimension) refers to the number of first-dimensional divisions when meshing coating components or environmentally related parameters. The number of mesh divisions for component or environmental parameters (second dimension) refers to the number of second-dimensional divisions when meshing coating components or environmentally related parameters. The default value is 128.

[0078] Step s22: Design a LAM-GRU as the encoder, with the input being the coating composition vector. (Content of each component), environmental vector (Temperature, salinity, etc.) in the grid-based memory tensor The corresponding grid coordinates are used to output the multidimensional feature vector of the coating through the read and write operations of the LAM-GRU encoder. The specific steps are as follows:

[0079] (1) Reading operation: based on the currently input grid coordinates Scan local bandwidth The neighborhood grid, through attention weights By integrating historical features from the memory tensor, historical states can be generated. ;

[0080] in, This is a vector of coating components, consisting of the content of various coating components (resin, rust-preventive pigment, antifouling agent, additives, etc.). This is an environmental vector, composed of parameters related to the coating's service environment (seawater temperature, salinity, pH, etc.). The current input mesh coordinates, i.e., the coating composition vector. or environment vector In grid-based memory tensors The corresponding grid coordinates. is the local bandwidth, i.e., the bandwidth used to scan the neighborhood grid during a read operation, and is fixed at 2. A is the attention weight, used to measure the contribution of historical features of different neighborhood grids in the memory tensor to the generation of the current feature during the read operation. This represents the historical state, generated by fusing historical features from the memory tensor during read operations, and is used to enhance the representational ability of the current coating features. Seed sample similarity matrix. For screening "highly similar seed sample pairs", their local features are preferentially written into the memory tensor. This provides a high-quality historical feature foundation for read operations;

[0081] (2) Write operation: using the bureau department Control the update intensity and the current feature state The weighted update to the memory tensor is calculated using the following formula:

[0082] ;

[0083] For local departments, it is used to control the update intensity of the memory tensor during write operations and to determine the proportion of the influence of the current feature state on the memory tensor; This represents the current feature state, i.e., the coating feature state generated by the LAM-GRU encoder at the current time step. The sigmoid activation function is used to convert local functions. The value is mapped to The interval enables normalized control over the update intensity of the memory tensor; grid coordinates The corresponding original feature values, that is, the local feature information of the seed sample stored in the grid before the update; grid coordinates The corresponding updated feature values ​​are obtained by fusing the current feature state through write operations. Subsequent feature information;

[0084] (3) The final hidden layer output of the LAM-GRU encoder is used as the multidimensional feature vector. To achieve a unified representation of multimodal data;

[0085] Multidimensional feature vectors The final hidden layer output of the LAM-GRU encoder is used to achieve a unified characterization of the coating's "composition-performance-environment" multimodal data.

[0086] ;

[0087] This is the weight matrix. For bias vectors, This is the final time step after the LAM-GRU encoder has traversed all input sequences. The hidden layer state, This enhances the nonlinear representation of features for activation functions.

[0088] Step 3, Construct a multi-task performance prediction network: Construct a multi-output fully connected network, taking the coating's multi-dimensional feature vector E as input, and outputting predicted values ​​for four types of performance: corrosion resistance, fouling resistance, mechanical properties, and environmental adaptability.

[0089] (1) Corrosion resistance branch: Predicted value of output polarization resistance Predicted salt spray corrosion resistance time Predicted values ​​of electrochemical impedance spectroscopy parameters , ;

[0090] (2) Antifouling performance branch: Output predicted value of biofouling coverage Predicted value of antifouling effectiveness Predicted value of antifouling agent leaching rate ;

[0091] (3) Mechanical properties branch: predicted value of output adhesion force abrasion resistance prediction value Flexibility prediction value Hardness prediction value , ;

[0092] (4) Environmental Adaptation Branch: Output the predicted performance retention rate after high and low temperature cycling / damp heat aging. .

[0093] Multidimensional feature vectors As input to a multi-output fully connected network, it is used to characterize the coating's "composition-performance-environment" multimodal data. This is the predicted polarization resistance value, which is the estimated coating polarization resistance value output by the corrosion resistance branch of the multi-task performance prediction network. This is the predicted salt spray corrosion resistance time, which is the estimated salt spray corrosion resistance time of the coating output by the corrosion resistance performance branch of the multi-task performance prediction network. This is the predicted low-frequency electrochemical impedance value, which is the estimated low-frequency impedance value of the coating electrochemical impedance spectrum output by the corrosion resistance branch of the multi-task performance prediction network. This is the predicted electrochemical phase angle, which is the estimated phase angle in the coating electrochemical impedance spectrum output by the corrosion resistance branch of the multi-task performance prediction network. This is the predicted biofouling coverage value, specifically the estimated 28-day marine biofouling coverage value of the coating output from the antifouling performance branch of the multi-task performance prediction network. This is the predicted antifouling duration, which is the estimated effective duration of the coating's antifouling function output by the antifouling performance branch of the multi-task performance prediction network. This is the predicted antifouling agent leaching rate, which is the estimated antifouling agent leaching rate in the coating output by the antifouling performance branch of the multi-task performance prediction network. This is the adhesion prediction value, which is the coating adhesion estimate output by the mechanical performance branch of the multi-task performance prediction network (corresponding to the cross-cut or pull-off test). The abrasion resistance prediction value (Taber abrasion index) is the estimated coating abrasion resistance value output by the mechanical property branch of the multi-task performance prediction network, in units of... change. The value is the flexibility prediction, which is the estimated diameter of the coating bending test output by the mechanical performance branch of the multi-task performance prediction network, in mm. This is the predicted value for pencil hardness, specifically the estimated value of coated pencil hardness output by the mechanical performance branch of the multi-task performance prediction network. The HV hardness prediction value is the estimated Vickers hardness of the coating output by the mechanical property branch of the multi-task performance prediction network. This refers to the predicted performance retention rate after high and low temperature cycling, i.e., the coating output from the environmental adaptation branch of the multi-task performance prediction network. Estimated performance retention rate after high and low temperature cycling. This is the predicted performance retention rate after damp heat aging, specifically the estimated performance retention rate of the coating after damp heat aging at 40℃ and 95%RH, output by the environmental adaptation branch of the multi-task performance prediction network.

[0094] Step s31: All performance branches share the input multidimensional feature vector. However, each has its own independent learnable parameters (weight matrix). Bias vector This avoids interference between different performance indicators. The general calculation process is as follows:

[0095] ;

[0096] For the predicted value of the k-th class of performance, These correspond to four performance categories: corrosion resistance, antifouling performance, mechanical properties, and environmental adaptability. , These are the weight matrix and bias vector of the fully connected layer 1 of the branch, respectively. For hidden layer dimensions , These are the weight matrix and bias vector of the branch output layer, respectively. The number of performance metrics for the k-th class Use the activation function of the hidden layer to enhance the nonlinear representation capability. The activation function for the branch output layer (choose according to the physical range of the performance metric, such as Sigmoid for proportional metrics and Linear for continuous numerical metrics).

[0097] Step s32, loss function design.

[0098] Construct a weighted multi-task loss function and assign differentiated weights to different performance metrics (such as weights for anti-pollution effective options). Polarization resistance weight Adhesion weight Environmental adaptability weight The formula is:

[0099] ;

[0100] In the formula, The weighted total loss for multiple tasks is used to measure the overall prediction error of the multi-output fully connected network for four performance metrics, and to guide the optimization of model parameters. For the first The weights for different performance categories are used to adjust their priority during model training. Corresponding antifouling performance, Corrosion resistance, Corresponding mechanical properties The default values ​​for corresponding environmental adaptability are as follows: , ; For the first The number of performance metrics for a class, i.e., the first The number of specific performance metrics included in class performance (e.g.) hour, (corresponding to three indicators of antifouling performance). For the first Class performance The true value of an indicator, that is, the actual value of the performance indicator obtained through experimental testing or historical data recording. For the first Class performance The predicted value of each indicator, that is, the estimated value of the performance indicator output by the multi-output fully connected network.

[0101] Step s33: The Adam optimizer is used to optimize the multi-task performance prediction network.

[0102] Using the Adam optimizer, set the learning rate. Number of iterations Rounds, each round is based on the seed sample similarity matrix We designed a distance-weighted sampling strategy and selected highly discriminative sample pairs (similar formulations, dissimilar formulations) for training to improve the sensitivity of the multi-task performance prediction network to subtle differences in formulations.

[0103] Adam optimizer: An optimization algorithm used for model training that accelerates model convergence and improves parameter optimization efficiency by adaptively adjusting the learning rate. The learning rate controls the step size for each parameter update by the Adam optimizer; the default value is [value to be filled in]. ; This is the maximum number of iterations for model training, i.e., the total number of rounds for optimizing model parameters. The default value is 100 rounds. This is the seed sample similarity matrix (same meaning as above), used to guide the implementation of the distance-weighted sampling strategy in each round of training, and to select sample pairs with strong discriminative power.

[0104] The Adam optimizer minimizes the weighted multi-task total loss through adaptive learning rate adjustment. To achieve multi-task performance prediction network parameters (weights of each branch) Bias Optimization of ).

[0105] Step 4: The coating formulation is efficiently optimized using the Hybrid Gray Wolf-Sunflower Optimization Algorithm (HGWSFO). The local search of the gray wolf optimization and the global search of the sunflower optimization are dynamically switched by the index threshold to balance the breadth of exploration and the accuracy of utilization of the algorithm and avoid the limitations of a single optimization algorithm.

[0106] Set index threshold This enables dynamic switching between Gray Wolf Optimization (GWO) local search and Sunflower Optimization (SFO) global search. The index value of the candidate recipe set is used to compare with the index threshold. The comparison determines the currently enabled optimized search mode:

[0107] When the index value of the recipe candidate set At that time, initiate SFO to perform a global exploration:

[0108] (1) Initialize the “Sunflower” population (recipe candidate set, population size) The formula with the best adaptability is set as "Sun". ; The size of the sunflower population, i.e. the number of recipe candidate sets during the global exploration phase of SFO, is fixed at 50. The "Sun" formulation, i.e. the coating formulation with the best adaptability in the SFO stage, is used as the search target for other candidate formulations.

[0109] (2) Calculate the step size based on the inverse square law: ( (These are the upper and lower limits of the content of each component). The maximum step size in the SFO stage is calculated based on the inverse square law and is used to control the maximum magnitude of the movement of the candidate formulation towards the "sun" formulation. This represents the upper limit of the content of each component, i.e., the highest allowable content of various components (resin, anti-rust pigment, etc.) in the coating in industrial applications. This represents the lower limit of the content of each component, i.e., the minimum allowable content of each component in the coating for industrial applications;

[0110] (3) Generate new formulas through pollination operations Remove "dead plants" with poor adaptability and update the position of the "sun". Expand the scope of formula search. The new formulations generated in the SFO stage are obtained by updating existing candidate formulations through pollination operations, thereby expanding the formulation search scope.

[0111] When the index value of the recipe candidate set At that time, start GWO to perform a local search:

[0112] (1) Initialize the "Grey Wolf" population (select from high-quality recipes searched by SFO, scale) ), The gray wolf population size, i.e., the number of candidate formulations in the GWO local search phase, is selected from the high-quality formulations searched in the SFO, and is fixed at 30. The top 3 coating formulations in terms of fitness are set as follows: (leader ), (Secondary Leader) ), (Executor) ); The (leader) recipe, namely the best recipe with the highest fitness in the GWO phase, dominates the local search direction. (Secondary Boss) Recipe, namely the second-best recipe in terms of adaptability during the GWO phase, is an auxiliary... Recipe-guided search. (Executor) Formula, namely the high-quality formula ranked third in fitness during the GWO phase, combined with... , Optimize the search path for the formula;

[0113] (2) Calculate the first coefficient vector Second coefficient vector ( Decrease linearly from 2 to 0. , for (random number); This is the coefficient decay factor for the GWO stage, which decreases linearly from 2 to 0, used to adjust the balance between exploration and exploitation during the search process. rand1 is... Random numbers within the interval are used to calculate the coefficient vector. This introduces search randomness. rand2 is... Random numbers within the interval are used to calculate the coefficient vector. This enhances search diversity. B is the first coefficient vector in the GWO stage, and the formula is... It is used to control the direction of candidate formulations to move closer to or further away from high-quality formulations, and to adjust the local search intensity. The second coefficient vector of the GWO stage is given by the formula: This is used to further enhance the randomness of the search and avoid getting trapped in local optima;

[0114] (3) Other "Grey Wolves" (common recipe) According to the leader Secondary leaders Executor Update its own recipe: ,when Focusing on local optimization and fine-tuning component content. The standard formula for the GWO stage (the first) (Recipe for each gray wolf), by reference , , The recipe updates its own location. This is the standard recipe after the GWO update; the formula is: That is, based on , , The average position of the formula is adjusted.

[0115] Step 5: Set the optimization objectives and constraints, and construct a weighted multi-objective optimization function.

[0116] Step s51, set constraints:

[0117] Set minimum thresholds for the core performance of the coating and screen formulations that meet engineering requirements:

[0118] (1) Antifouling performance: 28-day biofouling coverage Anti-fouling validity period Months, antifouling agent leaching rate ;

[0119] in, The 28-day biofouling coverage rate, which is the percentage of the coating area covered by marine organisms (diatoms, barnacles, etc.) after 28 days of service in the marine environment, is the core constraint indicator for antifouling performance. Antifouling effectiveness period, which is the duration for which a coating maintains its effective antifouling function, is a key constraint indicator of antifouling performance. Antifouling agent leaching rate, which is the rate at which the antifouling agent in the coating seeps into the external environment, is measured in units of... It needs to be within a reasonable range to ensure the anti-fouling effect and durability;

[0120] (2) Mechanical properties: adhesion (Pull-out method), abrasion resistance Turn, flexibility ,hardness ;

[0121] in, Adhesion, or the bonding strength between the coating and the substrate, is tested using the pull-off method and is a core constraint indicator of mechanical properties. Taber abrasion resistance is the coating's ability to resist wear, measured in feet (µm). Rotation is a key constraint indicator for mechanical properties. Flexibility refers to the coating's ability to withstand bending deformation without breaking. It is measured by the diameter of the bending test and is expressed in mm. It is an important constraint indicator of mechanical properties. Pencil hardness, or the coating's ability to resist scratching by hard objects, is one of the limiting indicators of mechanical properties.

[0122] (3) Cost constraints: Cost of coating materials per square meter Yuan;

[0123] in, The cost of materials per square meter of coating, i.e., the material cost required to prepare 1 square meter of coating, is an economic constraint indicator for formula screening.

[0124] Step s52, construct the weighted multi-objective optimization function:

[0125] ;

[0126] in, The weighted multi-objective optimization function value is used to comprehensively evaluate the quality of coating formulations, with the objective of maximizing this function value. The weight of the antifouling effectiveness period is used to adjust the importance of the antifouling effectiveness period in multi-objective optimization, and is fixed at 0.4. The weight of biofilm coverage is used to adjust the importance of biofilm coverage in multi-objective optimization, and is fixed at 0.2. The weight for wear resistance is used to adjust the importance of wear resistance in multi-objective optimization, and is fixed at 0.2. The weight of material cost is used to adjust the importance of material cost in multi-objective optimization, and is fixed at 0.2. The wear resistance constraint threshold, i.e., the maximum allowable wear of the coating, is fixed at a value. The normalization function is used to normalize the wear resistance index. The material cost constraint threshold, i.e. the maximum allowable material cost per square meter of coating, is fixed at 200 yuan and is used to normalize the cost index.

[0127] Step 6: The Hybrid Grey Wolf-Sunflower Optimization Algorithm (HGWSFO) iterates continuously, outputting a weighted multi-objective optimization function value that satisfies the constraints of Step 4. The largest and most optimal formula.

[0128] Algorithm Iteration After the round, the output satisfies the constraints and the objective function. The top 10 formulations are used as the candidate set, and the performance prediction for each formulation (including confidence intervals for each metric) is output. ) and recommended application sea areas (based on environmental adaptability matching). Among them, The number of iterations for algorithm optimization, i.e., the total number of iterations of the Hybrid Gray Wolf-Sunflower Optimization Algorithm (HGWSFO), is fixed at 50 rounds. This refers to the predicted value of a performance metric, which is an estimated value of a certain performance metric output by the multi-task performance prediction network. The standard deviation of the predicted values ​​of the performance index is used to measure the dispersion of the predicted values ​​and reflect the accuracy of the prediction. The confidence interval of a performance indicator is the range in which the actual value of the performance indicator is likely to fall, and it is used to assess the reliability of the prediction of the formulation performance.

[0129] Comparative Example

[0130] The model of this invention is compared with the Transformer, GRU, and LSTM models.

[0131] As shown in Figure 2, the accuracy of all four models steadily increases with the number of training iterations, consistent with the convergence law of model training. However, the algorithm of this invention, thanks to the synergistic advantages of seed-guided neurometric learning technology, Local Attention Memory (LAM) enhanced encoder, and Hybrid Gray Wolf-Sunflower Optimization Algorithm (HGWSFO), maintains the highest accuracy in each training stage. After 100 training iterations, the accuracy reaches 0.92, which is 9, 14, and 13 percentage points higher than Transformer (0.83), GRU (0.78), and LSTM (0.79), respectively. This fully verifies the technical superiority of this invention in multimodal data fusion, accurate feature representation, and formulation optimization, and can provide more reliable support for predicting the performance of the pollution barrier coating.

[0132] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for screening formulations of multifunctional anti-corrosion and anti-fouling coatings for debris-blocking nets, characterized in that, Includes the following steps: Step 1: Collect core data to build a database, perform preprocessing to obtain the preprocessed database, and use seed-guided neural metric learning to achieve deep fusion of multimodal data on "coating composition-performance-environment-failure". By selecting representative seed samples, a seed sample set with similarity labels is generated, and a similarity matrix S is constructed. Step 2: Construct a grid-based memory tensor M to store the local feature information of the seed samples. The local features of the similarity matrix S are preferentially written into the grid-based memory tensor M. Construct an LAM-GRU encoder with coating composition vectors... Environment vectors The corresponding grid coordinates in the grid-based memory tensor M The input is a coating multidimensional feature vector E with dimension d; Step 3: Construct a multi-task performance prediction network, using the coating multidimensional feature vector E as input, outputting predicted values ​​of four performance indicators, constructing a loss function to assign differentiated weights to the four performance indicators, and using the Adam optimizer to optimize the parameters of the multi-task performance prediction network; Step 4: Perform efficient optimization of the coating formulation using HGWSFO: dynamically switch between local search of gray wolf optimization and global search of sunflower optimization by using an index threshold; Step 5: Set optimization objectives and constraints, construct a weighted multi-objective optimization function, continuously iterate HGWSFO, and output the optimal formulation that satisfies the constraints and maximizes the objective function.

2. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 1, the data in the database includes coating formulation component data, performance data, environmental data, and failure data; preprocessing includes data standardization, outlier removal, and data completion and enhancement.

3. The method for screening the formulation of a multifunctional anti-corrosion and anti-fouling coating for a debris-blocking net as described in claim 1, characterized in that, In step 1, the similarity matrix ， This refers to the number of representative coating formulation components randomly sampled from the pre-processed database.

4. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 2, the grid base memory tensor P and Q represent the number of grid divisions for the component or environmental parameter in the first and second dimensions, respectively.

5. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 2, the multidimensional feature vector E of the coating is output through the read and write operation of the LAM-GRU encoder. The specific steps include the following: (1) Read operation: based on the coating component vector or environment vector The corresponding grid coordinates in the grid-based memory tensor M Scan the neighborhood grid of the local bandwidth w, and apply attention weights. Fusion grid-based memory tensor Historical characteristics in the process of generating historical states (2) Write operation: using the bureau department Control the update intensity and the current feature state The weighted update to the memory tensor is calculated using the following formula: ； For bureau departments; This represents the current feature state; It is the sigmoid activation function; grid coordinates The corresponding original feature values; grid coordinates The corresponding updated feature values; (3) Using the final hidden layer output of the LAM-GRU encoder as the multidimensional feature vector E, a unified representation of multimodal data is achieved; ； This is the weight matrix. For bias vectors, This is the final time step after the LAM-GRU encoder has traversed all input sequences. The hidden layer state, This enhances the nonlinear representation of features for activation functions.

6. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 3, the formula for the multi-task performance prediction network is expressed as follows: ； For the predicted value of the k-th type of performance index, These correspond to four performance categories: corrosion resistance, antifouling performance, mechanical properties, and environmental adaptability. These are the weight matrix and bias vector of the fully connected layer 1 of the branch, respectively. For the hidden layer dimension; These are the weight matrix and bias vector of the branch output layer, respectively. The number of metrics for the k-th performance category; For the hidden layer activation function, This is the activation function for the branch output layer.

7. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 3, the loss function is: ； To calculate the weighted total loss across multiple tasks, For the first Class performance weights For the first Class performance The true value of each indicator For the first Class performance The predicted values ​​of each indicator.

8. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 4, the index threshold is set. When the index value of the candidate recipe set When the index value of the recipe candidate set is reached, SFO is initiated to perform a global exploration. At that time, GWO will be launched to perform a local search.

9. The method for screening the formulation of a multifunctional anti-corrosion and anti-fouling coating for a debris-blocking net as described in claim 8, characterized in that, The specific steps for global exploration of SFO are as follows: (1) Initialize the "Sunflower" population and set the coating formula with the best fitness as "Sun". (2) Calculate the step size based on the inverse square law: ,in This is the maximum step size in the SFO stage, used to control the maximum magnitude by which the candidate formulation moves towards the "sun" formulation. These represent the upper and lower limits of the content of each component. For sunflower population size; (3) Generate new formulas through pollination operations. Remove "dead plants" with poor adaptability and update the "sun". The specific steps of the GWO local search are as follows: (1) Initialize the "gray wolf" population and set the top three coating formulas with the highest fitness as leaders. Secondary leaders Executor (2) Calculate the coefficient vector: the first coefficient vector of the GWO stage. The second coefficient vector of the GWO stage ， The coefficient decay factor for the GWO stage. The value decreases linearly from 2 to 0, rand1, All Random numbers within the interval; The first coefficient vector B is used to control whether the candidate formulation moves closer to or further away from the high-quality formulation. The second coefficient vector... Used to enhance search randomness; (3) Common formula According to the leader Secondary leaders Executor Update to obtain the updated standard recipe. 。 10. The method for screening the formulation of the multifunctional anti-corrosion and anti-fouling coating for the debris-blocking net as described in claim 1, characterized in that, In step 5, the weighted multi-objective optimization function is: ;in: To optimize the function value for weighted multi-objectives; Weighting of the effective period of anti-fouling treatment; The weight for bioattachment coverage; The bio-attachment coverage rate is 28 days. Weighting for wear resistance; Weighting for material costs; For wear resistance; The wear resistance constraint threshold; Cost of coating materials per square meter; This is the threshold for material cost constraints.

Images