Coating corrosion resistance prediction method and system based on reinforcement learning model

Abstract

The invention discloses a coating corrosion resistance prediction method and system based on a reinforcement learning model, and relates to the technical field of coating corrosion resistance prediction, and the method comprises the steps: collecting the data of a coating in a corrosion environment, forming a structured corrosion performance data table through a YOLO v5 algorithm, and carrying out the prediction of the corrosion resistance of the coating; a corrosion state of a time sequence is generated through an interpolation method in combination with a kinetic equation, an improved Double DQN algorithm is used for training a reinforcement learning model, and a corrosion resistance adjustment strategy is generated. The corrosion behavior of the coating in a corrosion environment can be accurately reflected, a corrosion resistance optimization adjustment strategy can be rapidly generated, and intelligence and high efficiency of corrosion performance adjustment are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coating corrosion resistance prediction, and in particular to a coating corrosion resistance prediction method and system based on reinforcement as the initial population input learning model of the NSGA-II algorithm. BACKGROUND

[0002] With the continuous development of the industrial field, coatings are widely used on metal surfaces to improve the corrosion resistance of materials. However, the uncontrollability under complex corrosion environment and long-term use makes it a research hotspot in the fields of material science and industrial engineering to accurately predict and effectively optimize the corrosion resistance of coatings. In traditional technology, the corrosion resistance of coatings is mainly predicted and optimized by laboratory experiments and empirical models. These methods usually require a large number of manual experiments and chemical composition adjustments, which are costly and inefficient. In addition, since corrosion is a complex physical and chemical process, its influencing factors are diverse, such as corrosion medium, time, temperature, and coating composition. The results under different experimental conditions are difficult to unify and induce, and the prediction ability for long-term corrosion durability is very limited. In recent years, with the rapid development of artificial intelligence and big data technology, deep learning and reinforcement learning technology have emerged in the field of industrial data processing, making it possible to extract key characteristics from complex experimental data and make predictions. However, there are still some technical obstacles in directly applying these emerging technologies to dynamic prediction and optimization of coating corrosion resistance, such as how to fully utilize the time series characteristics of corrosion data, how to realize the adaptive optimization of coating performance adjustment strategy, and how to provide multi-objective optimization scheme that can guide actual industrial production.

[0003] The existing technology has made some breakthroughs in the intelligent research of coating corrosion resistance analysis. For example, corrosion area detection and corrosion rate calculation based on traditional image processing technology can preliminarily realize the extraction of corrosion characteristic data, but due to the limited recognition ability of traditional image algorithms for corrosion features, especially in complex corrosion environments, they cannot adapt well to different types of noisy data and are inefficient. In addition, the existing intelligent prediction methods are less combined with physical and chemical kinetic mechanisms, making the prediction results lack theoretical support and unable to accurately reflect the gradual change process of corrosion behavior at the time scale. In addition, for most optimization strategy generation technologies, the existing optimization system often only focuses on a single indicator, such as the change of corrosion area or coating thickness, and fails to balance the needs of corrosion resistance, cost, resource efficiency, and other multiple objectives, which limits its practicality in actual industrial production. Therefore, the existing technology cannot form a complete closed-loop system for corrosion performance prediction and optimization, making the quantitative optimization and performance improvement of coating corrosion resistance still rely on a large number of manual trial-and-error experiments. SUMMARY

[0004] In view of the problems existing in the above-mentioned existing coating corrosion resistance prediction method and system based on a reinforcement learning model, the present application is proposed.

[0005] Therefore, the present application provides a coating corrosion resistance prediction method and system based on a reinforcement learning model to solve the problems of low data extraction efficiency, inaccurate time series modeling and single optimization strategy in coating corrosion resistance prediction.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a coating corrosion resistance prediction method based on a reinforcement learning model, which includes collecting data of coatings in a corrosion environment, using a YOLO v5 algorithm to form a structured corrosion performance data table; Based on the corrosion performance data table, a time series corrosion state is generated by an interpolation method combined with a kinetic equation, a reinforcement learning model is trained using the time series corrosion state combined with an improved Double DQN algorithm, an anti-corrosion performance adjustment strategy is obtained through the reinforcement learning model, a Pareto optimal initial population is generated based on the reinforcement learning model, and the adjustment strategy is optimized combined with an NSGA-II algorithm; The experience data generated in the process of generating the adjustment strategy and optimizing the adjustment strategy is uploaded to a database to form a closed-loop feedback.

[0007] As a preferred scheme of the coating corrosion resistance prediction method based on a reinforcement learning model, wherein: the reinforcement learning model is trained using the time series corrosion state combined with the improved Double DQN algorithm, and the anti-corrosion performance adjustment strategy is obtained through the reinforcement learning model, specifically, a state space vector is extracted from the time series corrosion state and the environmental state ; Determine the online network Online and the network parameters used by it And the target network Target and the network parameters used by it Use a deep neural network as a function approximator to calculate the Q value of the online network , the Q value of the target network ; Calculate the Q value Y of the next time step of the target network from the online network Select the optimal action for the next time step, and the target network Evaluate the Q value of the next time step; Based on the current Q value of the online network and the Q value of the next time step, obtain the loss calculation function ; Set the loss function threshold : If is greater than or equal to , calculate the gradient of the loss function , update the parameters according to the gradient of the loss function ; Update the target network parameters every K times of training ; If Less than , the reinforcement learning model training is completed, and the adjustment strategy is output according to the current state vector , the final output is the adjustment strategy.

[0008] As a preferred scheme of the paint corrosion resistance prediction method based on the reinforcement learning model, the structured corrosion performance data table is formed by using the YOLO v5 algorithm, specifically, based on the collected image data, the Labeling marking tool is used to mark the corrosion spots, the corrosion area formed by a large number of connections and the zinc coverage area, and the corrosion area is calculated based on the area of the marking frame And the zinc coverage area ; The corrosion area ratio of the paint is calculated using the corrosion area, and the structured data including the paint thickness, corrosion time, adhesion, corrosion area ratio, and zinc coverage area at each time point is finally output.

[0009] As a preferred scheme of the paint corrosion resistance prediction method based on the reinforcement learning model, the time series corrosion state is generated by the interpolation method combined with the kinetic equation based on the corrosion performance data table, specifically, according to the time point and state variable provided by the corrosion performance data table, the linear interpolation method is used to calculate the corrosion area corresponding to the time t And the zinc coverage area ; The corrosion rate is calculated based on the kinetic equation And the protection ability ; The corrosion state at a single time point is represented by the calculation results of the corrosion area, the zinc coverage area, the corrosion rate and the protection ability ; The time series is represented as a collection of multiple time point states The final output is the time series corrosion state.

[0010] As a preferred scheme of the paint corrosion resistance prediction method based on the reinforcement learning model, the data of the paint in the corrosion environment is collected, specifically, the thickness, corrosion time, adhesion and image data of the paint are collected, and the collected data is uploaded to the database.

[0011] As a preferred scheme of the coating corrosion resistance prediction method based on the reinforcement learning model, the initial population is generated based on the reinforcement learning model, and the NSGA-II algorithm is combined to optimize and adjust the strategy, specifically, a set of k multi-objective optimization solutions are generated as the initialization solution set ; Defining the maximum service life of the coating As the first optimization target of the adjustment strategy, the corrosion area is used As the second optimization target of the adjustment strategy, the adhesion test value is defined As the third optimization target of the adjustment strategy, the initialization solution set is input as the initial population of the NSGA-II algorithm; The solutions in the initialization solution set are non-dominantly sorted, and the crowding distance of all non-dominant solutions is calculated after sorting ; The crowding distance threshold is set If is less than , the e-th non-dominant solution is put into the preferred strategy set, otherwise it is not used; Based on the screened preferred strategy set, new child strategies are generated by adopting crossover and mutation operations , and the generated child strategies and the original preferred strategies are integrated into a new solution set ; The maximum number of iterations Iter is set, and after the number of iterations reaches the upper limit Iter, the strategy with the smallest crowding distance in the solution set is output as the optimal adjustment strategy.

[0012] As a preferred scheme of the coating corrosion resistance prediction method based on the reinforcement learning model, the experience data generated in the process of generating the adjustment strategy and optimizing the adjustment strategy is uploaded to the database to form a closed loop feedback, specifically, the coating life, corrosion area, and adhesion data of different coatings in different environments are recorded, and the experience of generating the adjustment strategy and optimizing the adjustment strategy based on the data is uploaded to the database to form a closed loop feedback.

[0013] In a second aspect, the application provides a coating corrosion resistance prediction system based on a reinforcement learning model, comprising, a data acquisition module for acquiring various data of a coating sample in a corrosion environment; a target detection module for detecting and identifying the collected coating corrosion image using the YOL v5 algorithm and extracting key corrosion features; a time series generation module configured to generate time series corrosion states using an interpolation method and a dynamic equation; a reinforcement learning model training module configured to train the reinforcement learning module using a DQN algorithm and output an adjustment strategy after the training is completed; a Pareto optimal initial population generation module configured to generate a plurality of adjustment strategies using the reinforcement learning model; an NSGA-II optimization module configured to optimize the set of strategies and output an optimal adjustment strategy; an experience storage and learning module configured to save the experience generated in the process of generating strategies and optimizing strategies and form a closed-loop feedback.

[0014] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the method for predicting the corrosion resistance of paint based on a reinforcement learning model according to the first aspect of the present application.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the method for predicting the corrosion resistance of paint based on a reinforcement learning model according to the first aspect of the present application.

[0016] The present application has the following beneficial effects: the improved YOLO v5 algorithm is used to accurately extract and structurally express corrosion data, and the interpolation method and the dynamic equation are used to generate time series corrosion state data, thereby solving the problems of dynamic complexity and multi-objective trade-off in the optimization of the corrosion resistance of paint. The data is input into the improved Double DQN algorithm to train the reinforcement learning model, so as to obtain a dynamic corrosion resistance adjustment strategy, realize the intelligentization, automation and high efficiency of the corrosion performance adjustment, further realize the multi-objective optimization by combining the NSGA-II algorithm, and iteratively optimize the adjustment strategy through the database closed-loop feedback mechanism, thereby forming an efficient data-driven strategy optimization system, realizing the circulation system from data acquisition to strategy generation, optimization and feedback, and significantly improving the efficiency and reliability of the optimization decision. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0018] Figure 1 FIG. 1 is a flowchart of the method for predicting the corrosion resistance of paint based on a reinforcement learning model according to Embodiment 1.

[0019] Figure 2 A structural schematic diagram of the coating corrosion resistance prediction system based on the reinforcement learning model in Embodiment 1. DETAILED DESCRIPTION

[0020] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0021] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited by the specific embodiments disclosed below.

[0022] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0023] Embodiment 1, Reference Figure 1 and Figure 2 , the first embodiment of the present application, the embodiment provides a coating corrosion resistance prediction method based on a reinforcement learning model, the coating corrosion resistance prediction method based on the reinforcement learning model comprises the following steps: S1: collecting data of the coating in the corrosion environment, and forming a structured corrosion performance data table using the YOLO v5 algorithm.

[0024] Specifically, collecting data of the coating in the corrosion environment means collecting the thickness, corrosion time, adhesion and image data of the coating, and uploading the collected data to a database.

[0025] By collecting data of the coating in the corrosion environment and uploading it to the database, centralized acquisition and standardized storage of coating data information are realized, providing a foundation support for subsequent data-driven optimization, transparency and accuracy of the coating usage in the corrosion environment, thereby improving the response efficiency of corrosion data analysis and adjustment strategy.

[0026] Further, forming a structured corrosion performance data table using the YOLO v5 algorithm means that based on the collected image data, the Labellmg labeling tool is used to label the corrosion spots, the corrosion area formed by a large number of connections and the zinc coverage area, and the corrosion area is calculated based on the area of the labeled frame and the zinc coverage area ; The area of the corrosion spot is calculated based on the area of the labeled box : , wherein, is the number of labels of the corrosion spot, is the number of labels of the corrosion area, is the labeled area of the i-th corrosion spot, is the labeled area of the i-th corrosion area; The zinc coverage area is calculated based on the area of the labeled box : , wherein, is the number of labels of the zinc coverage area, is the labeled area of the i-th zinc coverage area; The corrosion area ratio of the coating is calculated using the corrosion area, and finally the structured data containing the coating thickness, corrosion time, adhesion, corrosion area ratio, and zinc coverage area at each time point is output.

[0027] Through the YOLO v5 algorithm, the image data of the coating in the corrosion environment is subjected to target recognition, and the original and complex unstructured corrosion data is converted into structured corrosion performance data table which is easy to analyze, greatly improving the data processing efficiency and accuracy.

[0028] S2: Based on the corrosion performance data table, a time series of corrosion states is generated by an interpolation method combined with a kinetic equation, a reinforcement learning model is trained using the time series of corrosion states combined with an improved Double DQN algorithm, an anti-corrosion performance adjustment strategy is obtained through the reinforcement learning model, a Pareto optimal initial population is generated based on the reinforcement learning model, and the adjustment strategy is optimized by combining the NSGA-II algorithm.

[0029] Specifically, based on the corrosion performance data table, a time series of corrosion states is generated by an interpolation method combined with a kinetic equation, which is to calculate the corrosion area and the zinc coverage area : at time t using a linear interpolation method according to the time point and state variable provided by the corrosion performance data table. : , wherein, t is the time, is the i-th time, is the next time of the i-th time; The zinc coverage area at time t is calculated as : , Corrosion rate is calculated based on kinetic equation Protection ability , as follows: Corrosion rate is calculated according to dynamic change of corrosion area : , wherein, is time step; Protection ability is calculated according to dynamic change of zinc coverage area : , wherein, is initial zinc coverage area; Corrosion state at a single time point can be expressed as : , wherein, is corrosion area at t time, is corrosion rate, is zinc coverage area at t time, is protection ability at t time; Time sequence is expressed as a collection of states at multiple time points : , Output is corrosion state of time sequence.

[0030] By using interpolation method to generate corrosion state data with time sequence characteristics based on structured corrosion performance data, the problem of discontinuity of corrosion experimental data in time dimension is solved, the corrosion data becomes more complete, and the dynamic corrosion process can be more accurately reflected. Meanwhile, with the aid of kinetic equation, the time sequence data is more consistent with the corrosion mechanism, and key input variables are provided for the input of subsequent reinforcement learning model.

[0031] Further, the reinforcement learning model is trained by using time sequence corrosion state and improved Double DQN algorithm, and the adjustment strategy of corrosion resistance is obtained by the reinforcement learning model from the state space vector extracted from the corrosion state and the environment state of time sequence : , wherein, is environment temperature, is environment humidity, is environment salt concentration; Action space is defined : , wherein, is the adjustment strategy, in particular as follows: , wherein, is the zinc powder content value, is the coating thickness value, is the proportion value of specific auxiliary agents, including but not limited to dispersing auxiliary agents, anti-corrosion agents, anti-oxidation and stabilizing agents, etc.; define the reward function : , wherein, is the maximum service life of the coating, is the corrosion area, is the adhesion test value, , is the set weight coefficient; construct online network Online and target network Target, define neural network parameters and , and use deep neural network as function approximator to calculate the Q values of the two networks and : the use of deep neural network as function approximator calculates the Q value of the online network : , wherein, DNN is a multi-layer neuron that completes the input variable to Q value mapping, is the network parameter used by the online network; the use of deep neural network as function approximator calculates the Q value of the target network : , wherein, is the network parameter used by the target network; calculate the Q value Y of the next time step of the target network: , wherein, is the immediate reward, is the discount factor, is the state space vector of the next time step, is the action used by the Q network of the next time step, selected by the online network to select the optimal action of the next time step, the target network evaluate the Q value of the next time step; Based on the current Q value of the online network and the Q value of the next time step, a loss calculation function is obtained : , In each training step, a random action is selected from the action space with a probability Exploration, and a maximum Q value action is selected with a probability ; Set the loss function threshold : If is greater than or equal to , the gradient of the loss function is calculated : , Then update the parameters according to the gradient of the loss function: , Wherein, is the learning rate weight set; Update the target network parameters every K training : , Repeat the training of the reinforcement learning model until the loss function is less than the threshold value; If is less than , it means that the reinforcement learning model training is completed, and the latest online network parameters are used Output the adjustment policy according to the current state vector : , The final output is the adjustment policy.

[0032] By taking the time series of corrosion states as the state input of the reinforcement learning model, the reinforcement learning model is trained, and the Double DQN algorithm is used to overcome the defect that the traditional DQN algorithm easily overestimates the Q value, making the policy evaluation more accurate. By separating the processes of selecting actions and evaluating target values, the problem of complex state transition and large noise in the corrosion scene is solved. Through multiple rounds of training, the reinforcement learning model can autonomously explore and generate high-quality adjustment policies, solving the problem of relying on manual experience adjustment in traditional methods.

[0033] Furthermore, based on the reinforcement learning model, a Pareto optimal initial population is generated, and the NSGA-II algorithm is used to optimize the adjustment policy. Through multiple interactions with the reinforcement learning model, a set of k multi-objective optimization solutions are generated as the initial solution set : , Maximum usage time of the coating As the first optimization objective of the adjustment strategy, the corrosion area is used As a second optimization target for the adjustment strategy, use the adhesion measurement values As the third optimization goal of the adjustment strategy, the solution set As the initial population input of the NSGA-II algorithm; The solution set Perform non-dominated sorting on the solutions in 、 ,like: , but Non-dominant , h is the h-th optimization target; Calculate the crowding distance for all non-dominated solutions : , in, is the e-th solution among all non-dominated solutions, 、 are the solutions on the left and right adjacent to the e-th solution in the non-dominated solution, is the value of the solution on the left adjacent to the e-th solution in the non-dominated solution on the h-th optimization objective, is the value of the solution to the right of the e-th solution in the non-dominated solution on the h-th optimization target, g is the total number of optimization targets, 、 are the maximum and minimum values ​​of the h-th optimization objective respectively; Set the crowding distance threshold ,like Less than , then the e-th non-dominated solution Put it into the preferred strategy set, otherwise it will not be adopted; Based on the selected optimal strategy set, crossover and mutation operations are performed to randomly select two strategies from the optimal strategy set. 、 , exchange the two strategies and adjust the strategy value to generate a new sub-strategy , and randomly select an indefinite number of optimal strategies, randomly adjust the original strategy value to generate a new sub-strategy, and integrate the generated sub-strategy with the original optimal strategy into a new solution set ; Set the maximum number of iterations Iter, for the solution set The adjustment strategies in the non-dominated set are sorted, the non-dominated strategies obtained after sorting are subjected to congestion distance calculation, the strategies with congestion distance less than a threshold value are set as a new solution set, new sub-strategies are repeatedly generated and subjected to non-dominated sorting before the iteration number reaches a threshold value, and the strategy with the minimum congestion distance in the solution set is output after the iteration number reaches an upper limit Iter as the optimal adjustment strategy.

[0034] By combining multi-objective optimization technology on the basis of the reinforcement learning model, the prediction ability of reinforcement learning for dynamic corrosion adjustment is absorbed, and the ability of the multi-objective optimization algorithm to further improve the ability of the decision scheme to meet complex actual needs is utilized. A plurality of initial adjustment schemes are generated through a Pareto optimal set, and the adjustment strategies are further optimized by combining the NSGA-II algorithm, so that the system can find a plurality of trade-off adjustment schemes, and the bottleneck of the limitation of a single target scheme is solved.

[0035] S3: uploading experience data generated in the process of generating and optimizing the adjustment strategy to the database to form a closed loop feedback.

[0036] Specifically, the experience data generated in the process of generating and optimizing the adjustment strategy is uploaded to the database to form a closed loop feedback. The coating life, corrosion area, and adhesion data of different coatings in different environments are recorded, and the experience of generating and optimizing the adjustment strategy based on the data is uploaded to the database to form a closed loop feedback.

[0037] By uploading the adjustment strategies, reward values, and key data of corrosion state evolution generated in the optimization process to the database, the experience in reinforcement learning and multi-objective optimization is formed into a knowledge base through a closed loop feedback mechanism, repeated training and calculation are reduced, and more efficient iteration is realized in subsequent optimization.

[0038] The embodiment also provides a coating corrosion resistance prediction system based on a reinforcement learning model, which comprises: A data acquisition module is configured to acquire various data of a coating sample in a corrosion environment. A target detection module is configured to use a YOL v5 algorithm to detect and identify the collected coating corrosion image and extract key corrosion features. A time series generation module is configured to use an interpolation method and a dynamics equation to generate a time series corrosion state. A reinforcement learning model training module is configured to train the reinforcement learning module using a DQN algorithm and output an adjustment strategy after the training is completed. A Pareto optimal initial population generation module is configured to generate a plurality of adjustment strategies using the reinforcement learning model. An NSGA-II optimization module is configured to optimize a strategy set and output an optimal adjustment strategy. An experience storage and learning module is configured to save experience generated in the generation and optimization of the strategy and form a closed-loop feedback.

[0039] The embodiment also provides a computer device suitable for the case of the coating corrosion resistance prediction method based on the reinforcement learning model, which comprises a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the coating corrosion resistance prediction method based on the reinforcement learning model proposed in the above embodiment.

[0040] The computer device can be a terminal, which comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0041] The embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to implement the coating corrosion resistance prediction method based on the reinforcement learning model proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0042] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A coating corrosion resistance prediction method based on a reinforcement learning model, characterized by: include, Collect data on coatings in corrosive environments and use the YOLO v5 algorithm to generate a structured corrosion performance data table. Based on the corrosion performance data table, the interpolation method is combined with the kinetic equation to generate the time series corrosion state. The time series corrosion state is combined with the improved Double DQN algorithm to train the reinforcement learning model. The reinforcement learning model is used to obtain the corrosion resistance adjustment strategy. Based on the reinforcement learning model, the Pareto optimal initial population is generated, and the adjustment strategy is optimized with the NSGA-II algorithm. The experience data generated in the process of generating and optimizing adjustment strategies is uploaded to the database to form a closed-loop feedback.

2. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 1, characterized in that: The time series corrosion state is combined with the improved Double DQN algorithm to train the reinforcement learning model, and the corrosion resistance adjustment strategy is obtained through the reinforcement learning model. Specifically, the state space vector is extracted from the corrosion state and environmental state of the time series. ; Determine the online network and the network parameters it uses and the target network Target and its network parameters , using a deep neural network as a function approximator to calculate the Q value of the online network , the Q value of the target network ; Calculate the Q value Y of the target network at the next time step, by the online network Select the optimal action for the next time step, the target network Evaluate the Q value for the next time step; Based on the current Q value of the online network and the Q value of the next time step, the loss calculation function is obtained ; Setting the loss function threshold : like Greater than or equal to , calculate the gradient of the loss function , update the parameters according to the gradient of the loss function ; Update the target network parameters every K training times ; like Less than , reinforcement learning model training is completed, and the strategy is adjusted according to the current state vector output , the final output To adjust strategy.

3. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 2, characterized in that: The YOLO v5 algorithm is used to form a structured corrosion performance data table. Specifically, based on the collected image data, the Labelmg annotation tool is used to annotate the corrosion spots, the corrosion areas formed by a large number of connections, and the zinc-covered areas, and the corrosion area is calculated based on the area of ​​the annotated box. Covering area with zinc ; The corrosion area is used to calculate the corrosion area ratio of the coating, and the final output includes structured data of coating thickness, corrosion time, adhesion, corrosion ratio area, and zinc coverage area at each moment.

4. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 3 is characterized in that: The corrosion state of the time series is generated by combining the interpolation method with the kinetic equation based on the corrosion performance data table. Specifically, the corrosion area corresponding to time t is calculated using the linear interpolation method according to the time points and state variables provided in the corrosion performance data table. Covering area with zinc ; Calculation of corrosion rate based on kinetic equations and protection capabilities ; The corrosion state at a single time point is expressed as ; Representing a time series as a collection of states at multiple time points , the final output is the corrosion state of the time series.

5. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 4 is characterized in that: The data collected about the coating in the corrosive environment specifically includes collecting the coating thickness, corrosion time, adhesion and image data, and uploading the collected data to a database.

6. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 5, characterized in that: The initial population is generated based on the reinforcement learning model, and the optimization adjustment strategy is combined with the NSGA-II algorithm to generate a set of k multi-objective optimization solutions as the initialization solution set. ; Defines the maximum usage time of the coating As the first optimization objective of the adjustment strategy, the corrosion area is used As the second optimization goal of the adjustment strategy, define the adhesion measurement value As the third optimization objective of the adjustment strategy, the solution set is initialized As the initial population input of the NSGA-II algorithm; Initialize the solution set The solutions in the non-dominated sort are sorted, and the crowding distance is calculated for all non-dominated solutions after sorting. ; Set the crowding distance threshold ,like Less than , then the e-th non-dominated solution Put it into the preferred strategy set, otherwise it will not be adopted; Based on the selected optimal strategy set, crossover and mutation operations are performed to generate new sub-strategies , and integrate the generated sub-strategies with the original optimal strategy into a new solution set ; Set the maximum number of iterations Iter. When the number of iterations reaches the upper limit Iter, output the strategy with the minimum congestion distance in the solution set. as the optimal adjustment strategy.

7. The coating corrosion resistance prediction method based on the reinforcement learning model according to claim 6, characterized in that: The experience data generated in the process of generating and optimizing the adjustment strategy is uploaded to the database to form a closed-loop feedback. Specifically, the coating life, corrosion area, and adhesion data of different coatings in different environments are recorded, and the experience of generating and optimizing the adjustment strategy based on the data is uploaded to the database to form a closed-loop feedback.

8. A coating corrosion resistance prediction system based on a reinforcement learning model, based on the coating corrosion resistance prediction method based on a reinforcement learning model according to any one of claims 1 to 7, characterized in that: include, Data acquisition module, used to collect various data of coating samples in corrosive environment; The target detection module is used to detect and identify the collected coating corrosion images using the YOL v5 algorithm and extract key corrosion features; Time series generation module, used to generate time series corrosion states using interpolation methods and dynamic equations; Reinforcement learning model training module, which is used to train the reinforcement learning module using the DQN algorithm and output the adjustment strategy after training is completed; Pareto-optimal initial population generation module, used to generate multiple adjustment strategies using reinforcement learning models; NSGA-II optimization module, used to optimize the strategy set and output the optimal adjustment strategy; The experience storage and learning module is used to save the experience generated in the process of generating and optimizing strategies and form a closed-loop feedback.

9. A computer device comprising: memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the coating corrosion resistance prediction method based on the reinforcement learning model described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the coating corrosion resistance prediction method based on the reinforcement learning model according to any one of claims 1 to 7 are implemented.

Images