Steel plate pitting corrosion simulation method and system based on cellular automaton model

Abstract

The present invention relates to the technical field of computer-aided design, and in particular to a steel plate pitting corrosion simulation method and system based on a cellular automaton model. The method comprises: obtaining each inclined surface curve segment corresponding to each pit morphology curve, and determining a directional variation index at each depth position according to the depth position of the inclined surface curve segment; obtaining all inclined surface areas inside the pit under different solution concentrations and different corrosion times to analyze corrosion anomalies, and determining a local erosion progression coefficient at each depth position in combination with an expansion coefficient; obtaining the number of intersections between the depth positions of each type of inclined surface and each type of local erosion progression coefficient, and generating a dynamic rule for the evolution of cells toward the corresponding type of inclined surface in combination with the directional variation index; and obtaining a cellular automaton model with the best fit by training an LSTM network based on the dynamic rule, which avoids the problem of directional evolution under finite element rules and provides powerful assistance for studying the real pitting corrosion process of steel plates.

Description

Technical Field

[0001] The present invention relates to the technical field of computer-aided design, and in particular to a steel plate pitting simulation method and system based on a cellular automaton model. Background Art

[0002] Metal corrosion damage is a major focus of durability research on steel-concrete composite structures in civil engineering, and pitting corrosion is the primary form of steel plate corrosion. Pitting corrosion is both hidden and sudden, especially in industries such as construction, bridges, and nuclear power. Pitting corrosion can easily lead to perforation of steel plates, which in turn can cause stress concentration, a sharp drop in bearing capacity, and even catastrophic accidents such as collapse. Simulation and prediction of pitting corrosion are therefore necessary. Traditional experimental methods are costly and time-consuming, making it difficult to dynamically observe the evolution of pitting corrosion. Therefore, in recent years, cellular automata (CAs) have been widely used to simulate the mesoscopic dynamic behavior of metal corrosion due to their discretization and rule-driven properties.

[0003] The simulation of steel plate pitting corrosion based on the cellular automaton model focuses on simulation and mesoscopic scale. The metal and corrosion environment are divided into multiple "cells". Each cell updates its state according to local rules (such as electrochemical reactions and stress distribution). The development process of the pit surface and the thickness direction of the steel can be simulated separately, which is used to predict the in-service life of the structure.

[0004] During the cellular automaton evolution of steel plate pitting corrosion, complex dynamic competition between rules may occur within each cell. Because pitting corrosion initiation and propagation involve multiple competing processes, including electrochemical dissolution, stress concentration, and passive film rupture, a single cell may simultaneously evolve under multiple rules. However, finite element rules are insufficient to capture the random influence of various factors on the corrosion process, resulting in a fixed evolutionary pattern. This inability to describe the irregular evolution of pits in time and space is detrimental to steel plate pitting corrosion simulation. Summary of the Invention

[0005] In order to solve the technical problem that the evolution direction of the pits in the above-mentioned existing pitting simulation is a fixed pattern and cannot describe its irregular evolution in time and space, which is not conducive to the simulation of steel plate pitting corrosion, the purpose of the present invention is to provide a steel plate pitting corrosion simulation method and system based on a cellular automaton model. The technical solutions adopted are as follows:

[0006] One embodiment of the present invention provides a method for simulating pitting corrosion of a steel plate based on a cellular automaton model, the method comprising the following steps:

[0007] From the CA-FEM coupling model, the slope curve segments corresponding to the pit morphology curves at several corrosion times under the conditions without and with tensile stress generated by the CA model are obtained, where the CA rule is driven by the tensile stress field calculated by FEM.

[0008] According to the depth position of each inclined curve segment, analyzing the morphological anomalies of the inclined curve segments at different depth positions and the rule constraints at different depth positions, and determining the directional variation index at each depth position;

[0009] Determine the expansion coefficient of the pit surface at each of two consecutive etching times for the same solution concentration; analyze the corrosion anomaly of all inclined surface areas within the pit at different solution concentrations and different etching times, and determine the local erosion progression coefficient at each depth position at different solution concentrations based on the expansion coefficient;

[0010] Obtaining the number of intersections between each type of slope and each type of depth position of the local erosion progression coefficient, and combining the directional variation index of each depth position to generate a dynamic rule for each cell to evolve toward the corresponding type of slope;

[0011] On the basis of the dynamic rules, the cellular automaton model with the best fit for realizing the pitting corrosion simulation of steel plates is obtained by training the LSTM network.

[0012] Furthermore, the depth positions of the respective inclined curve segments are analyzed based on the morphological anomalies of the inclined curve segments at different depth positions and the rule constraints at different depth positions, and the directional variation index of each depth position is determined, including:

[0013] Sorting all the inclined surface curve segments according to a preset order of average depth to obtain an inclined surface curve segment set; calculating the difference between the slope of each inclined surface curve segment other than the first inclined surface curve segment and the slope of the previous inclined surface curve segment in the inclined surface curve segment set, to obtain the degree of morphological abnormality of the inclined surface curve segment at the corresponding depth position;

[0014] Obtain the change in the horizontal width difference between the same depth positions in the etch pit morphology curves of the cell under the set rule in the non-tensile stress environment and the tensile stress environment for each two consecutive corrosion times, and determine the rule constraint coefficient for each depth position;

[0015] The degree of morphological anomaly at each depth position is adjusted using the rule constraint coefficient to obtain a directional variation index at each depth position; wherein the degree of morphological anomaly and the rule constraint coefficient are both positively correlated with the directional variation index.

[0016] Furthermore, the determining of the rule constraint coefficient for each depth position includes:

[0017] Taking any depth position as the target depth position, the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each two consecutive corrosion times of the cell under the tensile stress environment is calculated, and recorded as the first abnormality index of the pitting pit at the target depth position under the natural rule set in the tensile stress environment;

[0018] The horizontal width difference is the difference between the horizontal widths of the target depth positions in the etch pit profile curves of two consecutive etch times, and the etch time is the time interval corresponding to the etch start time point to the etch end time point;

[0019] Calculate the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each two consecutive corrosion times in the cell under the non-tensile stress environment, and record it as the second abnormality index of the pitting pit at the target depth position under the natural rule set in the non-tensile stress environment;

[0020] A difference value between the first anomaly index and the second anomaly index is calculated, the difference value is normalized to obtain a normalized value, and the normalized value is used as a rule constraint coefficient of the target depth position.

[0021] Furthermore, the determination of the expansion coefficient corresponding to the etch pit surface at every two consecutive etching times for the same solution concentration includes:

[0022] For any two consecutive corrosion times, the pit surface is segmented using triangular meshes of the same specifications.

[0023] The curvature, torsion and coordinates of each vertex position on the grid are extracted, and the Euclidean norm of the coordinates, curvature and torsion between different positions on the two pit surfaces is calculated. The average Euclidean norm is used as the matching result of the corresponding positions.

[0024] A minimum matching result is selected from the matching results at all positions, and the minimum matching result is used as the expansion coefficient corresponding to the etch pit surface under two consecutive etching times.

[0025] Furthermore, the analysis of the corrosion anomaly in all the slope areas inside the etch pit under different corrosion times and different solution concentrations, combined with the expansion coefficient, determines the local corrosion progression coefficient at each depth position under different solution concentrations, including:

[0026] Obtaining three-dimensional morphological data of the etch pit at each etching time under several different solution concentrations, and segmenting the etch pit according to the three-dimensional morphological data to obtain the corresponding inclined surface areas inside the etch pit;

[0027] Acquire all normal angles between each slope region and all adjacent slope regions along the depth position sequence from top to bottom, and determine the dynamic rule competition index of each slope region in combination with the expansion coefficient;

[0028] Projecting the dynamic rule competition index onto a two-dimensional plane, and obtaining a relationship curve between the dynamic rule competition index and the depth position by curve fitting;

[0029] According to the relationship curve, the average derivative of all dynamic rule competition indicators at each depth position of each solution concentration in time series is calculated to determine the local erosion progression coefficient at each depth position at different solution concentrations.

[0030] Furthermore, obtaining all normal angles between each bevel region and all bevel regions adjacent to its edges includes:

[0031] A coordinate system is created with the vertical direction as the Y axis and the horizontal line where the centroid of the target slope area is located as the X axis, and a normal of each slope area passing through the centroid is obtained in the coordinate system;

[0032] According to the normal of each slope region passing through the centroid, all normal angles between each slope region and all the slope regions adjacent to its edges are obtained.

[0033] Furthermore, determining the dynamic rule competition index of each slope area includes:

[0034] Take any slope area as the target slope area, and take all slope areas adjacent to the edges of the target slope area as the neighboring slope areas;

[0035] The dynamic rule competition index of the target slope area is determined according to the cosine value of the normal angle between the target slope area and each neighboring slope area, the Z-axis coordinate value of the centroid of each neighboring slope area, and the expansion coefficient.

[0036] Furthermore, determining the dynamic rule competition index of the target slope area according to the cosine value of the normal angle between the target slope area and each neighboring slope area, the Z-axis coordinate value of the centroid of each neighboring slope area, and the expansion coefficient includes:

[0037] Calculate the product of the cosine value of the normal angle between the target slope area and each neighboring slope area and the Z-axis coordinate value of the centroid of the corresponding neighboring slope area, and use the average of all products as the first dynamic rule competition factor of the target slope area;

[0038] exponentially amplifying the expansion coefficient, and using the amplified value as a second dynamic rule competition factor of the target slope area;

[0039] The first dynamic rule competition factor and the second dynamic rule competition factor of the target slope area are fused to obtain a dynamic rule competition index of the target slope area.

[0040] Furthermore, the generation of dynamic rules for each cell to evolve toward the corresponding class slope includes:

[0041] Calculate the product of the number of intersections between each type of slope and the depth position of each type of local erosion progression coefficient and the directional variation index of the corresponding depth position to obtain the membership degree of each type of slope to each type of local erosion progression coefficient;

[0042] For any cell, the membership degree of each type of slope to the weighted mean of the local erosion progressive coefficients of all cells above the cell is determined, and the type of slope corresponding to the largest membership degree is taken as the evolution direction of the cell, so that the cell generates a dynamic rule for evolving towards the corresponding type of slope.

[0043] Another embodiment of the present invention also provides a steel plate pitting corrosion simulation system based on a cellular automaton model, comprising a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement the steel plate pitting corrosion simulation method based on the cellular automaton model.

[0044] The present invention has the following beneficial effects:

[0045] The present invention provides a method and system for simulating steel plate pitting corrosion based on a cellular automaton model. This method considers the influence of tensile stress and facilitates the prediction of corrosion fatigue life under complex loads. Based on this, the finite element method (FE) is insufficient to represent the random influence of various factors on the corrosion process, resulting in a fixed evolutionary pattern. The present invention proposes optimizing the cellular evolution rules by combining a long short-term memory (LSTM) network and local erosion progression. Specifically, the directional variation index at different depths under the established finite element method is calculated, as well as the local erosion progression coefficient at different concentrations and depths under an accelerated corrosion experiment. The membership of each type of inclined surface to different local erosion progression coefficients is calculated. Combined with the LSTM network, the inclined surfaces in the CA model, which evolves gradually in time, are dynamically matched to each other, increasing the likelihood of their evolution in different directions. This avoids the directional evolution problem under the finite element method, facilitates the accurate description of the irregular evolution in time and space, and provides powerful assistance for studying the real pitting corrosion process of steel plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 Schematic diagram of the CA model for pitting corrosion of steel plates;

[0048] Figure 2 A flowchart of a method for simulating steel plate pitting corrosion based on a cellular automaton model according to an embodiment of the present invention;

[0049] Figure 3 Schematic diagram of the morphology curve of the side profile of the corrosion pit under the same stress and different simulated corrosion times obtained by finite element model simulation in an embodiment of the present invention;

[0050] Figure 4 Schematic diagram of the morphology curve of the side profile of the corrosion pit at different stress levels under the same simulated corrosion time obtained by finite element model simulation in an embodiment of the present invention;

[0051] Figure 5 An example of etch pit morphology Figure 1 ;

[0052] Figure 6 An example of etch pit morphology Figure 2 . DETAILED DESCRIPTION

[0053] To further illustrate the technical means and effects employed by the present invention to achieve its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementations, structures, features, and effects of the technical solutions proposed by the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0054] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0055] The application scenarios targeted by the present invention may be:

[0056] Simulating steel pitting corrosion solely based on idealized discretized causal relationships (local rules) and corrosion expansion rates (corrosion concentrations) is unable to describe the irregular evolution of pits in time and space, hindering effective simulation. This is because pitting corrosion initiation and expansion involves multiple competing processes, including electrochemical dissolution, stress concentration, and passive film rupture. Multiple rules may evolve simultaneously within a single cell, and simulating the dynamic competition between these rules is currently a research challenge.

[0057] In order to overcome the defects of the existing steel plate pitting corrosion simulation, an embodiment of the present invention provides a steel plate pitting corrosion simulation method based on a cellular automaton model.

[0058] First, preparation before creating the cellular automaton.

[0059] A cellular automaton is a discrete dynamical system consisting of regularly arranged basic units (cells), whose states evolve synchronously in discrete time and space according to local rules.

[0060] A cellular automaton consists of four elements: cells, cellular space, neighbors, and rules.

[0061] Cells, the basic units, have finite discrete states (such as 0 / 1, color, etc.) and are distributed on a regular grid; cellular space, the grid on which cells are distributed, can be one-dimensional, two-dimensional, or multi-dimensional; neighbors, the set of neighboring cells that determines the state of a cell at the next moment, for example, the left and right neighbors of a one-dimensional cell (radius r = 1), the Moore neighborhood (8-neighborhood) of a two-dimensional cell; rules, local transition functions, calculate the state at the next moment based on the state of the current cell and its neighbors.

[0062] Cellular automata emerge with global complexity through the iteration of simple local rules. Figure 1 The CA model of steel plate pitting corrosion is shown in Figure 1, which consists of several two-dimensional cells. Figure 1 In the figure, H represents hydrogen ions, M represents metal matrix, F represents oxide film layer, R represents iron ions, and P represents iron hydroxide.

[0063] Based on the steel's application and field, the external environmental factors that cause pitting corrosion in steel plates are simulated, including salinity, temperature, humidity, light, pH, and other factors. This allows for the pre-determined corrosion rate, including both the lateral expansion rate and the longitudinal corrosion rate. Furthermore, the electrochemical reaction equation is clarified to provide a mechanistic basis for the deduction of cellular automata.

[0064] It should be noted that the metal hydroxide is attached to the substrate in the form of precipitation, forming a passivation protective film. In addition to the metal cells, all other cells in the model can move randomly.

[0065] Second, define the initialization cell settings and distribution.

[0066] In the 2D drawing software that comes with Windows, the various elements involved in the pitting corrosion process are converted into each cell in the cellular space to simulate the pitting corrosion process of metal materials in a corrosive environment. The molecules, ions or atoms represented by the actual chemical reactions are represented by cells. A total of 10 types of cells are set: W represents water, H represents hydrogen ions; M1 represents the oxide film layer; R1 represents iron ions; P1 represents iron hydroxide generated by the reaction; M2 represents the base steel plate; R2 represents ferrous ions; P2 represents ferrous hydroxide; O represents oxygen; and C represents chloride ions.

[0067] It is worth noting that the appearance of oxide film in CA simulation will shift the simulation from simple corrosion expansion to passivation to activation of dynamic competition system, significantly improving the realism.

[0068] Then import the PNG image (Portable Network Graphics) and grayscale it to get a 0, 1 matrix. The initially imported image does not include the oxide layer and the solution layer, so the construction layer and the solution layer are required.

[0069] Specifically, a double loop assignment statement is used. Because the line of one layer of thickness is dim, a three-layer thickness is used. ( ) is used to represent the thickness of the solution layer, which is 150 layers. ) to represent. Define the initial sentinel to record the corrosion area, with an initial value of 0. Define the upper and lower limits of vertical corrosion expansion, which are 40 and 350 respectively. Define the center position of horizontal corrosion expansion ( ), define the initial expansion range ( ), the number of expansion time steps is 5.

[0070] Third, set the cellular chemical corrosion reaction and diffusion parameters.

[0071] Among them, the parameters involved in the cellular reaction rule include: when there are at least six acidic sites H near the metal site M1, the reaction occurs with a corrosion probability of P1_corr; the M1 site will be replaced by the R1 substance.

[0072] When at least three neighbors of the active metal R1 are neutral sites W, Hydrolyzed to , the hydrolysis probability is P2_Hyd2, and the product is replaced by P1; the above is the law of electrochemical reaction of oxide film.

[0073] There are at least two acidic sites H near the metal site M2, and Fe oxidation occurs to generate For the forward reaction, the corrosion probability is P2_corr; the M2 site will be replaced by the R2 species.

[0074] When there is at least one acidic site H near the active metal site R2, Oxidized to , the oxidation probability is P2_ox; the R2 site will be replaced by D substance. When at least two neighbors of the active metal R2 are neutral sites W, Hydrolyzed to , the hydrolysis probability is P1_Hyd1; R2 and the two W sites are replaced by P2 and two H substances, respectively.

[0075] The final reaction is Hydrolyzed to , the hydrolysis probability is P2_Hyd2. When at least three of the adjacent positions of the D site are neutral sites W, the D and three W sites will be replaced by P1 and three H substances, respectively.

[0076] As an exemplary embodiment, the setting of diffusion parameters includes:

[0077] The corrosion generation probabilities of seven substances (water, hydrogen ions, oxide film layer, ferric hydroxide, ferrous hydroxide, oxygen, and chloride ions) were initially set (0.1433, 0.2218, 0.183, 0.1943, 0.096, 0.1252, and 0.1252, respectively); five diffusion coefficients were set (0.5, 0.3, 0.08, 0.06, and 0.01, respectively); the stress factor was set to 0.2, then the tensile coefficient was 1.2; the sedimentation factor was 2, then the sedimentation coefficient was 1.2; two hydrolysis probabilities were defined as 0.25 and 0.3; and the analysis step was defined as 30.

[0078] Fourth, build a bridge between cellular automata and finite element models.

[0079] First, in the cellular automaton model before stress is added, the position of each cell is recorded to facilitate one-to-one correspondence with the units in the finite element model;

[0080] Secondly, the finite element method calculates the stress and strain of the cell, so the establishment Create a finite element model, create materials, assign elastic-plastic constitutive properties to components, and define assemblies;

[0081] Then, create the component mesh in advance. The mesh is A small square grid corresponds to each cell;

[0082] Next, create two analysis steps. The first step is to define the cellular corrosion model, and the second step is to define the loading and perform finite element calculations of stress and strain.

[0083] Finally, the interactions are defined and, during the initial analysis, a cellular automaton model of the corrosion morphology of the electrochemical corrosion formation is created. The component becomes the shape of a pit, and the model is imported into the finite element model. After the mesh is divided, the finite element pit is created.

[0084] The transition from metastable to stable states in steel plate pitting corrosion is random. Cells at the pit edge may preferentially dissolve due to high stress, while the diffusion of surrounding corrosive substances slows corrosion and reforms the metal oxide film. Cellular automata rules consider the random walk of individual cells. Introducing dissolution and passivation probabilities can influence the direction of the cell's walk in three-dimensional space, meaning each direction has a weight. Furthermore, under finite element rules alone, each cell is likely to experience directional variations, which is detrimental to simulating the pitting corrosion trend in steel plates. Metastable states are characterized by the possibility of repair, while stable states represent continuous development.

[0085] In order to overcome the defects of the above-mentioned cellular evolution rules, one embodiment of the present invention uses a combination of long short-term memory network and local erosion progression to simulate the dynamic rule competition that may occur at different cellular positions. Based on this evolution rule, a cellular automaton model with the best fit for steel plate pitting corrosion simulation is constructed, such as Figure 2 As shown, the following steps are included:

[0086] S1. Obtain the slope curve segments corresponding to the pit morphology curves at several corrosion times under the conditions without tensile stress and with tensile stress generated by the CA model from the CA-FEM coupling model, where the CA rule is driven by the tensile stress field calculated by FEM.

[0087] It should be noted that in the presence of tensile stress, tensile stress can promote the increase of corrosion rate and the expansion of corrosion morphology, which is mainly reflected in the expansion of vertical corrosion depth, that is, the change of depth position.

[0088] As an exemplary embodiment, obtaining the inclined curve segment includes:

[0089] The FEM (Finite Element Method) calculates the tensile stress field and maps it to a CA mesh. CA simulates the evolution of the pits, extracting topography curves at different corrosion times during the evolution process. Using a slope segmentation algorithm, the slope curve segment features are statistically analyzed to obtain the individual slope curve segments corresponding to each pit topography curve. Of course, topography curves can also be obtained through the aforementioned bridge step between the cellular automaton and the finite element model.

[0090] The slope of the pit needs to be clearly defined through normal vector definition or curvature calculation, such as the area where the angle between the normal and the vertical direction is greater than 15 degrees. For the slope curve segment, the slopes of the pit morphology curve are obtained, and the data points corresponding to the slopes are used as segmentation points to segment the pit morphology curve to obtain the curve segments. Since the morphology curve is the slope curve of the pit, the curve segment can be regarded as the slope curve segment. The corrosion time is the time interval from the corrosion start time point to the corrosion end time point, and the unit is day (d), such as 1 to 5 (5d), 1 to 10 (10d), 1 to 15 (15d), etc. The number of corrosion times can be set by the implementer according to the specific actual situation and is not specifically limited.

[0091] The tensile stress is downward by default. The morphology curve of the corrosion pit side profile obtained by finite element model simulation under the same stress and different simulated corrosion time is as follows: Figure 3 As shown in the figure, the morphology curve of the side profile of the corrosion pit at different stress levels under the same simulated corrosion time obtained by the finite element model is shown in Figure 4 As shown, in Figure 3 and Figure 4 In the figure, F represents the stress magnitude, T represents the simulated corrosion time, the horizontal axis represents the width position, and the vertical axis represents the depth position.

[0092] Different slopes may represent nonlinear pitting trends occurring at localized time periods and locations. Therefore, the erosion slope can be used as the basic load-bearing unit of dynamic rules. Each slope curve segment has only one slope, so the slope segment can be segmented based on whether the slope value changes. By analyzing the evolution of the slope curve segment, the degree of anisotropy of stress corrosion can be quantified, providing a basis for life prediction.

[0093] Thus, this embodiment has obtained several inclined curve segments for analyzing directional variations at different depths.

[0094] S2, according to the depth position of each inclined curve segment, analyze the morphological anomalies of the inclined curve segments at different depth positions and the rule constraints at different depth positions, and determine the directional variation index of each depth position.

[0095] In order to overcome the influence of the directional variation of cells on the simulation of pitting corrosion trend of steel plates, the directional variation index at different depth positions is quantified based on the depth position characteristics of the inclined curve segment. The directional variation index can characterize the missing evolutionary possibility of cells at different corrosion depth positions of the set regular cellular model. The larger the directional variation index, the more competition factors need to be given to the cells at the corresponding depth position to make them carry richer dynamic competition randomness.

[0096] As an exemplary embodiment, the above step S2 can be implemented through steps S21 to S23:

[0097] S21, sorting all the inclined surface curve segments according to a preset order of average depth to obtain an inclined surface curve segment set; in the inclined surface curve segment set, calculating the difference between the slope of each inclined surface curve segment other than the first inclined surface curve segment and the slope of the previous inclined surface curve segment, and obtaining the degree of morphological abnormality of the inclined surface curve segment at the corresponding depth position.

[0098] Here, the degree of morphological abnormality refers to the degree of abnormality of the pit corrosion corresponding to the inclined curve segment.

[0099] The cellular automaton has already been programmed with various rules for environmental parameters such as dissolution probability, solution concentration, and temperature. There may also be several uncaptured dynamic rules. These dynamic rules are generally concentrated at the initial stage of corrosion. Competition between dynamic rules decreases with increasing depth, so it is necessary to sort and analyze all inclined curve segments by average depth. Generally, in a two-dimensional finite element model, the formation of the lower inclined surface is caused by competition between dynamic rules from the corrosion process in the upper layer.

[0100] In this example, all slope curve segments are sorted from top to bottom according to their average depth to obtain a slope curve segment set. Each data point in a slope curve segment has a corresponding depth, and the average depth of all data points in the same slope curve segment (average depth) is used as the depth of the corresponding slope curve segment. The difference between two slopes generally refers to the absolute value of the difference between the two slopes. The slope difference can reveal the local dynamic characteristics of the pit morphological evolution. The larger the slope difference, the more significant the sudden change in the pit expansion speed or direction in the local area, reflecting the uneven distribution of the stress field, and the more abnormal the morphological characteristics of the slope curve segment at the depth position.

[0101] It should be noted that each depth position in the pit morphology curve has its corresponding morphology abnormality degree. For the morphology abnormality degree at the same depth position, the average value of the morphology abnormality degree can be used as the morphology abnormality degree at the corresponding depth position. The morphology abnormality degree of each depth position corresponding to the same inclined curve segment is the same; the morphology abnormality degree is obtained by analyzing the difference in slope between the inclined curve segment and its adjacent previous inclined curve segment. The first inclined curve segment in the inclined curve segment set does not have a previous inclined curve segment, so the morphology abnormality degree analysis is not performed on the first inclined curve segment in the sequence.

[0102] S22, obtaining the change difference of the horizontal width difference between the same depth position in the etch pit morphology curve of each cell under the set rule in the non-tensile stress environment and the tensile stress environment, and determining the rule constraint coefficient of each depth position.

[0103] Here, the rule constraint coefficient is used to quantify the extent to which the slope morphological anomaly at depth position is not driven by dynamic competition rules but is constrained by finite evolution rules; the depth position when analyzing the horizontal width and the depth position when analyzing the degree of morphological anomaly show a one-to-one correspondence, that is, each depth position can correspond to both a degree of morphological anomaly and a rule constraint coefficient.

[0104] As an exemplary embodiment, the above step S22 can be implemented through steps S221 to S223:

[0105] S221, taking any depth position as the target depth position, calculating the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each cell at every two consecutive corrosion times under the tensile stress environment, and recording it as the first variation index of the pitting pit at the target depth position under the natural rule set in the tensile stress environment.

[0106] Here, any depth position is any depth position on the pit morphology curve; the horizontal width difference can represent the evolution difference of each depth position in the curve of continuous corrosion time of the cell under a tensile stress environment under the set rules; the first anomaly index can represent the variability of different depth positions of the pitting pit under the set natural rules of tensile stress interference.

[0107] In this embodiment, generally, the horizontal width of a larger etching time is larger, while the horizontal width of a smaller etching time is smaller, so the horizontal width difference can be the difference between the horizontal widths between the target depth positions in the etch pit morphology curves of two consecutive etching times, or the absolute value of the difference between the two horizontal widths.

[0108] S222, calculating the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each cell at every two consecutive corrosion times in the non-tensile stress environment, and recording it as the second variation index of the pitting pit at the target depth position under the set natural rule in the non-tensile stress environment.

[0109] Here, the horizontal width can represent the degree of horizontal corrosion expansion, and the horizontal width difference can represent the evolution difference of the cell at each depth position under continuous tensile stress in a tensile stress-free environment under the set rules; the second variation index represents the variability of pitting pits at different depths under the set natural rules without tensile stress interference.

[0110] S223 , calculating a difference between the first anomaly index and the second anomaly index, normalizing the difference to obtain a normalized value, and using the normalized value as a rule constraint coefficient of the target depth position.

[0111] In this embodiment, the absolute value of the difference between the two variation indices is calculated as the difference value; the difference value is normalized in inverse proportion using the exp(-) function, and the normalized value is used as the rule constraint coefficient corresponding to the target depth position.

[0112] In the process of calculating the rule constraint coefficient, the smaller the difference between the two anomaly indicators, the more it indicates that the set rule will produce a fixed evolution direction even in the presence of tensile stress. Therefore, the morphological anomaly of the slope at the target depth position is not driven by the dynamic competition rule, but is constrained by limited evolutionary rules.

[0113] By referring to the calculation process of the rule constraint coefficient of the target depth position, the rule constraint coefficient of each depth position can be obtained. Each depth position in the etch pit profile curve has its corresponding rule constraint coefficient.

[0114] S23, using the rule constraint coefficient to adjust the degree of morphological anomaly at each depth position to obtain a directional variation index at each depth position; wherein the degree of morphological anomaly and the rule constraint coefficient are both positively correlated with the directional variation index.

[0115] In this embodiment, the product of the rule constraint coefficient of each depth position and the degree of abnormality of the morphology of the slope segment where it is located can be used as the directional variation index of each depth position under the set rule.

[0116] In the finite element model, the more unusual the morphology and the greater the constraint of the rules, the greater the probability that the established rules will undergo directional variation under finite element conditions at the corresponding depth. The greater the probability of directional variation at a certain depth, the more important it is to assign competition factors to the cells at that depth, so that they carry a richer dynamic competitive randomness.

[0117] So far, this embodiment has obtained the directional variation index of each depth position in the CA model.

[0118] S3, determine the expansion coefficient corresponding to the pit surface at every two consecutive corrosion times for the same solution concentration; analyze the corrosion anomaly based on all the inclined surface areas inside the pit at different corrosion times for several different solution concentrations, and determine the local corrosion progression coefficient at each depth position at different solution concentrations in combination with the expansion coefficient.

[0119] Here, the expansion coefficient refers to the expansion state of the etch pit morphology at one corrosion time relative to the etch pit morphology at the previous corrosion time under the same solution concentration. The larger the expansion coefficient, the greater the overall corrosion process; the local corrosion progression coefficient refers to the local corrosion progression relationship at different depth positions under different corrosion rates.

[0120] In this embodiment, when conducting an accelerated corrosion experiment, different concentrations of corrosion solutions are dripped onto the surfaces of different steel samples. The corresponding corrosion rate can be obtained in advance according to the concentration of the corrosion solution. Then, the experimental samples are placed in a non-laboratory environment, that is, an environment in the field of steel application. When the corrosion time reaches 5d, 10d, 15d, 20d, 25d, 30d..., a batch of samples are taken out for observation of the morphology of the corrosion pits. The size and number of the corrosion time can be consistent with the size and number of the corrosion time recorded in step S1 above. Among them, the morphology of the corrosion pits is shown in the following example: Figure 1 like Figure 5 As shown, an example of the etch pit morphology Figure 2 like Figure 6 It should be noted that Figure 5 and Figure 6 It is essentially a schematic diagram and is only used to show the morphological characteristics of the etch pits at different etching times.

[0121] Depend on Figure 5 and Figure 6 It can be seen that in the early stages of corrosion, the sample forms a majority of nearly circular pits. As the corrosion time increases, the individual pits gradually develop in the surrounding and depth directions, presenting a hemispherical shape, and eventually continue to develop deeper at the bottom of the hemisphere. Therefore, in theory, the direction of the tensile stress at each cell position should be divergent in the early stage and concentrated downward in the later stage. Based on this, it can be seen that: assuming that there are tensile stresses in several directions on each cell, the number and angle of the cell tensile stress directions should be a function relationship that shows an inversely proportional nonlinear decay with the corrosion time or corrosion depth. Among them, the pit morphology data can be obtained by laser scanning the inside of the pit of the experimental sample.

[0122] The above step S3 can be implemented through steps S31 to S32:

[0123] S31, determining the expansion coefficient corresponding to the etch pit surface at every two consecutive etching times under the same solution concentration.

[0124] Before determining the expansion coefficient, the pit surfaces of each sample at any concentration and at different etching times are first obtained. The pit surfaces can be obtained by laser scanning the pit topography. By analyzing the match between the pit surface at each etching time and the pit surface at the previous etching time, the corresponding expansion coefficient of the pit surface at each etching time for the same solution concentration is quantified. The worse the match between the two pit surfaces, the greater the expansion degree and the larger the expansion coefficient.

[0125] As an exemplary implementation, the above step S31 can be implemented through steps S311 to S313:

[0126] S311, for any two consecutive etching times, the etch pit surface is grid-segmented using triangular meshes of the same specification.

[0127] In this embodiment, the implementation process of grid segmentation is a prior art and is not within the scope of protection of the present invention, and will not be processed in detail here.

[0128] S312, extracting curvature, torsion and coordinates for each vertex position on the grid, calculating the Euclidean norms of coordinates, curvature and torsion between different positions in the two pit surfaces, and taking the average Euclidean norm as the matching result of the corresponding positions.

[0129] In this embodiment, the average Euclidean norm refers to the average value of the Euclidean norms of three dimensions at the same position. Analyzing the Euclidean norm is to analyze the matching of the three-dimensional morphologies at two corrosion times.

[0130] S313 , selecting a minimum matching result from the matching results at all positions, and using the minimum matching result as the expansion coefficient corresponding to the etch pit surface under two consecutive etch times.

[0131] In this embodiment, the minimum matching result can best characterize the overall matching of the two pit surfaces. The larger the minimum matching result, the less matching the two pit surfaces are. The larger the expansion coefficient, the higher the degree of increase in pit corrosion expansion. Each corrosion time is matched with the previous corrosion time for morphology analysis, and its corresponding expansion coefficient can be obtained. The expansion coefficient of the pit surface at the t-th corrosion time is recorded as .

[0132] S32, analyzing the corrosion anomaly of all slope areas inside the etch pit at different solution concentrations and different etching times, and determining the local erosion progression coefficient at each depth position at different solution concentrations in combination with the expansion coefficient.

[0133] As an exemplary implementation, the above step S32 can be implemented through steps S321 to S324:

[0134] S321 , obtaining three-dimensional morphological data of the etch pits at each etching time under different solution concentrations, and segmenting the etch pits according to the three-dimensional morphological data to obtain corresponding inclined surface areas inside the etch pits.

[0135] In this embodiment, segmenting the slope region from the 3D pit topography data requires a combination of feature extraction and region growing algorithms. Specifically, the slope region can be determined by performing normal vector analysis, curvature filtering, and region growing on the 3D pit topography data. The method for obtaining the slope region is prior art and falls outside the scope of this invention, so it will not be elaborated on here.

[0136] S322, along the depth position sequence from top to bottom, obtain all normal angles between each slope area and all the slope areas adjacent to its edges, and determine the dynamic rule competition index of each slope area in combination with the expansion coefficient.

[0137] Here, the dynamic rule competition index is determined by analyzing the matching between the corrosion progress of the slope area and the overall corrosion progress. The greater the mismatch between the corrosion progress of any slope area and the overall corrosion progress, the greater the probability of dynamic rule competition in the slope area.

[0138] In this embodiment, the normal angle has positive and negative values. The larger the sum of the absolute values ​​of the normal angles, the greater the difference between the inclined surface area and the surrounding inclined surface areas, and the steeper the slope. Conversely, the smaller the sum of the absolute values ​​of the normal angles, the closer the normal angles are to parallel, and the smaller the difference between the inclined surface area and the surrounding inclined surface areas, and the flatter the slope.

[0139] The higher the depth position, the steeper the slope area, which means that the tensile stress divergence is insufficient and dynamic competition of corrosion rules may have occurred; the lower the depth position, the flatter the slope area, which means that the tensile stress is not concentrated enough. At this time, the flat slope area is more likely to have entered the late stage of evolution.

[0140] Therefore, for the etch pit morphology data at each corrosion time under each solution concentration, the dynamic rule competition index of each slope area is calculated in sequence along the depth position from top to bottom.

[0141] As an exemplary implementation, the above step S322 can be implemented through steps S3221 to S3222:

[0142] S3221: Take any inclined surface area as a target inclined surface area, and take all inclined surface areas adjacent to the edges of the target inclined surface area as neighboring inclined surface areas.

[0143] S3222: Determine a dynamic rule competition index for the target slope region based on the cosine value of the normal angle between the target slope region and each neighboring slope region, the Z-axis coordinate value of the centroid of each neighboring slope region, and the expansion coefficient.

[0144] First, calculate the product of the cosine value of the normal angle between the target slope area and each neighboring slope area and the Z-axis coordinate value of the centroid of the corresponding neighboring slope area, and use the average of all products as the first dynamic rule competition factor of the target slope area.

[0145] Secondly, the expansion coefficient is exponentially amplified and the amplified value is used as the second dynamic rule competition factor of the target slope area;

[0146] Finally, the first dynamic rule competition factor and the second dynamic rule competition factor of the target slope area are fused to obtain the dynamic rule competition index of the target slope area.

[0147] As an example, the calculation formula for the dynamic rule competition index of the oth slope area can be:

[0148] Where, represents the dynamic rule competition index of the oth slope area (target slope area), e represents the natural constant, represents the expansion coefficient of the etch pit surface at the t-th etch time, represents the number of neighboring slope regions of the oth slope region at the tth corrosion time, Represents the Z-axis coordinate value of the centroid of the ith neighboring slope area of ​​the oth slope area, represents the cosine value of the normal angle between the oth slope area and the i-th neighboring slope area, represents the overall corrosion progress at the t-th corrosion time, Indicates the corrosion progress of the o-th slope region at the t-th corrosion time.

[0149] In the calculation formula of dynamic rule competition index, The larger the value, the greater the overall corrosion process. The larger the value, the more abnormal the corrosion is at the location of the o-th slope area, and abnormal corrosion essentially means that the corrosion process is slow; the larger the dynamic rule competition index is, the greater the probability that dynamic rule competition may exist at the location of the o-th slope area.

[0150] Furthermore, the step of obtaining the normal angle between the target slope area and each neighboring slope area includes: creating a coordinate system with the vertical direction as the Y-axis and the horizontal line where the center of mass of the target slope area is located as the X-axis, and obtaining the normal of each slope area passing through the center of mass in the coordinate system; based on the normal of each slope area passing through the center of mass, obtaining all normal angles between each slope area and the slope areas adjacent to all its edges.

[0151] Referring to the dynamic rule competition index of the target slope area, the dynamic competition index of all slope areas in the depth sequence from top to bottom is obtained.

[0152] S323 , projecting the dynamic rule competition index onto a two-dimensional plane, and obtaining a relationship curve between the dynamic rule competition index and the depth position through curve fitting.

[0153] In this example, the dynamic rule competition index was projected onto a two-dimensional plane and fitted using the least squares method to obtain a curve representing the relationship between the dynamic rule competition index and depth position. Furthermore, a curve representing the relationship between the dynamic competition probability and depth position within the accelerated corrosion pit at each corrosion time and its immediately preceding corrosion time was obtained under all solution concentration conditions. The implementation of the least squares method is prior art and falls outside the scope of this invention, so it will not be elaborated on here.

[0154] S324, calculating the average derivative of all dynamic rule competition indicators at each depth position of each solution concentration in time series based on the relationship curve, and determining the local erosion progression coefficient at each depth position under different solution concentrations.

[0155] In this embodiment, each solution concentration can represent a different corrosion rate, and each corrosion time can represent the corrosion progress. Then, the changes in multiple dynamic rule competition indicators at each depth position under continuous corrosion time reflect the progressive relationship of local corrosion at different depth positions under different corrosion rates.

[0156] Specifically, by calculating the average derivative of all dynamic rule competition indicators at each depth position under each solution concentration in time series, the local erosion progression coefficient at each depth position under different solution concentrations is obtained.

[0157] It should be noted that depth refers to the longitudinal corrosion progress of the corrosion process. The deeper the corrosion, the later the longitudinal corrosion progress. The depth position here is the average value of all depth positions in a single inclined surface area; the inclined surface is the different morphological results formed during the corrosion process, which can characterize the surface condition of the steel plate after corrosion; dynamic rule competition refers to the evolution of cells restricted by some rules under different corrosion progress. For example, the simultaneous existence of the two rules of metal surface oxide film regeneration and corrosion concentration dilution, as well as the simultaneous existence of more rules, may limit the final corrosion result bias.

[0158] Thus, this embodiment has determined the local erosion progression coefficient at each depth position under different solution concentrations.

[0159] S4, obtain the number of intersections between the depth positions of each type of slope and each type of local erosion progression coefficient, combine the directional variation index of each depth position, and generate the dynamic rules for each cell to evolve towards the corresponding type of slope.

[0160] Here, the directional variation index used to represent the degree of mutation in the corrosion direction at a depth position determines the missing evolution possibility of the cells at different corrosion depth positions in the set regular cellular model. The local erosion progression coefficient used to represent the relative speed of the corrosion rate at depth represents the dynamic competition probability change at different corrosion depth positions with different corrosion rates in the actual accelerated corrosion experiment.

[0161] Then, for some cells that lack the possibility of evolution, the probability of their evolution to different types of slopes is increased. The evolution to different types of slopes is determined by the local erosion progressive relationship of the upper cell position. To a certain extent, the dynamic rule competition of the upper cell position during the pitting process can be restored in the cellular automaton, and the nonlinear and random evolutionary impact on the lower cell position can be analyzed.

[0162] Among them, in order to facilitate the subsequent calculation of membership, the depth positions corresponding to the directional anomaly index and the local erosion progression coefficient can be kept consistent in size and number through numerical fitting.

[0163] In this embodiment, each type of inclined surface is a set of samples obtained through accelerated corrosion experiments. All inclined surfaces are classified according to slope, area, and depth range. The slope, area, and depth range of inclined surfaces of the same type are highly similar, and each type of inclined surface may be a direction evolved by the dynamic competition rule. Local erosion progression coefficients with the same numerical value are classified as a type of local erosion progression coefficient. Some different depth positions may have the same local erosion progression coefficient. A inclined surface is between the depth position at one end of the inclined surface and the depth position at the other end on the Z axis. Therefore, all depth positions included in each type of inclined surface can be obtained. For the depth position set of a type of local erosion progression coefficient and the depth position set of a type of inclined surface, the number of intersecting (identical) depth positions in the two depth position sets is taken as the intersection number. The classification of inclined surfaces can be achieved by using the dynamic time rule to measure the similarity between the inclined surface vectors. The vectors are composed of elements in three dimensions: slope, area, and depth range.

[0164] As an exemplary embodiment, the above step S4 can be implemented through steps S41 to S42:

[0165] S41, calculating the product of the number of intersections between each type of slope and each type of depth position of the local erosion progression coefficient and the directional variation index of the corresponding depth position, and obtaining the membership degree of each type of slope to each type of local erosion progression coefficient.

[0166] The greater the number of depth positions that intersect all depth positions where a type of slope appears and all depth positions where each type of local erosion progression coefficient appears, the greater the probability of overlapping depth positions. The greater the directional variation index at a depth position, the more it is necessary to increase the dynamic rule competition at that depth position, that is, to increase the membership of the slope to the local erosion progression coefficient at that depth position. Increasing the membership can match more possible slopes to that depth position, while decreasing the membership requires decreasing the membership.

[0167] As an example, the calculation formula for the membership degree of the local erosion progression coefficient of the r-th type slope belonging to the z-th position of the etch pit under the s-th solution concentration can be:

[0168] Where, Indicates that the rth type of inclined plane belongs to The membership degree of the local erosion progressive coefficient, represents the local erosion progression coefficient at the zth depth of the etch pit under the sth solution concentration, represents the directional variation index at the z-th depth position, Represents the rth type of inclined plane and the The number of intersections between depth locations of the type Local Erosion Progression Coefficient.

[0169] Refer to the above r-type inclined plane belongs to The calculation process of the membership degree of each type of slope to each type of local erosion progressive coefficient can obtain the membership degree of each type of slope to each type of local erosion progressive coefficient.

[0170] So far, this embodiment has obtained the degree of membership of each type of slope to each type of local erosion progression coefficient.

[0171] Furthermore, the long short-term memory network is introduced.

[0172] LSTMs (Long Short-Term Memory) excel at processing time-series data and long-term dependencies. In pitting evolution, the current state of each cell depends not only on the immediate state of neighboring cells but also on the historical evolutionary paths of itself and the surrounding area. For example, the corrosion history of the cell above (such as the accumulation of corrosion products and the frequency of passive film rupture) dynamically adjusts the evolution probability of the cell below through the LSTM's gating mechanism (input gate, forget gate, and output gate), thereby simulating local rule competition.

[0173] The input features of each cell include: Current state: such as corrosion probability, hydrolysis probability, sedimentation coefficient, sedimentation factor, diffusion probability, stress factor, tension coefficient, reaction threshold, corrosion products, etc. Neighborhood state: captures the state of the adjacent cells above each cell. Time series: state changes over the past several time steps (such as corrosion depth, solution concentration gradient, temperature change).

[0174] A cell position in the time series may correspond to different corrosion rates, and may also have multiple local erosion progression coefficients of different degrees; based on the depth position and current state index of the cell, the corresponding corrosion rate estimate is generated, and the local erosion progression coefficient at the concentration closest to the depth position is matched according to the corrosion rate; similarly, based on the depth position and state of all cells in the neighborhood above the cell, the local erosion progression coefficients of all cells in the neighborhood are obtained; the local erosion progression coefficients of all cells above the cell are weighted and averaged using the distance Gaussian function.

[0175] Each cell is associated with an LSTM unit, and the output is the probability of the evolution direction of the next time step (the probability of expanding to different slopes). Specifically:

[0176] S42, for any cell, determine the membership of each type of slope to the weighted mean of the local erosion progressive coefficients of all cells above the cell, and take the type of slope corresponding to the largest membership as the evolution direction of the cell, so that the cell generates a dynamic rule for evolving toward the corresponding type of slope.

[0177] At this point, referring to the above-mentioned process of determining the dynamic rule for a single cell to evolve toward a corresponding class slope, the dynamic rule for each cell to evolve toward a corresponding class slope is obtained.

[0178] S5, based on the dynamic rules, the LSTM network is trained to obtain the cellular automaton model with the best fit for realizing the pitting corrosion simulation of steel plates.

[0179] In this example, based on the dynamic rules governing the evolution of cells toward corresponding slopes, an LSTM network is trained to predict the slopes to which the lower cells may evolve based on the state of the upper cells. A cross-entropy loss function is then used to minimize the discrepancy between the LSTM-predicted evolution direction and the actual simulation or experimental results. A plotting function is set up to refresh the graph constantly, allowing observation of changes in pit morphology within a time step. Sentinels are used to record the pit corrosion area, width, depth, aspect ratio, and other information. Finally, the simulation results of the LSTM-CA model are compared with those of a purely probabilistic CA model to assess the fit between the pit morphology (e.g., equivalent radius, depth) and random distribution (e.g., pit density) and the experimental data, resulting in a cellular automaton model with optimal fit for steel plate pitting simulation.

[0180] The LSTM's forget gate controls the influence of historical rules, such as whether to retain the inhibitory effect of early corrosion products. The input gate determines the weight of the influence of current environmental parameters, such as solution concentration and temperature, on the evolution direction. The actual training process of the LSTM network is prior art and falls outside the scope of this invention, so it will not be elaborated here.

[0181] It is worth noting that the combination of long short-term memory networks and local erosion progression can break through the limitations of multi-dimensional rule finite element settings, dynamically adapt to the coupling of multiple factors, and reproduce the path dependence in real corrosion. For example, early corrosion pits guide the subsequent expansion direction, thereby improving the accuracy of steel plate pitting simulation.

[0182] An embodiment of the present invention also provides a steel plate pitting corrosion simulation system based on a cellular automaton model, comprising a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement a steel plate pitting corrosion simulation method based on a cellular automaton model.

[0183] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for simulating steel plate pitting corrosion based on a cellular automaton model, characterized in that: The following steps are involved: From the CA-FEM coupling model, the slope curve segments corresponding to the pit morphology curves at several corrosion times under the conditions without and with tensile stress generated by the CA model are obtained, where the CA rule is driven by the tensile stress field calculated by FEM. According to the depth position of each inclined curve segment, analyzing the morphological anomalies of the inclined curve segments at different depth positions and the rule constraints at different depth positions, and determining the directional variation index at each depth position; Determine the expansion coefficient of the pit surface at each of two consecutive etching times for the same solution concentration; analyze the corrosion anomaly of all inclined surface areas within the pit at different solution concentrations and different etching times, and determine the local erosion progression coefficient at each depth position at different solution concentrations based on the expansion coefficient; Obtaining the number of intersections between each type of slope and each type of depth position of the local erosion progression coefficient, and combining the directional variation index of each depth position to generate a dynamic rule for each cell to evolve toward the corresponding type of slope; Based on the dynamic rules, the cellular automaton model with the best fit for realizing the pitting corrosion simulation of steel plates is obtained by training the LSTM network. The method of analyzing the morphological anomalies of the slope curve segments at different depth positions and the rule constraints at different depth positions based on the depth positions of the slope curve segments to determine the directional variation index at each depth position includes: Sorting all the inclined surface curve segments according to a preset order of average depth to obtain an inclined surface curve segment set; calculating the difference between the slope of each inclined surface curve segment other than the first inclined surface curve segment and the slope of the previous inclined surface curve segment in the inclined surface curve segment set, to obtain the degree of morphological abnormality of the inclined surface curve segment at the corresponding depth position; Obtain the change in the horizontal width difference between the same depth positions in the etch pit morphology curves of the cell under the set rule in the non-tensile stress environment and the tensile stress environment for each two consecutive corrosion times, and determine the rule constraint coefficient for each depth position; The degree of morphological anomaly at each depth position is adjusted using the rule constraint coefficient to obtain a directional variation index at each depth position; wherein the degree of morphological anomaly and the rule constraint coefficient are both positively correlated with the directional variation index.

2. The steel plate pitting corrosion simulation method based on the cellular automaton model according to claim 1, characterized in that: The determining of the rule constraint coefficient for each depth position includes: Taking any depth position as the target depth position, the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each two consecutive corrosion times of the cell under the tensile stress environment is calculated, and recorded as the first abnormality index of the pitting pit at the target depth position under the natural rule set in the tensile stress environment; The horizontal width difference is the difference between the horizontal widths of the target depth positions in the etch pit profile curves of two consecutive etch times, and the etch time is the time interval corresponding to the etch start time point to the etch end time point; Calculate the variance of the horizontal width difference between the target depth positions in the pit morphology curve of each two consecutive corrosion times in the cell under the non-tensile stress environment, and record it as the second abnormality index of the pitting pit at the target depth position under the natural rule set in the non-tensile stress environment; A difference value between the first anomaly index and the second anomaly index is calculated, the difference value is normalized to obtain a normalized value, and the normalized value is used as a rule constraint coefficient of the target depth position.

3. The steel plate pitting corrosion simulation method based on the cellular automaton model according to claim 1, characterized in that: The step of determining the expansion coefficient corresponding to the etch pit surface at two consecutive etching times under the same solution concentration includes: For any two consecutive corrosion times, the pit surface is segmented using triangular meshes of the same specifications. The curvature, torsion and coordinates of each vertex position on the grid are extracted, and the Euclidean norm of the coordinates, curvature and torsion between different positions on the two pit surfaces is calculated. The average Euclidean norm is used as the matching result of the corresponding positions. A minimum matching result is selected from the matching results at all positions, and the minimum matching result is used as the expansion coefficient corresponding to the etch pit surface under two consecutive etching times.

4. The method for simulating steel plate pitting corrosion based on a cellular automaton model according to claim 1, wherein: The analysis of the corrosion anomaly in all slope areas inside the etch pit under different corrosion times and different solution concentrations, combined with the expansion coefficient, determines the local corrosion progression coefficient at each depth position under different solution concentrations, including: Obtaining three-dimensional morphological data of the etch pit at each etching time under several different solution concentrations, and segmenting the etch pit according to the three-dimensional morphological data to obtain the corresponding inclined surface areas inside the etch pit; Acquire all normal angles between each slope region and all adjacent slope regions along the depth position sequence from top to bottom, and determine the dynamic rule competition index of each slope region in combination with the expansion coefficient; Projecting the dynamic rule competition index onto a two-dimensional plane, and obtaining a relationship curve between the dynamic rule competition index and the depth position by curve fitting; According to the relationship curve, the average derivative of all dynamic rule competition indicators at each depth position of each solution concentration in time series is calculated to determine the local erosion progression coefficient at each depth position at different solution concentrations.

5. The method for simulating steel plate pitting corrosion based on a cellular automaton model according to claim 4, wherein: The step of obtaining all normal angles between each bevel region and all bevel regions adjacent to its edges includes: A coordinate system is created with the vertical direction as the Y axis and the horizontal line where the centroid of the target slope area is located as the X axis, and a normal of each slope area passing through the centroid is obtained in the coordinate system; According to the normal of each slope region passing through the centroid, all normal angles between each slope region and all the slope regions adjacent to its edges are obtained.

6. The method for simulating steel plate pitting corrosion based on a cellular automaton model according to claim 5, characterized in that: Determining the dynamic rule competition index of each slope area includes: Take any slope area as the target slope area, and take all slope areas adjacent to the edges of the target slope area as the neighboring slope areas; The dynamic rule competition index of the target slope area is determined according to the cosine value of the normal angle between the target slope area and each neighboring slope area, the Z-axis coordinate value of the centroid of each neighboring slope area, and the expansion coefficient.

7. The method for simulating steel plate pitting corrosion based on a cellular automaton model according to claim 6, wherein: Determining the dynamic rule competition index of the target slope area according to the cosine value of the normal angle between the target slope area and each neighboring slope area, the Z-axis coordinate value of the centroid of each neighboring slope area, and the expansion coefficient includes: Calculate the product of the cosine value of the normal angle between the target slope area and each neighboring slope area and the Z-axis coordinate value of the centroid of the corresponding neighboring slope area, and use the average of all products as the first dynamic rule competition factor of the target slope area; exponentially amplifying the expansion coefficient, and using the amplified value as a second dynamic rule competition factor of the target slope area; The first dynamic rule competition factor and the second dynamic rule competition factor of the target slope area are fused to obtain a dynamic rule competition index of the target slope area.

8. The method for simulating steel plate pitting corrosion based on a cellular automaton model according to claim 1, wherein: The dynamic rules for generating each cell to evolve toward the corresponding class slope include: Calculate the product of the number of intersections between each type of slope and the depth position of each type of local erosion progression coefficient and the directional variation index of the corresponding depth position to obtain the membership degree of each type of slope to each type of local erosion progression coefficient; For any cell, the membership degree of each type of slope to the weighted mean of the local erosion progressive coefficients of all cells above the cell is determined, and the type of slope corresponding to the largest membership degree is taken as the evolution direction of the cell, so that the cell generates a dynamic rule for evolving towards the corresponding type of slope.

9. A steel plate pitting corrosion simulation system based on a cellular automaton model, characterized in that: The method comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement the steel plate pitting corrosion simulation method based on the cellular automaton model as described in any one of claims 1 to 8.

Images