Structural material corrosion damage cross-scale evaluation method and system

Abstract

The invention provides a structural material corrosion damage cross-scale evaluation method and system, and belongs to the field of service performance cross-scale evaluation. The method comprises the following steps: firstly, collecting corrosion data of related structural materials, establishing a structural material service safety evaluation database based on the corrosion data, and then respectively extracting material scale characteristics and structural scale characteristics of corrosion damage of the structural materials; predicting the service life of the structural material according to the scale characteristics of the material; setting a service safety evaluation index of a material scale and measuring the service safety evaluation index; predicting the service life of the structural material according to the structural scale characteristics; setting a service safety evaluation index of a structure scale and measuring the service safety evaluation index; and carrying out cross-scale correlation on the material scale characteristics and the structure scale characteristics, constructing a transmission model for representing the service performance through material-structure cross-scale correlation, fusing the evaluation indexes of the material scale and the structure scale, generating the evaluation index of the structure material, and evaluating the service performance of the structure material. The accuracy and precision of corrosion damage evaluation are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of service performance cross-scale evaluation, and particularly relates to a structural material corrosion damage cross-scale evaluation method and system. BACKGROUND

[0002] For material application, the corrosion damage service performance of structural materials determines the service life of the materials, and the corrosion resistance of the materials is closely related to the application environment. For example, the sea is an extremely complex environment, the seawater has high salt content, strong conductivity and rich microorganisms, is a natural strong corrosive medium, has problems of long-term high temperature, high humidity, high salt fog and high radiation, and the corrosion of seawater on materials is a key factor restricting the safe service of deep sea equipment and facilities; meanwhile, the pressure problem also exists in the deep sea environment, and the pressure is more complex than that in the shallow sea environment, the pressure increases by about 1MPa for every 100m increase in water depth, and the huge pressure will cause fatigue damage of the applied materials, and also restricts the development and construction of the sea defense engineering and equipment service.

[0003] Stress corrosion of metal refers to a material damage process caused by the combined action of strain and corrosion of metal in a corrosive medium under residual or external stress. This corrosion generally penetrates the grain, that is, the so-called transgranular corrosion, and there is also intergranular corrosion. When the metal undergoes stress corrosion, only local area appears cracks from the surface to the inside. The cracks extend along the main crack, and at the same time, a number of branches also develop. The direction of the cracks is macroscopically perpendicular to the direction of tensile stress, and the micro fracture mechanism is generally intergranular fracture, and it can also be transgranular cleavage fracture or a mixture of the two, and the fracture surface can see the "mud pattern" corrosion product and corrosion pits. In order to provide basic data for deep sea engineering, it is necessary to obtain material performance data through experiments, simulations or monitoring means to study and clarify the relationship between the environment and stress corrosion.

[0004] In the prior art, when corrosion damage analysis and life prediction are performed on structural materials, a matching relationship between scales is established on the basis of traditional analysis methods at respective scales. With the emergence of new generation information technologies such as high-performance computing, big data, artificial intelligence and digital twinning, traditional engineering material service safety evaluation technology is developing towards digitization and intelligentization. With the development of machine learning methods, researchers attempt to embed the nonlinear relationship between corrosion influencing factors and corrosion results in the topology structure of neural networks, and propose methods such as using generalized regression neural network (GRNN) and radial basis (GM-RBF) neural network based on error compensation principle to predict the corrosion rate of metals in marine environment, but the pure data-driven method still has the problem of explainability; in addition, the coupling fields corresponding to field experiments, physical experiments and simulation experiments are different, and a large amount of environmental factor data in the actual environment cannot be measured, so that the data has the characteristics of multi-source heterogeneity, which is not conducive to the analysis of the corrosion behavior of structural materials. SUMMARY

[0005] In order to solve the above problems, the present application provides a structural material corrosion damage cross-scale evaluation method and system, a structural material corrosion service safety evaluation database is constructed, the service performance of the structural material corrosion damage is characterized, a material-structure-system cross-scale correlation and service performance characterization parameter transmission model is established, and finally the material layer structure layer data correlation and transmission combined with computer simulation and physical experiment is carried out, the structural material performance degradation law and life prediction model are established, and the accuracy and precision of the corrosion damage evaluation are improved.

[0006] In order to achieve the above purpose, the technical scheme adopted by the embodiments of the present application is as follows:

[0007] In a first aspect, the embodiments of the present application provide a structural material corrosion damage cross-scale evaluation method, which comprises the following steps:

[0008] Step S1, collect corrosion data of related structural materials, and establish a structural material service safety evaluation database based on the corrosion data;

[0009] Step S2, according to the structural material service safety evaluation database, respectively extract material scale features and structure scale features of structural material corrosion damage based on a deep learning method;

[0010] Step S3, according to the material scale features, performing life prediction on the structural material; setting a service safety evaluation index of the material scale, and measuring the evaluation index according to the predicted life;

[0011] Step S4, according to the structure scale features, performing life prediction on the structural material; setting a service safety evaluation index of the structure scale, and measuring the evaluation index according to the predicted life;

[0012] Step S5, cross-scale correlation of material scale features and structure scale features is performed, and a material-structure cross-scale correlation transmission model for characterizing service performance is constructed;

[0013] Step S6, based on the material-structure cross-scale correlation transmission model for characterizing service performance, material scale and structure scale evaluation indexes are fused, a structure material evaluation index is generated, and service performance of the structure material is evaluated.

[0014] As a preferred embodiment of the present application, the corrosion data includes: environmental experiment data, key position test data, indoor single-multi typical factor experiment data, full-size structure test data and material-structure simulation analysis data.

[0015] As a preferred embodiment of the present application, the service safety evaluation index of the material scale in step S3 includes: corrosion rate change rate, crack initiation probability, microstructure integrity index and acoustic emission parameter stability index.

[0016] As a preferred embodiment of the present application, the service safety evaluation index of the structure scale in step S4 includes: equivalent stress intensity factor, residual strength ratio and structure stability index.

[0017] As a preferred embodiment of the present application, in step S5, when the cross-scale correlation is performed:

[0018] Firstly, according to the results of corrosion damage detection of the structure material, the corrosion damage is projected on the fatigue crack propagation plane, and according to the corrosion damage geometric feature, the equivalent crack size model is constructed.

[0019] Secondly, the projection process is modeled, and the equivalent crack size model is constructed.

[0020] Finally, according to the extracted material scale features and structure scale features, the multi-scale features are fused by means of the adaptive neural fuzzy inference system ANFIS, and the input multi-scale features and the corresponding component operating parameters are associated.

[0021] As a preferred embodiment of the present application, when the equivalent crack size model is constructed:

[0022] In the simulation framework of material-structure cross-scale correlation, firstly, the structure surface corrosion pit point cloud is obtained by three-dimensional scanning or image correlation technology and the curved surface is reconstructed, then the point cloud {x i} is projected on the fatigue crack propagation plane n T x+d=0, the projection point is x i ′=x i -(nT x i +d)n; Extract the closed contour within the two-dimensional projection region Ω. And calculate the area A p With maximum depth d p ;

[0023] Then, based on the equivalent semiellipse assumption, let... And take b = d p The equivalent crack semi-axis a = πA is obtained. p / (2d p );

[0024] Finally, the desired geometric parameters a, b, or r are embedded into the finite element model to calculate the stress intensity factor, and combined with Paris's rule da / dN=C(ΔK). m Crack propagation and life prediction are performed, and a quantitative equivalent crack size model is constructed to provide corrosion-fatigue coupled failure analysis.

[0025] As a preferred embodiment of the present invention, step S5, which involves constructing a transfer model for characterizing service performance across material and structure scales, is as follows:

[0026] Based on the operational status of structural materials, including their failure or non-failure states, data is categorized and labeled according to the two states.

[0027] Establish the connection between the microscopic activities and macroscopic performance of the material surface, find the feature vectors of cross-scale correlation, and finally use the macroscopic detection results to characterize the degree of microscopic corrosion. Then, predict the corrosion of the material surface through the corrosion model. Utilize the characterization and evolution results of the service performance of structural material corrosion damage to identify the service status of the observation unit, perform reliability analysis and safety assessment, and at the same time, inversely determine the degree of influence of the unit in the system on the overall equipment service performance.

[0028] Simulation methods are employed to directly acquire structural corrosion damage simulation results at the structural scale. At the material scale, based on the corrosion damage mechanism of materials, online non-destructive testing methods are used, with the obtained online monitoring data serving as detection signals for various sensors. At the structural scale, simulation analysis identifies the key structural locations of corrosion pit damage, while online real-time monitoring of data collected at these key structural locations is conducted. A corrosion-mechanical coupling model is used to analyze stress redistribution and residual strength evolution under corrosion, thereby predicting corrosion damage and lifetime. At the material scale, damage states are analyzed, lifetime is predicted, and corrosion damage at each key structural location is pre-judged.

[0029] Finally, the corrosion pit morphology is input by a material scale model, combined with a fatigue loading condition, a corrosion fatigue crack propagation model is introduced, the remaining life of the material is quantitatively evaluated, the damage information is transmitted from the material scale to the structure scale, the cross-scale correlation of the material and the structure from the material damage to the structural strength failure to the remaining life prediction is realized, and the transmission model of the cross-scale correlation of the material-structure is established to characterize the service performance.

[0030] As a preferred embodiment of the present application, the evaluation index of the structural material is generated based on the transmission model in step S6, comprising:

[0031] In step S61, the observed characteristics obtained from the structural stress corrosion test and the macro damage variables output by the multi-scale finite element simulation and the stress intensity factor {K )，i} are used as input characteristics together, and the posterior distribution is solved; the fatigue life N life is integrated and solved by using the Weibull distribution model, and a fatigue life prediction model combining computer simulation and physical experiment is constructed;

[0032] In step S62, a service safety evaluation index system of the structural material itself is constructed based on the indexes of the material scale and the structure scale;

[0033] In step S63, the indexes in the service safety evaluation index system are predicted according to the fatigue life prediction model, and the service grade is divided according to the prediction result;

[0034] In step S64, the service safety evaluation result of the structural material is obtained by using the Bayesian classifier safety decision according to the divided service grade.

[0035] As a preferred embodiment of the present application, the service safety evaluation index system of the structural material itself comprises: failure unit distribution density, service remaining life ratio and system level corrosion risk score.

[0036] In a second aspect, the embodiments of the present application also provide a structural material corrosion damage cross-scale evaluation system, the system comprising: a database construction module, a feature extraction module, a material scale prediction module, a structure scale prediction module, a cross-scale correlation transmission module and an evaluation module; wherein,

[0037] The database construction module is used to collect the corrosion data of the related structural material, and establish a structural material service safety evaluation database based on the corrosion data;

[0038] The feature extraction module is used to extract the material scale characteristics and the structure scale characteristics of the structural material corrosion damage respectively based on the deep learning method according to the structural material service safety evaluation database;

[0039] The material scale prediction module is configured to predict the service life of the structural material according to the material scale characteristics, set a service safety evaluation index of the material scale, and measure the evaluation index according to the predicted service life.

[0040] The structure scale prediction module is configured to predict the service life of the structural material according to the structure scale characteristics, set a service safety evaluation index of the structure scale, and measure the evaluation index according to the predicted service life.

[0041] The cross-scale correlation transmission module is configured to correlate the material scale characteristics and the structure scale characteristics across scales, and construct a transmission model of the material-structure cross-scale correlation characterization service performance.

[0042] The evaluation module is configured to fuse the evaluation indexes of the material scale and the structure scale based on the transmission model of the material-structure cross-scale correlation characterization service performance, generate the evaluation index of the structural material, and evaluate the service performance of the structural material.

[0043] The scheme of the embodiment of the present application has the following beneficial effects:

[0044] The structural material corrosion damage cross-scale evaluation method and system provided by the embodiment of the present application construct a structural material corrosion service safety evaluation database, structural material corrosion damage service performance characterization, material corrosion damage multi-scale feature extraction based on deep learning, and carry out material service information fusion research; cross-scale service performance research and service safety intelligent decision-making, establish a material-structure-system cross-scale correlation and service performance characterization parameter transmission model; based on multi-scale modeling, construct a corrosion damage evolution model, and calibrate through finite element simulation and experimental data, realize corrosion life prediction, on the basis of the life prediction result, establish a service safety evaluation index system, and obtain the structural service safety evaluation result through the Bayesian classifier safety decision.

[0045] Of course, implementing any product or method of the present application does not necessarily require achieving all the advantages described above at the same time. DETAILED DESCRIPTION

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Figure 1 The flowchart of the structural material corrosion damage cross-scale evaluation method provided by the embodiment of the present application is shown in the figure.

[0048] Figure 2A schematic diagram of the evaluation method according to an embodiment of the present application;

[0049] Figure 3 A flowchart of a process of constructing a corrosion life prediction model according to a material scale in an embodiment of the present application;

[0050] Figure 4 A schematic diagram of a principle of cross-scale correlation between a material scale and a structure scale in an embodiment of the present application;

[0051] Figure 5 A schematic diagram of a principle of extracting multi-scale features and representing service performance in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The present inventors have found the above problems and have conducted a detailed study on the existing structural material corrosion loss evaluation method. It has been found that some existing studies have clustered data, analyzed and extracted feature substructures in each cluster using association rules, and integrated various data based on the feature substructures. Therefore, intelligent service safety evaluation based on service safety multi-source data fusion and evaluation model is an inevitable development trend. For example, in the construction of marine environment, the safety management level of offshore titanium-steel facilities will be greatly improved.

[0053] For data, taking the marine environment as an example, there are few basic data of titanium alloy corrosion in deep-sea environment, but there are many data sources, and there is strong heterogeneity between multi-source experimental data. Therefore, it can be classified as sparse multi-source heterogeneous characteristics. On the one hand, it is necessary to estimate and fill in the missing values in the experimental data, i.e. to realize data enhancement. On the other hand, it is necessary to effectively integrate multi-source heterogeneous data to realize data complementation and correction. Basic data of titanium alloy materials in deep-sea environment, physical test data, real sea test data, simulation analysis data, etc. are used to align and correlate various monitoring and testing data of titanium alloy materials under single factor and multi-factor change conditions, to provide a basis for subsequent data fusion and data modeling.

[0054] The research on cross-scale correlation mechanism and algorithm is a major fundamental problem in the current cross field of material science and information science. The current research mainly focuses on the corrosion fatigue and life prediction of mechanical structure, and the main idea is to establish the matching correlation between scales on the basis of traditional analysis methods in each scale. In the aspect of cross-level correlation of multi-level system, a research team carried out cross-scale research on the damage behavior of high-speed milling cutter, and proposed a macro-micro damage cross-scale correlation analysis method based on entropy; some scholars proposed a finite element method based on lagging hydraulic fracture simulation based on the basic design and mathematical analysis of material model and numerical algorithm; in addition, a researcher proposed a method of structure and material-structure damage information cross-scale correlation and decision-making for long-term service structure that cannot be repaired, which fully utilizes the information of service structure at the material scale and the structure scale, and uses equivalent stress for cross-scale registration correlation. From the current research work, most of them are to solve the correlation between material micro / micro damage and macro damage.

[0055] With the rapid development of information technology such as big data and deep learning, more and more traditional research fields begin to cross disciplines, and can draw on data-driven and deep learning methods to analyze and explain the uncertain and uninterpretable parts in traditional corrosion behavior research. For example, for the corrosion environment conditions of titanium clad steel structure materials, the feature level and decision level fusion of multi-source corrosion behavior data can be carried out to carry out evaluation research on the corrosion behavior of titanium clad steel in the open sea environment, which can effectively support the design and construction of open sea island coastal defense engineering.

[0056] It should be noted that the defects in the above prior art solutions are the result of the inventor's practice and careful study, therefore, the discovery process of the above problems and the solutions proposed by the embodiments of the present application to solve the above problems should be the contribution of the inventor to the present application.

[0057] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. It should be noted that the embodiments and features in the embodiments can be combined with each other without conflict.

[0058] It should be noted that: similar reference numbers and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In the description of the present application, the terms "first", "second", "third", "fourth" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0059] Based on the above in-depth analysis, the embodiment of the present application provides a structural material corrosion damage cross-scale evaluation method and system. First, a structural material corrosion service safety evaluation database is constructed, material corrosion damage multi-scale feature extraction based on deep learning is performed, and structural material corrosion damage service performance characterization is performed. A material-structure cross-scale correlation and service performance characterization parameter transmission model is established. Then, the structural material cross-scale performance degradation law and life prediction model and service safety evaluation are realized. The embodiment of the present application takes structural materials with corrosion damage as the object, analyzes the corrosion damage of structural materials, performs multi-source data fusion on various test, simulation, monitoring and detection data, analyzes the internal problems of cross-scale system service safety research such as service performance cross-scale correlation between materials and constituent structures, computer simulation and physical experiment, and life prediction method combining monitoring and detection information, and multi-scale coupling analysis can avoid the limitations of single-scale research, significantly improve the accuracy of the corrosion prediction model, and provide a reliable basis for engineering life assessment.

[0060] As shown in Figure 1 and Figure 2 , the structural material corrosion damage cross-scale evaluation method comprises the following steps:

[0061] Step S1, collect corrosion data of related structural materials, and establish a structural material service safety evaluation database based on the corrosion data.

[0062] In this step, when establishing the database, collect and organize related data, establish a service safety evaluation database of structural materials, including environmental experiment data, key position test data, indoor single-multi typical factor experiment data, full-size structure test data and material-structure simulation analysis data.

[0063] Step S2, according to the structural material service safety evaluation database, respectively extract material scale features and structure scale features of structural material corrosion damage based on deep learning method.

[0064] In this step, the material scale features and structure scale features form multi-scale features. When extracting multi-scale features, first, pre-process the data, including noise reduction, etc., and then use the extreme random tree algorithm to extract multi-scale features from the noise-reduced data.

[0065] Specifically, when extracting material scale features and structure scale features, at each scale, based on the extreme random tree (Extra-Trees) algorithm, Extra-Trees is used to randomly select features at each leaf node, multiple decision trees are constructed and threshold values are set, leaf node indexes are generated as high-dimensional original features through formulaic random splitting and impurity reduction, and one-hot encoding is performed on the leaf node indexes to form the first layer high-level features Finally, the mapping is realized to the actual running state y by means of a downstream learner or a multi-layer iterative decision tree model, so as to realize accurate prediction.

[0066] The algorithm formula of the decision tree model in feature extraction is as follows:

[0067]

[0068] In formula (1), h m is the mth random decision tree; each decision tree randomly selects K features at the leaf node, and uniformly samples the threshold value θ j for each feature j in the value range [min(x j ), max(x j )].

[0069] High-dimensional feature coding is to do one-hot coding for the leaf indexes {l1(x), …, l M (x)} of all decision trees, to generate a high-dimensional vector where L M is the total number of leaf nodes of the Mth tree.

[0070] In feature extraction, the importance of the feature is evaluated, and the importance of the jth original feature is defined as:

[0071]

[0072] In formula (2), Imp(j) represents the importance value of the jth feature; M represents the total number of decision trees trained (for example, in a random forest or a gradient boosting tree); u represents a certain internal leaf node in the decision tree; split on j represents that the leaf node u is divided using the jth feature; ΔI u represents the information gain (Information Gain) or the reduction of impurity (such as the reduction of Gini index, the reduction of entropy, or the reduction of squared error) at the leaf node u, that is, the degree of reduction of sample uncertainty before and after division.

[0073] Features that have corresponding contributions to life prediction are screened through formula (2).

[0074] In step S3, the life of the structural material is predicted according to the material size characteristics; the service safety evaluation index of the material size is set, and the evaluation index is measured according to the predicted life.

[0075] In this step, the service safety evaluation index of the material size reflects the micro-damage evolution behavior under the synergistic action of corrosion and stress, and the index includes:

[0076] The corrosion rate change rate Δr, and the index calculation formula is: ​

[0077]

[0078] In formula (3), r t is the average corrosion depth at time t, in mm / a, used to judge the corrosion development trend.

[0079] Crack initiation probability P c , the index is output by the adaptive neural fuzzy system Initiation risk probability distribution, ranging from [0, 1], to reflect the possibility of local crack initiation and development of materials.

[0080] Microstructure integrity index MCI, the index calculation formula is:

[0081]

[0082] In formula (4), N void and N crack are the number of micro defects, N total is the total number of feature points in the scanning area, reflecting the integrity of the microstructure.

[0083] Acoustic emission parameter stability index AE σ , the index is normalized by standard deviation of acoustic emission signal statistics (such as energy E, duration T, etc.), reflecting the corrosion activity, and the index calculation formula is:

[0084]

[0085] In formula (5), σ E represents the standard deviation of acoustic emission energy, reflecting the amplitude of energy fluctuation of each acoustic emission event; R mean represents the average value of acoustic emission energy, as a normalization factor, to facilitate comparison between different samples; σ T represents the standard deviation of acoustic emission duration, indicating the degree of variation of acoustic emission event signal duration; T mean represents the average value of acoustic emission duration, also as a reference for normalization processing.

[0086] Step S4, according to the structural scale characteristics, the service life of the structural material is predicted; the service safety evaluation index of the structural scale is set, and the evaluation index is measured according to the predicted service life.

[0087] In this step, the structural scale index is used to characterize the manifestation of corrosion damage on the macro mechanical response, including:

[0088] Equivalent stress intensity factor K ep , the index is calculated by extended finite element analysis or crack mechanics model, and the index calculation formula is:

[0089]

[0090] In equation (6), a is the equivalent crack half-axis, σ is the principal stress, and Y is the geometric correction factor. This can be used to determine whether the critical failure criterion K has been reached. IC .

[0091] The residual strength ratio (RSR) is calculated using the following formula:

[0092]

[0093] In equation (7), σ res For the remaining load-bearing capacity of the corroded structure, σ yield This represents the original yield strength. RSR < 0.7 is generally considered to indicate a significant safety risk.

[0094] The Component Stability Index (CSI) measures the changing trends of parameters such as vibration response and dynamic stiffness. CSI is used to describe the overall stability evolution of a structure during corrosion development and is modeled using a normalized rate of change.

[0095] Step S5: Construct a transfer model for the cross-scale correlation between materials and structures to characterize service performance.

[0096] In this step, such as Figure 3 As shown, a transfer model for characterizing service performance across material and structure is constructed using computational simulation methods. Based on the results of structural corrosion damage detection, computer graphics techniques are used to project the corrosion damage onto the fatigue crack propagation plane. According to the geometric morphology of the corrosion damage, it is equivalently converted into fracture mechanical defects with regular dimensions, such as semi-elliptical or circular shapes. This process is modeled to obtain an equivalent crack size model. Finally, based on the extracted multi-scale features and combined with the equivalent crack size model, an Adaptive Neuro-Fuzzy Inference System (ANFIS) algorithm is used to construct a cross-scale correlation model between material and structure. This model fuses multi-scale features and links the input multi-scale features with the corresponding component operating parameters, thereby characterizing the service performance of the structural materials.

[0097] In constructing the equivalent crack size model, within the simulation framework of material-structure cross-scale correlation, the point cloud of corrosion pits on the structural surface is first obtained through 3D scanning or image correlation techniques, and the surface is reconstructed. Then, the point cloud {x i Projected onto the fatigue crack propagation plane n T On x+d=0, the projection point is x. i ′=xi -(n T x i +d)n; Extract the closed contour within the two-dimensional projection region Ω. And calculate the area A p With maximum depth d p Next, based on the equivalent semiellipse assumption, let... And take b = d p The equivalent crack semi-axis a = πA is obtained. p / (2d p Finally, the desired geometric parameters a, b, or r are embedded into the finite element model (XFEM / SBFEM) to calculate the stress intensity factor and combine it with Paris's rule da / dN=C(ΔK). m Crack propagation and lifetime prediction are performed, providing a quantitative equivalent crack size model for corrosion-fatigue coupled failure analysis.

[0098] The ANFIS is a hybrid intelligent algorithm that combines the advantages of fuzzy logic and artificial neural networks. It utilizes the self-learning and adaptive capabilities of neural networks to optimize the membership functions and rules in fuzzy systems, achieving a powerful ability to model and predict complex nonlinear systems. The basic structure of ANFIS used in this embodiment is built upon a Takagi-Sugeno type fuzzy inference system, including an input layer, a rule layer, a normalization layer, a consequent layer, and an output layer.

[0099] In the input layer (Layer 1), each node represents a fuzzy membership function (such as a triangle or Gaussian function) of an input variable, mapping the input to a fuzzy space. For example: in, The input x is in the fuzzy set A i The membership degree in the rule layer (Layer 2) corresponds to a fuzzy rule, and the output is the product of the membership degrees of each premise (i.e., fuzzy AND operation), such as: The normalization layer (Layer 3) normalizes the weights of each rule; in the consequent layer (Layer 4), each node represents a linear consequent function of a Sugeno-type rule, such as: The output layer (Layer 5) performs a weighted sum of all rule outputs to generate the final system output. For example:

[0100] In this embodiment, the ANFIS training adopts a hybrid learning algorithm, i.e., combining the least square method (LS) and the gradient descent (GD). Forward propagation: input data are calculated through the network to output, and the consequent parameters are updated by the least square method; backward propagation: the parameters of the input membership function are adjusted by the gradient descent method to minimize the error function (usually MSE); thus, the high efficient learning of neural network and the interpretability of fuzzy system are combined.

[0101] Specifically, the steps of constructing the transfer model of the material-structure cross-scale associated performance service performance are as follows:

[0102] According to the operating state of the structural material, including its failure or non-failure state, the data are classified and labeled according to the two states. The connection between the micro-activity of the material surface and the macro-performance is established, the cross-scale associated feature vector is found, and finally the micro-corrosion degree is represented by the macro-detection result, and then the corrosion model is used to predict the corrosion of the material surface. The corrosion damage service performance of the structural material is used to represent and evolve the results, to identify the service state of the observation unit, to analyze the reliability and to evaluate the safety, and at the same time, the influence degree of the unit on the overall equipment service performance in the system can be inverted.

[0103] As shown in Figure 4 , the simulation method is adopted to directly obtain the simulation result data of the structural corrosion damage at the structure scale; at the material scale, the online non-destructive monitoring method is adopted to obtain the online monitoring data as the detection signals of various sensors according to the corrosion damage mechanism of the material. At the structure scale, the simulation analysis determines the key structure position of the corrosion pit damage, and the data are collected by the online real-time monitoring of the key structure position distribution, the stress redistribution and the residual strength evolution under the corrosion action are analyzed by using the corrosion-mechanical coupling model, so as to predict the corrosion damage and perform the life prediction; at the material scale, the damage state is analyzed and the life prediction is performed to predict the corrosion damage of each key structure position. Finally, the corrosion pit morphology is driven by the material scale model and input, the corrosion fatigue crack propagation model is introduced under the condition of fatigue loading, the residual life of the material is quantitatively evaluated, the damage information is transferred from the material scale to the structure scale, the cross-scale association of the material and the structure from the material damage to the structural strength failure to the residual life prediction is realized, and the transfer model of the material-structure cross-scale associated performance service performance is established.

[0104] As shown in Figure 5 , the specific operation at the material scale is as follows: first, the acoustic emission signals that can reflect the slight changes of the internal damage of the material in the material experiment are collected by the sensor.

[0105] Based on the principle that the micro-fracture events such as metal film rupture and hydrogen evolution in the corrosion mechanism will produce transient stress waves, piezoelectric acoustic emission sensors (PZT / PVDF) are arranged, the sampling frequency is generally 0.1-2 MHz, and the threshold Vth To filter out environmental noise. After pre-amplification and band-pass filtering, key parameters are extracted: Counts, peak amplitude A max , energy and duration Δt = t2-t1. Further time-frequency analysis (short-time Fourier transform or wavelet packet decomposition) is performed to obtain instantaneous spectrum S(f, t) and energy distribution, to identify different damage mechanisms such as corrosion nucleation, propagation and micro-crack initiation. Through multi-sensor time difference positioning algorithm (TDOA), the sound source position x s is estimated:

[0106] ‖x s -x i ‖-||x s -x j ||=v(t i -t j ), (8)

[0107] In equation (8), v is the sound velocity, x i is the position of the i-th sensor, and t i is the firing time.

[0108] The acquired online monitoring data set is spatio-temporally synchronized with the structural scale simulation output {h i (t k ), D i (t k )} to model and calibrate the corrosion-fatigue coupling failure process.

[0109] With the help of wavelet analysis method, characteristic quantities reflecting the material degradation state are extracted, and discrete wavelet transform (DWT) is used to decompose s(t) into multi-scale sub-band coefficients:

[0110] W j，k =∫s(t)ψ j，k (t)dt, (9)

[0111] In equation (9), ψ j，k (t) is the wavelet basis function with scale j and translation k. In each sub-band, energy (E j =∑ k |W j,k | 2 ), entropy (H j = -∑ k p j,k logp j,k ) and statistical moments (such as variance, skewness) and other features are extracted to form a feature vector x = [E1, …, E J , H1, …, H J , …].

[0112] And using support vector machine, realize the damage identification in the process of stress corrosion of base material, input the characteristic vector x into the radial basis kernel SVM, and the classification decision function is:

[0113]

[0114] In formula (10), K(x i ,X)=exp(-γ‖x i -X‖ 2 ), α i And b are obtained by convex quadratic programming from training samples. The SVM outputs the damage state category, and the damage severity can be estimated based on the distance of the decision function.

[0115] Based on the damage identification result, the degradation state of the material is determined, the performance degradation trajectory of the material is fitted, the degradation index D(N) (such as the elastic modulus reduction rate or the fatigue strength loss) of the sample under the cycle number N or the stress level σ is determined according to the SVM identification result, and the degradation model is fitted:

[0116] D(N)=1-exp(-αN β ), (11)

[0117] In formula (11), the parameters α and β can be solved by least squares method or Bayesian estimation.

[0118] The boundary tracking method in the free boundary problem is applied to evaluate the failure critical interval of the structural material, the failure prediction model is studied, based on the fitted D(N), the crack propagation and stress concentration evolution of the material are simulated by using the boundary tracking algorithm (Boundary Element Method, BEM) of the free boundary problem, and the number of times that the critical stress intensity factor threshold K IC Reaches the boundary of the region is determined to determine the failure critical cycle number N c Range; find the fouling-corrosion, erosion-corrosion material property change behavior and corresponding parameter change trend, that is, the failure standard of material corrosion damage;

[0119] Then use the random forest regression method to build the corrosion life prediction model, take the cycle number N, stress amplitude Δσ, temperature T, failure standard index, etc. As input characteristics, and use random forest regression to build the life prediction model:

[0120]

[0121] In formula (12), z=[Δσ, T, Ra, …], each regression tree h m Split nodes by minimizing mean square error on the training set; the model can be used for life prediction under unknown working conditions after cross-validation evaluation.

[0122] Step S6, based on the transmission model of material-structure cross-scale correlation characterization of service performance, the evaluation indexes of material scale and structure scale are fused to generate the evaluation index of structure material, and the service performance of the structure material is evaluated.

[0123] In this step, based on the stress corrosion experiment results of the structure, combined with the life prediction results of the material scale and the structure scale, the feature level fusion is performed by using the method based on Bayesian inference, and a fatigue life prediction model combining computer simulation and physical experiment is constructed. Based on the life prediction result, a service safety evaluation index system is established, and the structure service safety evaluation result is obtained through the Bayesian classifier safety decision.

[0124] Specifically, the evaluation index of the structure material is generated based on the transmission model to evaluate the service performance of the structure material, including:

[0125] Step S61, the implementation of the life prediction model. In the model, the observed features (such as cycle number N, damage index D, stress amplitude Δσ and temperature T) obtained from the structure level stress corrosion test and the macro damage variables and stress intensity factor {K I，i} output by multi-scale finite element simulation are used as input features together, and a hierarchical Bayesian model is constructed:

[0126] p(θ∣f exp ,f sim )∝p(f exp ∣θ)p(f sim ∣θ)p(θ) (13)

[0127] In formula (13), θ is a latent parameter.

[0128] The latent parameter θ (including material sensitive coefficient and environmental influence factor) is inferred jointly; in this framework, the experiment and simulation features are assumed to follow a multivariate Gaussian distribution, and the Markov chain Monte Carlo (MCMC) or variational inference (VI) method is used to solve the posterior distribution. Subsequently, by using the Weibull distribution model, the fatigue life N life is integrated and solved, and the 95% confidence interval is determined by Bayesian decision theory to evaluate the model reliability; finally, this method realizes the feature level information fusion based on the physical test and computer simulation data, and constructs a fatigue life prediction model combining computer simulation and physical experiment.

[0129] Step S62, construction of service safety evaluation index system.

[0130] In this step, based on the material scale and structure scale indicators, the service safety evaluation index system of the structure material itself is constructed, and the evaluation index of the structure material includes:

[0131] Failure unit distribution density p f The index calculation formula is:

[0132]

[0133] In formula (14), N f is the number of units reaching the failure criterion in the finite element model, and A is the total evaluation area, which can be used for heat map labeling of key areas.

[0134] Service residual life ratio (Service Residual Index, SRI), the index calculation formula is:

[0135]

[0136] In formula (14), T remain is the predicted residual life, T design is the original design life; SRI<0.4 is recommended for maintenance or replacement.

[0137] System-level corrosion risk score (Corrosion Risk Severity Index, CRSI), the index calculation formula is:

[0138] CRSI=w1·K ep +w2·P c +w3·RSR+w4p f (15)

[0139] In formula (15), w i is the weight, which can be adjusted according to the minimum error of the training data set.

[0140] Index normalization and fusion calculation, here all the indexes are denoised by wavelet smoothing, and then unitized and normalized:

[0141]

[0142] In formula (16), x i : the original i-th index value; x min : the minimum value of the index in all samples; x max : the maximum value of the index in all samples; The normalized i-th index value is in the range [0, 1]. The normalized index enters the comprehensive evaluation function to construct the structure corrosion service safety score (SCSEI) as follows:

[0143]

[0144] In formula (17), a i is a weighted coefficient for each index, which can be determined automatically by historical evaluation data or principal component analysis (PCA).

[0145] Step S63, according to the SCSEI score range, the service level is divided.

[0146] In this step, according to the indexes of the structural material, the service level is classified as follows, as shown in Table 1:

[0147] Table 1

[0148] SCSEI range Safety class Meaning description ≥0.85 Safe Normal operation, no intervention required 0.65～0.85 Pre-warning Risk exists, regular detection recommended 0.45～0.65 Warning Local repair or reduced load operation recommended ≤0.45 Danger High risk, immediate repair or retirement required

[0149] According to the SCSEI score and the multi-source input features, the corrosion service safety level C is divided as:

[0150] C∈{Safe, Warning, Alert, Danger}

[0151] Corresponding label C k , respectively, 0 / 1 / 2 / 3 represent different safety levels.

[0152] Step S64, the service safety evaluation result of the structural material is obtained by adopting the Bayesian classifier safety decision.

[0153] In this step, the Bayesian classifier is a supervised learning algorithm based on probability statistics theory, which is suitable for fast classification problems under high-dimensional features, and is especially suitable for processing multi-source heterogeneous data. This method has good robustness, strong interpretability and is easy to integrate into the engineering safety evaluation system. Specifically, it includes:

[0154] Step S641, selecting input features.

[0155] The key variables in the constructed service safety evaluation index set are selected as the input feature vector X=[x1, x2,..., x n ], for example:

[0156] Material scale features: corrosion rate Δr, crack initiation probability P_c, acoustic emission statistics AE_σ;

[0157] Structural scale features: equivalent stress intensity K_eq, RSR;

[0158] System scale features: ρ_f, SRI, SCSEI;

[0159] After normalization and redundant feature elimination, the standardized input feature space is formed

[0160] Step S642: Construct a classification model based on a Bayesian classifier.

[0161] In this step, the Bayesian classifier calculates the posterior probability of each category given the features, based on Bayes' theorem:

[0162]

[0163] In equation (18), P(C) k () indicates security level C k The prior probability can be obtained from historical evaluation data; P(X|C k ) indicates that in the known category C k Given the condition, the conditional probability of the feature vector X; P(X) represents the overall probability of the feature, which is a normalization factor and can be ignored for decision-making.

[0164] Naive Bayes ( If we assume the Bayesian assumption, that is, that features are conditionally independent, then we have:

[0165]

[0166] In equation (19), for the continuous variable x i Assume it follows a Gaussian distribution:

[0167]

[0168] In equation (20), μ ik ,σ ik Features x i In category C k The mean and standard deviation are given; for discrete variables, Bernoulli distribution or multinomial model can be used.

[0169] Step S643: Input the material-scale and structural-scale features into the model and output the evaluation results.

[0170] In this step, for each set of feature input X, the posterior probability of all categories is calculated, and the category corresponding to the maximum value is taken as the final judgment result. At the same time, the decision confidence (i.e. the maximum posterior probability) and the candidate level distribution can be output for subsequent visualization and early warning.

[0171] Step S644: Verify the output results.

[0172] In this step, the classifier training phase uses the existing historical lifetime dataset (labeled corrosion samples) to perform maximum likelihood estimation (MLE) to obtain the parameter μ. ik ,σ ikThe classification performance is evaluated by 10-fold cross-validation. The evaluation indexes include: accuracy (Accuracy); confusion matrix (Confusion Matrix); average F1 score (Macro-F1); warning accuracy (Warning Sensitivity).

[0173] The real-time monitoring data is input into the classifier, and the current service safety level, failure risk trend graph, recommended maintenance time window and safety level evolution sequence are output.

[0174] The system can be linked with a predicted life module, and when the predicted life is lower than a threshold value and the SCSEI is less than 0.6, an early warning is automatically triggered.

[0175] Based on the same idea, the embodiment of the application also provides a structural material corrosion damage cross-scale evaluation system, which comprises a database construction module, a feature extraction module, a material scale prediction module, a structure scale prediction module, a cross-scale correlation transmission module and an evaluation module.

[0176] The database construction module is used to collect corrosion data of related structural materials, and establish a structural material service safety evaluation database based on the corrosion data;

[0177] The feature extraction module is used to extract material scale features and structure scale features of structural material corrosion damage based on deep learning method according to the structural material service safety evaluation database;

[0178] The material scale prediction module is used to predict the life of the structural material according to the material scale features; set the service safety evaluation index of the material scale, and measure the evaluation index according to the predicted life;

[0179] The structure scale prediction module is used to predict the life of the structural material according to the structure scale features; set the service safety evaluation index of the structure scale, and measure the evaluation index according to the predicted life;

[0180] The cross-scale correlation transmission module is used to correlate the material scale features and the structure scale features across scales, and construct a transmission model of material-structure cross-scale correlation characterization service performance;

[0181] The evaluation module is used to fuse the evaluation indexes of the material scale and the structure scale based on the transmission model of material-structure cross-scale correlation characterization service performance, generate the evaluation index of the structural material, and evaluate the service performance of the structural material.

[0182] It can be seen from the above technical solutions that the structural material corrosion damage cross-scale evaluation method and system provided by the embodiments of the present application takes the structural material with corrosion damage as the object, analyzes the corrosion damage of the structural material, performs multi-source data fusion on various test, simulation, monitoring and detection data, analyzes the internal problems of the cross-scale system service safety research such as the cross-scale correlation of the service performance between the material and the structure, the life prediction method combining the computer simulation and the physical experiment and the monitoring and detection information, the multi-scale coupling analysis can avoid the limitation of the single-scale research, significantly improves the accuracy of the corrosion prediction model, and provides a reliable basis for the engineering life evaluation.

[0183] The above description is merely preferred embodiments of the present application and a description of the principles of the technology used, and is not intended to limit the scope of the claimed present application, but merely represents the preferred embodiments of the present application. Those skilled in the art should understand that the scope of the present application is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or equivalent features without departing from the inventive concept. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

Claims

1. A method for evaluating corrosion damage of structural materials across scales, characterized in that, The method comprises the following steps: Step S1, collecting corrosion data of relevant structural materials, and establishing a structural material service safety evaluation database based on the corrosion data; Step S2, according to the structural material service safety evaluation database, extracting material scale features and structure scale features of structural material corrosion damage based on a deep learning method respectively; Step S3, according to the material scale features, predicting the service life of the structural material; Setting the service safety evaluation index of the material scale, and measuring the evaluation index according to the predicted life; Step S4, according to the structure scale features, predicting the service life of the structural material; Setting the service safety evaluation index of the structure scale, and measuring the evaluation index according to the predicted life; Step S5, cross-scale correlation of material scale features and structure scale features, and construction of material-structure cross-scale correlation characterization transfer model of service performance; Step S6, based on the material-structure cross-scale correlation characterization transfer model of service performance, fusion of material scale and structure scale evaluation indexes, generation of structural material evaluation index, and evaluation of service performance of structural material.

2. The method of claim 1, wherein, The corrosion data includes: environmental experiment data, key position test data, indoor single-multi typical factor experiment data, full-size structure test data and material-structure simulation analysis data.

3. The method of claim 1, wherein, The service safety evaluation index of the material scale in step S3 includes: corrosion rate change rate, crack initiation probability, microstructure integrity index and acoustic emission parameter stability index.

4. The method of claim 1, wherein, The service safety evaluation index of the structure scale in step S4 includes: equivalent stress intensity factor, residual strength ratio and structure stability index.

5. The method of claim 1, wherein, In step S5, when cross-scale correlation is performed: Firstly, according to the results of corrosion damage detection of the structural material, the corrosion damage is projected on the fatigue crack propagation plane, and according to the corrosion damage geometric feature, it is equivalent to a fracture mechanics defect with regular size; Secondly, the projection process is modeled, and an equivalent crack size model is constructed; Finally, according to the extracted material scale features and structure scale features, combining the equivalent crack size model, with the help of adaptive neuro-fuzzy inference system ANFIS, the multi-scale features are fused, and the input multi-scale features and the corresponding component operating parameters are associated.

6. The method of claim 5, wherein, When constructing the equivalent crack size model: In the simulation framework of material-structure cross-scale correlation, firstly the structure surface corrosion pit point cloud is obtained by three-dimensional scanning or image correlation technology and the curved surface is reconstructed, then the point cloud {x i} is projected onto the fatigue crack propagation plane n T x+d=0, the projection point is x′ i =x i -(n T x i +d)m; the closed contour is extracted in the two-dimensional projection area Ω and the area A p and the maximum depth d p are calculated; Again, according to the equivalent half-ellipse hypothesis, let and take b = d p , we get the equivalent crack length half-axis a = πA p / (2d p ); Finally, the calculated geometric parameters a, b or r are embedded into the finite element model to calculate the stress intensity factor and combine with Paris law da / dN=C(ΔK) m The crack propagation and life prediction are carried out, and an equivalent crack size model is constructed to provide quantitative analysis for corrosion-fatigue coupling failure.

7. The method of claim 1, wherein, The steps of step S5 for constructing the material-structure cross-scale correlation characterization transfer model of service performance are as follows: According to the operating state of the structural material, including its failure or non-failure state, according to the two states for classification data labeling; Establish the connection between the micro activity of the material surface and the macro performance, find the feature vector of cross-scale correlation, finally use the macro detection results to represent the micro corrosion degree, and then predict the material surface corrosion through the corrosion model, use the representation and evolution results of the service performance of the structural material corrosion damage to identify the service state of the observation unit, analyze the reliability and safety, and at the same time, the influence degree of the unit on the overall equipment service performance in the system can be inverted; The simulation method is adopted to directly obtain the simulation result data of the structural corrosion damage in the structural scale; in the material scale, the online nondestructive monitoring method is adopted to obtain the online monitoring data as the detection signals of various sensors according to the corrosion damage mechanism of the material; In the structural scale, the simulation analysis determines the key structural positions of the corrosion pit damage, and the data of the key structural positions are collected by the online real-time monitoring, the stress redistribution and the residual strength evolution under the corrosion effect are analyzed by using the corrosion-mechanical coupling model, so as to predict the corrosion damage and perform the life prediction; In the material scale, the damage state is analyzed, and the life prediction is performed to predict the corrosion damage of each key structural position; Finally, the corrosion pit morphology is driven and input by the material scale model, the corrosion fatigue crack propagation model is introduced under the condition of fatigue loading, the residual life of the material is quantitatively evaluated, the damage information is transmitted from the material scale to the structural scale, the cross-scale correlation of the material and the structure is realized from the material damage to the structural strength failure to the residual life prediction, and the transmission model of the cross-scale correlation of the material-structure is established to represent the service performance.

8. The method of claim 1, wherein, In step S6, the evaluation index of the structural material is generated based on the transmission model, including: Step S61: Combine the observed characteristics obtained from the structural stress corrosion test with the macroscopic damage variables output from the multi-scale finite element simulation. and stress intensity factor {K I，i } are used together as input features, and the posterior distribution is solved; the fatigue life N is calculated using the Weibull distribution model. life By performing integral solutions, a fatigue life prediction model combining computer simulation and physical experiments is constructed. In step S62, the service safety evaluation index system of the structural material itself is constructed based on the indexes of the material scale and the structural scale; In step S63, the indexes in the service safety evaluation index system are predicted according to the fatigue life prediction model, and the service grade is divided according to the prediction result; In step S64, the service safety evaluation result of the structural material is obtained by using the Bayesian classifier safety decision according to the divided service grade.

9. The method of claim 8, wherein, In the service safety evaluation index system of the structural material itself, the evaluation indexes include: failure unit distribution density, service residual life ratio and system level corrosion risk score.

10. A structural material corrosion damage cross-scale evaluation system, comprising: The system includes: a database construction module, a feature extraction module, a material scale prediction module, a structural scale prediction module, a cross-scale correlation transmission module and an evaluation module; wherein, The database construction module is used to collect the corrosion data of the related structural material, and establish a structural material service safety evaluation database based on the corrosion data; The feature extraction module is used to extract the material scale features and the structural scale features of the structural material corrosion damage respectively based on the deep learning method according to the structural material service safety evaluation database; The material scale prediction module is used to perform life prediction on the structural material according to the material scale features; The service safety evaluation indexes of the material scale are set, and the evaluation indexes are measured according to the predicted life; The structural scale prediction module is used to perform life prediction on the structural material according to the structural scale features; The service safety evaluation indexes of the structural scale are set, and the evaluation indexes are measured according to the predicted life; The cross-scale correlation transmission module is used to perform cross-scale correlation on the material scale features and the structural scale features, and construct a transmission model of the cross-scale correlation of the material-structure to represent the service performance; The evaluation module is used for generating the evaluation index of the structure material and evaluating the service performance of the structure material based on a transmission model of material-structure cross-scale correlation characterization of the service performance, and fusing the evaluation indexes of the material scale and the structure scale.

Images