CN-122020324-A - Online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupling corrosion

Abstract

The invention discloses an online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupling corrosion, and relates to the technical field of metal corrosion monitoring and evaluation. The data acquisition module acquires the multiplex time sequence corrosion data in real time through the sensor array, and the characteristic processing module acquires the corrosion dynamic characteristic vector through interpolation of the missing values, abnormal fluctuation filtering and time synchronization alignment processing of the multisource data. The state evolution module inputs the vector into a pre-training deep corrosion behavior prediction model, a microscopic corrosion state evolution map is generated through a multi-layer feature fusion network, the path matching module matches corresponding corrosion dynamics evolution paths, and the performance evaluation module calculates performance degradation curves of different depth layers and generates corrosion resistance level quantification reports. The system can realize accurate characterization and layered corrosion resistance online quantitative evaluation of the high-nitrogen steel corrosion microscopic evolution under the multi-factor coupling environment.

Inventors

• SUN CHUANGUO
• TAN MING

Assignees

• 常熟市长江不锈钢材料有限公司

Dates

Publication Date

20260512

Application Date

20260415

Claims (10)

1. 1. The online evaluation system for the corrosion resistance of the high-nitrogen steel under multi-factor coupling corrosion is characterized by comprising the following components: the data acquisition module is used for deploying the sensor array in the service environment and acquiring a multi-element time sequence data set of the surface of the high-nitrogen steel material relative to the corrosion environment in real time; the characteristic processing module is used for preprocessing the multi-element time sequence data set to obtain a corrosion dynamic characteristic vector set, wherein the preprocessing comprises missing value interpolation, abnormal fluctuation point filtering and multi-source data time synchronization alignment; The state evolution module is used for inputting the corrosion dynamic feature vector set into a pre-trained deep corrosion behavior prediction model, and generating a microscopic corrosion state evolution map of the high-nitrogen steel surface in the current service environment through a multi-layer feature fusion network in the deep corrosion behavior prediction model; The path matching module is used for identifying potential corrosion damage modes of the high-nitrogen steel material at the current moment according to the microscopic corrosion state evolution map, calling a corrosion damage mode knowledge base and matching corresponding corrosion dynamics evolution paths for each identified corrosion damage mode; The performance evaluation module is used for calculating and obtaining material performance degradation prediction curves of the high-nitrogen steel material at different depth levels in a preset evaluation period based on the corrosion dynamics evolution path corresponding to the corrosion damage mode, and generating a corrosion resistance level quantification report of the high-nitrogen steel material according to the material performance degradation prediction curves.
2. 2. The online evaluation system for corrosion resistance of high nitrogen steel under multi-factor coupling corrosion according to claim 1, wherein the preprocessing of the multi-element time series data set to obtain a corrosion dynamic feature vector set comprises: the multi-element time sequence data set comprises an ambient temperature sequence, an electrolyte concentration sequence, a stress load sequence and an oxidation-reduction potential sequence; respectively executing missing point detection on the environment temperature sequence, the electrolyte concentration sequence, the stress load sequence and the oxidation-reduction potential sequence, and interpolating and supplementing the missing points by using the data median in the local time window to generate an interpolated multi-element time sequence data set; Respectively applying sliding window filters to each sequence in the interpolated multi-element time sequence data set, wherein the sliding window filters adaptively adjust window widths according to preset corrosion response characteristic time scales so as to filter short-time noise fluctuation and obtain a smoothed multi-element time sequence data set; Performing time stamp alignment on different sequence data in the smoothed multi-element time sequence data set to ensure that each moment point simultaneously contains the synchronous measured value of the ambient temperature, the electrolyte concentration, the stress load and the oxidation-reduction potential to form a synchronous time sequence data matrix; And extracting statistical characteristics and variation trend characteristics of each corrosion factor from the synchronous time sequence data matrix, wherein the statistical characteristics comprise mean values, variances and skewness, the variation trend characteristics comprise linear fitting slopes and local extreme point densities, and all the characteristics are combined to form the corrosion dynamic characteristic vector set.
3. 3. The system for online evaluation of corrosion resistance of high nitrogen steel under multi-factor coupled corrosion according to claim 2, wherein extracting statistical features and trend features of each corrosion factor from the synchronized time series data matrix comprises: calculating average temperature and temperature fluctuation variance in a preset feature extraction time section for an environment temperature sequence in the synchronous time sequence data matrix, and taking the average temperature and temperature fluctuation variance as a first statistical feature; Performing piecewise linear fitting on the environmental temperature sequence, calculating absolute value differences of adjacent piecewise fitting slopes, and taking an average value of the absolute value differences as a change intensity characteristic of the environmental temperature sequence; calculating the product of the concentration accumulation increasing quantity and the instantaneous concentration gradient change rate of the electrolyte concentration sequence in the preset characteristic extraction time section as an electrolyte concentration impact factor for representing the concentration impact strength; Detecting the times of crossing a preset potential threshold point of the oxidation-reduction potential sequence, dividing the crossing times by the length of a feature extraction time section to obtain the oscillation frequency feature of the oxidation-reduction potential sequence; Calculating the ratio of the peak value to the valley value of the stress load sequence, and identifying the loading-holding cycle times exceeding a preset stress threshold value in the stress load sequence to form a stress cycle fatigue characteristic; And combining the first statistical characteristic, the variation intensity characteristic, the electrolyte concentration impact factor, the oscillation frequency characteristic and the stress cycle fatigue characteristic into a multidimensional vector to generate the corrosion dynamic characteristic vector set.
4. 4. The online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupled corrosion according to claim 3, wherein the input of the corrosion dynamic feature vector set into a pre-trained deep corrosion behavior prediction model generates a microscopic corrosion state evolution map of the high-nitrogen steel surface under the current service environment through a multi-layer feature fusion network inside the deep corrosion behavior prediction model comprises the following steps: Inputting the corrosion dynamic feature vector set to a bottom time sequence coding layer of the multi-layer feature fusion network, and carrying out time domain local correlation extraction on the dynamic feature of each corrosion factor through one-dimensional convolution check to generate a time sequence coding feature vector containing local correlation; inputting the time sequence coding feature vectors containing the local correlation into a spatial attention layer, calculating real-time interaction weights among different corrosion factors, and carrying out weighted fusion on the time sequence coding feature vectors by using the real-time interaction weights to generate coupling factor fusion feature vectors; Inputting the coupling factor fusion feature vector into a multi-layer perceptron network, wherein the multi-layer perceptron network is used for simulating a joint action mechanism of corrosion electrochemistry and stress corrosion crack propagation, and outputting a microscopic state parameter set representing the corrosion front advancing rate, the pit geometric form evolution trend and the passivation film stability change trend; Mapping the microscopic state parameter set to a two-dimensional or three-dimensional discrete grid on the surface of the high-nitrogen steel material, filling microscopic state parameter values of corresponding position points in each unit of the discrete grid, and generating the microscopic corrosion state evolution map containing time dimension.
5. 5. The online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupled corrosion according to claim 4, wherein identifying a potential corrosion damage mode of the high-nitrogen steel material at the current moment according to the microscopic corrosion state evolution map comprises: Performing spatial cluster analysis on the microscopic corrosion state evolution map, identifying spatial region clusters with similar microscopic state parameter values, and defining each spatial region cluster as a potential local corrosion damage unit; for each local corrosion damage unit, calculating a correlation index between the corrosion front advancing rate, the pit depth gradient and the passivation film thickness change rate, and determining a dominant corrosion damage mechanism of the local corrosion damage unit according to a numerical interval of the correlation index; invoking mapping rules of corrosion damage mechanisms and damage modes from a corrosion damage mode knowledge base, and matching corresponding damage modes for each dominant corrosion damage mechanism, wherein the damage modes at least comprise uniform corrosion, pitting corrosion, crevice corrosion, stress corrosion cracking and corrosion fatigue; And integrating the damage modes of all the local corrosion damage units, generating a global potential corrosion damage mode list of the current high-nitrogen steel material, and marking the spatial distribution range and severity index of each damage mode in the microscopic corrosion state evolution map.
6. 6. The system for online evaluation of corrosion resistance of high nitrogen steel under multi-factor coupled corrosion according to claim 5, wherein said invoking the knowledge base of corrosion damage patterns matches a corresponding corrosion kinetics evolution path for each identified corrosion damage pattern, comprising: Inquiring a corrosion damage mode knowledge base by taking the identified corrosion damage mode as an index, and extracting a dynamic equation prototype corresponding to each damage mode and key control parameters, wherein the key control parameters comprise a corrosion current density threshold value, a crack expansion driving force factor and a repassivation rate constant; extracting the average value of the environmental temperature sequence, the electrolyte concentration sequence, the stress load sequence and the oxidation-reduction potential sequence in a space region corresponding to the current damage mode from the microscopic corrosion state evolution map, and taking the average value as the current environmental load condition; Substituting the current environmental load condition into the dynamic equation prototype, and carrying out numerical calculation on key control parameters in the dynamic equation prototype to generate a concrete corrosion dynamic evolution path which is applicable to the current evaluation scene; The erosion dynamics evolution path describes the mathematical relationship of the erosion depth, crack length or damage area over time for a given damage pattern in the form of a functional curve or parameterized equation.
7. 7. The online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupling corrosion according to claim 6, wherein the calculation of the material performance degradation prediction curves of the high-nitrogen steel material at different depth levels in a preset evaluation period based on the corrosion dynamics evolution path corresponding to the corrosion damage mode comprises the following steps: setting a group of material depth layers, wherein each depth layer corresponds to a specific depth range inside the high-nitrogen steel material; Predicting a predicted value reached by the corrosion damage in the preset evaluation period by utilizing a corresponding corrosion dynamics evolution path of each identified corrosion damage mode; Comparing the predicted value with material depth levels to determine which material depth levels are penetrated or affected by the corrosion damage at the end of a preset evaluation period; Calculating the performance degradation degree of each affected depth level according to a material performance degradation model stored in a corrosion damage mode knowledge base, wherein the material performance degradation model takes damage amount as input, and outputs the performance degradation rate of material yield strength, fracture toughness and fatigue limit; and generating material performance degradation prediction curves of different depth layers by taking time as a horizontal axis and the performance degradation rate of each depth layer as a vertical axis.
8. 8. The online evaluation system for corrosion resistance of high nitrogen steel under multi-factor coupled corrosion according to claim 7, wherein generating a corrosion resistance level quantification report of the high nitrogen steel material according to the material performance degradation prediction curve comprises: extracting performance degradation rate data of the material performance degradation prediction curves of different depth layers at the end of a preset evaluation period; Comparing the performance degradation rate of each depth level to a predefined threshold criteria defining a plurality of level ranges from mild, moderate, severe to failure; counting the number of depth layers with performance degradation rate exceeding a 'serious' level threshold, and taking the ratio of the number of depth layers to the total number of depth layers as an overall damage depth index; The average degradation rate of all depth layers with the performance degradation rate being in a medium level and above is counted and used as a material performance degradation index; Based on the overall damage depth index and the material performance degradation index, obtaining a comprehensive corrosion resistance grade quantization score through weighted calculation; Generating a corrosion resistance grade quantification report comprising the comprehensive corrosion resistance grade quantification score, the overall damage depth index, the material performance degradation index and the detailed degradation of each depth layer.
9. 9. The online evaluation system for corrosion resistance of high nitrogen steel under multi-factor coupled corrosion according to claim 8, wherein the pre-trained deep corrosion behavior prediction model is constructed in a manner comprising: Collecting a historical experimental data set of the high-nitrogen steel material in a multi-factor coupling corrosion environment, wherein the historical experimental data set comprises accelerated corrosion experimental records under different environment temperatures, electrolyte concentrations, stress loads and oxidation-reduction potential combination conditions, and corresponding microcosmic corrosion morphology image sequences and electrochemical impedance spectrum test data; Performing data cleaning and standardization processing on the historical experimental data set, removing invalid data samples caused by experimental equipment faults, performing standardization processing on environmental temperature, electrolyte concentration, stress load and oxidation-reduction potential, and performing size unification and gray scale normalization on a microcosmic corrosion morphology image sequence to generate a standardized training data set; the method comprises the steps of constructing a basic network architecture of the deep corrosion behavior prediction model, wherein the basic network architecture comprises an input layer, the multi-layer feature fusion network and an output layer which are sequentially connected, and the multi-layer feature fusion network consists of the bottom time sequence coding layer, a spatial attention layer and a multi-layer perceptron network cascade; Dividing the standardized training data set into a training set and a verification set, performing iterative training on the basic network architecture by adopting a supervised learning mode, inputting a training sample consisting of the statistical characteristics and the change trend characteristics of a standardized environmental temperature sequence, an electrolyte concentration sequence, a stress load sequence and an oxidation-reduction potential sequence into an input layer in each training iteration, and optimizing network weight parameters by using corrosion front propulsion rate, pit geometric form parameters and passivation film thickness variation marked in a corresponding microscopic corrosion morphology image sequence as supervision labels through a counter-propagation algorithm; introducing a physical constraint loss function in the training process, wherein the physical constraint loss function is based on the conventional mean square error loss, and superposing a polarization curve fitting residual error item based on the corrosion electrochemical principle and a crack propagation rate consistency penalty item based on fracture mechanics to force the network output to conform to the basic rule of corrosion physics and chemistry; And periodically evaluating the model performance in the training process by using the verification set, terminating the training when the comprehensive loss function value on the verification set is not reduced any more after continuous preset iteration times, and storing the network weight parameter at the moment as the pre-trained deep corrosion behavior prediction model.
10. 10. The online evaluation system for corrosion resistance of high nitrogen steel under multi-factor coupled corrosion according to claim 9, further comprising an incremental learning module for online incremental learning of the pre-trained deep corrosion behavior prediction model: after a corrosion resistance grade quantification report of each evaluation is generated, the complete corrosion dynamic characteristic vector set used in the evaluation and the calibrated actual corrosion damage data obtained from the subsequent actual detection are used as a training sample pair together; Adding the training sample pairs into an incremental training sample library, and sampling the sample library at equal intervals to control the scale of the sample library; Periodically using samples in an incremental training sample library to perform parameter fine tuning training on the deep corrosion behavior prediction model, wherein the fine tuning training adopts a small learning rate strategy to protect the existing knowledge of the model, and simultaneously learns corrosion characteristics related to the current service environment in a new sample; after each fine tuning training, updating the recommended value of the key control parameter of the prototype of the dynamic equation in the corrosion damage mode knowledge base, so that the key control parameter is closer to the actual corrosion rate of the current service environment.

Description

Online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupling corrosion Technical Field The invention belongs to the technical field of metal corrosion monitoring and evaluation, and particularly relates to an online evaluation system for corrosion resistance of high-nitrogen steel under multi-factor coupling corrosion. Background The corrosion resistance of the conventional high-nitrogen steel in the multi-factor coupling corrosion environment is evaluated, environmental parameters are collected by adopting a single sensor, material corrosion data are obtained by means of an off-line detection means, simple filtering treatment is only carried out in data preprocessing, time synchronization alignment operation of missing value interpolation and multi-source data is not carried out, the evaluation process depends on a traditional corrosion prediction model to output single indexes such as macroscopic corrosion rate, potential and the like, and a corrosion state representation mode of a microscopic level is not constructed. In the prior art, under a multi-factor coupling working condition, the problems of time sequence dislocation, abnormal fluctuation interference and data deletion exist in multi-source time sequence data, a depth model does not adopt a multi-layer characteristic fusion network structure, a high-nitrogen steel surface microscopic corrosion state evolution spectrum cannot be generated, no special knowledge base support exists after corrosion damage mode identification, the overall performance is estimated only by adopting a unified corrosion dynamics model, the corresponding corrosion dynamics evolution path cannot be matched, and the performance degradation changes of different depth layers of the material cannot be calculated. Under a multi-factor coupling corrosion environment, high-nitrogen steel multi-element time sequence corrosion data cannot finish accurate pretreatment and time synchronization, a microscopic corrosion state evolution map is difficult to generate through a multi-layer feature fusion network of a depth model, a corrosion damage mode cannot be identified according to the microscopic corrosion state evolution map and an exclusive dynamics evolution path cannot be matched, material performance degradation curves of different depth layers in a preset evaluation period cannot be calculated based on the corresponding paths, and an accurate corrosion resistance grade quantification result cannot be formed. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; Therefore, the invention provides an online evaluation system for corrosion resistance of high nitrogen steel under multi-factor coupling corrosion, which comprises the following components: the data acquisition module is used for deploying the sensor array in the service environment and acquiring a multi-element time sequence data set of the surface of the high-nitrogen steel material relative to the corrosion environment in real time; the characteristic processing module is used for preprocessing the multi-element time sequence data set to obtain a corrosion dynamic characteristic vector set, wherein the preprocessing comprises missing value interpolation, abnormal fluctuation point filtering and multi-source data time synchronization alignment; The state evolution module is used for inputting the corrosion dynamic feature vector set into a pre-trained deep corrosion behavior prediction model, and generating a microscopic corrosion state evolution map of the high-nitrogen steel surface in the current service environment through a multi-layer feature fusion network in the deep corrosion behavior prediction model; The path matching module is used for identifying potential corrosion damage modes of the high-nitrogen steel material at the current moment according to the microscopic corrosion state evolution map, calling a corrosion damage mode knowledge base and matching corresponding corrosion dynamics evolution paths for each identified corrosion damage mode; The performance evaluation module is used for calculating and obtaining material performance degradation prediction curves of the high-nitrogen steel material at different depth levels in a preset evaluation period based on the corrosion dynamics evolution path corresponding to the corrosion damage mode, and generating a corrosion resistance level quantification report of the high-nitrogen steel material according to the material performance degradation prediction curves. Further, preprocessing the multi-element time sequence data set to obtain a corrosion dynamic characteristic vector set, including: the multi-element time sequence data set comprises an ambient temperature sequence, an electrolyte concentration sequence, a stress load sequence and an oxidation-reduction potential sequence; respectively executing missing point detection on the environment t