Multi-parameter monitoring method and device for material stress corrosion test

Abstract

According to the multi-parameter monitoring method and device for the material stress corrosion test, the flow rate, temperature and ion concentration of corrosive liquid are adjusted in real time through closed-loop control, and the real service environment of a material is simulated; in combination with a constant strain rate and constant load mixed loading mode, loading parameters are dynamically adjusted according to material failure and environment change, and the behavior of deviating from real failure is avoided. Electrochemical, mechanical and corrosion environment data are synchronously collected, a ridge regression algorithm is utilized to fuse multi-modal features, a Granger causal test and mutual information entropy are combined to analyze a corrosion-mechanical-environment interaction effect, and early-stage tiny damage is captured. And calculating the number of residual cycles, generating a three-dimensional probability cloud picture to quantify a failure risk level, and realizing dynamic regulation and control through a grading strategy instead of traditional fixed threshold judgment. According to the method, the real-time performance, the reliability and the automation level of stress corrosion testing are remarkably improved, and a systematic solution is provided for deep exploration of a material failure mechanism in a complex service environment.

Description

Technical Field

[0001] The present invention relates to the technical field of material corrosion testing, and particularly to a multi-parameter monitoring method and device for material stress corrosion testing. Background Art

[0002] Existing material stress corrosion testing technologies mostly adopt constant strain rate or constant load loading modes, lacking an adaptive adjustment mechanism and being unable to adjust loading parameters in real time according to material degradation such as plastic deformation accumulation or corrosion medium concentration changes, resulting in the test process deviating from the true failure behavior of the material. In addition, existing technologies mostly analyze electrochemical signals or mechanical signals in isolation, ignoring the interaction of corrosion-mechanics-environment, and it is difficult to capture early micro damages under the synergistic action of corrosion and stress, such as local corrosion acceleration and film layer rupture.

[0003] Furthermore, traditional detection means (such as OCP drift, EIS impedance change) are not sensitive enough to the early signs of crack initiation. Especially under the complex coupling action of corrosion and stress, it is difficult to accurately identify the precursor signals of crack formation. In the case of relying on fixed thresholds to determine failure, it is impossible to dynamically quantify and predict uncertainties, and the control measures require manual intervention with a slow response speed, easily leading to excessive damage to the specimen or loss of key data. These technical bottlenecks limit the in-depth analysis of the true failure process of materials in complex service environments, and there is an urgent need to improve the accuracy and reliability of stress corrosion testing through dynamic regulation, intelligent detection, and multi-field coupling analysis means. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention proposes a multi-parameter monitoring method and device for material stress corrosion testing to solve the technical problem that the material failure mechanism in the existing technology is difficult to accurately evaluate.

[0005] The technical solution adopted by the present invention is a multi-parameter monitoring method for material stress corrosion testing.

[0006] In a first realizable manner, it includes: applying a load to a material specimen according to a loading mode and the number of loading cycles set by a user; performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen to generate a fused feature vector; identifying whether a mutation occurs in the corrosion behavior based on the temporal variation of the fused feature vector. If no mutation occurs, maintaining the step of performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen. If a mutation occurs, predicting the change trends of the electrochemical signal, mechanical signal, and corrosion environment data, and adjusting the number of loading cycles; Generate a confidence level according to the change trend. If the confidence level is lower than 85%, terminate the loading and corrosion process. If the confidence level is not lower than 85%, trigger the hierarchical regulation mechanism.

[0007] Further, apply a load to the material specimen according to the loading mode and the number of loading cycles set by the user. It includes: Select one of the constant strain rate loading mode and the constant load loading mode according to the user settings, and apply a load to the material specimen. When setting the constant strain rate loading mode, apply a load to the material specimen according to the number of loading cycles. The material specimen is stretched at a constant strain rate during the loading process. In the constant strain rate loading mode, each loading-unloading is regarded as a complete loading cycle. When setting the constant load loading mode, apply a load to the material specimen. The tensile of the material specimen rapidly rises to the set load value at the initial stage of loading, and then the load remains constant. In the constant load loading mode, the continuous loading per unit time is equivalent to a complete loading cycle.

[0008] Further, perform multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data to generate a fused feature vector, including: Extract the corrosion process sensitive features and mechanical response correlation features according to the electrochemical signal, mechanical signal, and corrosion environment data. Extract cross-modal coupling features by cross-modal analyzing the synergistic effect of the electrochemical signal, mechanical signal, and corrosion environment data. Optimize the weights of the corrosion process sensitive features, mechanical response correlation features, and cross-modal coupling features through the ridge regression algorithm to construct a fused feature vector.

[0009] Further, the electrochemical signal includes electrochemical impedance spectroscopy, current density, open circuit potential, local impedance modulus decay rate, the corrosion environment data includes ion concentration, and the mechanical signal includes displacement rate, load, main frequency offset of load fluctuation, mechanical strain energy density. Extract the corrosion process sensitive features and mechanical response correlation features according to the electrochemical signal, mechanical signal, and corrosion environment data, including: Obtain the relationship curve between the phase angle and frequency in the electrochemical impedance spectroscopy, and extract the slope change rate of the phase angle. Take the slope change rate as a corrosion process sensitive feature. Use the method shown in the following difference calculation formula to calculate the difference according to the current density in the scanning area of the scanning electrochemical microscopy probe and the current density in the adjacent scanning cycle, and take the difference as a corrosion process sensitive feature: where D represents the difference degree between adjacent scanning periods, represents the current density measured in the t-th scanning period, represents the current density measured by the SECM probe in the previous period; An association model between ion concentration and open-circuit potential drift rate is established based on the Nernst equation. The open-circuit potential is obtained according to the ion concentration and the association model, and the open-circuit potential is used as a corrosion process sensitive feature; According to the displacement rate and the load, a displacement-load curve is generated. The first derivative of the displacement-load curve is obtained as the strain rate, and the second derivative of the displacement-load curve is obtained as the curvature value. The curvature value is used as a mechanical response correlation feature; A mapping relationship between the strain rate and the offset of the main frequency of load fluctuation is established according to the Paris crack growth model. The offset of the main frequency of load fluctuation is obtained according to the strain rate and the mapping relationship, and the offset of the main frequency of load fluctuation is used as a mechanical response correlation feature; The correlation coefficient of stress corrosion acceleration is calculated according to the mechanical strain energy density and the attenuation rate of the local impedance modulus, and the correlation coefficient is used as a mechanical response correlation feature.

[0010] Further, the electrochemical signal includes electrochemical impedance spectroscopy, and the mechanical signal includes displacement rate, load. By cross-modal analysis of the synergistic effects of the electrochemical signal, the mechanical signal and the corrosion environment data, cross-modal coupling features are extracted, including: calculating the temperature change rate according to the displacement rate and the Granger causality test model, and using the temperature change rate as a cross-modal coupling feature; Obtain the mutation event sequence of the phase angle of the electrochemical impedance spectroscopy and the event sequence of the load fluctuation amplitude exceeding the threshold, and use windowed mutual information entropy to analyze the time alignment degree between the mutation event of the phase angle of the electrochemical impedance spectroscopy and the load fluctuation event, and use the time alignment degree as a cross-modal coupling feature.

[0011] Further, based on the time series change of the fusion feature vector, it is identified whether the corrosion behavior generates a mutation, including: using the method shown in the following standardized feature vector calculation formula, normalizing the fusion feature vector according to the mean of the 30-minute sliding window and the standard deviation of the sliding window to generate a standardized feature vector: where represents the standardized feature vector, X(t) represents the fusion feature vector, represents the mean of the sliding window, represents the standard deviation of the sliding window; Using the method shown in the following cumulative deviation calculation formula, introducing an adjustment coefficient to perform anti-noise compensation on the standardized feature vector and calculating the cumulative deviation: ) Among them, represents the cumulative deviation, t represents, represents the normalized eigenvector, and k represents the adjustment coefficient; Adopt the method shown in the following dynamic threshold calculation formula to calculate the dynamic threshold according to the cumulative deviation sequence in the previous 30 minutes: Among them, represents the dynamic threshold, represents the mean value of the sliding window, represents the standard deviation of the sliding window; When the cumulative deviation exceeds the dynamic threshold, it is determined as an abnormal event of the mutation of the corrosion behavior.

[0012] Furthermore, predict the change trends of the electrochemical signal, mechanical signal and corrosion environment data according to the prediction mechanism, and adjust the number of loading cycles, including: using the exponential smoothing method to extrapolate the trends of the electrochemical signal, mechanical signal and corrosion environment data, and output the short-term prediction results of the above parameters within the next 10s; Call the pre-trained long short-term memory network, the input layer receives the fused eigenvector and the real-time environment parameters, and the output layer outputs the predicted value of the strain rate within the next 1 minute; adopt the method shown in the following Paris crack growth formula to calculate the crack growth rate: Among them, represents the crack growth rate, C represents the material constant reflecting the basic crack growth rate, represents the stress intensity factor range, and m represents the Paris exponent; Calculate the remaining number of cycles of the material specimen under the current corrosion-load coupling condition according to the crack growth rate, strain energy density and the change amount of the local corrosion potential; Update the number of loading cycles according to the remaining number of cycles.

[0013] Furthermore, generate a confidence level according to the change trend, including: adopting the method shown in the following confidence level calculation formula to generate a confidence level according to the predicted value of the strain rate: Among them, z(t) represents the standardized residual, represents the actual observed value of the strain rate, represents the time, represents the predicted value of the strain rate output by the long short-term memory network, denotes the historical prediction error calculated based on the autoregressive model, denotes the confidence level, denotes the standard normal distribution function; Three-dimensionally correlate the confidence level with the spatial position and test time process of the material specimen to generate a probability cloud map, and display the failure risk level with a color gradient.

[0014] Furthermore, if the confidence level is not less than 85%, trigger a hierarchical regulation mechanism, including: based on the predicted values of the crack growth rate, remaining cycle number, and strain rate, train a classification model in combination with historical data, and output a failure probability value; When the failure probability value is in the range of 0.7 - 0.85, trigger an audible and visual alarm; When the failure probability value is in the range of 0.85 - 0.95, reduce the loading rate according to the user settings; When the failure probability value is greater than 0.95, immediately pause the loading and inject an inhibitor into the corrosive solution environment.

[0015] Combined with the first implementation manner, in the second implementation manner, it includes a device for multi-parameter monitoring of material stress corrosion testing, including: a load control module that applies a load to the material specimen according to the loading mode and loading cycle number set by the user; a feature extraction and recognition algorithm unit that performs multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen to generate a fused feature vector; a mutation detection and pre-judgment analysis unit that identifies whether a mutation occurs in the corrosion behavior based on the temporal change of the fused feature vector. If no mutation occurs, maintain the step of performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen. If a mutation occurs, predict the change trend of the electrochemical signal, mechanical signal, and corrosion environment data, and adjust the loading cycle number; an alarm and feedback control unit that generates a confidence level according to the change trend. If the confidence level is less than 85%, terminate the loading and corrosion process. If the confidence level is not less than 85%, trigger a hierarchical regulation mechanism.

[0016] As can be seen from the above technical solutions, the beneficial technical effects of the present invention are as follows: 1. Optimize and quantify damage accumulation through the loading cycle number, and adjust the loading parameters in real time according to material degradation and environmental changes to avoid deviating from the true failure behavior.

[0017] 2. Synchronously extract sensitive features of electrochemistry, mechanics, and corrosion environment, optimize the weights through the ridge regression algorithm, and combine Granger causality test and mutual information entropy to analyze the interaction of corrosion-mechanics-environment to capture early minor damage.

[0018] 3. Predict the change trend of material corrosion according to the prediction mechanism, generate a three-dimensional probability cloud map to quantify the failure risk level, and achieve dynamic regulation through a grading strategy, replacing the traditional fixed threshold determination, reducing manual intervention and specimen damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0020] Figure 1 It is a flowchart of Embodiment 1 of the present invention; Figure 2 It is a structural diagram of the material stress corrosion test platform of Embodiment 2 of the present invention; In the above-mentioned drawings: 1. Pressure sensor; 2. Cavity; 3. Tensile device; 4. Servo motor. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The following will describe in detail the embodiments of the technical solutions of the present invention in conjunction with the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, so they are only examples and cannot be used to limit the protection scope of the present invention.

[0022] It should be noted that unless otherwise specified, the technical terms or scientific terms used in this application should have the ordinary meanings understood by those skilled in the art to which the present invention belongs.

[0023] Embodiment 1 This embodiment provides a multi-parameter monitoring method for material stress corrosion testing. The following will detail the working principle of Embodiment 1: The flowchart of this embodiment is as Figure 1 shown and includes: Fix and install the material specimen to be tested in the loading fixture, ensure the initial position of the specimen is centered, and connect the working electrode interface of the specimen to the electrochemical detection module to ensure good electrical contact.

[0024] Set the corrosion liquid environment of the material specimen according to the target service conditions, including injecting the corrosion liquid into the cavity for corrosion simulation, setting the flow rate, temperature, and ion concentration of the corrosion liquid through the cavity, and maintaining the dynamic parameters stable through closed-loop control. Among them, the flow rate is adjusted by an electronically controlled micro pump, the temperature is controlled by a micro heating / cooling unit, and the ion concentration is adjusted by switching the flow path of the electronically controlled solenoid valve and the concentrated / diluted liquid.

[0025] Select one of the constant strain rate loading mode and the constant load loading mode according to the user settings, and apply a load to the material specimen according to the number of loading cycles; the process of applying the load includes presetting the loading mode in the central control platform, starting the servo motor, and loading through the cooperation of a high-precision lead screw and a loading fixture, and using a displacement-load double-closed-loop control system for real-time adjustment to ensure that the loading process is highly controllable.

[0026] Synchronously collect and store the electrochemical signals, mechanical signals, and corrosion environment data of the material specimen; perform multi-modal feature extraction on the electrochemical signals, mechanical signals, and corrosion environment data respectively, and generate a fused feature vector; Based on the temporal variation of the fused feature vector, identify whether the corrosion behavior has a mutation; if there is no mutation, maintain the step of synchronously collecting and storing the electrochemical signals, mechanical signals, and corrosion environment data of the material specimen; if there is a mutation, predict the change trend of the electrochemical signals, mechanical signals, and corrosion environment data according to the prediction mechanism, and adjust the loading and corrosion processes; Generate a confidence level according to the change trend, and generate a three-dimensional probability cloud map; Judge whether the confidence level is lower than 85%; If the confidence level is lower than 85%, terminate the loading and corrosion processes. At this time, start a scanning electrochemical microscope for subsequent manual auxiliary verification such as post-fracture analysis and corrosion morphology observation of the specimen. If the confidence level is not lower than 85%, obtain the material failure probability value and trigger a hierarchical regulation mechanism.

[0027] Select one of the constant strain rate loading mode and the constant load loading mode according to the user settings, including: When setting the constant strain rate loading mode, the material specimen is stretched at a constant strain rate during the loading process, and the stress gradually increases with deformation until fracture or the set end point; When setting the constant load loading mode, the tensile of the material specimen rapidly rises to the set load value at the initial stage of loading, and then the load remains constant, and the stress gradually decreases with the corrosion or creep of the material, and the strain will gradually increase.

[0028] The number of loading cycles includes: in the constant strain rate loading mode, each loading-unloading is regarded as a complete cycle, and in the constant load loading mode, the continuous loading per unit time is equivalent to a complete cycle, which is used to quantify the damage accumulation.

[0029] Collect and store the electrochemical signals, mechanical signals, and corrosion environment data of the material specimen to be tested. The electrochemical signals include electrochemical impedance spectroscopy, current density, open circuit potential, and the decay rate of local impedance modulus. The corrosion environment data includes ion concentration, and the mechanical signals include displacement rate, load, main frequency offset of load fluctuation, and mechanical strain energy density.

[0030] Perform multi-modal feature extraction on the electrochemical signals and corrosion environment data respectively to generate a fused feature vector, including: extracting corrosion process sensitive features and mechanical response related features according to the electrochemical signals, mechanical signals, and corrosion environment data; extracting cross-modal coupling features by cross-modal analysis of the synergistic effects of the electrochemical signals, mechanical signals, and corrosion environment data; and optimizing the weights of the corrosion process sensitive features, mechanical response related features, and cross-modal coupling features through the ridge regression algorithm to construct a fused feature vector. Synchronously extract the sensitive features of electrochemistry, mechanics, and corrosion environment, optimize the weights through the ridge regression algorithm, and combine Granger causality test and mutual information entropy to analyze the interaction of corrosion-mechanics-environment and capture early micro damages.

[0031] Extracting corrosion process sensitive features and mechanical response related features according to the electrochemical signals, mechanical signals, and corrosion environment data includes: applying a low-frequency alternating current signal of 0.1–1 Hz to the material to be tested, obtaining the relationship curve between the phase angle and frequency in the electrochemical impedance spectroscopy, and extracting the slope change rate of the phase angle. The slope change rate is used as a corrosion process sensitive feature. The slope change rate is used to characterize the stability of the corrosion film structure. When the absolute value of the slope change rate drops by more than a set threshold, it is determined as a precursor to the rupture of the film layer.

[0032] Extract the standard deviation of the current density in the scanning area of the scanning electrochemical microscopy (SECM) probe And the difference degree from the adjacent scanning period, and use the difference degree as a corrosion process sensitive feature. When the standard deviation of the current density and the difference degree from the adjacent scanning period increase synchronously by more than a specified multiple of the baseline value, it is determined that local corrosion is accelerating; The calculation formula for the difference degree from the adjacent scanning period is: where D represents the difference degree from the adjacent scanning period, represents the current density measured by the SECM probe in the t-th scanning period, represents the current density measured by the SECM probe in the previous period; Based on the Nernst equation, an association model between ion concentration and the drift rate of the open circuit potential (OCP) is established to obtain the open circuit potential, and the open circuit potential is used as a corrosion process sensitive feature. When the actual open circuit potential deviates from the theoretical value by more than a set threshold, it is determined that there is local enrichment of ion concentration; Generate a displacement-load curve according to the mechanical signal, obtain the second derivative ( ) of the displacement-load curve, calculate the curvature value according to the second derivative, and use the curvature value as a mechanical response correlation feature. When the absolute value of the curvature value reaches the curvature threshold, it is determined as the yield inflection point. The curvature threshold is in the order of magnitude range of to times the Young's modulus of the material. The specific value can be adjusted according to different materials through pre-experiments or historical data, and it is applicable to various material types including metals, ceramics, and composites. In the elastic stage of the material, the load and displacement are linearly related, the first derivative is the constant stiffness, and the second derivative is zero; near the yield point, that is, in the non-linear section, strain hardening or local plastic deformation causes the load growth rate to decrease, the second derivative becomes negative, and the absolute value increases rapidly with the expansion of the plastic zone. Therefore, the curvature value is calculated according to the second derivative, and when the absolute value of the curvature value reaches the curvature threshold, it is determined as the yield inflection point.

[0033] Establish a mapping relationship between the strain rate and the offset of the main frequency of load fluctuation according to the Paris crack growth model, obtain the offset of the main frequency of load fluctuation according to the strain rate and the mapping relationship, and use the offset of the main frequency of load fluctuation as a mechanical response correlation feature. When the offset of the main frequency of load fluctuation exceeds 15%, it means that the main frequency migrates to the low-frequency band, and it is determined as the crack growth period; Calculate the correlation coefficient between the mechanical strain energy density ( ) and the decay rate of the local impedance modulus ( ), and use the correlation coefficient as a mechanical response correlation feature. When the correlation coefficient shows a negative correlation and exceeds -0.8, it is determined that the stress corrosion acceleration stage is entered.

[0034] Extract cross-modal coupling features by cross-modal analysis of the synergistic effects of the electrochemical signal, mechanical signal, and corrosion environment data, including: calculating the time-delay effect of the temperature change rate on the displacement rate according to the Granger causality test model, and using the temperature change rate as a cross-modal coupling feature. When the temperature change rate satisfies: dT / dt ≥ 0.2 °C / s and the lag time exceeds 5 s, it is determined as the material creep-dominated stage; Obtain the mutation event sequence of the phase angle of the electrochemical impedance spectrum and the event sequence of the load fluctuation amplitude exceeding the threshold, and use windowed mutual information entropy to analyze the time alignment between the phase angle mutation event of the electrochemical impedance spectrum and the load fluctuation event. Take the time alignment as a cross-modal coupling feature, and when the time alignment (i.e., the synchronous occurrence probability) exceeds 70%, it is determined as the corrosion-mechanical coupling failure critical point.

[0035] Identify whether the corrosion behavior has a mutation based on the temporal variation of the fusion feature vector, including: Normalize the fusion feature vector according to the mean of the 30-minute sliding window and the standard deviation of the sliding window to generate a standardized feature vector. The calculation formula is: Where, represents the standardized feature vector, X(t) represents the fusion feature vector, represents the mean of the sliding window, represents the standard deviation of the sliding window; Introduce a regulation coefficient to perform anti-noise compensation on the standardized data and calculate the cumulative deviation. The calculation formula is: ) Where, represents the cumulative deviation, t represents, represents the standardized feature vector, and k represents the regulation coefficient; Calculate the dynamic threshold according to the cumulative deviation sequence in the first 30 minutes. The calculation formula is: Where, represents the dynamic threshold, represents the mean of the sliding window, represents the standard deviation of the sliding window; the mean and standard deviation of the sliding window are updated by sliding with time to adapt to material degradation or environmental changes.

[0036] When the cumulative deviation exceeds the dynamic threshold, it is determined as an abnormal event of the corrosion behavior mutation.

[0037] According to the prediction mechanism, predict the change trends of the electrochemical signal, mechanical signal, and corrosion environment data, and adjust the loading and corrosion processes, including: using the exponential smoothing method to perform trend extrapolation on the electrochemical signal, mechanical signal, and corrosion environment data, strengthening the contribution degree of recent data through the weighting coefficient, and outputting the short-term prediction results of the above parameters within the next 10s; calling the pre-trained long short-term memory network (LSTM) to output the strain rate prediction value within the next 1 minute. Its network architecture configuration is: The input layer receives the fused feature vector and real-time environmental parameters; the hidden layer includes 64 GRU units, and the time window is set to 60 seconds; the output layer outputs the predicted value of the strain rate; Calculate the crack growth rate based on the Paris crack growth formula, and the calculation formula is: Where, represents the crack growth rate, C represents the material constant reflecting the basic crack growth rate, represents the stress intensity factor range, m represents the Paris exponent, reflecting the crack growth rate's sensitivity to ; Calculate the remaining cycle number of the material specimen under the current corrosion-load coupling condition according to the crack growth rate, strain energy density, and local corrosion potential change amount. This remaining cycle number can indirectly evaluate the remaining service life of the material; update the loading cycle number according to the remaining cycle number. Optimize and quantify damage accumulation through the loading cycle number, and adjust the loading parameters in real-time according to material degradation and environmental changes to avoid deviating from the true failure behavior.

[0038] Generate a confidence level according to the change trend and generate a three-dimensional probability cloud map, including generating a confidence level according to the predicted strain rate value. The calculation formula is: Where, z(t) represents the standardized residual (characterizing the degree of deviation from the prediction center), represents the actual observed value of the strain rate, represents time, represents the predicted strain rate value output by the long short-term memory network, represents the historical prediction error calculated based on the autoregressive model, represents the confidence level, represents the standard normal distribution function; Three-dimensionally correlate the confidence level with the spatial position and test time process of the material specimen to generate a probability cloud map, and display the failure risk level with a color gradient to visually present the damage evolution trend of the material in the stress corrosion environment.

[0039] Obtain the material failure probability value in real-time and trigger a hierarchical control mechanism, including: Train a classification model based on the crack growth rate, remaining cycle number, and predicted strain rate value, combined with historical data, and output the failure probability value; When the failure probability value is in the range of 0.7 - 0.85, trigger an audible and visual alarm; When the failure probability value is in the range of 0.85 - 0.95, automatically reduce the loading rate by 10% - 50% according to user settings; When the failure probability value is greater than 0.95, the loading is immediately paused, and an inhibitor is injected into the corrosive liquid environment. According to the prediction mechanism, the change trend of material corrosion is predicted, a three-dimensional probability cloud map is generated to quantify the failure risk level, and dynamic regulation is achieved through a grading strategy, replacing the traditional fixed threshold determination, which can reduce manual intervention and specimen damage.

[0040] Example 2 This embodiment provides a multi-parameter monitoring device for a material stress corrosion test box. The working principle of Example 2 is described in detail below: The device of this embodiment includes: A load control module that applies a load to a material specimen according to the loading mode and the number of loading cycles set by the user.

[0041] In this embodiment, the load control module is a material stress corrosion test platform, as Figure 2 shown, where a pressure sensor 1 is used to obtain the mechanical signal of the material specimen; a cavity 2 is used to provide a corrosive liquid environment, a stretching device 3 is used to apply a load to the material specimen according to the loading mode and the number of loading cycles set by the user, and a servo motor 4 is used to control the stretching device 3.

[0042] A feature extraction and recognition algorithm unit that performs multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen to generate a fused feature vector; A mutation detection and pre-judgment analysis unit that identifies whether a mutation occurs in the corrosion behavior based on the temporal change of the fused feature vector. If no mutation occurs, the step of performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen is maintained. If a mutation occurs, the change trends of the electrochemical signal, mechanical signal, and corrosion environment data are predicted, and the number of loading cycles is adjusted; An alarm and feedback control unit that generates a confidence level according to the change trend. If the confidence level is lower than 85%, the loading and corrosion processes are terminated. If the confidence level is not lower than 85%, a grading regulation mechanism is triggered.

[0043] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the various embodiments of the present invention, and they should all be covered by the scope of the claims and the specification of the present invention.

Claims

1. A multi-parameter monitoring method for material stress corrosion testing, characterized in that, Including: Applying a load to a material specimen according to a loading mode and a number of loading cycles set by a user; Performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen to generate a fused feature vector; Identifying whether a mutation occurs in the corrosion behavior based on the temporal variation of the fused feature vector. If no mutation occurs, maintaining the step of performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data of the material specimen. If a mutation occurs, predicting the change trends of the electrochemical signal, mechanical signal, and corrosion environment data, and adjusting the number of loading cycles; Generating a confidence level according to the change trends. If the confidence level is lower than 85%, terminating the loading and corrosion processes. If the confidence level is not lower than 85%, triggering a hierarchical control mechanism.

2. The multi-parameter monitoring method for material stress corrosion test according to claim 1, wherein Applying a load to a material specimen according to a loading mode and a number of loading cycles set by a user, including: Selecting one of a constant strain rate loading mode and a constant load loading mode according to the user setting, and applying a load to the material specimen; When setting the constant strain rate loading mode, applying a load to the material specimen according to the number of loading cycles. The material specimen is stretched at a constant strain rate during loading, and each loading-unloading is regarded as a complete loading cycle in the constant strain rate loading mode; When setting the constant load loading mode, applying a load to the material specimen. The tensile of the material specimen rapidly rises to a set load value at the initial stage of loading, and then the load remains constant. Each continuous loading per unit time is regarded as a complete loading cycle in the constant load loading mode.

3. The multi-parameter monitoring method for material stress corrosion testing according to claim 1, characterized in that Performing multi-modal feature extraction on the electrochemical signal, mechanical signal, and corrosion environment data to generate a fused feature vector, including: Extracting corrosion process sensitive features and mechanical response correlation features from the electrochemical signal, mechanical signal, and corrosion environment data; Extracting cross-modal coupling features by analyzing the synergistic effect of the electrochemical signal, mechanical signal, and corrosion environment data through cross-modal analysis; Optimizing the weights of the corrosion process sensitive features, mechanical response correlation features, and cross-modal coupling features through a ridge regression algorithm to construct a fused feature vector.

4. The multi-parameter monitoring method for material stress corrosion test according to claim 3, wherein The electrochemical signal includes electrochemical impedance spectroscopy, current density, open circuit potential, local impedance modulus decay rate, the corrosion environment data includes ion concentration, and the mechanical signal includes displacement rate, load, main frequency offset of load fluctuation, mechanical strain energy density; Extracting corrosion process sensitive features and mechanical response correlation features from the electrochemical signal, mechanical signal, and corrosion environment data, including: Obtaining a relationship curve between the phase angle and frequency in the electrochemical impedance spectroscopy, and extracting the slope change rate of the phase angle. Taking the slope change rate as a corrosion process sensitive feature; Using the method shown in the following difference calculation formula, calculating the difference according to the current density in the scanning area of the scanning electrochemical microscopy probe and the current density in the adjacent scanning period, and taking the difference as a corrosion process sensitive feature: where D represents the difference degree between adjacent scanning cycles, represents the current density measured in the t-th scanning cycle, represents the current density measured by the SECM probe in the previous cycle; Based on the Nernst equation, an association model between ion concentration and the drift rate of open-circuit potential is established. The open-circuit potential is obtained according to the ion concentration and the association model, and the open-circuit potential is used as a corrosion process sensitive feature; According to the displacement rate and the load, a displacement-load curve is generated. The first derivative of the displacement-load curve is obtained as the strain rate, and the second derivative of the displacement-load curve is obtained as the curvature value. The curvature value is used as a mechanical response association feature; According to the Paris crack growth model, a mapping relationship between the strain rate and the offset of the main frequency of load fluctuation is established. The offset of the main frequency of load fluctuation is obtained according to the strain rate and the mapping relationship, and the offset of the main frequency of load fluctuation is used as a mechanical response association feature; According to the mechanical strain energy density and the attenuation rate of the local impedance modulus, the correlation coefficient of stress corrosion acceleration is calculated, and the correlation coefficient is used as a mechanical response association feature.

5. A multi-parameter monitoring method for material stress corrosion testing according to claim 3, characterized in that, The electrochemical signal includes electrochemical impedance spectroscopy, and the mechanical signal includes displacement rate, load. By cross-modal analysis of the synergistic effects of the electrochemical signal, mechanical signal and corrosion environment data, cross-modal coupling features are extracted, including: According to the displacement rate and the Granger causality test model, the temperature change rate is calculated, and the temperature change rate is used as a cross-modal coupling feature; The mutation event sequence of the phase angle of the electrochemical impedance spectroscopy and the event sequence of the load fluctuation amplitude exceeding the threshold are obtained, and the windowed mutual information entropy is used to analyze the time alignment degree between the mutation event of the phase angle of the electrochemical impedance spectroscopy and the load fluctuation event. The time alignment degree is used as a cross-modal coupling feature.

6. The multi-parameter monitoring method for material stress corrosion test according to claim 3, characterized in that, Based on the temporal variation of the fused feature vector, it is identified whether the corrosion behavior undergoes a mutation, including: Using the method shown in the following standardized feature vector calculation formula, the fused feature vector is normalized according to the mean of the 30-minute sliding window and the standard deviation of the sliding window to generate a standardized feature vector: Among them, represents the standardized eigenvector, and X(t) represents the fused eigenvector, represents the mean of the sliding window, represents the standard deviation of the sliding window; Using the method shown in the following cumulative deviation calculation formula, an adjustment coefficient is introduced to perform anti-noise compensation on the standardized feature vector, and the cumulative deviation is calculated; ) Among them, represents the cumulative deviation, t represents, represents the normalized eigenvector, and k represents the adjustment coefficient; Using the method shown in the following dynamic threshold calculation formula, the dynamic threshold is calculated according to the cumulative deviation sequence in the previous 30 minutes; Among them, represents the dynamic threshold, represents the mean value of the sliding window, represents the standard deviation of the sliding window; When the cumulative deviation exceeds the dynamic threshold, it is determined as an abnormal event of the mutation of the corrosion behavior.

7. A multi-parameter monitoring method for material stress corrosion testing according to claim 1, characterized in that, According to the prediction mechanism, the change trends of the electrochemical signal, mechanical signal and corrosion environment data are predicted, and the number of loading cycles is adjusted, including: The exponential smoothing method is used to perform trend extrapolation on the electrochemical signal, mechanical signal and corrosion environment data, and the short-term prediction results of the above parameters within the next 10 s are output; Call the pre-trained long short-term memory network. The input layer receives the fused feature vector and real-time environment parameters, and the output layer outputs the predicted value of the strain rate within the next 1 minute. Using the method shown in the following formula based on the Paris crack growth formula, the crack growth rate is calculated; Among them, represents the crack growth rate, C represents the material constant reflecting the basic crack growth rate, represents the stress intensity factor range, and m represents the Paris exponent; According to the crack growth rate, the strain energy density and the change amount of the local corrosion potential, the remaining number of cycles of the material specimen under the current corrosion-load coupling condition is calculated; The number of loading cycles is updated according to the remaining number of cycles.

8. The multi-parameter monitoring method for material stress corrosion test according to claim 7, characterized in that, According to the change trend, a confidence level is generated, including: Using the method shown in the following confidence level calculation formula, the confidence level is generated according to the predicted value of the strain rate: Among them, z(t) represents the standardized residual, represents the actual observed value of the strain rate, represents time, represents the predicted value of the strain rate output by the long short-term memory network, represents the historical prediction error calculated based on the autoregressive model, represents the confidence level, represents the standard normal distribution function; Three-dimensionally correlate the confidence level with the spatial position and test time course of the material specimen to generate a probability cloud map, and display the failure risk level with a color gradient.

9. The multi-parameter monitoring method for material stress corrosion testing according to claim 8, characterized in that, If the confidence level is not less than 85%, trigger a hierarchical control mechanism, including: Based on the predicted crack growth rate, remaining cycle number, and strain rate, train a classification model in combination with historical data, and output a failure probability value; When the failure probability value is in the range of 0.7 - 0.85, trigger an audible and visual alarm; When the failure probability value is in the range of 0.85 - 0.95, reduce the loading rate according to user settings; When the failure probability value is greater than 0.95, immediately pause the loading and inject a corrosion inhibitor into the corrosive liquid environment.

10. An apparatus for implementing a multi-parameter monitoring method for a material stress corrosion test according to any one of claims 1 to 9, characterized in that, Including: A load control module that applies a load to the material specimen according to the loading mode and number of loading cycles set by the user; A feature extraction and recognition algorithm unit that performs multi-modal feature extraction on the electrochemical signals, mechanical signals, and corrosion environment data of the material specimen to generate a fused feature vector; A mutation detection and pre-judgment analysis unit that identifies whether a mutation occurs in the corrosion behavior based on the temporal variation of the fused feature vector. If no mutation occurs, maintain the step of performing multi-modal feature extraction on the electrochemical signals, mechanical signals, and corrosion environment data of the material specimen. If a mutation occurs, predict the change trends of the electrochemical signals, mechanical signals, and corrosion environment data, and adjust the number of loading cycles; An alarm and feedback control unit that generates a confidence level based on the change trends. If the confidence level is less than 85%, terminate the loading and corrosion processes. If the confidence level is not less than 85%, trigger a hierarchical control mechanism.

Images