A method for predicting local corrosion environment severity according to surrounding environment

Abstract

This invention discloses a method for predicting the severity of localized corrosion environments based on the surrounding environment. The method includes: constructing a semi-closed simulation test system to simulate the internal local environment of equipment; conducting accelerated corrosion tests under different external environmental conditions, collecting external environmental parameters, internal local environmental parameters, and corrosion response data of metal samples; constructing a time-series feature matrix from the external environmental parameters; calculating a comprehensive corrosion risk index based on the corrosion response data; constructing and training a correlation prediction model from the time-series feature matrix to the comprehensive corrosion risk index using an attention mechanism long short-term memory network; collecting external environmental parameters in real time and inputting them into the trained correlation prediction model; and outputting a predicted value for the severity of localized corrosion environments at the target location of the equipment. This invention enables accurate and rapid assessment of internal corrosion risk based solely on externally measurable data, solving the problem of difficulty in directly monitoring the internal or localized corrosion environment of equipment.

Description

Technical Field

[0001] This invention belongs to the field of environmental corrosion assessment and material service safety technology, specifically relating to a method for predicting the severity of localized corrosion environments based on the surrounding environment. Background Technology

[0002] In aerospace, power, petrochemical, and energy equipment industries, critical equipment (such as engine hot-end components, reaction vessels, and boiler pipes) often operates for extended periods in complex environments characterized by high temperatures, high humidity, and corrosive media. The corrosive environment (temperature, humidity, contaminant concentration, condensate chemistry, etc.) within the interior or localized crevices of such equipment is often impossible to measure directly and in real-time due to spatial constraints, difficulty in sensor placement, or extreme operating conditions. However, the severity of this localized environment directly determines the corrosion rate of materials and the safe lifespan of the equipment. Currently, engineering practices typically rely on periodic shutdowns for inspection, empirical estimations, or overly conservative design margins, lacking a scientific, online, and non-invasive method for assessing the severity of localized corrosive environments.

[0003] Traditional methods primarily rely on installing corrosion monitoring probes (such as electrochemical noise and resistance probes) inside the equipment, but this is typically costly, difficult to install, and may interfere with equipment operation. Another approach is to measure atmospheric environmental parameters surrounding the equipment (such as ambient temperature, relative humidity, SO2, and Cl-). - While concentration can be used to indirectly assess the process, simple empirical formulas or linear models cannot accurately describe the complex heat transfer, mass transfer, condensation, and chemical reaction processes between the "external environment" and the "internal local microenvironment," resulting in large prediction errors and insufficient reliability.

[0004] Therefore, there is an urgent need to develop a method that can use readily available external environmental parameters of equipment to accurately predict the severity of internal, locally unmeasurable corrosive environments through data-driven models, so as to achieve proactive early warning and intelligent management of equipment corrosion status. Summary of the Invention

[0005] The purpose of this invention is to provide a method for predicting the severity of localized corrosion environments based on the surrounding environment, which can accurately and quickly assess internal corrosion risks based solely on externally measurable data.

[0006] To achieve the above objectives, one aspect of the present invention provides a method for predicting the severity of a localized corrosive environment based on the surrounding environment, comprising: Step S1: Construct a simulation test system. The test system includes a main test chamber and an environmental control chamber. The main test chamber is used to simulate the local environment inside the equipment and contains metal samples. The environmental control chamber surrounds or connects to the main test chamber and is used to control and measure external environmental parameters. Step S2: Run the test system under different external environmental conditions to conduct accelerated corrosion tests, collect external environmental parameters, internal local environmental parameters and corrosion response data of metal samples, construct a time-series feature matrix of external environmental parameters, and calculate a comprehensive corrosion risk index based on corrosion response data. The rows of the time-series feature matrix represent time points or test cycles, and the column features include external environmental parameters, derived parameters characterizing the relationship between internal and external environments, and historical parameter statistics. Step S3: Using the temporal feature matrix as input and the comprehensive corrosion risk index as the target output, construct and train the correlation prediction model from the temporal feature matrix to the comprehensive corrosion risk index using an attention mechanism long short-term memory network. Step S4: Collect external environmental parameters of the equipment in real time, construct a time-series feature matrix and input it into the trained correlation prediction model, and output the predicted value of the local corrosion environment severity of the target part of the equipment.

[0007] According to the method for predicting the severity of localized corrosion environment based on the surrounding environment of the present invention, by establishing a nonlinear correlation model between the external macroscopic environmental parameters of the equipment and the severity of internal localized corrosion, it is possible to accurately and quickly assess the internal corrosion risk based solely on externally measurable data, thus solving the problem that it is difficult to directly monitor the internal or localized corrosion environment of equipment. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort: Figure 1 This is a flowchart illustrating a method for predicting the severity of localized corrosion environments based on the surrounding environment, according to an embodiment of the present invention. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0010] One embodiment of the present invention provides a method for predicting the severity of a localized corrosive environment based on the surrounding environment, as described below. Figure 1 The process shown below provides a detailed description of the method in this embodiment of the invention.

[0011] Step S1: Construct a simulation test system and collect data Design and build a semi-enclosed high-temperature corrosion test system to simulate the working conditions of equipment under the "external measurable environment" and "internal unmeasurable local environment".

[0012] The system includes: (1) a main test chamber, which simulates the high-temperature local environment inside the equipment and houses the metal sample; (2) an environmental control chamber, which surrounds or connects to the main test chamber to form a semi-closed simulation test system, used to accurately control and measure "external" environmental parameters (such as temperature T_out, relative humidity RH_out, specific pollutant gas concentration C_out, etc.); (3) a heating and temperature control device, used to form a stable high-temperature zone on the inner wall of the main test chamber or near the sample, forming a controllable temperature difference (ΔT) with the external environment; (4) a data acquisition system, used to synchronously and continuously record external environmental parameters, internal local environmental parameters (such as sample surface temperature T_surf, electrochemical characteristics of liquid film that may condense), and corrosion response parameters of the metal sample (such as corrosion potential E_corr, polarization resistance). Or the corrosion weight loss rate V_loss obtained by periodic sampling.

[0013] Step S2: Construct a multi-dimensional parameter dataset The test system was run under different external environmental conditions (variable T_out, RH_out, C_out, and ΔT) to conduct accelerated corrosion tests.

[0014] Collect time series data and construct the equipment environment parameter matrix X and corrosion response parameter matrix Y: (1) The X matrix is ​​a time series feature matrix. Each row represents a time point or a test cycle. The column features include: external environmental parameters (T_out, RH_out, C_out...), derived parameters characterizing the relationship between internal and external environments (such as ΔT, calculated dew point temperature difference), and parameter statistics within the historical time window (mean, variance, slope). (2) The Y matrix corresponds to the local corrosion severity quantification index for each time point / cycle. The following can be selected: ① Electrochemical parameters, such as corrosion current density i_corr; ② Corrosion rate, such as annual corrosion depth converted from weight loss; ③ Comprehensive corrosion severity index, which is obtained by weighting multiple corrosion parameters after normalization.

[0015] To condense the complex corrosion state into a scalar that can be learned by a model, this invention defines a Comprehensive Corrosion Risk Index (CRI). This index is obtained by weighted fusion of measurable parameters reflecting different aspects of the corrosion process from experiments.

[0016] a. Parameter normalization: First, normalize the measurement parameters of different dimensions to the interval [0, 1].

[0017] in, Represents the original parameters (such as polarization resistance) Maximum corrosion pit depth , thermocouple current ), and These are the minimum and maximum values ​​of this parameter across the entire training dataset.

[0018] b. CRI calculation:

[0019] in: p Normalized polarization resistance (the higher the resistance, the lower the corrosion tendency, hence 1). p (Indicates corrosion activity); depth : The maximum depth of corrosion pits after normalization; galv Normalized thermocouple current density; w 1. w 2. w 3 represents the weighting coefficients for each parameter, satisfying... w 1+ w 2 + w 3 = 1; weights can be determined through principal component analysis (PCA) or in combination with expert experience (e.g., in the early stages of crevice corrosion, pit depth weight). w 2 is likely higher; in scenarios dominated by galvanic corrosion, w 3. Higher weight). In this embodiment, the following can be adopted: w 1 = 0.4 w 2 = 0.4 w 3 = 0.2.

[0020] The above formula ensures that the CRI value is continuously distributed between 0 (no risk of corrosion) and 100 (extreme risk of corrosion), quantitatively characterizing the corrosion severity of the local environment.

[0021] Step S3: Construction and Training of Neural Network Correlation Model Using the environmental parameter matrix X as input and the corrosion response parameter matrix Y as the target output, a deep learning correlation prediction model is constructed and trained. The model input is a multi-dimensional time series, and the model accepts a time series of length [length missing]. The historical window sequence is used as input to predict the CRI for the current period. Represented as:

[0022] in The sequence length (e.g., 10 periods). For feature dimensions.

[0023] The preferred model is a Long Short-Term Memory (LSTM) network with an attention mechanism or a Temporal Convolutional Network (TCN) to capture the dynamic cumulative effect of environmental parameters on corrosion over time. The attention-mechanism LSTM prediction model captures the long-term dependencies in the evolution of corrosion severity and identifies key historical periods. It includes LSTM layers, attention layers, and an output layer, with the attention mechanism used to assign weights to the hidden states at historical time steps to generate context vectors.

[0024] 1. LSTM layer: This layer processes the input sequence... Feed it into the stacked LSTM cells. For the first... Layer LSTM at time steps Calculation:

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] in, These are the forget gate, input gate, and output gate, respectively. In cellular state; It is in a hidden state; It is the sigmoid function; For element-wise multiplication; These are trainable parameters.

[0031] After passing through two LSTM layers, a hidden state sequence containing temporal information is obtained. .

[0032] 2. Attention layer: An attention mechanism is introduced to assign different weights to the hidden states at each historical time step, focusing on the period that has the greatest impact on the current prediction, and combining the historical hidden states with the latest information to generate a context vector.

[0033]

[0034]

[0035]

[0036] in, It is a time step Attention score It is the hidden state of the last time step (representing the latest information). These are trainable parameters. These are normalized attention weights. It is the context vector after weighted summation.

[0037] 3. Output layer: context vector The hidden state of the last time step The CRI values ​​are predicted by splicing the data and then using a fully connected network for regression analysis.

[0038]

[0039] in, For the first Forecast CRI for each cycle, This represents vector concatenation. These are the output layer parameters.

[0040] The model is trained and optimized. The training process includes: (1) dividing the dataset into training set, validation set and test set; (2) defining loss function (such as mean squared error MSE); (3) training with optimization algorithm (such as Adam) and preventing overfitting by early stopping strategy on validation set; (4) the trained model can learn the complex mapping relationship from external measurable environmental sequence to internal local corrosion severity.

[0041] Mean squared error (MSE) is used as the loss function to measure the predicted value. Compared with the true value The difference between (the CRI calculated by the experiment).

[0042]

[0043] in, This represents the number of training samples.

[0044] Step S4: Use the model for on-site prediction and early warning In actual equipment deployment, external environmental parameters and derived parameters defined in step S2 are collected in real time or periodically at locations easily accessible for sensor installation on the outside of the equipment, forming a time-series data stream. This data stream is then input into a pre-trained correlation prediction model, which outputs real-time predictions or trend warnings of the severity of localized corrosion in target areas within the equipment. Based on the prediction results, targeted maintenance strategies can be developed, such as adjusting process parameters, planning maintenance in advance, or replacing components.

[0045] In this embodiment, preferably, in step S1, the metal sample can represent different materials (such as 304, 316L) to expand the applicability of the model. In step S3, Bayesian optimization method can be used to automatically optimize the hyperparameters of the neural network model. In step S4, the predicted corrosion severity and fatigue performance data of the equipment material can be combined to further predict the remaining corrosion fatigue life.

[0046] The following describes the method of this invention in detail using typical aircraft corrosion scenarios. This example uses the lower fuselage skin connection of narrow-body passenger aircraft and other aircraft operating at coastal airports as the prediction object, predicting the severity of the corrosion environment inside the gaps at this connection. This part is exposed to high salt spray and high humidity ground environment for a long time, and due to the prominent risks of structural gaps and galvanic corrosion (aluminum alloy skin and steel / titanium alloy fasteners), the corrosion microenvironment inside its gaps is difficult to monitor directly.

[0047] 1. Problem and Operating Condition Simulation When aircraft are parked at coastal airports, salty atmospheric pollutants can easily accumulate at the seams and rivets of the lower fuselage skin. Furthermore, due to diurnal temperature variations and humid air, a corrosive microenvironment (electrolyte film) forms within these gaps. The severity of this microenvironment (e.g., Cl) can be significant. - Enrichment levels, electrolyte film conductivity, and pH value directly accelerate pitting corrosion of aluminum alloys and galvanic corrosion between them and fasteners, but sensors cannot be directly deployed for real-time monitoring. Using the method of this invention, the severity of the corrosive microenvironment in the concealed location is predicted by collecting macroscopic environmental data from the airport where the aircraft is located and flight operation data.

[0048] 2. Construction of a semi-closed simulation test system and data acquisition Design a simulation test device for aviation crevice corrosion that can simulate alternating flight segments and ground parking.

[0049] Main test chamber (simulating the microenvironment of fuselage gaps): The interior is equipped with classic "aluminum alloy plate-stainless steel fastener" lap joint samples to simulate the skin connection structure. The chamber can be programmed to control the temperature, simulating the alternating cycle of the flight phase (low temperature, dry) and the ground parking phase (high temperature, high humidity).

[0050] Environmental control chamber (simulating the external atmospheric environment of an airport): Encloses the main test chamber, and can precisely control and monitor temperature (-10℃ to 50℃), relative humidity (30% to 98%), and atomize and spray specific concentrations of NaCl solution (simulating salt spray deposition rate, such as 0.1-10 mg / cm² / day) and low concentrations of SO2 (simulating industrial pollution).

[0051] Data Synchronous Acquisition: (1) External Parameters: Continuously record the temperature T_out, relative humidity RH_out, salt spray deposition rate S_out, and SO2 concentration C_SO2_out of the control cabin. (2) Internal (Gap) Parameters: Monitor the electrochemical impedance spectroscopy (EIS) at the overlapping sample, and periodically obtain the polarization resistance R_p and double-layer capacitance C_dl (reflecting the formation of electrolyte film and corrosion activity); embed miniature temperature and humidity sensors at specific locations on the sample to measure the temperature T_in and relative humidity RH_in (approximate values) in the gap. (3) Corrosion Response Terminal Data: After each set "flight cycle + ground parking" cycle (e.g., simulating 7 days), disassemble the sample, perform three-dimensional morphology scanning, and quantify the maximum corrosion pit depth P. depth And measure the thermocouple current I galv .

[0052] 3. Construction of multi-dimensional parameter dataset Simulation experiments were conducted, covering typical coastal environmental spectra across different seasons (e.g., high-temperature, high-salinity summers and low-temperature, moderate-salinity winters) and varying "flight duration / parking duration" ratios. Each "ground parking period" was used as a data unit to construct the input feature matrix X and the output target vector Y. X-matrix features (one row for each parking period): Macro-environmental parameters: During the parking period, the daily average values ​​and fluctuation range of T_out, RH_out, S_out, and C_SO2_out in the control cabin; Operational parameters: simulated altitude (corresponding to low temperature) and duration of the preceding flight phase; total duration of this parking; Cumulative effect parameters: number of consecutive simulated parking days; cumulative total of S_out over the past N periods; Key derived parameters: Calculate the internal and external temperature difference ΔT = T_in - T_out (approximate); the difference between the external dew point and T_in calculated based on T_out and RH_out.

[0053] Y-vector target (quantification of corrosion severity): A comprehensive corrosion risk index (CRI, range 0-100) is used as the output. This index is determined by the electrochemical parameters at the end of the cycle. Normalized reciprocal, weight 0.4), morphological parameters ( Normalized value, weight 0.4) and galvanic corrosion intensity ( The CRI value is obtained by linear weighting (normalized value, weight 0.2). The higher the CRI value, the more severe the corrosion environment inside the crevice during the storage period.

[0054] 4. Construction and Training of Neural Network Association Models To address the aforementioned time-series data characteristics, a Long Short-Term Memory (Attention-LSTM) network model based on an attention mechanism is constructed.

[0055] Input layer: Receives an X matrix data of sequence length M (e.g., M=10, representing 10 consecutive parking cycles). For the t-th ground parking cycle, its input feature vector xt contains not only the current cycle parameters but also historical information.

[0056] Example of the construction of the feature vector xt (d dimensions in total): , : The mean value of external environmental parameters in period t. Approximate value of the internal and external temperature difference. Flight and parking duration. The cumulative salt spray deposition over the past N cycles. . : The mean and standard deviation of the external temperature over the past N periods.

[0057] Network structure: Sequence data passes through two LSTM layers to capture long-term dependencies; then an attention layer is added to enable the model to focus on the historical stage that has the greatest impact on corrosion in the current cycle (e.g., a period of continuous high-salt storage); finally, a fully connected layer outputs the predicted CRI value for the current cycle.

[0058] Training and Validation: 70% of the total dataset was used for training, 15% for validation, and 15% for testing. Mean squared error (MSE) was used as the loss function, and the Adam optimizer was used for training. Early stopping was performed on the validation set to prevent overfitting. After training, the model achieved a coefficient of determination (R²) of 0.94 between predicted and measured CRI values ​​on the test set, and a mean absolute percentage error (MAPE) of less than 8%, demonstrating good fit.

[0059] 5. On-site prediction and engineering application In actual aircraft operation and maintenance, the following steps should be followed: Data Acquisition: Deploy weather stations at the airports where the target aircraft operates to collect real-time data such as T_out, RH_out, and precipitation (used to estimate salt spray deposition). Simultaneously, obtain the aircraft's post-flight report from the flight operation system, including the previous flight altitude and time, and the planned parking duration.

[0060] Feature generation and prediction: The above data is used to generate a feature sequence in real time according to the definition in step S3. Whenever the aircraft enters the ground parking state, the system automatically calls the pre-trained prediction model deployed in the cloud, inputs the feature sequence of the most recent M cycles, and outputs in real time the corrosion environment severity index (predicted CRI) that the lower skin connection of the aircraft will face during this parking period.

[0061] The above examples fully illustrate the innovation, effectiveness, and engineering practical value of the method of the present invention in the field of aircraft corrosion protection. Based on the principle of the present invention, it can be adapted and implemented for different aircraft types, different parts (such as landing gear bays and engine pylons) or different environments (high-altitude airports and industrial airports).

[0062] In summary, the method for predicting the severity of localized corrosion based on the surrounding environment in this invention first constructs a semi-closed simulation test system to collect external macroscopic environmental parameters, internal local environmental parameters, and material corrosion response data. Through feature engineering, the external parameters are constructed into a time-series feature matrix, and the Comprehensive Corrosion Risk Index (CRI) is calculated based on the corrosion response data to quantify the severity of localized corrosion. An Attention-LSTM network is used to establish a correlation prediction model from external time-series features to the internal CRI. In practical applications, only real-time collection of environmental and operational data from the equipment's periphery is needed, and the trained model can be input to predict the real-time corrosion severity of concealed parts inside the equipment online and non-invasively. This invention solves the problem of directly monitoring localized corrosion environments under special working conditions, and is particularly suitable for assessing the actual corrosion risk faced by metallic materials (such as stainless steel) under special working conditions such as high temperature and semi-closed environments.

[0063] The method of this invention, through the construction of a semi-closed simulation test system, can simultaneously simulate both the "externally measurable environment" and the "internal unmeasurable local environment," achieving synchronous acquisition and correlation analysis of internal and external environmental parameters, particularly through the introduction of key derived parameters such as temperature difference (ΔT) and salt spray deposition rate. Based on the normalization and weighted fusion of multiple corrosion response parameters (such as polarization resistance, maximum corrosion pit depth, and galvanic current), this invention constructs a continuously quantified comprehensive corrosion risk index to characterize the severity of local corrosion environments. This invention employs attention-enhanced long short-term memory networks or temporal convolutional networks as temporal prediction models to capture the nonlinear, temporal correlation between external environmental parameters and internal corrosion risk, achieving high-precision prediction from externally measurable parameters to internally unmeasurable corrosion severity. The multi-source temporal feature engineering of this invention not only includes current external environmental parameters but also introduces statistical features (such as mean, variance, and slope) and operational mode parameters (such as preceding flight altitude and parking duration) within historical time windows, enhancing the model's temporal modeling capabilities.

[0064] The method of this invention eliminates the need to install sensors in high-risk or hard-to-reach areas inside equipment. It achieves internal corrosion risk assessment using only external environmental data, making it safe and cost-effective. Employing a deep learning model, it can capture the complex nonlinear and temporal relationships between the external environment and the internal localized corrosion microenvironment, achieving prediction accuracy far exceeding that of traditional empirical formulas.

[0065] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the severity of a localized corrosive environment based on the surrounding environment, characterized in that, include: Step S1: Construct a simulation test system. The test system includes a main test chamber and an environmental control chamber. The main test chamber is used to simulate the local environment inside the equipment and contains metal samples. The environmental control chamber surrounds or connects to the main test chamber and is used to control and measure external environmental parameters. Step S2: Run the test system under different external environmental conditions to conduct accelerated corrosion tests, collect external environmental parameters, internal local environmental parameters and corrosion response data of metal samples, construct a time-series feature matrix of external environmental parameters, and calculate a comprehensive corrosion risk index based on corrosion response data. The rows of the time-series feature matrix represent time points or test cycles, and the column features include external environmental parameters, derived parameters characterizing the relationship between internal and external environments, and historical parameter statistics. Step S3: Using the temporal feature matrix as input and the comprehensive corrosion risk index as the target output, construct and train the correlation prediction model from the temporal feature matrix to the comprehensive corrosion risk index using an attention mechanism long short-term memory network. Step S4: Collect external environmental parameters of the equipment in real time, construct a time-series feature matrix and input it into the trained correlation prediction model, and output the predicted value of the local corrosion environment severity of the target part of the equipment.

2. The method as described in claim 1, characterized in that, The comprehensive corrosion risk index is calculated as follows: in, To determine the comprehensive corrosion risk index, Polarization resistor, The maximum depth of the corrosion pit For thermocouple current, p The normalized polarization resistance. depth This represents the normalized maximum corrosion pit depth. galv This is the normalized thermocouple current density; w 1. w 2. w 3 is the weighting coefficient, which satisfies w 1 + w 2 + w 3 = 1.

3. The method as described in claim 1 or 2, characterized in that, The attention mechanism long short-term memory network includes a long short-term memory network layer, an attention layer, and an output layer: The Long Short-Term Memory (LSTM) network layer takes the temporal feature matrix as a multi-dimensional time series input and obtains a hidden state sequence containing temporal information through two layers of LSM network. The attention layer assigns different weights to the hidden states at each historical time step through an attention mechanism, and generates a context vector by combining the historical hidden states with the latest information. The output layer concatenates the context vector with the hidden state at the last time step, and uses a fully connected network to regress and predict the value of the comprehensive corrosion risk index.

4. The method as described in claim 3, characterized in that, The context vector is generated as follows: in, It is a time step Attention score It is the hidden state of the last time step. For trainable parameters, These are normalized attention weights. It is a context vector.

5. The method as described in claim 4, characterized in that, The following regression analysis predicts the value of the comprehensive corrosion risk index: in, For the first The comprehensive corrosion risk index is predicted for each cycle. This represents vector concatenation. These are the output layer parameters.

6. The method as described in claim 1 or 2, characterized in that, External environmental parameters include external temperature, relative humidity, and pollutant gas concentration; internal local environmental parameters include sample surface temperature and the electrochemical characteristics of the condensed liquid film; derived parameters include the internal and external temperature difference and the dew point temperature difference calculated based on the temperature difference; historical parameter statistics include the mean, variance, and slope of external environmental parameters and derived parameters within the historical time window.

7. The method as described in claim 6, characterized in that, The equipment is an aircraft, and the target areas are the lower fuselage skin connection, landing gear bay, and engine pylon area. External environmental parameters also include salt spray deposition parameters.

8. The method as described in claim 7, characterized in that, The column features of the time series feature matrix also include the aircraft's operating mode parameters, which include the aircraft's previous flight altitude, time, and current ground parking duration.

9. The method as described in claim 1 or 2, characterized in that, The test system also includes a heating and temperature control device and a data acquisition system. The heating and temperature control device is used to create a high-temperature zone on the inner wall of the main test chamber or near the sample, forming a controllable temperature difference with the external environment. The data acquisition system is used to record external environmental parameters, internal local environmental parameters, and corrosion response parameters of the metal sample.

Images