Steel structure corrosion prevention evaluation system based on environmental data and AI

Abstract

The present application relates to the technical field of structural health monitoring and artificial intelligence, and discloses a steel structure corrosion prevention evaluation system based on environmental data and AI, which comprises a data sensing module, an environmental index generating module, an impact function generating module, a stress-corrosion coupling effect model constructing module, a steel structure life prediction module, a grading early warning module and a system self-learning module; the present application can quantify the corrosion impact of environmental emergencies and realize advanced early warning; for the first time, real-time stress monitoring data is directly incorporated into a corrosion life evaluation model, and the coupling effect of stress corrosion is accurately depicted; modeling and prediction are carried out for the unique environment and stress state of each specific structure, and the results are more instructive; the present application realizes the transformation from planned maintenance to condition-based maintenance, avoids waste caused by excessive maintenance, prevents risks caused by insufficient maintenance, and significantly reduces the life cycle cost.

Description

Technical Field

[0001] This invention relates to the fields of structural health monitoring and artificial intelligence technology, and more specifically to a steel structure corrosion assessment system based on environmental data and AI. Background Technology

[0002] Steel structures are widely used in bridges, ports, wind turbine towers, chemical equipment and other fields, but corrosion is a core problem affecting structural safety and service life.

[0003] Traditional corrosion assessment methods mainly rely on a combination of regular manual inspections and local non-destructive testing. Conventional practices include sampling and assessing the corrosion status of structures through visual inspection, coating thickness gauges, corrosion test strips, and ultrasonic testing. Meanwhile, some advanced systems have introduced environmental sensors such as temperature, humidity, and salt spray concentration monitoring to collect corrosion environment data and establish the correlation between corrosion rate and environmental factors based on statistical methods or empirical models, thereby achieving preliminary prediction of structural corrosion trends and life estimation.

[0004] Existing evaluation methods also have the following shortcomings:

[0005] Existing technologies lack the ability to effectively identify and quantify environmental abrupt events. Their assessment models are mostly based on long-term average environmental data, which cannot capture and quantify the impact of short-term drastic environmental changes such as rainstorms, salt spray attacks, and chemical leaks on the corrosion process. This makes them insensitive to the risk of accelerated corrosion failure caused by sudden events, resulting in delayed early warnings. Often, damage can only be detected through regular inspections after it has occurred.

[0006] Existing life prediction methods generally fail to incorporate real-time stress monitoring data directly into the evaluation model. They typically use conservative, fixed safety factors to roughly estimate the stress impact or completely ignore the coupling effect between stress and the environment. This results in an inability to accurately characterize the actual risks of stress corrosion cracking and corrosion fatigue, leading to significant biases in life prediction results. These results are either too conservative or too optimistic, failing to provide accurate basis for decision-making.

[0007] Existing assessment systems rely heavily on industry-standard norms, specifications, or average empirical models derived from large amounts of data. These models fail to adequately consider the differences in the microenvironment of specific structures, the uniqueness of their load history, and the individualized characteristics of their material states. Consequently, the assessment results tend to be generalized and cannot accurately reflect the true state of a specific object, thus offering limited guidance.

[0008] The existing maintenance strategy mainly adopts a planned maintenance model based on fixed cycles. This model is prone to two drawbacks: first, over-maintenance, which involves unnecessary repairs and replacements of components in good condition, resulting in a waste of human and material resources; second, under-maintenance, which fails to detect hidden damage in time, misses the best maintenance window, and ultimately causes minor problems to develop into major problems, significantly increasing repair costs and safety risks.

[0009] Therefore, methods that are forward-looking, accurately predictive, consider individual differences, and have the characteristics of cost reduction and efficiency improvement are needed to solve the above problems. Summary of the Invention

[0010] In order to overcome the above-mentioned defects of the prior art, the present invention provides a steel structure corrosion assessment system based on environmental data and AI to solve the problems existing in the background art.

[0011] To achieve the above objectives, the present invention provides the following technical solution: a steel structure corrosion assessment system based on environmental data and AI, comprising:

[0012] Data sensing module: used to collect real-time environmental data and steel structure status data of the steel structure;

[0013] Environmental index generation module: used to generate a real-time environmental corrosion index based on the current environmental data of the steel structure through weighted calculation;

[0014] Impact function generation module: used to identify environmental abrupt changes, and combine the intensity and duration of changes in the environmental corrosivity index to construct an environmental abrupt change impact function to calculate the additional corrosion amount;

[0015] Stress-corrosion coupling effect model construction module: It is used to select an algorithm and train it to obtain a stress-corrosion coupling effect model based on the real-time status data of steel structure, real-time environmental corrosion index and material properties, and output the actual corrosion rate by combining the impact function.

[0016] Steel structure life prediction module: used to calculate the cumulative corrosion damage of steel structure at the current moment based on the actual corrosion rate and to predict its life.

[0017] Tiered early warning module: Used to combine the output results of each model and sensor data to provide tiered early warning and visual decision support.

[0018] System self-learning module: used to periodically calculate and compare the predicted corrosion thickness reduction with the measured value of the ultrasonic thickness gauge, and automatically calibrate the corrosion rate model.

[0019] The technical effects and advantages of this invention are as follows:

[0020] 1. This invention can quantify the corrosive impact of sudden environmental events, achieve early warning, and avoid structural safety accidents caused by sudden corrosion events.

[0021] 2. This invention is the first to directly incorporate real-time stress monitoring data into the corrosion life assessment model, accurately characterizing the coupling effect of stress corrosion and significantly improving the accuracy of life prediction.

[0022] 3. The system of this invention models and predicts the unique environment and stress state of each specific structure, rather than using a general empirical model, and the results are more instructive.

[0023] 4. This invention realizes the transformation from planned maintenance to condition-based maintenance, avoiding the waste caused by over-maintenance and preventing the risks caused by insufficient maintenance, thus significantly reducing the total life cycle cost. Attached Figure Description

[0024] Figure 1 This is a structural block diagram of the present invention.

[0025] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The steel structure corrosion assessment system based on environmental data and AI involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Reference Figure 1 This invention provides a steel structure corrosion assessment system based on environmental data and AI, including a data perception module, an environmental index generation module, an impact function generation module, a stress-corrosion coupling effect model construction module, a steel structure life prediction module, a graded early warning module, and a system self-learning module.

[0028] Reference Figure 2 The specific implementation steps of the present invention include the following steps:

[0029] S1. Real-time acquisition of environmental data and structural status data of the steel structure.

[0030] It should be specifically noted that the environmental data collection for the steel structure includes:

[0031] Temperature and humidity sensor: monitors atmospheric temperature and humidity to assess the risk of condensation;

[0032] Corrosive ion sensor: monitors chloride ion (Cl⁻) and sulfur dioxide (SO₂) concentrations;

[0033] pH sensor: monitors the acidity or alkalinity of the surface electrolyte;

[0034] Weather stations: monitor wind speed, wind direction, and rainfall to analyze sudden environmental events.

[0035] The steel structure condition data acquisition specifically includes:

[0036] Strain gauge / fiber grating sensor (FBG): Real-time monitoring of micro-strain and stress changes;

[0037] Corrosion rate sensors, such as electrochemical noise (EN) and linear polarization resistance (LPR), directly measure the instantaneous corrosion rate;

[0038] Ultrasonic thickness gauge: Periodically and automatically measures the remaining thickness of the base material under the coating.

[0039] S2. Based on the current environmental data of the steel structure, a real-time environmental corrosion index is generated through weighted calculation.

[0040] It should be specifically noted that the real-time environmental corrosivity index is as follows:

[0041] ;

[0042] Where E is the environmental corrosivity index at the i-th time point, which is calculated by comprehensively considering various complex environmental factors to quantify environmental corrosivity; the closer E is to 1, the stronger the current environment's corrosivity to the steel structure; the closer it is to 0, the weaker the corrosivity.

[0043] j is the index variable, representing the j-th environmental factor, and m is the total number of environmental factors, indicating that the summation calculation process will traverse m environmental factors;

[0044] For example: j=1 represents chloride ion concentration (Cl⁻), j=2 represents sulfur dioxide concentration (SO2), j=3 represents relative humidity (RE), j=4 represents temperature (T), j=5 represents pH value; then m=5.

[0045] W j Let W be the weight of the j-th environmental factor, ranging from 0 to 1, and the sum of the weights of all factors equals 1. Different environmental factors contribute differently to corrosion, and their weights are W. j The larger the value, the more important the factor is in the comprehensive index, and the greater the impact of its changes on the final environmental corrosivity. The entropy weight method is used for calculation, which is determined by the dispersion of the data itself. The greater the fluctuation of a factor's data, the more information it contains, and the system will automatically assign it a higher weight.

[0046] X ij This refers to the standardized value of the j-th environmental factor at time point i. Standardization maps all factor values ​​to the range of 0 to 1, eliminating differences in units and orders of magnitude and making them comparable. In corrosion analysis, a larger value indicates stronger corrosivity (called a positive indicator), while a smaller value indicates stronger corrosivity (called a negative indicator). The standardization process includes conversion formulas to handle these two cases, ensuring that all X values ​​are ultimately standardized. ij The larger the value, the stronger the corrosiveness, and the logical direction is consistent.

[0047] S3. Identify environmental abrupt changes, and combine the intensity and duration of changes in the environmental corrosivity index to construct an environmental abrupt change impact function to calculate the additional corrosion amount.

[0048] It should be noted that the system automatically identifies environmental abrupt events (such as the onset of acid rain or salt spray) by monitoring the E value in real time and using the CUSUM method of change point detection. Once an abrupt event is identified, the system will quantify the intensity ΔE and duration Δt of the event and construct an environmental abrupt event impact function.

[0049] The specific environmental abrupt change shock function is as follows:

[0050] ;

[0051] Where C is the environmental abrupt change impact function, which represents the portion of corrosion exceeding the normal corrosion amount caused by environmental abrupt change events. This portion is added to the total corrosion amount to make the life prediction more accurate.

[0052] α is an empirical constant, a calibration parameter of the model, which determines the scale of the entire function. Adjusting the value of the function output to match the actual observed value requires fitting historical data or laboratory experiments. The most suitable α value can be derived by analyzing the actual measured corrosion increments after multiple past mutation events.

[0053] ΔE is the intensity of change in the environmental corrosivity index, that is, the intensity of the environmental abrupt event, which quantifies the severity of the event. The larger ΔE is, the more severe the environmental degradation and the more intense the event, resulting in more additional corrosion. The calculation formula is: ΔE = Es - Eb, where Es is the peak value of E during the abrupt event, and Eb is the baseline value of the environment before the abrupt event, which is the average value of the previous stage.

[0054] β is the intensity index, an empirical index used to adjust the degree of influence of intensity ΔE on the final result. If β > 1, it means that the influence of intensity is superlinear, that is, if the intensity increases by a little, the additional corrosion will increase a lot. It is determined by data fitting and reflects the sensitivity of the corrosion process to changes in environmental intensity.

[0055] Δt represents the duration of an environmental abrupt event, i.e., the length of time it takes for the E value to rise abnormally and then fall back to normal levels. It quantifies the persistence of the event. The longer Δt is, the longer the structure is exposed to the harsh environment, and therefore the more additional corrosion accumulates.

[0056] γ is the duration index, an empirical index used to adjust the degree of influence of duration Δt on the final result. The relationship between corrosion amount and time is linear, that is, γ=1 means that if the time is doubled, the corrosion amount is also doubled. If γ<1, it means that there is a saturation effect and the influence of time is decreasing. If >1, it means that the longer the time, the more severe the corrosion acceleration, because the fresh metal is continuously exposed after the corrosion products are washed away. It is determined by data fitting and reflects the characteristics of the corrosion process accumulating over time.

[0057] S4. Based on the real-time status data of the steel structure, the real-time environmental corrosion index, and material properties, select an algorithm and train it to obtain a stress-corrosion coupling effect model, and output the actual corrosion rate by combining the impact function.

[0058] It should be specifically noted that the construction steps of the stress-corrosion coupling effect model are as follows:

[0059] A1. Define the input features, which are the model's input vector X constructed from the factors influencing the model. Specifically:

[0060] X = [σ, f, E, M];

[0061] Where σ is the real-time stress value, the static or quasi-static stress level is the main factor leading to stress corrosion cracking (SCC). The higher the stress, the stronger the atomic activity, the easier it is for the protective film to break, and the easier it is for corrosion to occur.

[0062] F is the stress variation frequency, which is obtained by performing Fourier transform (FFT) time-frequency analysis on the stress signal. Alternating stress is the main cause of corrosion fatigue.

[0063] E is the real-time environmental corrosion index. The same stress level produces different acceleration effects in different corrosive environments.

[0064] M represents material properties, including yield strength, chemical composition, and heat treatment state. Different materials have different resistance to stress corrosion. By using these properties as inputs, the model can distinguish structures made of different materials and achieve personalized predictions.

[0065] A2. Select an algorithm and train it; choose Gradient Boosting Tree (GBRT). The relationship between stress, environment and corrosion rate is complex. GBRT can automatically capture these complex nonlinear interactions. After training, the model reflects which feature has a greater influence and is interpretable.

[0066] The specific steps for training the algorithm are as follows:

[0067] B1. Collect a large amount of historical data, including the input feature X and the corresponding stress corrosion factor S=R / Rb at each time point;

[0068] Where S is the stress corrosion factor, representing the amplification factor of the current stress state on the corrosion rate of the base; R is the actual measured corrosion rate, which is the true corrosion rate under the combined action of stress and environment; Rb is the corrosion rate caused solely by the environment, a baseline value calculated from the current environmental corrosivity index; the stress corrosion factor quantifies the accelerating effect of stress. When S=1, it means that stress does not accelerate corrosion and corrosion is entirely caused by the environment. When S>1, it means that stress is accelerating corrosion. For example, S=2.5 means that the current stress level makes the corrosion rate 2.5 times that under pure environmental action. The larger the S value, the more severe the stress-corrosion coupling effect and the more dangerous the structure.

[0069] B2. Input the dataset (X, S) into the GBRT algorithm for training;

[0070] B3. The algorithm minimizes the mean square error between the predicted value and the actual value of the stress corrosion factor by continuously iterating, that is, adding decision trees one by one.

[0071] B4. After training, a stress corrosion factor prediction model that can be put into use is obtained.

[0072] A3. Model application: The actual corrosion rate Ra is predicted in real time based on the predicted stress corrosion factor, real-time environmental corrosion rate and impact function.

[0073] The system packages the latest sensor data (σ,f,E,M) into a feature vector X every second / minute, and inputs X into the trained GBRT model. The model instantly outputs the current Sa. The system finally calculates the actual corrosion rate Ra=Re*Sa+C, where Re is the real-time environmental corrosion rate, Sa is the predicted stress corrosion factor, and C is the real-time impact function. If no environmental change occurs at the current moment, then C=0.

[0074] S5. Calculate the cumulative corrosion damage of the steel structure at the current moment based on the actual corrosion rate and predict its lifespan.

[0075] It should be specifically noted that the cumulative damage refers to:

[0076] ;

[0077] Where D(t) represents the cumulative damage, which is the total amount of damage accumulated by the steel structure due to corrosion from the start of its use, i.e., time 0, to the current time t. The health status of the structure is judged by monitoring the value of D(t) in real time.

[0078] 0 is the starting point of integration, which usually represents the moment when the structure starts to be used or the initial moment when the system starts monitoring. t is the ending point of integration, which represents the current moment and defines the time range for calculating cumulative damage. By integrating, the corrosion rate at every instant from time 0 to time t is added together to obtain the accurate total damage.

[0079] Ra(τ) is the actual corrosion rate, that is, the real-time corrosion rate at a specific instant τ. It is the integrand in the integral formula, which comprehensively considers the impact of normal environment, stress acceleration effect and environmental abrupt event. τ is the integration variable, which takes over every time point between 0 and t, so that the corrosion rate at each point is included in the calculation. After the integration is completed, τ itself will not appear in the result D(t).

[0080] The specific steps for lifetime prediction are as follows:

[0081] C1. Set the critical damage value Dc, which is the maximum corrosion damage that the structure can withstand before failure. It is determined by design specifications, laboratory experiments or historical data.

[0082] C2. Based on the current state D(t), input the future environmental prediction scenario and the future stress prediction scenario;

[0083] C3. Using the trained model, simulate and calculate the actual corrosion rate at each future time step by step, and calculate the cumulative future damage Df.

[0084] C4. When Df≥Dc, the corresponding future time point is the predicted failure time T, and the remaining lifetime = failure time - current time.

[0085] S6. Combine the output results of each model with sensor data to provide graded early warning and visualized decision support.

[0086] It should be specifically noted that the tiered early warning system refers to:

[0087] Level 1 warning is a long-term trend warning; the specific triggering conditions include:

[0088] Environmental trend: EC7 > E30*1.3, meaning that the average environmental corrosion index over the past 7 days is more than 1.3 times the average index over the past 30 days, indicating that environmental corrosion has a clear deterioration trend, rather than just a single-day fluctuation;

[0089] Damage progress: D(t) / Dc > 0.4, that is, the cumulative damage has reached 40% of the critical damage value. Based on the design life progress, a macro warning is given, indicating that the structure has passed the initial stage.

[0090] System response: Log entries are made, and the status indicator light for the structure turns yellow on the visual dashboard. The system automatically generates a weekly report and adds the structure to the priority monitoring list.

[0091] Level 2 warnings are short-term, time-limited warnings; the specific triggering conditions include:

[0092] Environmental mutation: ΔE > E30*0.8 and Δt > 2 hours, the intensity of a single environmental mutation event exceeds 0.8 times the 30-day average ECI, and the duration of the event exceeds 2 hours;

[0093] Stress surge: S(t) / SCF24h>1.8 and S(t)>2.0, meaning that the current stress corrosion factor has surged by 80% compared to the average of the past 24 hours, and its absolute value is greater than 2.0. This indicates that the stress level is not only high, but has also changed abnormally in the short term, and corrosion is being accelerated rapidly.

[0094] Lifespan warning: T < Ts, meaning the system's predicted remaining lifespan is shorter than the safety margin of Ts required by the design specifications;

[0095] System Response: Automatically send alert emails / SMS to the operations and maintenance team leader; the structure icon on the visual dashboard turns orange and flashes slowly; the system automatically generates a diagnostic report, indicating whether the main cause is an environmental event or an abnormal stress.

[0096] Level 3 alarms are immediate emergency warnings, and the specific triggering conditions include:

[0097] The damage threshold D(t) / Dc > 0.9 means that the cumulative damage has exceeded 90% of the threshold value, the structure is on the verge of failure, and the risk is high.

[0098] Crack initiation: The acoustic emission sensor detected high-frequency acoustic emission signals (>100 times / second) with high signal intensity (>80 dB), which is a typical characteristic of the rapid propagation of microcracks and the most direct precursor to macroscopic fracture of the structure.

[0099] Structural response: The strain value monitored in real time exceeded 80% of the material's yield strain. This indicates that the structure has entered the nonlinear stage, and local plastic deformation may be occurring, resulting in a significant decrease in load-bearing capacity.

[0100] System Response: Triggers audible and visual alarms, automatically dials the maintenance duty room, the visual dashboard flashes red across the entire screen and a highest priority alarm pops up, the system immediately locks the data and generates an emergency report, and it is strongly recommended to immediately carry out on-site handling and safety assessment.

[0101] The visual decision support specifically includes:

[0102] Global Overview: A geographic information map that marks all monitored steel structures. The real-time health status of each structure is displayed intuitively using yellow / orange / red colors. Clicking on any structure allows you to drill down and view its detailed data.

[0103] Real-time data panel: Digital displays and dashboards show current key indicators such as E, S, Ra and maximum stress values ​​in real time; trend graphs show historical curves of key indicators over time.

[0104] Remaining life prediction curve: A chart predicting future time, with the X-axis representing time and the Y-axis representing cumulative damage D(t). The chart includes historical damage curves, predicted damage curves, a critical line, and predicted failure points. The historical damage curve is the actual trajectory of D(t) from the past to the present. The predicted damage curve is the predicted growth path of D(t) based on current environmental trends. The critical line is a horizontal line marked Dc. The predicted failure point is the intersection of the predicted curve and the critical line, marked with the expected date.

[0105] Corrosion risk heatmap: Based on the corrosion rate or remaining life data at each sensor location, a color-gradient heatmap is generated at the corresponding location on the structural model, with blue representing low risk and red representing high risk.

[0106] Maintenance Action Recommendation List: An automatically generated to-do list where the system pushes specific, actionable suggestions based on risk identification and diagnosis results.

[0107] S7. Periodically calculate and compare the predicted corrosion thickness reduction with the measured value of the ultrasonic thickness gauge to automatically calibrate the corrosion rate model.

[0108] It should be specifically noted that the steps for automatically calibrating the corrosion rate model are as follows:

[0109] D1. Select the calibration time period, choosing the cycle from the last ultrasonic measurement time t1 to the current ultrasonic measurement time t2; obtain the measured data, the thickness value Tt1 measured ultrasonically at time t1, and the thickness value Tt2 measured ultrasonically at time t2; calculate the actual thickness reduction: ΔD=Tt1-Tt2;

[0110] D2. Obtain the prediction data. Extract the time series data of the instantaneous corrosion rate Ra(τ) predicted by the model from the database during the time period from t1 to t2, and calculate the predicted thickness reduction: This is accomplished by accumulating the sampling intervals (e.g., one data point per minute), where ΔDp≈Σ[Ra(ti)*Δt], and Δt is the sampling time interval.

[0111] D3. Calculate and analyze the absolute and relative errors; Absolute error A = ΔDp - ΔD, relative error B = A / ΔD, where ΔDp is the predicted reduction in corrosion thickness and ΔD is the actual reduction in corrosion thickness; If B > 0, the model prediction is too conservative, and the predicted corrosion amount is more than the actual measurement; If B < 0, the model prediction is too aggressive, and the actual corrosion amount is underestimated; If |B| < 0.1, the model accuracy is high, the error is within 10%, and no adjustment or fine-tuning is required.

[0112] D4. Perform model calibration. The system mainly calibrates the output scale of the environmental reference corrosion rate model Re and the stress corrosion factor S.

[0113] A calibration factor K is introduced to correct the entire corrosion rate model. The new calibration factor is calculated as: Knew = ΔD / ΔDp, where ΔDp is the predicted corrosion thickness reduction and ΔD is the actual corrosion thickness reduction. The new calibration factor is applied to future predictions: Ra′(τ) = K*Ra. Initially, K = 1. After this calibration, it is updated to K = 0.7*Kold + 0.3*Knew, where Kold is the previous K value and Knew is the new K value. A smooth update is used to avoid drastic fluctuations caused by a single error.

[0114] D5. Write the new calibration factor K into the system's configuration database for all subsequent prediction calculations, and record all information for this calibration: calibration time t2, time period Δt=t2-t1, ΔD, ΔDp, A, B, old Kold value, new Knew value, and updated K value.

[0115] It is important to note that if |B| exceeds 25% twice consecutively, it indicates that a sensor has failed or the model has completely deviated. The system should then trigger an alert regarding the system's own health status, notifying engineers to conduct a manual check.

[0116] Through the above description of the embodiments, those skilled in the art can clearly understand that the various embodiments of this application can be implemented by means of software or software combined with necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware functions. Based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The software product is stored in a storage medium and includes several instructions to cause a computer device, such as including but not limited to a personal computer, server, or network device, to execute all or part of the steps of the method described in any embodiment of this application.

[0117] The foregoing has described exemplary embodiments of this application. It should be understood that the above exemplary embodiments are not restrictive but illustrative, and the scope of protection of this application is not limited thereto. It should be understood that those skilled in the art can make modifications and variations to the embodiments of this application without departing from the spirit and scope of this application, and such modifications and variations should be within the scope of protection of this application.

Claims

1. An environment data and AI-based steel structure corrosion prevention evaluation system, characterized in that, Specifically comprising: Data sensing module: for real-time acquisition of environment data and steel structure state data; Environment index generation module: for generating real-time environmental corrosivity index through weighted calculation according to the current environment data of the steel structure; Impact function generation module: for identifying environmental mutations, combining the change intensity and duration of the environmental corrosivity index, and constructing an environmental mutation impact function to calculate the additional corrosion amount; Stress-corrosion coupling effect model construction module: for selecting an algorithm and training to obtain a stress-corrosion coupling effect model according to the real-time state data of the steel structure, the real-time environmental corrosivity index, and the material properties, and combining the impact function to output the actual corrosion rate; The construction steps of the stress-corrosion coupling effect model are specifically: A1, define the input features, and form the input vector X of the model through the model influencing factors, specifically: X=[σ, f, E, M]; wherein σ is the real-time stress value, F is the stress change frequency, E is the real-time environmental corrosivity index, and M is the material property; A2, select an algorithm and perform algorithm training, select gradient boosting tree GBRT to capture complex nonlinear interactions, and the trained model can predict the stress corrosion factor; A3, model application, according to the predicted stress corrosion factor, real-time environmental corrosion rate and impact function, real-time prediction of actual corrosion rate Ra, Ra=Re*Sa+C, wherein Re is the real-time environmental corrosion rate, which is calculated according to the real-time environmental corrosivity index, Sa is the predicted stress corrosion factor, and C is the real-time impact function. If there is no environmental mutation at the current time, C=0; The steps of the algorithm training are specifically: B1, collect a large amount of historical data, including the input features X and the corresponding stress corrosion factor S=R / Rb at each time point; wherein S is the stress corrosion factor, R is the actual measured corrosion rate, and Rb is the corrosion rate caused only by the environment, which is calculated through the current real-time environmental corrosivity index; B2, input the data set (X, S) into the GBRT algorithm for training; B3, the algorithm minimizes the mean square error between the predicted value of the stress corrosion factor and the true value of the stress corrosion factor through continuous iteration; B4, the training is completed, and the model can predict the stress corrosion factor; Steel structure life prediction module: for calculating the current corrosion cumulative damage of the steel structure according to the actual corrosion rate and predicting the life; Graded early warning module: for combining the output results of each model and the sensor data to perform graded early warning and provide visual decision support; System self-learning module: for periodically comparing the predicted corrosion thickness reduction with the measured value of the ultrasonic thickness gauge to automatically calibrate the corrosion rate model.

2. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 1, characterized in that: The environment data of the steel structure specifically includes: Atmospheric temperature and humidity, chloride ion and sulfur dioxide concentration, surface electrolyte acidity, wind speed, wind direction and rainfall; The steel structure state data specifically includes: strain and stress change, instantaneous corrosion rate and residual thickness of the base metal under the coating.

3. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 2, characterized in that: The real-time environmental corrosivity index is specifically: ； Wherein E is the environmental corrosivity index at the ith time point; j is an index variable representing the jth environmental factor; m is the total number of environmental factors, indicating that the summation calculation process will traverse m environmental factors; W j is the weight of the jth environmental factor, which is calculated by the entropy weight method, and the weight of each environmental factor is between 0 and 1, and the sum of the weights of all factors is equal to 1; X ij is the value of the jth environmental factor after standardization at the ith time point.

4. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 3, characterized in that: The environmental mutation impact function is specifically: ； Wherein C is the environmental mutation impact function, indicating the part of the corrosion amount caused by environmental mutation events beyond the normal corrosion amount; a is the calibration parameter of the model, which is derived by analyzing the actual measured corrosion increment after multiple mutation events; ΔE is the environmental corrosion index change intensity, the calculation formula is: ΔE = Es - Eb, Es is the peak value of E during the mutation period, Eb is the baseline value of the environment before the mutation, taking the average value of the previous stage; β is the intensity index, which is determined by data fitting; Δt is the duration of the environmental mutation event, that is, the time length experienced by E from the beginning of the abnormal rise to the normal level; γ is the duration index, which is determined by data fitting.

5. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 1, characterized in that: The cumulative damage is specifically: ； Wherein D(t) is the cumulative damage, from the time 0 when the steel structure starts to be used to the current time t, the total damage amount of the steel structure accumulated due to corrosion; 0 represents the time when the structure starts to be used, t represents the current time; Ra(τ) is the actual corrosion rate, τ is the integral variable, τ takes every time point between 0 and t, and the corrosion rate of each point is included in the calculation.

6. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 5, characterized in that: The steps of life prediction are specifically: C1, set the critical damage value Dc, that is, the maximum corrosion damage that the structure can withstand before failure, which is determined by design specifications, laboratory experiments or historical data; C2, based on the current state D(t), input the future environmental prediction scenario and the future stress prediction scenario; C3, use the trained model to simulate and calculate the actual corrosion rate at each future time, and calculate the cumulative future damage Df; C4, when Df≥Dc, the corresponding future time point is the predicted failure time T, and the remaining life = failure time - current time.

7. The steel structure corrosion prevention evaluation system based on environmental data and AI according to claim 6, characterized in that: The graded early warning is specifically: The first level warning is long-term trend warning; the triggering conditions specifically include: environmental trend: EC7>E30*1.3, that is, the average environmental corrosion index in the past 7 days is more than 1.3 times of the average index in the past 30 days; damage progress: D(t) / Dc>0.4, that is, the cumulative damage has reached 40% of the critical damage value; The second level warning is short-term time warning; the triggering conditions specifically include: environmental mutation: ΔE>E30*0.8 and Δt>2 hours, the intensity of a single environmental mutation event is more than 0.8 times of the average ECI in the past 30 days, and the duration of the event is more than 2 hours; stress surge: S(t) / SCF24h>1.8 and S(t)>2.0, that is, the current stress corrosion factor has increased by 80% compared with the average value in the past 24 hours, and the absolute value is greater than 2.0; life warning: T The third level warning is immediate emergency warning, the triggering conditions specifically include: damage criticality D(t) / Dc>0.9, that is, the cumulative damage has exceeded 90% of the critical value; crack initiation: the acoustic emission sensor monitors more than 100 times / second of high frequency acoustic emission signals and more than 80 decibels of high intensity signals; structural response: the real-time monitored strain value exceeds 80% of the material yield strain, indicating that the structure has entered the nonlinear stage.

Images