Bridge structure corrosion fatigue coupling test method and system based on artificial intelligence

Abstract

The invention relates to the technical field of bridge engineering material tests, and discloses a bridge structure corrosion fatigue coupling test method and system based on artificial intelligence. The bridge structure corrosion fatigue coupling test method based on artificial intelligence comprises the following steps: accurately expressing interaction among physical fields by constructing a multi-physical field coupling mathematical model; a deep reinforcement learning algorithm is applied, an environment parameter cooperative control model is built, and multi-environment-parameter high-precision self-adaptive regulation and control are achieved; building an environment acceleration test equivalence mapping model, and balancing the optimal acceleration ratio and equivalence of the acceleration test; constructing a generalized predictive control model to achieve multi-physical field cooperative regulation and control; a real-time system identification and parameter updating model is constructed, multi-physical field interaction influence is compensated, and stable control is achieved; according to the method, the test environment control precision is improved, multi-environment-condition cooperative control is realized, the test period is shortened, the test authenticity and the result stability are improved, and reliable data support is provided for bridge structure safety assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge engineering material tests, and particularly to an artificial intelligence-based bridge structure corrosion fatigue coupling test method and system. Background Art

[0002] During the service process of a bridge structure, it is under the combined action of a corrosion environment and fatigue loads. The corrosion fatigue coupling effect is an important factor affecting the safety of the bridge structure. Accurately understanding the evolution law of corrosion fatigue coupling damage is crucial for bridge structure life prediction and safety assessment, and conducting corrosion fatigue coupling tests is the main way to obtain relevant data.

[0003] However, there are several key technical problems in the existing corrosion fatigue coupling test methods: during the service process of a bridge structure, it is affected by multiple physical fields such as a mechanical field, an electrochemical field, and a temperature field. These physical fields interact with and influence each other, forming a complex coupling effect. It is difficult for traditional test methods to accurately simulate this complex interaction relationship; during the actual service process, the development of corrosion and fatigue damage often takes several years or even decades, and it is difficult for traditional test methods to complete effective accelerated tests within a reasonable period; traditional test methods have low control accuracy and poor stability, and it is difficult to reproduce the complex and changeable actual service environment conditions; traditional test methods usually control environmental conditions and load conditions separately, and cannot reflect the dynamic coupling relationship between the two, resulting in insufficient test authenticity.

[0004] With the rapid development of artificial intelligence technology, applying advanced artificial intelligence algorithms and control technologies to corrosion fatigue coupling tests is expected to break through the above technical bottlenecks and improve the accuracy, efficiency, and reliability of tests. However, there is currently a lack of a systematic and effective solution to integrate key technologies such as multi-physical field modeling, coordinated control of environmental parameters, and optimization of accelerated tests to achieve the overall optimization of bridge structure corrosion fatigue coupling tests. Summary of the Invention

[0005] The present invention discloses an artificial intelligence-based bridge structure corrosion fatigue coupling test method and system, aiming to solve the problems of multi-physical field interaction, too long test period, insufficient environmental control accuracy, and separation of environmental conditions and load conditions existing in the prior art.

[0006] The present invention provides an artificial intelligence-based bridge structure corrosion fatigue coupling test method, including the following steps:

[0007] Construct a multi-physical field coupling mathematical model including a mechanical field, an electrochemical field, and a temperature field to accurately express the interaction between physical fields;

[0008] Apply the deep reinforcement learning algorithm to construct an environmental parameter collaborative control model, realizing high-precision control of multiple environmental parameters and adaptive regulation based on material response;

[0009] Construct an environmental accelerated test equivalence mapping model, optimize the accelerated test parameter combination through a multi-objective optimization algorithm, and achieve the balance between the optimal acceleration ratio and equivalence;

[0010] Construct a generalized predictive control model, and achieve collaborative regulation of multiple physical fields through a rolling optimization strategy, improving control precision and response speed;

[0011] Construct a real-time system identification and parameter update model, realize online identification of system parameters based on the recursive least squares algorithm, and actively compensate and stably control the interactive effects of multiple physical fields.

[0012] Furthermore, the multi-physical field coupling mathematical model is established through a partial differential equation system, which includes state variables of each physical field, time derivatives of state variables, spatial gradients of state variables, interaction functions between physical fields, and influence functions of control inputs on physical fields, and is used to characterize the coupling relationship between different physical fields.

[0013] Furthermore, the deep reinforcement learning environmental parameter collaborative control model uses a deep Q-network to realize policy learning in complex environments, and realizes the learning of the optimal control policy by iteratively updating the Q-value function, where the Q-value function represents the expected cumulative reward value of executing a specific action in a specific state.

[0014] Furthermore, the reward function of the deep reinforcement learning environmental parameter collaborative control model is designed as a weighted combination of environmental parameter control precision and material response compliance, where the environmental parameter control precision is calculated by the deviation between the actual value and the target value, and the material response compliance is quantified by a special evaluation function.

[0015] Furthermore, the environmental accelerated test equivalence mapping model realizes the precise setting of accelerated test conditions by establishing the mapping relationship between accelerated test environmental parameters and actual service environmental parameters, and introducing the time compression ratio and time-varying acceleration factor.

[0016] Furthermore, the optimization of the accelerated test parameter combination uses a multi-objective optimization algorithm, and finds the optimal accelerated test parameter combination by balancing two objectives of the time compression ratio and the equivalence deviation, where the equivalence deviation evaluates the difference degree between the accelerated test environment and the actual service environment through a special metric function.

[0017] Furthermore, the generalized predictive control model realizes the collaborative regulation of multiple physical fields by constructing a weighted cost function including the output tracking error and the change amount of the control input, and performing rolling optimization within the prediction time domain and the control time domain.

[0018] Furthermore, the real-time system identification adopts the recursive least squares algorithm to realize the online identification of the system model by continuously updating the parameter estimation values. In the parameter update process, the Kalman gain is used to weight and correct the error between the system output and the predicted output.

[0019] Furthermore, the active compensation method for the interactive influence of multiple physical fields is implemented by a feedforward-feedback composite structure. An interactive influence decoupling matrix is constructed based on the multi-physical field coupling model obtained from real-time system identification, and the compensation control input is calculated according to the change amount of the target physical field state, so as to effectively compensate for the interactive influence of the physical fields.

[0020] Furthermore, an artificial intelligence-based bridge structure corrosion fatigue coupling test system is used to execute the above-mentioned artificial intelligence-based bridge structure corrosion fatigue coupling test method, and includes:

[0021] A multi-physical field coupling model module for constructing a coupling mathematical model including a mechanical field, an electrochemical field, and a temperature field;

[0022] A deep reinforcement learning control module for realizing the collaborative control of environmental parameters and the adaptive regulation based on material response;

[0023] An accelerated test mapping module for establishing an equivalence mathematical model between environmental accelerated test parameters and the actual service environment;

[0024] A generalized predictive control module for realizing the collaborative regulation of multiple physical fields and the rolling optimization control;

[0025] A real-time system identification module for continuously updating the coupling model parameters and actively compensating for the interactive influence of multiple physical fields.

[0026] The beneficial effects of the present invention are as follows:

[0027] By virtue of the deep reinforcement learning environmental parameter collaborative control strategy and the generalized predictive control algorithm, the present invention significantly improves the performance of the test system. In terms of control accuracy, the environmental parameter fluctuation control of the present invention exceeds the traditional method, improving the data reliability; by constructing a multi-physical field coupling mathematical model, any combination and smooth transition of various environmental conditions are realized, approaching the actual service environment of the bridge; relying on the environmental accelerated test equivalence mapping model, while ensuring the reliability of the data, the test cycle is shortened and the test cost is reduced; using the multi-physical field collaborative control and real-time system identification, the consistency between the test and the actual bridge service conditions is improved; through real-time system identification and parameter update, the interactive influence of multiple physical fields is actively compensated, the fluctuation of multi-round test data is reduced, and the comparability of results between different laboratories is improved, providing more reliable data support for the safety assessment of bridge structures. Description of the Drawings

[0028] Figure 1 This is a flowchart of a method for coupling corrosion fatigue tests of bridge structures based on artificial intelligence according to the present invention. Specific implementation mode

[0029] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0030] In at least one embodiment of the present invention, a method for coupling corrosion fatigue tests of bridge structures based on artificial intelligence is disclosed. As Figure 1 shown, it includes the following steps:

[0031] Step 1: Construct a multi - physical - field coupling mathematical model including a mechanical field, an electrochemical field, and a temperature field to accurately express the interaction between physical fields;

[0032] Specifically, it includes:

[0033] Step 1.1: Collect response data of the bridge structure under the action of different physical fields;

[0034] Collect response data of the bridge structure under the individual and combined actions of a mechanical field (such as dynamic loads with different amplitudes and frequencies), an electrochemical field (such as corrosion media with different concentrations), and a temperature field (such as different temperature conditions), including information such as stress - strain relationships, corrosion current densities, and material temperature distributions.

[0035] Step 1.2: Construct a partial differential equation system for physical - field coupling;

[0036] Based on the collected data, establish a partial differential equation system describing the multi - physical - field coupling relationship. Its calculation formula is:

[0037] ;

[0038] Among them, , represent the state variables of the , th physical field, , represent the spatial gradients of the state variables of the , th physical field, is the total number of physical fields; represents the time variable; Represents the interaction function between physical fields; Represents the influence function of control inputs on physical fields; 、 Represents the 、 th control input variable, is the total number of control input variables.

[0039] Step 1.3, discretize the coupled equations by the finite element method;

[0040] Spatially discretize the coupled partial differential equations established in Step 1.2 by the finite element method, transform the continuous domain problem into a numerical calculation problem on discrete nodes, and obtain the discretized equations:

[0041] ;

[0042] Where, Represents the vector of physical field state variables at discrete nodes; and represent its first and second time derivatives respectively; 、 and represent the mass matrix, damping matrix and stiffness matrix respectively; Represents the resultant force vector of external loads and internal coupling effects, which is a function of the control input vector and the state variable vector .

[0043] Step 1.4, verify the accuracy of the multi-physical field coupling model;

[0044] Evaluate the accuracy and reliability of the model by comparing the model prediction results with the measured data, and adjust the model parameters if necessary to ensure that the model can accurately describe the interaction relationship between multi-physical fields.

[0045] Step 2, apply the deep reinforcement learning algorithm to construct an environmental parameter collaborative control model to achieve high-precision control of multi-environmental parameters and adaptive regulation based on material response;

[0046] Specifically include:

[0047] Step 2.1, construct the test environment state space and action space;

[0048] Define the test environment state space , including environmental parameters such as temperature, humidity, salinity, pH value, and state variables such as the stress state and corrosion potential of the test sample; define the action space , including the adjustment parameters of each environmental control device, such as heating power, humidity control value, salt mist concentration, etc.

[0049] Step 2.2, construct a Q-learning model based on deep reinforcement learning;

[0050] Use a deep Q-network (DQN) to implement policy learning in complex environments. The core is the iterative update of the Q-value function, and its calculation formula is:

[0051] ;

[0052] where, represents the expected cumulative reward value of executing action in state ; represents the immediate reward; represents the discount factor, which is used to balance the immediate reward and the long-term benefit; represents the next state after executing action ; represents the maximum Q-value of the next state.

[0053] Step 2.3, construct a deep neural network to approximate the Q-value function;

[0054] To handle the high-dimensional continuous state space, use a deep neural network to approximate the Q-value function. The network structure includes an input layer (state variables), multiple hidden layers (using the ReLU activation function), and an output layer (Q-values of each action). By randomly sampling batch empirical data , minimize the loss function for training:

[0055] ;

[0056] where, is the loss function, represents the current network parameters; represents the target network parameters, which are periodically copied and updated from the current network; represents the expected operation.

[0057] In this embodiment, the specific structure of the constructed deep neural network is as follows: The input layer contains 24 neurons, corresponding to the recorded values of 4 types of environmental parameters such as temperature, humidity, salinity, and pH value at the current moment and the past 5 moments; the hidden layer contains 3 layers, with each layer containing 128, 64, and 32 neurons respectively, and all use the ReLU activation function to enhance the non-linear expression ability of the network; the output layer contains 15 neurons, corresponding to different regulation actions of environmental parameters such as temperature, humidity, salinity, and pH value. During training, this network uses an experience replay buffer with a storage capacity of 10,000 transition samples. Each time, 512 samples are randomly sampled for mini-batch training, and the target network parameters are updated every 200 steps during the training process. This neural network has achieved high-precision environmental regulation with a temperature control accuracy of ±0.5°C, a relative humidity control accuracy of ±0.8%, and a salt concentration control accuracy of ±0.05% in the corrosion fatigue test environment control of bridge reinforced concrete specimens under a multi-salt fog - dynamic load composite environment.

[0058] Step 2.4, implement an adaptive regulation model based on material response;

[0059] Use the material response signals of the test samples (such as strain growth rate, corrosion current change rate, etc.) as feedback inputs, and dynamically adjust the environmental parameter control strategy through the reinforcement learning model to achieve adaptive regulation of environmental parameters based on the real-time response of materials, and improve the precise control ability during the test.

[0060] In this embodiment, the adaptive regulation model adopts a dual-input strategy. In addition to the environmental parameter state, it also introduces material response characteristics as inputs to the model. Specifically, a deep network structure with dual-channel inputs is constructed: The first channel receives environmental parameter information, and the second channel receives material response signals (including strain change rate, corrosion potential change rate, corrosion current density, etc.). After passing through their respective feature extraction layers, the two channels are merged in the fusion layer, and finally, the regulation actions of environmental parameters are output. The design of the reward function of this model includes two parts: environmental parameter control accuracy reward and material response compliance reward, and its calculation formula is:

[0061] ;

[0062] Among them, represents the reward value obtained by executing action in state ; and respectively represent the actual value and the target value of the i-th environmental parameter; represents the total number of environmental parameters; represents the material response compliance evaluation function; and Weight coefficients for environmental control accuracy and material response compliance, respectively.

[0063] In practical applications, this adaptive regulation model has been successfully applied to the corrosion fatigue coupling test of reinforced concrete components of the Hangzhou Bay Sea-Crossing Bridge. Based on the detected sudden changes in the corrosion current density of the steel bars, the model can automatically adjust the salt spray concentration and relative humidity to keep the corrosion process consistent with the corrosion rate under the actual environment. At the same time, according to the changes in the fatigue crack growth rate of the material, the model can intelligently adjust the load frequency and amplitude, ensuring that the consistency between the test process and the actual service conditions is increased by 31%, far superior to the traditional fixed-parameter test method.

[0064] Step 3: Construct an equivalence mapping model for environmental accelerated tests, and optimize the accelerated test parameter combinations through a multi-objective optimization algorithm to achieve the balance between the optimal acceleration ratio and equivalence;

[0065] Specifically, it includes:

[0066] Step 3.1: Collect actual bridge service environment data and accelerated test data;

[0067] Collect long-term monitoring data of the actual bridge under different service environments, including the variation laws of environmental parameters and structural response data; at the same time, collect short-term test data under different accelerated test conditions to provide a data basis for establishing the mapping relationship between the accelerated test and the actual service environment.

[0068] Step 3.2: Establish a mathematical model for the equivalence of environmental accelerated tests;

[0069] Based on the collected data, establish a mathematical model for the equivalence between the parameters of environmental accelerated tests and the actual service environment. Its calculation formula is:

[0070] ;

[0071] Where, represents the vector of environmental parameters of the accelerated test; represents the vector of actual service environment parameters; represents the time compression ratio; represents the environmental parameter mapping function; represents the time-varying acceleration factor, which is used to dynamically adjust the accelerated test parameters to maintain equivalence with the actual service environment.

[0072] Step 3.3: Optimize the accelerated test parameter combinations;

[0073] Based on the equivalence mathematical model established in Step 3.2, use a multi-objective optimization algorithm to solve the optimal accelerated test parameter combinations. The objective functions include maximizing the time compression ratio and minimizing the equivalence deviation. Its calculation formula is:

[0074] ;

[0075] Among them, and respectively represent the time compression ratio and the weight coefficient of the equivalence deviation; represents the time compression ratio; represents the actual service environment parameter vector; represents the equivalence deviation measurement function between the accelerated test environment and the actual service environment; represents the norm operation.

[0076] In this embodiment, the multi-objective optimization problem is solved by an improved non-dominated sorting genetic algorithm (NSGA-II). The specific implementation of this algorithm is as follows: First, define the decision variable as the accelerated test environment parameter vector , including parameters such as the temperature cycle amplitude, humidity change range, salt spray concentration, pH value range, and loading frequency in the accelerated test; Second, construct a population (the population size is 100) and perform encoding; Then, define the fitness function, including two objectives:

[0077] Maximize the time compression ratio , and the value range is [5, 20];

[0078] Minimize the equivalence deviation , which is calculated by comparing key indicators such as the corrosion rate and fatigue crack growth rate in the accelerated test and the actual service environment.

[0079] After the algorithm runs for 500 generations, screen the non-dominated solution set, and select the optimal compromise solution through the fuzzy set theory method.

[0080] This optimization algorithm has been successfully applied in the corrosion fatigue coupling test planning of the steel box girder of the Wuhan Yangtze River Highway Bridge. Through the global optimization of 16 environmental parameters, on the premise of ensuring that the equivalence deviation is less than 8%, the original 3-year corrosion fatigue test cycle is shortened to 9 months, and the time compression ratio reaches 4 times. The optimized accelerated test parameters include expanding the temperature fluctuation of ±15°C in the service environment to ±35°C, increasing the original stress amplitude of 10 MPa to 25 MPa, and increasing the salt spray concentration from 3.5% to 6.8%. The test results show that the damage morphology and failure mechanism of the accelerated test samples are highly consistent with the long-term on-site monitoring results, verifying the effectiveness and reliability of this optimization algorithm.

[0081] Step 3.4, verify the equivalence of the accelerated test;

[0082] By comparing the accelerated test results with the actual service data, verify the equivalence and reliability of the accelerated test, and adjust the equivalence model parameters if necessary to ensure that the accelerated test can accurately reflect the corrosion fatigue coupling effect under actual service conditions.

[0083] Step 4: Construct a generalized predictive control model to achieve coordinated regulation of multiple physical fields through a rolling optimization strategy, improving control accuracy and response speed;

[0084] Specifically, it includes:

[0085] Step 4.1: Establish a system prediction model;

[0086] Based on the multi-physical field coupling mathematical model constructed in Step 1, establish a system prediction model for control to predict the response of the system over multiple future time steps. Its calculation formula is:

[0087] ;

[0088] where, represents the prediction horizon length, represents the system output prediction vector for the next future time steps starting from the current time , , respectively represent the predicted values of the system output at the current time for the future times , , ; represents the transpose operation.

[0089] Step 4.2: Construct a generalized predictive control cost function;

[0090] Design a cost function that includes the output tracking error and the change in the control input. Its calculation formula is:

[0091] ;

[0092] where, represents the cost function of generalized predictive control, represents the prediction horizon length; represents the control horizon length ( ); represents the reference trajectory; represents the change in the control input; and respectively represent the weighting matrices of the output tracking error and the change in the control input; represents the time step index.

[0093] Step 4.3: Solve the optimal control sequence;

[0094] Based on the cost function constructed in Step 4.2, use the quadratic programming method to solve the optimal control sequence. Its calculation formula is:

[0095] ;

[0096] Among them, represents the optimal control sequence; represents the optimal control input value that should be immediately applied to the system calculated at the current moment ; , respectively represent the optimal control inputs for future moments at the current moment , .

[0097] Step 4.4, construct a receding horizon optimization control model;

[0098] Adopt a receding horizon optimization strategy, and only execute the first control input of the optimal control sequence in each control period , and then update the state according to the actual response of the system, re-solve the optimal control sequence, and implement closed-loop feedback control to improve the robustness of the system to disturbances.

[0099] In this embodiment, the receding horizon optimization control model is implemented in the following specific steps:

[0100] At the current moment , based on the current state of the system and the multi-physics field coupling mathematical model, predict the system output at the next moments;

[0101] Solve the optimization problem to obtain the optimal control sequence at the next moments;

[0102] Only execute the first control input ;

[0103] Measure the actual state of the system at the next moment ;

[0104] Move the time pointer forward one step to , and repeat the above process.

[0105] The key parameters of this model are set as follows: the prediction horizon , the control horizon , the control period is 200 ms, and the quadratic programming solver qpOASES in the MATLAB environment is used to solve the optimization problem, which can complete an optimization calculation within 50 ms to meet the real-time control requirements.

[0106] The rolling optimization control model has been successfully applied to the corrosion fatigue coupling test system of stay cable anchorage area of Sutong Bridge, realizing the coordinated control of multiple parameters such as temperature, humidity, salt concentration, and mechanical load. In the face of sudden disturbances (such as a sudden temperature change of ±5°C), the control model can suppress the influence of the disturbance within 3 control cycles (600 ms) and quickly restore the system to the target state, while the traditional PID control requires 12 control cycles to achieve the same effect. In practical applications, the control model improves the stability of the corrosion fatigue test environment by 78% and the smoothness of the environmental conditions during the key transition period by 65%, effectively avoiding the overshoot and oscillation problems common in traditional control methods, providing a reliable guarantee for accurately simulating the corrosion fatigue coupling behavior of bridge structures under actual service conditions.

[0107] Step 5: Construct a real-time system identification and parameter update model, and realize the online identification of system parameters based on the recursive least squares algorithm to actively compensate and stably control the interactive influence of multiple physical fields.

[0108] Specifically, it includes:

[0109] Step 5.1: Construct a parameterized system model.

[0110] To realize the online identification of system parameters, a parameterized system model is constructed, and its calculation formula is:

[0111] ;

[0112] Where, represents the system output; represents the regression vector, including past input and output data; represents the parameter vector to be identified; represents the model error.

[0113] Step 5.2: Design a recursive least squares algorithm for parameter identification.

[0114] The recursive least squares (RLS) algorithm is used to realize the online identification and parameter update of system parameters, and its calculation formula is:

[0115] ;

[0116] Where, , respectively represent the parameter estimation values at times , ;

[0117] represents the Kalman gain, and its calculation formula is:

[0118] ;

[0119] Among them, represents the moment parameter estimation covariance matrix of.

[0120] Covariance matrix update of, the calculation formula is:

[0121] ;

[0122] In this embodiment, the specific implementation of the recursive least squares algorithm includes the following steps: First, initialize the parameter estimation value to a zero vector, initialize the covariance matrix , where is a relatively large positive constant (such as 1000), is an identity matrix; then within each control period (typical value is 100ms), based on the newly acquired system input-output data, construct the regression vector , including the input-output data of the past 4 control periods; then calculate the parameter update amount according to equations (11)-(13). To improve the robustness of the algorithm, this embodiment also introduces a forgetting factor (value range 0.95 - 0.99), and modify the covariance matrix update formula to:

[0123] ;

[0124] In practical applications, this algorithm has been successfully used for multi-physical field interaction parameter identification of the main cable corrosion fatigue coupling test system of large suspension bridges. The algorithm can converge the model parameters to within ±3% of the true values within 30 iterations, and can quickly track the changes in the dynamic characteristics of the system under sudden changes in environmental conditions (such as temperature, humidity), and complete the adaptive update of the model parameters within 200ms. This real-time identification ability enables the system to accurately capture the non-linear interaction relationships between different physical fields, providing an accurate model basis for subsequent active compensation control.

[0125] Step 5.3, construct an adaptive control method based on model parameters;

[0126] Substitute the system parameters identified in step 5.2 into the generalized predictive control model in step 4, dynamically update the control model, and achieve adaptive control based on model parameters to improve the adaptability of the system to parameter changes.

[0127] Step 5.4, construct an active compensation method for multi-physical field interaction effects;

[0128] Based on the identified system parameters, analyze the interaction laws between multiple physical fields, construct an active compensation method to offset the system instability factors caused by the interaction, and improve the stability of the test conditions and the control accuracy.

[0129] In this embodiment, the active compensation method for the interaction of multiple physical fields is implemented by using a feedforward-feedback composite structure. First, based on the multi-physical-field coupling model obtained by real-time system identification, construct an interaction decoupling matrix , where is the control input dimension, is the physical field state dimension. The decoupling matrix is generated by the following algorithm:

[0130] Based on the identified system model, calculate the interaction sensitivity matrix of each physical field, and the element represents the influence degree of the th physical field on the th physical field;

[0131] Based on the sensitivity matrix, construct the decoupling matrix , where represents the pseudo-inverse matrix of the sensitivity matrix ;

[0132] Use the singular value decomposition (SVD) method to handle the condition number problem of the decoupling matrix to ensure the stability of numerical solution.

[0133] The implementation process of the compensation control is as follows:

[0134] Through the decoupling matrix , map the change amount of the target physical field state to the compensation control input :

[0135] ;

[0136] Superimpose the compensation control input and the basic control input to form the final control command;

[0137] Based on the actual response, dynamically adjust the decoupling matrix to adapt to the changes in the system characteristics.

[0138] This active compensation method has been successfully applied in the corrosion fatigue coupling test of the steel structure of the Nanjing Yangtze River Bridge. For example, during the test, when the mechanical load change caused the temperature of the specimen to rise by 3.2 °C, it took 20 minutes for the traditional method to stabilize the temperature. However, with the active compensation control of this method, the system can predict and compensate for this interaction in advance, control the temperature fluctuation within the range of ±0.6 °C, and shorten the stabilization time to 3 minutes. Similarly, for the interaction between the electrochemical field and the mechanical field, this method reduces the corrosion potential fluctuation from ±25 mV in traditional control to ±5 mV, significantly improving the accuracy and reliability of the test data. The application of this method reduces the inter-field interaction interference in the multi-physical field coupling test by 87%, providing technical support for accurately reproducing the complex physical field coupling effect under the actual service environment.

[0139] This embodiment proposes a method for corrosion fatigue coupling test of bridge structures based on artificial intelligence. By means of key technologies such as multi-physical field coupling modeling and deep reinforcement learning environment control, it solves the problems of existing test methods and achieves remarkable technical effects. In terms of the control accuracy of the test environment, by using the deep reinforcement learning environment parameter collaborative control strategy and the generalized predictive control algorithm, the environmental parameter fluctuation is controlled within ±0.8%, far better than ±5% of the traditional method, improving the reliability of test data; in terms of the collaborative control of multiple environmental conditions, by constructing a multi-physical field coupling mathematical model and the generalized predictive control algorithm, any combination and smooth transition of more than 10 environmental conditions such as temperature and humidity are realized, approaching the actual service environment of the bridge; relying on the environmental accelerated test equivalence mapping model, while ensuring the reliability of data, the test cycle is shortened by 65% and the test cost is reduced; through multi-physical field collaborative control and real-time system identification, the intelligent matching of environmental and load conditions is realized, improving the consistency between the test and the actual bridge service conditions by 90%; through real-time system identification and parameter update, the active compensation of multi-physical field interaction is carried out, reducing the multi-round test data fluctuation by 75% and improving the comparability of results between different laboratories by 85%, providing reliable data for the safety assessment of bridge structures and overcoming the defects of traditional test methods in terms of accuracy, authenticity, efficiency and stability. 4. Feedforward feedback Real application examples of this embodiment

[0140] In an embodiment of the present invention, an example of the aforementioned method for corrosion fatigue coupling test of bridge structures based on artificial intelligence is provided:

[0141] Application scenario description: The method of this embodiment is applied to the durability research project of steel bridges on heavy-haul railways in coastal areas. The steel bridge is located in the southeast coastal area of China, and the service environment is characterized by high humidity, high salt mist concentration, and periodic temperature changes. At the same time, it bears high-frequency fatigue loads caused by heavy-haul trains. The research goal is to evaluate the corrosion-fatigue coupling damage evolution law of the key structures of the steel bridge and its impact on the remaining service life of the bridge, providing a scientific basis for bridge maintenance decisions. Traditional research methods require 8 to 10 years of on-site tests, or use simplified accelerated tests with poor equivalence, and cannot accurately predict the structural performance evolution under actual service conditions. To solve the above problems, this project adopts the corrosion-fatigue coupling test method based on artificial intelligence proposed in this embodiment to establish an equivalent accelerated test method to obtain valuable research data within a reasonable period. The test objects are 25 standard specimens intercepted from the steel bridge, including key structural forms such as welded joints, bolt connections, and main girder sections. The test device includes an intelligent environment control system (temperature, humidity, salt mist concentration, pH value, etc.) and a multi-channel electro-hydraulic servo loading system, which is designed and controlled entirely based on the method of this embodiment.

[0142] In this application example, based on one-year on-site monitoring data, a three-field coupling mathematical model including mechanical mechanics field, electrochemical corrosion field, and temperature field is established. For the special environment of coastal heavy-haul railway steel bridges, the key variables and their interaction relationships included in the model are shown in Table 1:

[0143] Table 1: Key variables and interaction relationships of multi-physical fields in the corrosion-fatigue coupling test of feedforward-feedback steel bridges:

[0144]

[0145] Based on the actual monitoring data and material test results, the coupling function of the mechanics field and electrochemical field is established as follows:

[0146] ;

[0147] Where, and respectively represent the corrosion current density with and without strain action; and respectively represent the current strain and the initial strain; and respectively represent the current temperature and the reference temperature; and respectively represent the strain sensitivity coefficient and the temperature sensitivity coefficient, which are determined through tests , .

[0148] For the particularity of this application scenario, a stress corrosion crack growth rate correction function is introduced on the basis of the traditional model:

[0149] ;

[0150] wherein, represents the crack growth rate; represents the effective stress intensity factor amplitude; and are material constants; is the corrosion temperature correction function, and the expression is:

[0151] ;

[0152] wherein, and are the reference corrosion current density and the reference temperature respectively; , and are fitting parameters, which are determined by comparative experiments , .

[0153] The prediction accuracy of this model in practical applications is shown in Table 2:

[0154] Table 2: Evaluation of the prediction accuracy of the feedforward-feedback multi-physical field coupling model:

[0155]

[0156] In this application example, based on the deep reinforcement learning environment control model constructed in Step 2, high-precision coordinated control of the environmental parameters of the steel bridge corrosion fatigue coupling test is achieved. Aiming at the special service environment of coastal steel bridges, the state space of environmental parameters is expanded, and targeted optimization is carried out on the basis of the basic network architecture.

[0157] Aiming at the corrosion fatigue characteristics of steel bridges, a dual reward function is designed, including environmental control accuracy reward and material response matching reward:

[0158] ;

[0159] wherein, and represent the actual value and the target value of the i-th environmental parameter respectively; represents the matching degree function of the material corrosion current and the crack growth rate with the on-site data; and are the weight coefficients of control accuracy and material response matching respectively.

[0160] After 200 hours of training on the actual test system, the deep reinforcement learning model achieved high-precision collaborative control of environmental parameters. In the face of rapidly changing environmental scenarios, the environmental control effects of the traditional PID control method and this method are compared in Table 3 as follows:

[0161] Table 3: Comparison of the effects of feedforward-feedback traditional PID and deep reinforcement learning control methods:

[0162]

[0163] In practical applications, the system can accurately simulate the diurnal and seasonal changes in the coastal environment. The material response evaluation shows that the consistency between the obtained corrosion current density changes and the data monitored in the actual service environment reaches 93.2%, and the consistency of the crack propagation rate reaches 91.5%, which is much higher than the traditional control method.

[0164] In this application example, based on the environmental accelerated test equivalence mapping model constructed in Step 3, the time compression of the corrosion-fatigue coupling damage process of the steel bridge is realized, and the test cycle is greatly shortened on the premise of ensuring equivalence.

[0165] First, based on the 3-year historical monitoring data collected on-site, the typical change patterns of actual service environment parameters and the development laws of corrosion-fatigue damage are established. Then, the improved non-dominated sorting genetic algorithm (NSGA-II) is used for multi-objective optimization to solve the optimal accelerated test parameter combination. For the steel bridge application scenario, the following key parameters are used in the optimization process, as shown in Table 4:

[0166] Table 4: Key parameters for optimizing the equivalence of feedforward-feedback accelerated tests:

[0167]

[0168] Based on the multi-physical field coupling model and a large amount of test data, an equivalence evaluation standard between the accelerated test and the actual service environment is established, including indicators such as damage mechanism consistency, corrosion morphology similarity, and fatigue crack propagation behavior matching. The genetic algorithm uses a population size of 150 and runs for 800 generations. The finally determined optimal accelerated parameter combination is shown in Table 5:

[0169] Table 5: Accelerated test parameter combination after feedforward-feedback optimization and its time compression effect:

[0170]

[0171] The comprehensive optimization results show that, on the premise of maintaining 90.5% equivalence, a time compression ratio of 5.3 times is achieved. That is, the corrosion-fatigue coupling damage process that originally took 10 years can be completed in only 23 months in this application example, greatly improving the research efficiency. The comparative analysis of the damage morphology shows that the consistency of the corrosion-fatigue crack morphology between the accelerated test and the steel structures in on-site service reaches 92.7%, and the prediction accuracy of the fatigue life reaches a high-reliability level of ±8.5%.

[0172] In this application example, based on the generalized predictive control model and the real-time system identification model constructed in Steps 4 and 5, a highly coordinated regulation of environmental conditions and load conditions is achieved, effectively solving the problem of insufficient test authenticity caused by the separate control of environment and load in traditional test systems.

[0173] The hardware configuration of the coordinated regulation system is as follows: Intel feedforward-feedback Core feedforward-feedback i7 processor, 16GB of memory, NVIDIA feedforward-feedback GeForce feedforward-feedback RTX feedforward-feedback 3080 feedforward-feedback GPU acceleration, and the control cycle is set to 50ms. For the test scenario of coastal steel bridges, the following key parameters are configured, as shown in Table 6:

[0174] Table 6: Key parameter configuration of the feedforward-feedback multi-physical field coordinated regulation system:

[0175]

[0176] The application effect of this system in the corrosion-fatigue coupling test of steel bridges is manifested in the following aspects:

[0177] A highly coordinated change of the temperature field, electrochemical field, and mechanical field is achieved, accurately reproducing the complex coupling relationship of the three physical fields in the actual service environment. For example, when the temperature rises, the salt spray concentration is automatically increased to maintain the consistency of the corrosion current density, and when the load increases, the temperature and humidity parameters are actively adjusted to compensate for the environmental disturbance caused by the load.

[0178] By continuously updating the model parameters through system identification, the nonlinear changes and uncertainties during the test process are effectively addressed.

[0179] Technical effect verification: This application example fully verifies the technical effects of Embodiment 1, and particularly achieves remarkable results in the following two aspects:

[0180] The improvement effect of the consistency between the feedforward-feedback test and actual service: Compared with the traditional corrosion-fatigue test method, this embodiment significantly improves the consistency between the test and actual service conditions, providing more reliable data support for the remaining life assessment of steel bridge structures. The detailed comparison results are shown in Table 7:

[0181] Table 7: Comparison of the consistency between the feedforward-feedback test and actual service:

[0182]

[0183] Feedforward-feedback test efficiency and cost optimization effect: In the first implementation method, while ensuring the test quality, the test efficiency was significantly improved and the research cost was reduced. The efficiency and cost of three different-scale corrosion fatigue coupling tests are compared as shown in Table 8:

[0184] Table 8: Comparison of efficiency and cost under different test scales of feedforward-feedback:

[0185]

[0186] It can be seen from the above technical effect verification data that in the corrosion fatigue coupling test of coastal heavy-haul railway steel bridges applied in this implementation method, compared with the traditional method, the consistency between the test and the actual service conditions has increased by 39.6%, while the test cost has decreased by 64.7%, fully verifying the technical effect of the present invention. This technical achievement has been successfully applied in 3 steel structure bridge life assessment projects of different scales.

[0187] The embodiments of the present invention have been described above, but these embodiments are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of this embodiment.

Claims

1. An artificial intelligence-based corrosion-fatigue coupling test method for bridge structures, characterized in that Including the following steps: Construct a multi-physics field coupling mathematical model including a mechanical field, an electrochemical field, and a temperature field to accurately express the interaction between physical fields; Apply a deep reinforcement learning algorithm to construct an environmental parameter collaborative control model to achieve high-precision control of multiple environmental parameters and adaptive regulation based on material responses; Construct an environmental accelerated test equivalence mapping model, and optimize the accelerated test parameter combination through a multi-objective optimization algorithm to achieve the balance between the optimal acceleration ratio and equivalence; Construct a generalized predictive control model, and achieve collaborative regulation of multiple physical fields through a rolling optimization strategy to improve control accuracy and response speed; Construct a real-time system identification and parameter update model, and achieve online identification of system parameters based on the recursive least squares algorithm to actively compensate for and stably control the interactive effects of multiple physical fields.

2. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 1, wherein The multi-physics field coupling mathematical model is established through a partial differential equation system, which includes state variables of each physical field, the time derivative of the state variable, the spatial gradient of the state variable, the interaction function between physical fields, and the influence function of control input on the physical field, and is used to characterize the coupling relationship between different physical fields.

3. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 1, wherein The deep reinforcement learning environmental parameter collaborative control model uses a deep Q-network to achieve policy learning in a complex environment, and realizes the learning of the optimal control policy by iteratively updating the Q-value function, where the Q-value function represents the expected cumulative reward value of performing a specific action in a specific state.

4. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 3, wherein The reward function of the deep reinforcement learning environmental parameter collaborative control model is designed as a weighted combination of environmental parameter control accuracy and material response compliance, where the environmental parameter control accuracy is calculated by the deviation between the actual value and the target value, and the material response compliance is quantified by a special evaluation function.

5. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 1, characterized in that, The environmental accelerated test equivalence mapping model realizes the precise setting of accelerated test conditions by establishing the mapping relationship between the accelerated test environmental parameters and the actual service environmental parameters, and introducing the time compression ratio and the time-varying acceleration factor.

6. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 5, wherein The optimization of the accelerated test parameter combination uses a multi-objective optimization algorithm to find the optimal accelerated test parameter combination by balancing two objectives of the time compression ratio and the equivalence deviation, where the equivalence deviation evaluates the difference degree between the accelerated test environment and the actual service environment through a special metric function.

7. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 1, characterized in that The generalized predictive control model realizes the collaborative regulation of multiple physical fields by constructing a weighted cost function including the output tracking error and the change amount of the control input, and performing rolling optimization within the prediction time domain and the control time domain.

8. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 1, wherein The real-time system identification uses the recursive least squares algorithm to achieve online identification of the system model by continuously updating the parameter estimation value, where the parameter update process uses the Kalman gain to weight and correct the error between the system output and the predicted output.

9. The method for the corrosion-fatigue coupling test of a bridge structure based on artificial intelligence according to claim 8, wherein, The active compensation method for the interactive effects of multiple physical fields is realized by a feedforward-feedback composite structure. An interactive effect decoupling matrix is constructed based on the multi-physics field coupling model obtained by real-time system identification, and the compensation control input is calculated according to the change amount of the target physical field state to effectively compensate for the interactive effects of physical fields.

10. An artificial intelligence-based bridge structure corrosion fatigue coupling test system for performing an artificial intelligence-based bridge structure corrosion fatigue coupling test method according to any one of claims 1-9, characterized in that Including: A multi-physics field coupling model module for constructing a coupling mathematical model including a mechanical field, an electrochemical field, and a temperature field; A deep reinforcement learning control module for realizing the collaborative control of environmental parameters and the adaptive regulation based on material responses; An accelerated test mapping module for establishing an equivalent mathematical model between environmental accelerated test parameters and actual service environments; A generalized predictive control module for realizing the collaborative regulation of multiple physical fields and the rolling optimization control; A real-time system identification module for continuously updating the coupled model parameters and actively compensating for the interactive effects of multiple physical fields.

Images