Method for long-term monitoring and detection of the effect of protection against chloride salt corrosion of steel reinforcement in concrete

Abstract

This invention discloses a long-term monitoring and detection method for the protective effect against chloride erosion of steel reinforcement in concrete, specifically relating to the field of material durability testing technology. This method involves the in-situ synchronous acquisition of multi-source data, including chloride ion, electrochemical, strain, and temperature data, using a pre-embedded sensing module. After fusion, a standardized dataset is generated and input into a physical information neural network evaluation model. This model is trained by embedding the residuals of Fick's law and the Butler-Volmer equation as regularization terms into a loss function, utilizing automatic differentiation to ensure physical constraints. The model can output a dynamic index of the protective effect, the spatial distribution of the erosion state, and a probability prediction of the remaining life in real time. This invention represents a leap from single-point failure monitoring to in-situ intelligent prediction, providing accurate and reliable evaluation, and can be directly applied to preventive maintenance and life management of engineering structures.

Description

Technical Field

[0001] This invention relates to the field of material durability testing technology, and more specifically, to a long-term monitoring and testing method for the protective effect of steel reinforcement in concrete against chloride erosion. Background Technology

[0002] Chloride-induced steel corrosion is the primary factor leading to the deterioration of concrete structure performance and shortened service life in marine and de-icing salt areas. Accurate and long-term monitoring and evaluation of the effectiveness of protective coatings, rust inhibitors and other protective measures are crucial to ensuring the safe operation and economical maintenance of critical infrastructure.

[0003] Currently, the technical means in this field can be mainly divided into the following categories: destructive detection methods, such as core sampling, which have the disadvantages of damaging the structure and being unable to be continuously observed; conventional in-situ monitoring methods, such as anodic ladders and corrosion sensors, which have limitations such as single information dimension, susceptibility to interference, and incomplete characterization of protection effect; and traditional model prediction methods, which are mostly based on simplified formulas, whose parameters depend on short-term data or accelerated tests, making it difficult to accurately reflect long-term real evolution and resulting in insufficient prediction reliability.

[0004] In summary, existing technologies generally suffer from prominent problems such as single-dimensional monitoring information, difficulty in long-term continuous monitoring, and disconnect between assessment and prediction. There is a lack of an integrated solution that can acquire multi-parameter information in situ, synchronously, and over a long period of time, and can dynamically and intelligently assess the protective effect and predict its lifespan. Therefore, this invention provides a long-term monitoring and detection method for the chloride erosion protection effect of steel bars in concrete. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a long-term monitoring and detection method for the protection effect of steel bars in concrete against chloride erosion, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a long-term monitoring and detection method for the protective effect of steel reinforcement against chloride erosion in concrete, comprising the following steps:

[0007] S1. In-situ synchronous acquisition of sensing data: A sensing module is pre-embedded inside the concrete structure. The sensing module includes a first sensor unit for monitoring the chloride ion concentration gradient, a second sensor unit for monitoring the electrochemical and environmental parameters of the steel reinforcement interface, and a third sensor unit for monitoring the internal response of the concrete. The sensing module performs in-situ, long-term, synchronous and continuous acquisition of multi-physics information to obtain a multi-source heterogeneous monitoring data stream.

[0008] S2. Multi-source data fusion and feature extraction: The monitoring data stream is spatiotemporally synchronized and aligned, and multiple time-varying feature parameters related to the chloride salt erosion process and protection effect are extracted to generate a standardized fusion feature dataset.

[0009] S3. Intelligent Evaluation and Dynamic Prediction: The fused feature dataset is input into a pre-trained evaluation model, which is a physical information neural network. This model is trained by embedding the residuals of the partial differential equations of Fick's second law and the Butler-Volmer equation as regularization terms into a loss function. The input layer of the neural network receives the standardized fused feature dataset, and the hidden layer outputs the chloride ion concentration field. and corrosion potential The spatiotemporal distribution of the residual regularization term is obtained by automatically differentiating the output of the hidden layer. The evaluation model calculates and outputs a quantitative evaluation index that characterizes the dynamic changes in the protection effect in real time, as well as the spatial distribution information of the steel reinforcement erosion state.

[0010] S4. Long-term performance and remaining service life prediction: Based on historical and current fused feature datasets, the erosion process is extrapolated through the evaluation model to predict the time to reach the preset critical corrosion state, and the remaining service life assessment results of the structure under chloride salt erosion are output.

[0011] Preferably, step S1 specifically includes:

[0012] A sensing array is formed by multiple solid chloride ion sensors arranged in a gradient along the thickness direction of the concrete protective layer. The concentration of free chloride ions at different depths is captured in real time, and a diffusion profile of chloride ions in the concrete is dynamically generated.

[0013] By using an integrated sensor patch placed on the surface of the steel reinforcement, the corrosion potential, electrochemical impedance spectrum, pH value and temperature and humidity parameters of the steel reinforcement / concrete interface are collected simultaneously.

[0014] By embedding strain and temperature sensors around the reinforcing bars, changes in the micro-strain and temperature field inside the concrete caused by chloride ion intrusion and potential steel corrosion are monitored.

[0015] Preferably, in step S2, the time-varying characteristic parameters include at least: the time-varying diffusion coefficient calculated based on the chloride ion concentration profile, the instantaneous corrosion current density calculated based on electrochemical data, the characteristic frequency parameters extracted based on impedance spectroscopy, and the micro-strain energy accumulation rate calculated based on strain data.

[0016] Preferably, the evaluation model in step S3 is a physical information neural network model, which is constructed in the following way:

[0017] Establish a parameterized multiphysics coupling mechanism model framework, which integrates at least Fick's second law describing chloride ion transport and the Butler-Volmer equation describing electrochemical corrosion reaction;

[0018] A deep neural network is constructed as a surrogate model, with the fused feature dataset as input and the chloride ion concentration field as output. and corrosion potential During training, the loss function is composed of a weighted average of the data fitting term Ldata and the physical residual regularization term Lphysics, where Lphysics is obtained by automatically differentiating the output of the neural network to calculate the residual norms of the Fick equation and the Butler-Volmer equation.

[0019] Preferably, the machine learning agent model is a long short-term memory network or a time-series prediction model based on the Transformer architecture.

[0020] Preferably, the evaluation index output in step S3 is a normalized scalar value, which is calculated by integrating one or more of the following characteristic parameters: the relative time for the chloride ion front to reach the surface of the steel bar, the ratio of the current corrosion rate of the steel bar interface to the baseline value, and the rate of change of strain energy density of the concrete cover due to erosion.

[0021] Preferably, step S3 further includes outputting visualization results, which include at least: a spatiotemporal cloud map of chloride ion concentration, a probability distribution map of corrosion activity on the surface of reinforcing steel bars, and a zoning map of the risk level of rust expansion inside concrete.

[0022] Preferably, in step S4, when predicting the remaining service life, a probabilistic prediction method is used to output the time range and confidence interval for reaching the critical state under different environmental scenarios.

[0023] Preferably, the method further includes a physical information neural network model update step: when monitoring data accumulates to a new stage or structure is repaired and intervened, the parameters of the proxy model are incrementally learned or fine-tuned using the new data to achieve dynamic evolution of the model's evaluation capability.

[0024] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned long-term monitoring and detection method for the protection effect of steel reinforcement in concrete against chloride erosion.

[0025] The technical effects and advantages of this invention are as follows:

[0026] 1. This invention achieves a fundamental leap from damage, single-point, offline monitoring to in-situ, fusion, and intelligent prediction. By implanting multi-source sensors to simultaneously collect information from multiple physical fields such as chloride ions, electrochemistry, and mechanics, and combining it with a physical information neural network model, it realizes dynamic quantitative evaluation of protection effectiveness and probabilistic prediction of remaining lifespan, providing an advanced decision-making basis for durability maintenance.

[0027] 2. This invention ensures the accuracy and reliability of long-term evaluation by deeply integrating physical mechanisms and data-driven approaches. It embeds physical laws as constraints into neural network training, making the model predictions conform to both data patterns and physical laws. This effectively overcomes the distortion problem of long-term predictions by pure data models and ensures the scientific credibility of the full life cycle evaluation results.

[0028] 3. This invention forms a closed-loop engineering solution, which significantly improves the economy and safety of operation and maintenance. The system integrates sensing, edge computing and cloud intelligent analysis to achieve unmanned long-term monitoring. Its output quantitative index and risk map can directly support the preventive maintenance of infrastructure, transforming passive emergency response into proactive maintenance, reducing the total life cycle cost and ensuring long-term safety. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the overall steps of the present invention. Detailed Implementation

[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0031] This embodiment provides a long-term monitoring and detection method for the chloride erosion protection effect of steel reinforcement in concrete. The monitoring object is a pier in the splash zone of a cross-sea bridge. The pier surface is coated with a high-performance anti-corrosion coating, as shown in the attached figure. Figure 1 As shown, the implementation steps of the method of the present invention are as follows:

[0032] S1, Synchronous acquisition of in-situ sensing data

[0033] During the prefabrication stage of the bridge piers, embedded multi-source sensing modules are pre-embedded in typical locations of the pier columns. These modules simultaneously collect three types of physical field information:

[0034] 1. Chloride ion concentration gradient monitoring: Four solid chloride ion selective microsensors are used, arranged at equal intervals along the direction perpendicular to the surface of the steel bar, starting from 10 mm away from the concrete surface and extending to the surface of the steel bar (with a protective layer thickness of 50 mm). Each sensor measures the free chloride ion concentration in its micro-region every 6 hours to form time-series data of chloride ion concentration depth profile.

[0035] 2. Multi-parameter monitoring of rebar interface: A flexible integrated sensor patch is used, which is directly and tightly attached to the surface of the selected main rebar after the rebar is tied. The patch integrates a micro three-electrode system, a micro pH sensing point and a temperature and humidity sensing chip. It automatically performs a measurement cycle every 12 hours: first, the open circuit potential is measured, then a linear polarization resistance scan is performed to calculate the instantaneous corrosion current density, then a set of electrochemical impedance spectroscopy data is collected, and finally the pH value and temperature and humidity data are recorded synchronously. All data are aligned with the chloride ion measurement timestamp.

[0036] 3. Internal response monitoring of concrete: A fiber optic grating sensor string containing 3 grating measuring points is embedded in the concrete about 20mm away from the surface of the above-mentioned steel bars to measure longitudinal strain. A miniature temperature sensor is also embedded at the same location. The fiber optic grating demodulator collects strain data once per hour, and the temperature sensor records the temperature synchronously to monitor strain changes caused by temperature change effect and possible early rust expansion.

[0037] The power supply, control, and primary data aggregation of all the above sensors are completed by a sealed embedded edge data acquisition box, which ensures the synchronous triggering of the three types of data acquisition commands and generates a multi-source heterogeneous monitoring data stream with a unified timestamp.

[0038] S2, Multi-source data fusion and feature extraction

[0039] The edge data acquisition box has a built-in preprocessing algorithm that fuses and extracts features from the raw data stream:

[0040] 1. Spatiotemporal synchronization and alignment: The system uses absolute time as a reference to align the chloride ion profile (4 depth points), steel interface parameters (potential, corrosion current, impedance spectrum, pH, temperature and humidity), concrete strain and temperature data collected every hour, and package them into a spatiotemporal data snapshot.

[0041] 2. Feature Parameter Extraction: Calculate a set of time-varying feature parameters from each data snapshot:

[0042] Chloride ion transport characteristics: Chloride ion concentration profile data at the current moment. Substituting the values ​​into the unsteady Fick second law for inversion analysis, the time-varying apparent chloride ion diffusion coefficient of the concrete at that moment is fitted. Fick's second law is expressed as:

[0043]

[0044] in, Chloride ion concentration, The depth from the concrete surface. For time;

[0045] Electrochemical corrosion characteristics: Instantaneous corrosion current density directly calculated using linear polarization resistance measurement. Simultaneously, the electrochemical impedance spectroscopy data were fitted using an equivalent circuit model to extract characteristic frequency parameters characterizing the reaction kinetics at the steel-reinforcement interface. and film resistance ;

[0046] Concrete response characteristics: strain time-series data measured by fiber Bragg grating First, using synchronously measured temperature data Thermal strain compensation is performed, and the mechanical strain is obtained after subtracting the temperature effect. Calculate the micro-strain energy accumulation rate of its strain energy. Its approximate expression is:

[0047]

[0048] in, The elastic modulus of concrete;

[0049] 3. Generate a standardized dataset: The diffusion coefficients extracted above... Corrosion current density Characteristic frequencies Film resistance Strain energy change rate Parameters such as temperature, humidity, and pH value are normalized and formatted to form a standardized fusion feature dataset, which is then uploaded to a cloud server via a wireless network.

[0050] S3, Intelligent Assessment and Dynamic Prediction

[0051] The core of cloud deployment is a pre-trained physical information neural network evaluation model:

[0052] 1. Model Building:

[0053] Physical Mechanism Framework: A parameterized mechanistic framework is established, whose core equations include Fick's second law describing chloride ion transport and the Butler-Volmer equation describing electrode reaction kinetics, the latter describing current density. With overpotential The relationship between them:

[0054]

[0055] in, For exchange current density, The charge transfer coefficient, It is Faraday's constant. Let be the ideal gas constant. Absolute temperature;

[0056] Physical Information Neural Network Construction and Training: A deep neural network is constructed as a surrogate model, with a standardized fused feature dataset as input and a chloride ion concentration field as output. and corrosion potential The spatiotemporal distribution of the intermediate physical quantities is directly output by the hidden layer of the neural network, including the chloride ion concentration field. and interface overpotential ;

[0057] Physical residual regularization term Defined in the space-time domain The residuals of the following two equations Norm integral:

[0058] Fick's second law residual:

[0059]

[0060] Butler-Volmer equation residuals:

[0061]

[0062] The integral expression is:

[0063]

[0064] Automatic differentiation uses a deep learning framework to analyze the network output. and Gradient calculation is performed during training, and the loss function is used in this process. It is specially designed as a weighted sum of data prediction error and physical equation residuals:

[0065]

[0066] in, The mean square error between the network-predicted value and the actual monitored value. To substitute the intermediate physical quantities predicted by the network into the residual norm calculated using Fick's law and the Butler-Volmer equation, To balance the hyperparameters, the loss function is optimized so that the model's predictions conform to both data patterns and physical laws.

[0067] 2. Online assessment and prediction: The cloud server will integrate the feature dataset into the pre-trained assessment model in real time. The model outputs a dynamic assessment index of the protection effect (a scalar value between 0 and 1), as well as a probability distribution map of corrosion activity on the steel surface, a spatiotemporal cloud map of chloride ion concentration, and a zoning map of the risk level of concrete rust expansion, so as to realize the visualization assessment and early warning of the corrosion status.

[0068] S4. Long-lasting performance and remaining life prediction

[0069] 1. Dynamic Model Evolution: When monitoring data accumulates to a new stage (such as a bridge having served for 3 years) or after structural maintenance intervention, the parameters of the model are incrementally learned or fine-tuned using the new data to achieve dynamic evolution of the model's evaluation capabilities.

[0070] 2. Remaining life probability prediction: Set a critical state, such as "the chloride ion concentration on the steel reinforcement surface reaches a critical threshold (e.g., 0.4% of the cementitious material mass) and the corrosion current density..." Continue to exceed "Defined as the critical state of protection failure, the evaluation model takes the current fusion characteristics as the starting point, combines the future environmental action spectrum, and performs Monte Carlo simulation to finally output the probability prediction of the remaining service life. For example, "At a 95% confidence level, it is predicted that the time for the steel bars at this monitoring point to reach the critical corrosion state is 12 to 18 years."

[0071] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A long-term monitoring and detection method for the protective effect of steel reinforcement in concrete against chloride erosion, characterized by: Includes the following steps: S1. In-situ synchronous acquisition of sensing data: A sensing module is pre-embedded inside the concrete structure. The sensing module includes a first sensor unit for monitoring the chloride ion concentration gradient, a second sensor unit for monitoring the electrochemical and environmental parameters of the steel reinforcement interface, and a third sensor unit for monitoring the internal response of the concrete. The sensing module performs in-situ, long-term, synchronous and continuous acquisition of multi-physics information to obtain a multi-source heterogeneous monitoring data stream. S2. Multi-source data fusion and feature extraction: The monitoring data stream is spatiotemporally synchronized and aligned, and multiple time-varying feature parameters related to the chloride salt erosion process and protection effect are extracted to generate a standardized fusion feature dataset. S3. Intelligent Evaluation and Dynamic Prediction: The fused feature dataset is input into a pre-trained evaluation model, which is a physical information neural network. This model is trained by embedding the residuals of the partial differential equations of Fick's second law and the Butler-Volmer equation as regularization terms into a loss function. The input layer of the neural network receives the standardized fused feature dataset, and the hidden layer outputs the chloride ion concentration field. and corrosion potential The spatiotemporal distribution of the residual regularization term is obtained by automatically differentiating the output of the hidden layer. The evaluation model calculates and outputs a quantitative evaluation index that characterizes the dynamic changes in the protection effect in real time, as well as the spatial distribution information of the steel reinforcement erosion state. S4. Long-term performance and remaining service life prediction: Based on historical and current fused feature datasets, the erosion process is extrapolated through the evaluation model to predict the time to reach the preset critical corrosion state, and the remaining service life assessment results of the structure under chloride salt erosion are output.

2. The method according to claim 1, characterized in that: Step S1 specifically includes: A sensing array is formed by multiple solid chloride ion sensors arranged in a gradient along the thickness direction of the concrete protective layer. The concentration of free chloride ions at different depths is captured in real time, and a diffusion profile of chloride ions in the concrete is dynamically generated. By using an integrated sensor patch placed on the surface of the steel reinforcement, the corrosion potential, electrochemical impedance spectrum, pH value and temperature and humidity parameters of the steel reinforcement / concrete interface are collected simultaneously. By embedding strain and temperature sensors around the reinforcing bars, changes in the micro-strain and temperature field inside the concrete caused by chloride ion intrusion and potential steel corrosion are monitored.

3. The method according to claim 1, characterized in that: In step S2, the time-varying characteristic parameters include at least: the time-varying diffusion coefficient calculated based on the chloride ion concentration profile, the instantaneous corrosion current density calculated based on electrochemical data, the characteristic frequency parameters extracted based on impedance spectroscopy, and the micro-strain energy accumulation rate calculated based on strain data.

4. The method according to claim 1, characterized in that: The evaluation model in step S3 is a physical information neural network model, which is constructed in the following way: Establish a parameterized multiphysics coupling mechanism model framework, which integrates at least Fick's second law describing chloride ion transport and the Butler-Volmer equation describing electrochemical corrosion reaction; A deep neural network is constructed as a surrogate model, with the fused feature dataset as input and the chloride ion concentration field as output. and corrosion potential During training, the loss function is composed of a weighted average of the data fitting term Ldata and the physical residual regularization term Lphysics, where Lphysics is obtained by automatically differentiating the output of the neural network to calculate the residual norms of the Fick equation and the Butler-Volmer equation.

5. The method according to claim 4, characterized in that: The machine learning agent model is a time-series prediction model based on a long short-term memory network or a Transformer architecture.

6. The method according to claim 1, characterized in that: The evaluation index output in step S3 is a normalized scalar value, which is calculated by integrating one or more of the following characteristic parameters: the relative time for the chloride ion front to reach the surface of the steel bar, the ratio of the current corrosion rate of the steel bar interface to the baseline value, and the rate of change of strain energy density of the concrete cover due to erosion.

7. The method according to claim 1, characterized in that: Step S3 also includes outputting visualization results, which include at least: a spatiotemporal cloud map of chloride ion concentration, a probability distribution map of corrosion activity on the surface of reinforcing steel bars, and a zoning map of the risk level of rust expansion inside concrete.

8. The method according to claim 1, characterized in that: In step S4, when predicting the remaining service life, a probabilistic prediction method is used to output the time range and confidence interval for reaching the critical state under different environmental scenarios.

9. The method according to claim 4, characterized in that: The method also includes a physical information neural network model update step: when monitoring data accumulates to a new stage or structure is repaired and intervened, the parameters of the proxy model are incrementally learned or fine-tuned using the new data to achieve dynamic evolution of the model's evaluation capability.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the long-term monitoring and detection method for the protection effect of steel reinforcement in concrete against chloride erosion as described in any one of claims 1-9.

Images