Subway tunnel reinforced concrete structure chloride ion concentration detection system

Abstract

The invention relates to a subway tunnel reinforced concrete structure chloride ion concentration detection system, aims to solve the destructive problem of a traditional detection method and realize lossless and real-time monitoring, and comprises an electrochemical noise signal data acquisition unit, a data processing unit and a chloride ion concentration prediction unit, according to the method, electrochemical noise data of a reinforced concrete simulation test block is collected, a segmented polynomial fitting method is used for removing direct current drift, key electrochemical parameters are extracted, a prediction model is constructed in combination with an XGBoost algorithm, the chloride ion concentration is accurately predicted on the basis of electrochemical noise signal characteristics, and early warning of the corrosion risk of a subway tunnel structure is achieved. According to the system, electrochemical noise analysis, wavelet transform and machine learning technologies are creatively integrated, and the prediction accuracy and efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of chloride ion concentration detection technology, and in particular to a chloride ion concentration detection system for reinforced concrete structures in subway tunnels. Background Technology

[0002] In the development and utilization of underground space in modern cities, subway tunnels, as a critical infrastructure, directly impact the smooth flow of urban traffic and the safety of passengers. This is especially true in saline-alkali soil areas and coastal regions, where underground structures face particularly severe corrosion challenges. The extremely high chloride ion content in the soil and groundwater in these areas poses a serious threat to the reinforced concrete of the subway tunnel's main structure. Chloride ions are highly corrosive, capable of penetrating the concrete protective layer to reach the surface of the reinforcing steel, damaging its passivation film, and triggering an electrochemical corrosion process. This leads to a reduction in the cross-sectional area of ​​the reinforcing steel, causing rust expansion and cracking in the concrete structure, ultimately affecting its load-bearing capacity and long-term stability.

[0003] Traditional methods for detecting chloride ion concentration in reinforced concrete, such as colorimetry, potentiometric titration, and phosphate distillation-mercury titration, can determine chloride ion content to a certain extent, but they are all destructive tests and are not suitable for online monitoring in existing projects, thus limiting their practical application value. Therefore, it is necessary to propose a chloride ion concentration detection system for reinforced concrete structures in subway tunnels. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing traditional methods for detecting chloride ion concentration in reinforced concrete, such as colorimetry, potentiometric titration, and phosphate distillation-mercury titration. While these methods can determine chloride ion content to a certain extent, they are all destructive tests and not suitable for online monitoring in existing projects, thus limiting their practical application value. Therefore, this invention proposes a chloride ion concentration detection system for reinforced concrete structures in subway tunnels.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A system for detecting chloride ion concentration in reinforced concrete structures of subway tunnels, characterized in that it includes: an electrochemical noise signal data acquisition unit, used to acquire electrochemical noise signals from simulated reinforced concrete structure test blocks;

[0007] The electrochemical noise signal data processing unit is responsible for removing the DC drift component from the acquired signal, and using a piecewise polynomial fitting method to refine the signal and determine the optimal polynomial order m and window size r.

[0008] The chloride ion concentration prediction unit, in conjunction with electrochemical noise signal parameters, uses the XGBoost algorithm to construct a prediction model to predict chloride ion concentration. The parameters include, but are not limited to, the basic statistical parameters of the signal, the total energy of wavelet decomposition, and the key features selected through feature importance evaluation.

[0009] The above technical solution further includes:

[0010] Furthermore, the specific implementation steps of the electrochemical noise signal data processing unit include:

[0011] The original signal is divided into multiple non-overlapping time intervals, each containing r sampling points. The least squares method is applied to fit an m-order polynomial function to approximate the DC drift trend within the interval.

[0012] Based on time-frequency analysis and wavelet transform, the parameters m and r of the polynomial fitting in each interval are optimized and determined to ensure that the high-frequency characteristics of the signal are preserved and the low-frequency DC drift is effectively removed.

[0013] In signal processing based on time-frequency analysis and wavelet transform, optimizing polynomial fitting parameters requires considering the multi-resolution characteristics of wavelets. Firstly, this is achieved through the continuous wavelet transform (CWT) formula. The signal is decomposed into wavelets of different scales (a) and locations (b), where the scale parameter a controls the width of the time-frequency window, with larger scales corresponding to low-frequency components and smaller scales corresponding to high-frequency components. The scale-frequency transformation formula is then used. The scale is mapped to the actual frequency to identify low-frequency DC drift (e.g., a>2Fc corresponds to the 0-fs / 2 frequency band) and high-frequency features (a<2Fc corresponds to the high-frequency band). For the low-frequency band, a low-order polynomial (e.g., m=1-2) is used to fit the baseline drift and remove it. For the high-frequency band, the original data is retained or a high-order polynomial (e.g., r=3-5) is used to fit the local trend to ensure feature preservation. Finally, the signal is reconstructed through inverse wavelet transform to achieve a balance between high-frequency details and low-frequency correction.

[0014] Furthermore, the chloride ion concentration prediction unit employs the XGBoost algorithm for feature optimization and model construction. Through a four-stage process—feature optimization selection, model input setting, decision tree ensemble training, and model validation—it achieves model parameter tuning, ensuring high accuracy and strong generalization ability of the prediction model. The core formulas and steps are as follows:

[0015] Feature optimization selection: based on the information gain formula, Where H(D) is the entropy of dataset D, f is a feature, and Dv is a subset of the values ​​v of feature f. By filtering features with high information gain, redundancy is reduced.

[0016] Model input settings: The optimized features are input into XGBoost, and the objective function is: Where l is the loss function (e.g., mean squared error (yi-y^i)2), Ω is the regularization term, T is the number of leaf nodes, and w is the weight;

[0017] Decision tree ensemble training: Optimizes the tree structure using a greedy algorithm, generating a new tree ft in each iteration and updating the predicted values. Where η is the learning rate (shrinkage coefficient);

[0018] Model validation: Calculate the root mean square error (RMSE) or R² using cross-validation (e.g., K-fold). 2 Fractions, formula is By optimizing hyperparameters (such as tree depth and subsampling ratio) through grid search, XGBoost ensures the model's generalization ability. Through regularized objective function, gradient boosting strategy, and feature importance evaluation, it achieves high-precision prediction of chloride ion concentration.

[0019] Furthermore, the electrochemical noise signal acquisition uses a three-electrode measurement system to ensure the accuracy and stability of data acquisition, making it suitable for online monitoring of reinforced concrete structures in a simulated subway environment;

[0020] The three-electrode measurement system uses a reference electrode (R) to stably monitor the potential of the working electrode (WE1), and the counter electrode (WE2) serves as a current loop. Combined with power spectral density (PSD) analysis of electrochemical noise, it achieves chloride ion concentration detection. The core formula is the PSD calculation formula: Where T is the time interval, A is the instantaneous pulse frequency of potential and current per unit area (s-1·m-2), and E(f) is the frequency domain potential signal;

[0021] How to use:

[0022] System setup: A three-electrode system is adopted (WE1 is the working electrode, WE2 is the counter electrode, and R is the reference electrode). The potential is controlled by a potentiostat, and the electrochemical noise signal is collected by a low-noise preamplifier.

[0023] Signal processing: Perform a Fast Fourier Transform (FFT) on the acquired time-domain signal to convert it into a frequency-domain signal, and calculate the PSD curve.

[0024] Feature analysis: The corrosion type is determined by the slope of the high-frequency segment of the PSD curve (e.g., the slope of the high-frequency linear segment is less than 20dB / decade in the case of uniform corrosion). Combined with the slope of the flicker noise (1 / f feature) (-10dB / decade) and the effects of double-layer capacitance and charge transfer resistance (each increasing by -20dB / decade), the passivation or corrosion state of the steel reinforcement surface is distinguished.

[0025] Chloride ion concentration correlation: quantitative concentration analysis is achieved by calibrating the relationship between characteristic parameters of the PSD curve (such as high-frequency noise intensity) and chloride ion concentration.

[0026] Furthermore, the system uses Python 3.9 and the VSCode environment for data preprocessing, including data standardization, to ensure consistency and accuracy between model training and testing, and to improve the reliability of the prediction model.

[0027] The Python 3.9 and VSCode environment improves model reliability through data standardization, with the core formula being Z-score standardization: Where x is the original data, μ is the data mean, and σ is the standard deviation;

[0028] The usage methods include:

[0029] Data loading: Use pandas to read chloride ion concentration and electrochemical noise characteristics data in CSV or Excel format;

[0030] Standardized calculation: The mean and standard deviation of the training set are automatically calculated using sklearn.preprocessing.StandardScaler, and the above formula is applied uniformly to both the training and test sets to avoid data leakage;

[0031] Consistency verification: Ensure that the standardized parameters (μ, σ) of the training set and the test set are the same, and guarantee that the model input distribution is consistent;

[0032] Model ensemble: Input standardized data into XGBoost or neural network models to improve convergence speed and prediction accuracy.

[0033] This process enhances the generalization ability of chloride ion concentration prediction by eliminating dimensional differences and making the model focus more on the relative relationships between features.

[0034] The present invention has the following beneficial effects:

[0035] This system needs to overcome the limitations of traditional detection technologies and achieve non-destructive, real-time monitoring of chloride ion concentration in subway tunnel lining structures, providing a scientific basis for predicting corrosion risks and formulating maintenance strategies, thereby ensuring the safety and durability of subway tunnel structures.

[0036] This system adopts a three-electrode measurement system, combined with embedded sensors and a wireless transmission module, which can continuously collect electrochemical noise signals at the interface of steel bars or concrete for a long time, avoiding secondary damage to the structure.

[0037] By using piecewise polynomial fitting to remove DC drift and extracting key characteristic parameters such as power spectral density and flicker noise, and combining this with the XGBoost algorithm to construct a high-precision prediction model, this technology can quantitatively analyze chloride ion concentration in real time, provide early warning of corrosion risks, and provide a scientific basis for formulating maintenance strategies such as cathodic protection and local repair. This significantly improves the long-term safety and durability of subway tunnel structures and reduces the total life-cycle maintenance cost. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the detection system workflow of the present invention.

[0039] Figure 2 This is a schematic diagram illustrating the working principle of the chloride ion concentration prediction unit in the detection system of the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Example

[0042] like Figures 1-2 As shown, the present invention proposes a chloride ion concentration detection system for reinforced concrete structures in subway tunnels, characterized in that it includes:

[0043] An electrochemical noise signal data acquisition unit is used to acquire electrochemical noise signals from a reinforced concrete structure simulation test block.

[0044] The electrochemical noise signal data processing unit is responsible for removing the DC drift component from the acquired signal, and using a piecewise polynomial fitting method to refine the signal and determine the optimal polynomial order m and window size r.

[0045] The chloride ion concentration prediction unit, combining electrochemical noise signal parameters, uses the XGBoost algorithm to construct a prediction model to predict chloride ion concentration. These parameters include, but are not limited to, basic statistical parameters of the signal, wavelet decomposition energy sum, and key features selected through feature importance evaluation.

[0046] The specific implementation steps of the electrochemical noise signal data processing unit include:

[0047] The original signal is divided into multiple non-overlapping time intervals, each containing r sampling points. The least squares method is applied to fit an m-order polynomial function to approximate the DC drift trend within the interval.

[0048] Based on time-frequency analysis and wavelet transform, the parameters m and r of the polynomial fitting in each interval are optimized and determined to ensure that the high-frequency characteristics of the signal are preserved and the low-frequency DC drift is effectively removed.

[0049] In signal processing based on time-frequency analysis and wavelet transform, optimizing polynomial fitting parameters requires considering the multi-resolution characteristics of wavelets. Firstly, this is achieved through the continuous wavelet transform (CWT) formula. The signal is decomposed into wavelets of different scales (a) and locations (b), where the scale parameter a controls the width of the time-frequency window, with larger scales corresponding to low-frequency components and smaller scales corresponding to high-frequency components. The scale-frequency transformation formula is then used. The scale is mapped to the actual frequency to identify low-frequency DC drift (e.g., a>2Fc corresponds to the 0-fs / 2 frequency band) and high-frequency features (a<2Fc corresponds to the high-frequency band). For the low-frequency band, a low-order polynomial (e.g., m=1-2) is used to fit the baseline drift and remove it. For the high-frequency band, the original data is retained or a high-order polynomial (e.g., r=3-5) is used to fit the local trend to ensure feature preservation. Finally, the signal is reconstructed through inverse wavelet transform to achieve a balance between high-frequency details and low-frequency correction.

[0050] The chloride ion concentration prediction unit employs the XGBoost algorithm for feature optimization and model construction. Through a four-stage process—feature optimization selection, model input setting, decision tree ensemble training, and model validation—it achieves model parameter tuning, ensuring high accuracy and strong generalization ability of the prediction model. The core formulas and steps are as follows:

[0051] Feature optimization selection: based on the information gain formula, Where H(D) is the entropy of dataset D, f is a feature, and Dv is a subset of the values ​​v of feature f. By filtering features with high information gain, redundancy is reduced.

[0052] Model input settings: The optimized features are input into XGBoost, and the objective function is: Where l is the loss function (e.g., mean squared error (yi-y^i)2), Ω is the regularization term, T is the number of leaf nodes, and w is the weight;

[0053] Decision tree ensemble training: Optimizes the tree structure using a greedy algorithm, generating a new tree ft in each iteration and updating the predicted values. Where η is the learning rate (shrinkage coefficient);

[0054] Model validation: Calculate the root mean square error (RMSE) or R² using cross-validation (e.g., K-fold). 2 Fractions, formula is By optimizing hyperparameters (such as tree depth and subsampling ratio) through grid search, XGBoost ensures the model's generalization ability. Through regularized objective function, gradient boosting strategy, and feature importance evaluation, it achieves high-precision prediction of chloride ion concentration.

[0055] The electrochemical noise signal acquisition uses a three-electrode measurement system to ensure the accuracy and stability of data acquisition, and is suitable for online monitoring of reinforced concrete structures in a simulated subway environment;

[0056] The three-electrode measurement system uses a reference electrode (R) to stably monitor the potential of the working electrode (WE1), and the counter electrode (WE2) serves as a current loop. Combined with power spectral density (PSD) analysis of electrochemical noise, it achieves chloride ion concentration detection. The core formula is the PSD calculation formula: Where T is the time interval, A is the instantaneous pulse frequency of potential and current per unit area (s-1·m-2), and E(f) is the frequency domain potential signal;

[0057] How to use:

[0058] System setup: A three-electrode system is adopted (WE1 is the working electrode, WE2 is the counter electrode, and R is the reference electrode). The potential is controlled by a potentiostat, and the electrochemical noise signal is collected by a low-noise preamplifier.

[0059] Signal processing: Perform a Fast Fourier Transform (FFT) on the acquired time-domain signal to convert it into a frequency-domain signal, and calculate the PSD curve.

[0060] Feature analysis: The corrosion type is determined by the slope of the high-frequency segment of the PSD curve (e.g., the slope of the high-frequency linear segment is less than 20dB / decade in the case of uniform corrosion). Combined with the slope of the flicker noise (1 / f feature) (-10dB / decade) and the effects of double-layer capacitance and charge transfer resistance (each increasing by -20dB / decade), the passivation or corrosion state of the steel reinforcement surface is distinguished.

[0061] Chloride ion concentration correlation: quantitative concentration analysis is achieved by calibrating the relationship between characteristic parameters of the PSD curve (such as high-frequency noise intensity) and chloride ion concentration.

[0062] The system uses Python 3.9 and the VSCode environment for data preprocessing, including data standardization, to ensure consistency and accuracy between model training and testing, and to improve the reliability of the prediction model.

[0063] The Python 3.9 and VSCode environment improves model reliability through data standardization, with the core formula being Z-score standardization: Where x is the original data, μ is the data mean, and σ is the standard deviation;

[0064] The usage methods include:

[0065] Data loading: Use pandas to read chloride ion concentration and electrochemical noise characteristics data in CSV or Excel format;

[0066] Standardized calculation: The mean and standard deviation of the training set are automatically calculated using sklearn.preprocessing.StandardScaler, and the above formula is applied uniformly to both the training and test sets to avoid data leakage;

[0067] Consistency verification: Ensure that the standardized parameters (μ, σ) of the training set and the test set are the same, and guarantee that the model input distribution is consistent;

[0068] Model ensemble: Input standardized data into XGBoost or neural network models to improve convergence speed and prediction accuracy.

[0069] This process enhances the generalization ability of chloride ion concentration prediction by eliminating dimensional differences and making the model focus more on the relative relationships between features.

[0070] In this embodiment, firstly, an electrochemical noise signal data acquisition unit is used to acquire the electrochemical potential noise and current noise signals of the reinforced concrete structure in a chloride ion concentration corrosion environment, based on the three-electrode measurement system in the electrochemical workstation.

[0071] Secondly, the electrochemical noise signal data processing unit performs data processing on the collected electrochemical noise signal, specifically including DC drift removal processing and calculation of relevant time-frequency parameters of the electrochemical noise signal.

[0072] The DC drift removal processing sampling piecewise polynomial fitting method achieves refined processing of electrochemical noise signals through a local polynomial fitting strategy.

[0073] The specific implementation steps are as follows:

[0074] Signal segmentation and parameter setting: First, the continuously acquired electrochemical noise raw signal is divided into p non-overlapping time intervals, each interval contains r sampling points, and the sampling interval is . According to the signal characteristics and analysis requirements, the appropriate window size r and the highest order m of the polynomial for each signal segment are determined in advance to balance computational efficiency and fitting accuracy.

[0075] Polynomial fitting: For each time interval, an m-order polynomial function is applied to fit the signal within that interval. The polynomial coefficients are solved using optimization algorithms such as the least squares method to ensure that the fitted curve closely approximates the signal trend, which represents the DC drift component within that interval.

[0076] DC drift removal: After calculating the fitted polynomial for each segment, the fitted DC drift component is subtracted from the original signal to obtain the electrochemical noise signal with DC drift removed. This step ensures that the true dynamic characteristics of the signal are highlighted in subsequent analyses.

[0077] Parameter Optimization: To determine the optimal polynomial order *m* and window size *r*, this embodiment employs a combination of time-frequency analysis and wavelet transform. First, time-frequency analysis is used to observe the signal's spectral characteristics and identify the main frequency range of DC drift. Then, utilizing the multi-resolution characteristics of wavelet transform, the denoising effect of different parameter settings is evaluated, and the parameter combination that maximizes the preservation of high-frequency signal features while effectively removing low-frequency DC drift is selected.

[0078] The specific calculation unit for time-frequency parameters related to electrochemical noise signals is as follows: This system first calculates and extracts a series of indicators reflecting the corrosion characteristics of chloride ions from the electrochemical noise signals, including basic statistical parameters (mean, standard deviation, skewness, kurtosis, noise impedance, white noise level, power spectral density index) and wavelet decomposition energy sum (energy sum of layers D1 to D3, D4 to D6, and D7 to D8).

[0079] Finally, a chloride ion concentration prediction unit was constructed based on the XGBoost algorithm, including the following:

[0080] Feature Importance Assessment and Model Construction: Utilizing the built-in functions of the XGBoost algorithm, the contribution rate of each feature under different parameter combinations is quantitatively evaluated to guide feature selection. The feature set that contributes most to the model's predictive performance is identified, including standard deviation, white noise level, power spectral density exponent, and the total energy of layers D4–D6 and D7–D8, thereby constructing an efficient prediction model.

[0081] Model Building Process: The system employs a four-stage model building process: feature optimization and selection, model input setting, decision tree ensemble training, and model validation. Feature optimization involves selecting key features to construct the dataset; data preprocessing is performed using Python 3.9 and VSCode, including data standardization to eliminate the influence of unit dimensions; the dataset is split into training and test sets at 80%-20%; the XGBoost algorithm is applied, and the model is iteratively trained according to parameter tuning strategies, aiming to minimize the mean squared error and optimize model parameters; finally, the model performance is validated using the test set, and a comprehensive evaluation is conducted through diverse evaluation metrics, continuously iterating and optimizing the model until optimal predictive performance is achieved.

[0082] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A chloride ion concentration detection system for reinforced concrete structures in subway tunnels, characterized in that, include: An electrochemical noise signal data acquisition unit is used to acquire electrochemical noise signals from a reinforced concrete structure simulation test block. The electrochemical noise signal data processing unit is responsible for removing the DC drift component from the acquired signal, and using a piecewise polynomial fitting method to refine the signal and determine the optimal polynomial order m and window size r. The chloride ion concentration prediction unit, in conjunction with electrochemical noise signal parameters, uses the XGBoost algorithm to construct a prediction model to predict chloride ion concentration. The parameters include, but are not limited to, the basic statistical parameters of the signal, the total energy of wavelet decomposition, and key features selected through feature importance evaluation.

2. The chloride ion concentration detection system for reinforced concrete structures in subway tunnels according to claim 1, characterized in that, The specific implementation steps of the electrochemical noise signal data processing unit include: The original signal is divided into multiple non-overlapping time intervals, each containing r sampling points. The least squares method is applied to fit an m-order polynomial function to approximate the DC drift trend within the interval. Based on time-frequency analysis and wavelet transform, the parameters m and r of the polynomial fitting in each interval are optimized and determined to ensure that the high-frequency characteristics of the signal are preserved and the low-frequency DC drift is effectively removed. In signal processing based on time-frequency analysis and wavelet transform, optimizing polynomial fitting parameters requires considering the multi-resolution characteristics of wavelets. Firstly, this is achieved through the continuous wavelet transform (CWT) formula. The signal is decomposed into wavelets of different scales (a) and locations (b), where the scale parameter a controls the width of the time-frequency window, with larger scales corresponding to low-frequency components and smaller scales corresponding to high-frequency components. The scale-frequency transformation formula is then used. The scale is mapped to the actual frequency to identify low-frequency DC drift (e.g., a>2Fc corresponds to the 0-fs / 2 frequency band) and high-frequency features (a<2Fc corresponds to the high-frequency band). For the low-frequency band, a low-order polynomial (e.g., m=1-2) is used to fit the baseline drift and remove it. For the high-frequency band, the original data is retained or a high-order polynomial (e.g., r=3-5) is used to fit the local trend to ensure feature preservation. Finally, the signal is reconstructed through inverse wavelet transform to achieve a balance between high-frequency details and low-frequency correction.

3. The chloride ion concentration detection system for reinforced concrete structures in subway tunnels according to claim 1, characterized in that, The chloride ion concentration prediction unit employs the XGBoost algorithm for feature optimization and model construction. Through a four-stage process—feature optimization selection, model input setting, decision tree ensemble training, and model validation—it achieves model parameter tuning, ensuring high accuracy and strong generalization ability of the prediction model. The core formulas and steps are as follows: Feature Optimization Selection: Based on Information Gain The formula is given by H(D), where H(D) is the entropy of dataset D, f is a feature, and Dv is a subset of the values ​​v of feature f. By filtering features with high information gain, redundancy is reduced. Model input settings: The optimized features are input into XGBoost, and the objective function is: Where l is the loss function (e.g., mean squared error (yi-y^i)2), Ω is the regularization term, T is the number of leaf nodes, and w is the weight; Decision tree ensemble training: Optimizes the tree structure using a greedy algorithm, generating a new tree ft in each iteration and updating the predictions. The value, where η is the learning rate (shrinkage coefficient); Model validation: Calculate the root mean square error (RMSE) or R² using cross-validation (e.g., K-fold). 2 Fractions, formula is By optimizing hyperparameters (such as tree depth and subsampling ratio) through grid search, XGBoost ensures the model's generalization ability. Through regularized objective function, gradient boosting strategy, and feature importance evaluation, it achieves high-precision prediction of chloride ion concentration.

4. A chloride ion concentration detection system for reinforced concrete structures in subway tunnels according to any one of claims 1 to 3, characterized in that, The electrochemical noise signal acquisition uses a three-electrode measurement system to ensure the accuracy and stability of data acquisition, and is suitable for online monitoring of reinforced concrete structures in a simulated subway environment; The three-electrode measurement system uses a reference electrode (R) to stably monitor the potential of the working electrode (WE1), and the counter electrode (WE2) serves as a current loop. Combined with power spectral density (PSD) analysis of electrochemical noise, it achieves chloride ion concentration detection. The core formula is the PSD calculation formula: Where T is the time interval, A is the instantaneous pulse frequency of potential and current per unit area (s-1·m-2), and E(f) is the frequency domain potential signal.

5. A chloride ion concentration detection system for reinforced concrete structures in subway tunnels according to any one of claims 1 to 4, characterized in that, The VSCode and Python 3.9 environment are used for data preprocessing, including data standardization, to ensure consistency and accuracy between model training and testing, and to improve the reliability of the prediction model. The Python 3.9 and VSCode environment improves model reliability through data standardization, with the core formula being Z-score standardization: In the diagram, x represents the original data, μ represents the data mean, and σ represents the standard deviation.

Images