System and method for detecting permeability of recycled concrete

Abstract

The invention discloses a recycled concrete permeability detection system and method, and relates to the technical field of material science, and the system comprises an electric field auxiliary module which applies a preset voltage gradient through an electric field auxiliary device, obtains initial ion concentration distribution data and a current change signal from a recycled concrete sample, and obtains a reference parameter of a permeation process; the modeling and simulation module is used for constructing a pore structure model by adopting a finite element simulation method according to the reference parameters, fusing an ion migration rate and a non-uniform distribution characteristic, and determining a dynamic regulation and control scheme under the action of an electric field; according to the recycled concrete permeability detection system and method, dynamic regulation and control and real-time monitoring of ion permeability behaviors are achieved, the barrier property evaluation precision of recycled concrete is remarkably improved, and efficient technical support is provided for durability optimization of engineering materials.

Description

Technical Field

[0001] This invention relates to the field of materials science and technology, specifically to a system and method for testing the permeability of recycled concrete. Background Technology

[0002] In the fields of civil engineering and materials science, the durability of concrete is directly related to the safety and service life of infrastructure. Especially in the context of a circular economy, recycled concrete has attracted much attention due to its environmentally friendly characteristics. However, the penetration behavior of charged ions (such as chloride and sulfate ions) in recycled concrete can significantly affect its long-term performance and threaten structural stability. How to effectively control and monitor the ion penetration process has become a key issue in ensuring the durability of recycled concrete.

[0003] Current methods for studying ion permeation primarily rely on traditional diffusion experiments or chemical analysis. While these methods provide some data support, they have significant limitations. Traditional experiments often struggle to capture the dynamic migration behavior of ions in complex porous structures in real time, and lack precise characterization of responses to changes in the external environment (such as electric fields). Furthermore, existing technologies face technical bottlenecks in simultaneously monitoring ion concentration distribution and current changes, resulting in an insufficient comprehensive understanding of the permeation process and difficulty in accurately assessing material properties. The core challenge lies in how to achieve dynamic control and real-time monitoring of the ion permeation process under the assistance of an electric field. An electric field, as an external force intervention, can significantly alter the migration rate of charged ions, but its mechanism of action and its interaction with the pore structure remain unclear. This complexity makes it difficult to accurately calculate the effective permeability coefficient under the influence of an electric field, thus affecting the assessment of the barrier performance of recycled concrete. For example, in practical engineering, an electric field may induce non-uniform ion distribution in porous media, causing local concentration abrupt changes and affecting the reliability of monitoring data. Another related challenge is that the synchronous monitoring of current changes and ion concentration distribution requires high-precision equipment and algorithms, but existing technologies struggle to achieve efficient coupling between the two under dynamic conditions, limiting the comprehensive analysis of the permeation process. Summary of the Invention

[0004] The purpose of this invention is to provide a system and method for testing the permeability of recycled concrete, which accelerates the permeation process of charged ions in recycled concrete under the assistance of an electric field, simultaneously monitors the changes in current and the distribution of ion concentration, accurately calculates the effective permeability coefficient under the action of the electric field, and evaluates the blocking performance of recycled concrete against charged ions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a system for detecting the permeability of recycled concrete, comprising: an electric field-assisted module, which applies a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from recycled concrete samples, thereby obtaining baseline parameters for the permeation process; a modeling and simulation module, which constructs a pore structure model using the finite element method based on the baseline parameters, incorporates ion migration rate and non-uniform distribution characteristics, and determines a dynamic control scheme under the action of the electric field; a voltage optimization module, which obtains the voltage adjustment sequence in the dynamic control scheme, optimizes the local electric field strength for the concentration change region, and if the current change exceeds a preset threshold after optimization, re-simulates the pore structure to obtain distribution data; and a permeation path calculation module, which extracts synchronous monitoring indicators from the distribution data, calculates the ion permeation path using a diffusion equation algorithm, and obtains a preliminary estimate of the permeability coefficient. The system comprises the following modules: a coupling analysis module, which analyzes the coupling between the initial permeability coefficient value and the real-time current change signal; if the coupling deviation is less than a preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution; an iterative calculation module, which performs iterative calculations based on the corrected concentration distribution and incorporates migration rate parameters to determine the permeability coefficient under electric field assistance; a performance evaluation module, which obtains the permeability coefficient and performs a quantitative evaluation of the overall barrier performance of recycled concrete; if the evaluation results meet the preset standards, a final monitoring report is output; otherwise, the dynamic control scheme is looped back to obtain optimized iterative data; and an intelligent monitoring module, which extracts key features from the optimized iterative data, uses a neural network algorithm to process the correlation between current changes and concentration abrupt changes to obtain an enhanced synchronous monitoring framework. Based on this framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators.

[0006] Preferably, the electric field-assisted module applies a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from the recycled concrete sample, thereby obtaining the baseline parameters for the infiltration process. This includes: applying a preset voltage gradient to the recycled concrete sample through the electric field-assisted device, collecting initial ion concentration distribution data and current change signals, and determining the baseline parameters for the infiltration process; using a sensor array to monitor the ion concentration distribution in the recycled concrete sample in real time to obtain initial ion concentration distribution data; performing noise reduction processing on the collected ion concentration distribution data through a signal processing module to obtain a smooth ion concentration distribution curve; based on the smooth ion concentration distribution curve and the current change signal, using a support vector machine algorithm to extract features and determine the ion migration rate; if the ion migration rate exceeds a preset threshold, adjusting the voltage gradient through a control module to obtain a new current change signal; processing the new current change signal using a time series analysis method to obtain dynamic parameters for the infiltration process; and determining the baseline parameters for the infiltration process based on a comparative analysis of the dynamic parameters and the initial ion concentration distribution curve.

[0007] Preferably, the modeling and simulation module, based on benchmark parameters, constructs a pore structure model using the finite element method, incorporating ion migration rate and non-uniform distribution characteristics, and determines a dynamic control scheme under the action of an electric field. This includes: meshing the pore structure model using the finite element method; obtaining initial electric field distribution data by combining ion migration rate and non-uniform distribution characteristics; analyzing the correlation between ion migration rate and electric field strength using partial least squares regression based on the initial electric field distribution data to obtain an electric field control coefficient; if the electric field control coefficient exceeds a preset threshold, adjusting the electric field distribution in the pore structure model using an iterative optimization method to obtain optimized electric field distribution data; extracting the dynamic change trend of ion migration rate using time series analysis based on the optimized electric field distribution data to obtain dynamic control parameters; classifying ion distribution characteristics using a support vector machine algorithm based on the dynamic control parameters to determine the applicability of the control parameters; updating the pore structure model using numerical simulation to obtain the final dynamic control scheme; and analyzing the stability of ion migration rate using simulation verification based on the final dynamic control scheme to determine the control result under the action of the electric field.

[0008] Preferably, the voltage optimization module acquires the voltage adjustment sequence in the dynamic control scheme, optimizes the local electric field intensity in the concentration change region, and if the current change after optimization exceeds a preset threshold, re-simulates the pore structure to obtain distribution data, including: constructing a pore structure model using the finite element simulation method, generating an initial voltage adjustment sequence, and acquiring electric field intensity distribution data in the concentration change region; if the electric field intensity distribution data exceeds the preset threshold, adjusting the voltage sequence using the gradient descent algorithm to obtain the optimized voltage sequence; calculating the current change data using numerical simulation based on the optimized voltage sequence to obtain the current change trend; if the current change trend exceeds the preset threshold, updating the pore structure model using an iterative optimization method to obtain the adjusted structure model; recalculating the electric field intensity distribution data based on the adjusted structure model to obtain a stable electric field distribution; processing the stable electric field distribution using time series analysis to extract dynamic voltage adjustment parameters to obtain a control parameter set; and classifying the electric field distribution in the concentration change region using a support vector machine algorithm based on the control parameter set to determine the stability of the control parameters and obtain the final control scheme.

[0009] Preferably, the permeation path calculation module extracts synchronous monitoring indicators from the distribution data and calculates the ion permeation path using a diffusion equation algorithm to obtain a preliminary permeation coefficient estimate, including: extracting ion concentration gradient features from the distribution data and generating concentration distribution features using a data extraction method; calculating the ion migration path using a diffusion equation algorithm based on the concentration distribution features to obtain migration path features; if the migration path features do not match a preset threshold, adjusting the path optimization parameters through an iterative optimization method to obtain an optimized migration path; calculating the ion diffusion rate based on the optimized migration path to obtain a diffusion rate distribution; extracting the permeation coefficient estimate from the diffusion rate distribution and classifying the permeation coefficient distribution using a support vector machine algorithm to determine distribution stability; if the distribution stability analysis result does not reach a preset threshold, recalculating the ion migration path through a numerical simulation method to obtain an updated permeation coefficient distribution; and extracting the final path optimization parameters based on the updated permeation coefficient distribution to obtain a stable permeation coefficient estimate.

[0010] Preferably, the coupling analysis module, through coupling analysis of the preliminary permeability coefficient value and the real-time current change signal, confirms the barrier performance evaluation model if the coupling deviation is less than a preset threshold; otherwise, it adjusts the dynamic control scheme to obtain the corrected concentration distribution. This includes: extracting dynamic features from the real-time current change signal; generating a current fluctuation sequence using Fourier transform to obtain frequency distribution features; calculating the coupling deviation between the preliminary permeability coefficient estimate and the current fluctuation sequence based on the frequency distribution features to obtain deviation distribution features; if the deviation distribution features are less than a preset threshold, determining the barrier performance evaluation model using logistic regression to obtain the model evaluation result; if the deviation distribution features are greater than a preset threshold, adjusting the concentration distribution parameters using gradient descent to obtain an optimized concentration distribution; re-extracting the dynamic features of the real-time current change signal based on the optimized concentration distribution to generate an updated current fluctuation sequence; determining whether the preset threshold is met through coupling analysis of the updated current fluctuation sequence and the preliminary permeability coefficient estimate to obtain the final barrier performance evaluation model; and extracting key performance parameters based on the final barrier performance evaluation model to generate the corrected concentration distribution features.

[0011] Preferably, the iterative calculation module, based on the corrected concentration distribution and incorporating migration rate parameters, performs iterative calculations to determine the permeability coefficient under electric field assistance, including: extracting migration rate parameters based on the corrected concentration distribution and generating a preliminary permeability coefficient estimate using a numerical integration method; calculating the migration rate distribution under the electric field using the preliminary permeability coefficient estimate and the electric field-assisted parameters, employing a finite difference method; if the deviation of the migration rate distribution is greater than a preset threshold, adjusting the electric field-assisted parameters using a gradient descent method to obtain an optimized migration rate distribution; re-extracting dynamic features based on the optimized migration rate distribution to generate an updated permeability coefficient estimate; constructing a permeability coefficient model using a logistic regression algorithm based on the updated permeability coefficient estimate to obtain the model output result; if the deviation of the model output result is less than a preset threshold, obtaining key performance parameters using a signal extraction method to determine the final effective permeability coefficient; and generating corrected concentration distribution features based on the final effective permeability coefficient and outputting optimized distribution data.

[0012] Preferably, the performance evaluation module acquires the permeability coefficient and performs a quantitative evaluation of the overall barrier performance of recycled concrete. If the evaluation result meets the preset standard, a final monitoring report is output; otherwise, the dynamic control scheme is cyclically returned to obtain optimized iterative data, including: acquiring the barrier performance data of recycled concrete; analyzing the material composition distribution using X-ray fluorescence spectroscopy to obtain composition distribution characteristics; if the deviation of the composition distribution characteristics is greater than a preset threshold, adjusting the detection parameters of X-ray fluorescence spectroscopy using a particle swarm optimization algorithm to obtain optimized composition distribution characteristics; calculating the preliminary evaluation value of barrier performance using finite element analysis based on the optimized composition distribution characteristics to obtain preliminary barrier performance data; if the deviation of the preliminary barrier performance data is greater than a preset threshold, classifying the preliminary barrier performance data using a support vector machine algorithm to obtain classified performance characteristics; processing the performance characteristics using a data standardization method based on the classified performance characteristics to obtain standardized performance data; if the deviation of the standardized performance data is less than a preset threshold, extracting key performance parameters using a signal processing method to determine the final barrier performance data; and adjusting the analysis parameters of the material composition distribution using a dynamic control scheme based on the final barrier performance data to obtain optimized material distribution data.

[0013] Preferably, the intelligent monitoring module extracts key features from the optimization iteration data, uses a neural network algorithm to process the correlation between current changes and concentration abrupt changes, and obtains an enhanced synchronous monitoring framework. Based on the enhanced synchronous monitoring framework, it simulates ion permeation behavior under various electric field scenarios to determine the final barrier performance indicators, including: extracting key features from the optimization iteration data, using principal component analysis to reduce the dimensionality of the key features, obtaining dimensionality-reduced feature data; if the variance of the dimensionality-reduced feature data is less than a preset threshold, then using time series analysis to perform time-series modeling on the dimensionality-reduced feature data, obtaining a time-series feature model; based on the time-series feature model, using a convolutional neural network to analyze the correlation between current change data and concentration abrupt changes, obtaining a correlation feature vector; if the classification accuracy of the correlation feature vector is greater than a preset threshold, then using a data standardization method to process the correlation feature vector, obtaining standardized correlation data; based on the standardized correlation data, using signal processing methods to extract key parameters for synchronous monitoring, determining the synchronous monitoring parameter set; extracting time-series synchronous features from the synchronous monitoring parameter set, and using cluster analysis to process the time-series synchronous features... The system is divided into groups to obtain a synchronous monitoring framework. Based on this framework, a data fusion method is used to integrate multi-source monitoring data streams to obtain enhanced synchronous monitoring results. Dynamic electric field signals are obtained from the multi-source electric field data, and time-series analysis is used to model these signals, resulting in a dynamic electric field model. If the fluctuation amplitude of the dynamic electric field model exceeds a preset threshold, feature extraction is performed on the model using signal decomposition to obtain an ion permeation feature set. Based on this feature set, a convolutional neural network is used to analyze the correlation between ion permeation behavior and electric field strength, resulting in a permeation correlation vector. The permeation correlation vector is normalized using a data standardization method to obtain standardized permeation data. Cluster analysis is used to group the standardized permeation data to determine a preliminary barrier performance index set. If the classification consistency of the preliminary barrier performance index set exceeds a preset threshold, a data fusion method is used to integrate the multi-source electric field data with the preliminary barrier performance index set to obtain enhanced barrier performance indices. Based on these enhanced barrier performance indices, a time-series prediction method is used to analyze the future trend of ion permeation behavior and determine the final barrier performance index set.

[0014] A method for testing the permeability of recycled concrete, using the aforementioned recycled concrete permeability testing system, includes: applying a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from the recycled concrete sample, thus obtaining baseline parameters for the permeation process; based on the baseline parameters, constructing a pore structure model using a finite element simulation method, incorporating ion migration rate and non-uniform distribution characteristics, and determining a dynamic control scheme under the action of the electric field; obtaining the voltage adjustment sequence in the dynamic control scheme, optimizing the local electric field strength for the concentration change region, and if the current change exceeds a preset threshold after optimization, resimulating the pore structure to obtain distribution data; extracting synchronous monitoring indicators from the distribution data, calculating the ion permeation path using a diffusion equation algorithm, and obtaining a preliminary estimate of the permeability coefficient. The process involves coupling analysis of the initial permeability coefficient value with real-time current change signals. If the coupling deviation is less than a preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution. Based on the corrected concentration distribution, migration rate parameters are incorporated for iterative calculation to determine the permeability coefficient under electric field assistance. The permeability coefficient is then used to quantitatively evaluate the overall barrier performance of the recycled concrete. If the evaluation results meet preset standards, a final monitoring report is output; otherwise, the dynamic control scheme is looped back to obtain optimized iterative data. Key features are extracted from the optimized iterative data, and a neural network algorithm is used to process the correlation between current changes and concentration abrupt changes, resulting in an enhanced synchronous monitoring framework. Based on this framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators.

[0015] As can be seen from the above technical solution, the present invention has the following beneficial effects:

[0016] This system and method for testing the permeability of recycled concrete addresses the challenge of dynamically controlling and accurately assessing ion permeation in complex porous structures. Addressing the difficulties of uneven ion concentration distribution and unpredictable permeation behavior under electric fields in recycled concrete, this invention obtains initial ion concentration and current change signals by applying a preset voltage gradient, establishes a finite element model of the pore structure, incorporates migration rate and non-uniform distribution characteristics, and generates a dynamic control scheme. The scheme optimizes the local electric field strength through voltage adjustment sequences, calculates the permeation path using a diffusion equation algorithm, and iteratively derives the effective permeability coefficient. If coupling deviation or barrier performance fails to meet standards, the control scheme is iteratively optimized. Finally, an enhanced synchronous monitoring framework is constructed using a neural network algorithm to simulate ion permeation behavior in multiple scenarios and accurately quantify barrier performance. This invention achieves dynamic control and real-time monitoring of ion permeation behavior, significantly improving the accuracy of barrier performance assessment in recycled concrete and providing efficient technical support for optimizing the durability of engineering materials. Attached Figure Description

[0017] Figure 1 This is a system connection diagram of the present invention;

[0018] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] 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.

[0020] like Figure 1 As shown, this invention provides a technical solution: a system for detecting the permeability of recycled concrete, comprising an electric field-assisted module, which applies a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from recycled concrete samples, thereby obtaining baseline parameters for the permeation process; a modeling and simulation module, which constructs a pore structure model using the finite element method based on the baseline parameters, incorporating ion migration rate and non-uniform distribution characteristics, and determines a dynamic control scheme under the action of the electric field; a voltage optimization module, which obtains the voltage adjustment sequence in the dynamic control scheme, optimizes the local electric field strength for the concentration change region, and if the current change after optimization exceeds a preset threshold, re-simulates the pore structure to obtain distribution data; and a permeation path calculation module, which extracts synchronous monitoring indicators from the distribution data, calculates the ion permeation path using a diffusion equation algorithm, and obtains a preliminary estimate of the permeability coefficient; and a coupling module; The analysis module, through coupled analysis of the preliminary permeability coefficient value and real-time current change signal, confirms the barrier performance evaluation model if the coupling deviation is less than a preset threshold; otherwise, it adjusts the dynamic control scheme to obtain the corrected concentration distribution. The iterative calculation module, based on the corrected concentration distribution, incorporates migration rate parameters for iterative calculation to determine the permeability coefficient under electric field assistance. The performance evaluation module obtains the permeability coefficient and performs a quantitative evaluation of the overall barrier performance of recycled concrete. If the evaluation results meet the preset standards, it outputs the final monitoring report; otherwise, it loops back to the dynamic control scheme to obtain optimized iterative data. The intelligent monitoring module extracts key features from the optimized iterative data, uses a neural network algorithm to process the correlation between current changes and concentration abrupt changes, and obtains an enhanced synchronous monitoring framework. Based on the enhanced synchronous monitoring framework, it simulates ion permeation behavior under various electric field scenarios to determine the final barrier performance indicators.

[0021] The system utilizes an electric field-assisted module to apply a preset voltage gradient across the ends of a recycled concrete sample, inducing ion migration within the sample. Sensors simultaneously collect ion concentration distribution and current change signals to establish baseline parameters for the permeation process. Based on these parameters, the modeling and simulation module employs the finite element method to create a three-dimensional model of the concrete's internal pore structure, coupling the ion migration rate and its spatially non-uniform distribution characteristics into the model to generate an adjustable dynamic electric field application scheme. The voltage optimization module fine-tunes the electric field strength locally based on this scheme to match the concentration variation characteristics of different regions, and judges the optimization effect by detecting the current change amplitude. When the optimized current change exceeds a preset threshold, the system reruns the finite element simulation to update the concentration distribution data. The permeation path calculation module calculates the ion trajectory and permeation path using the diffusion equation to obtain a preliminary permeation coefficient. The coupling analysis module fits this coefficient with the real-time current change signal; if the deviation is less than a threshold, the effectiveness of the barrier performance evaluation model is confirmed; otherwise, the system returns to the dynamic control module to update the scheme. The iterative calculation module repeatedly calculates the permeation coefficient based on the corrected concentration distribution and migration rate parameters until a convergent permeation coefficient is obtained. The performance evaluation module quantifies and scores the overall barrier capacity of recycled concrete based on this coefficient to determine whether it meets the preset standards; if the standards are not met, the optimization is repeated. The intelligent monitoring module uses neural network analysis to optimize the correlation between current changes and concentration abrupt changes in the iterative data, generates an enhanced monitoring framework, and performs multiple simulations based on different electric field scenarios to determine the final barrier performance indicators.

[0022] This system enables high-precision, dynamic detection of the permeability of recycled concrete, significantly improving temporal resolution, spatial resolution, and data fitting accuracy compared to traditional single-sample static measurement methods. An electric field-assisted method accelerates ion migration, shortening the experimental cycle, while finite element simulation combined with real-time data feedback enhances the reliability of model predictions. A voltage optimization module adaptively adjusts to local structural differences, reducing errors caused by concrete heterogeneity. Coupled analysis and iterative calculation ensure the stability and accuracy of the permeability coefficient. The intelligent monitoring module incorporates neural networks for feature extraction and pattern recognition, adapting to various working conditions and enhancing predictive capabilities. The overall system boasts advantages such as automation, high repeatability, and high detection efficiency, providing a more scientific basis for the quality assessment and engineering application of recycled concrete.

[0023] The electric field-assisted module applies a preset voltage gradient through an electric field-assisted device to acquire initial ion concentration distribution data and current change signals from recycled concrete samples, obtaining baseline parameters for the infiltration process. This includes: applying a preset voltage gradient to the recycled concrete samples using an electric field-assisted device, collecting initial ion concentration distribution data and current change signals, and determining the baseline parameters for the infiltration process; using a sensor array to monitor the ion concentration distribution in the recycled concrete samples in real time to obtain initial ion concentration distribution data; using a signal processing module to denoise the collected ion concentration distribution data to obtain a smooth ion concentration distribution curve; using the smooth ion concentration distribution curve and the current change signal, using a support vector machine algorithm to extract features and determine the ion migration rate; if the ion migration rate exceeds a preset threshold, the voltage gradient is adjusted through a control module to acquire a new current change signal; using a time series analysis method to process the new current change signal to obtain dynamic parameters for the infiltration process; and comparing the dynamic parameters with the initial ion concentration distribution curve to determine the baseline parameters for the infiltration process.

[0024] In one possible implementation, a preset voltage gradient is first applied to both ends of the recycled concrete sample using an electric field-assisted device. The sample is a cylinder, 10 cm long and 5 cm in diameter, with the electric field direction aligned with the sample's axis. The voltage gradient is 1.0 volt per centimeter, corresponding to a voltage of 10 volts at the corresponding end. The voltage linearly increases to the set value within 10 seconds to avoid transient impacts. Simultaneously, a sensor array is formed by evenly spaced eight sensing points, with one sensing point arranged 1.11 cm along the axial direction starting from the liquid inlet end. The array sampling frequency is 10 times per second, synchronously acquiring the ion concentration signal and loop current signal of each sensing point and aligning them with timestamps. The time alignment method is to record all samples with the initial rise time of the power supply as the zero point.

[0025] The electric field-assisted device consists of a constant-voltage DC power supply module, electrode assembly, and insulating support structure. It is used to establish a uniform and controllable voltage gradient across the two ends of the recycled concrete sample to drive ion migration. The constant-voltage DC power supply module has an output voltage range of 0V to 30V with an accuracy of no less than 0.01V, a maximum output current of 2A, short-circuit protection, and output stability within a load impedance range of 20 ohms to 200 ohms, with voltage fluctuations not exceeding ±0.05V. It also supports programmable voltage ramp-up time settings; in this embodiment, it is set to linearly ramp to the target voltage within 10 seconds. The electrode assembly is made of high-purity graphite material with a purity of no less than 99.9%. The disc has a diameter of 5 cm and a thickness of 5 mm. The contact surface is mechanically polished to a surface roughness of no more than 0.8 micrometers. A 50N clamping force is applied to ensure tight contact with the sample end face using stainless steel clamps. A 1.5 mm² tin-plated copper wire is welded to the back and connected to the power output terminal via bolt terminals. The insulating support structure uses a 10 mm thick polytetrafluoroethylene (PTFE) plate to prevent leakage paths other than those at the end faces. The voltage gradient is determined by dividing the target terminal voltage by the sample length. In this embodiment, the terminal voltage is 10 volts, the sample length is 10 centimeters, and the gradient is 1.0 volt per centimeter. When the gradient needs to be adjusted, the control module completes linear voltage adjustment within 5 seconds. During operation, the sample is placed on an insulating support and connected to the electrode assembly. After confirming that the power output is zero, the system starts up. The power supply rises from 0 volts to the target value according to the set boost curve and then enters constant voltage mode, outputting voltage and current signals in real time for the host computer to collect and analyze.

[0026] Subsequently, initial ion concentration distribution data calibration was performed using a two-point method. The zero point was a deionized water background, and the span points were standard electrolyte solutions with a mass concentration equivalent to 100 mmol / L. At each sensing point, the zero-point output and the span-point output were recorded. The difference between the two outputs was divided by the known concentration difference between the two points to obtain the conversion coefficient for that sensing point. Then, the zero-point output was subtracted from the actual sensing point output during the testing period, and multiplied by this conversion coefficient to obtain the concentration value of that sensing point at each sampling time, thus forming the initial ion concentration distribution data.

[0027] Denoising and smoothing were performed point-by-point on the above distribution data. Denoising employed median filtering with a window length of 5 sampling points. For each sampling time, the concentration values ​​of that time and the two sampling points before and after it were sorted numerically, and the third value in the sorted sequence was selected as the denoised result for that time. Smoothing employed moving average with a window length of 7 sampling points. For each sampling time, the denoised result was calculated using the arithmetic mean of the concentration values ​​of that time and the three sampling points before and after it, and this average value replaced the original value, resulting in a smoothed ion concentration distribution curve.

[0028] The smoothed ion concentration distribution curve and the current change signal are registered on a unified time axis. If there is uneven sampling of the current signal, it is resampled at equal intervals of 0.1 seconds on the unified time axis using linear interpolation. After registration, feature extraction and ion migration rate determination are performed. Feature extraction is strictly performed according to the following steps with fixed parameters: The first feature is the concentration growth rate. For the smoothed concentration sequence of each sensing point, the concentration value of the next moment is subtracted from the concentration value of the previous moment and divided by the time interval between the two moments to obtain the rate sequence. The maximum value of the rate sequence within each sensing point is selected as the peak growth rate of that sensing point. Then, the arithmetic mean of the peak growth rates of the 8 sensing points is calculated to obtain the candidate value of the concentration growth rate of the sample.

[0029] The second feature is the concentration gradient. For each time point within the initial 30 seconds, the concentration difference between two adjacent sensing points is taken on the same time cross section and divided by the distance between the two sensing points to obtain the gradient for each segment. The arithmetic mean of each gradient segment is calculated to obtain the average concentration gradient for that time point. Then, the arithmetic mean of the average concentration gradients for all time points within the 30 seconds is calculated to obtain the candidate concentration gradient value. The third feature is the current change rate. For each time step in the current sequence, the current value of the next time step is subtracted from the current value of the previous time step and divided by the time interval to obtain the change rate sequence. The maximum absolute value of this sequence is taken as the candidate current change rate value.

[0030] The three candidate values ​​are input into the Support Vector Machine (SVM) algorithm to determine the migration rate. The SVM algorithm uses a radial basis function kernel with a penalty parameter of 100 and a kernel width of 0.5. The model is trained on 300 historical samples before equipment delivery, including 100 slow, 100 medium, and 100 fast samples. After training, it is validated with 30 independent samples, achieving a classification accuracy of no less than 90%. In this embodiment, during the inference phase, the three features are combined with fixed weights to form a numerical estimate of the ion migration rate. The weights are assigned as follows: concentration growth rate weight 0.6, concentration gradient weight 0.3, and current change rate weight 0.1. The three are added together to obtain the migration rate estimate, and the ranking results for the slow, medium, and fast categories are given simultaneously.

[0031] The threshold was set using a deterministic strategy. The calculation method used the estimated migration rate sequence for the first 60 seconds after voltage stabilization as the baseline sequence. First, the arithmetic mean of this sequence was calculated to obtain the baseline average. Then, the fluctuation amplitude was obtained by dividing the difference between the maximum and minimum values ​​of the sequence by 2. The threshold was calculated by adding three times the fluctuation amplitude to the baseline average. Simultaneously, a sample protection limit of 0.20 mmol / L / s was set. The final threshold was the smaller of the two values ​​mentioned above, and this threshold remained constant throughout the test. If the estimated migration rate was greater than the threshold, the control module reduced the voltage gradient by a fixed ratio, which was the current gradient multiplied by 0.80. The new setting was linearly adjusted to the target value within 5 seconds. After the adjustment, a 10-second wait was completed to ensure system stability before acquiring new current change signals at the same sampling frequency as the initial stage.

[0032] Time series analysis was performed on the new current change signal. The analysis method was the autoregressive moving average method. First, stationarity was tested. The current series was divided into 10 windows of equal duration, and the arithmetic mean and variance of each window were calculated. If the maximum difference between the means of all windows did not exceed 5% of the total series mean and the maximum difference between the variances did not exceed 10% of the total series variance, it was considered stationary. Otherwise, first-order differencing was performed, that is, the current value at each time step was subtracted from the current value at the previous time step to obtain a new series, and the above stationarity test was repeated. If it still did not meet the requirements, differencing was performed again, with a maximum of 2 differencing times.

[0033] After stationarity is satisfied, the model order is selected, with candidate orders ranging from 1 to 5. For each candidate order, the most recent 300 data points are used for parameter estimation, and the next 100 data points are used for one-step prediction. The absolute value of the difference between the predicted and actual values ​​is calculated, and the arithmetic mean of the 100 differences is taken to obtain the mean absolute error (MAO) for that order. The order with the smallest MAO is selected as the final order. After the final order is determined, the dynamic parameters of the permeation process are output. These dynamic parameters are fixed and include three items: the predicted maximum current value in the next 60 seconds, the time to reach that maximum value, and the average rise rate in the next 60 seconds. The average rise rate is calculated by subtracting the first value from the last value of the predicted current sequence in the next 60 seconds and then dividing by 60 seconds.

[0034] The dynamic parameters were compared with the initial ion concentration distribution curve. Before comparison, the initial ion concentration distribution curve was aligned with the new current signal on the time axis, and the current signal was normalized. The normalization method was to subtract the minimum value from the maximum value in the current sequence to obtain a range. Each current value was then subtracted from the minimum value and divided by this range, so that the normalized current value was between 0 and 1. During comparison, the difference between the sample average concentration value of the initial ion concentration distribution curve and the normalized current value was taken at the same time point. The square of each difference was then calculated as the arithmetic mean over the entire time period to obtain the deviation value. The deviation tolerance was 0.02. If the deviation value was not greater than 0.02, the baseline parameters of the permeation process were confirmed.

[0035] The baseline parameters are fixed and include seven items: the smoothed ion concentration distribution curve, the initial current change curve, the estimated ion migration rate, the migration rate threshold, the current voltage gradient value, three time-series dynamic parameters, and the deviation value of the concentration-current comparison. These seven parameters are recorded as the baseline parameter set for this sample. If the deviation value is greater than 0.02, no new module or branch is introduced. Instead, the above steps of time alignment, stationarity check, order selection, and dynamic parameter calculation are repeated once using the last voltage gradient in the current process and the corresponding new current change signal. The deviation value is then recalculated using the new dynamic parameters and the initial ion concentration distribution curve. The deviation value obtained from the second calculation is used as the final basis for determining the baseline parameters of the permeation process. A third calculation is not performed, thus ensuring process closure, parameter uniqueness, and avoiding uncertain conclusions.

[0036] The preset voltage gradient was determined to be 1.0 V / cm using a three-level comparison method, with the three levels being 0.5 V / cm, 1.0 V / cm, and 1.5 V / cm. The stable currents on samples with the same mix ratio were 20 mA, 35 mA, and 70 mA, respectively. The target range was 20 mA to 50 mA, hence the choice of 1.0 V / cm. The sensor array was chosen to have 8 sensors, evenly distributed, to ensure that each segment length did not exceed 1.5 cm to cover the heterogeneous structure of the sample. The sampling frequency of 10 times per second was chosen to meet the requirement of ten times oversampling for changes on the order of 1 second. The window lengths for median filtering and moving average were determined through quantitative evaluation of pre-test data. A window length of 5 achieved a 90% peak elimination rate, while a window length of 7 provided the optimal smoothness score. The kernel type, penalty parameter, and kernel width of the support vector machine algorithm were derived from offline cross-validation results, and this combination achieved a classification accuracy of 92% on the validation set. The migration rate threshold is determined strictly using a combination rule of the baseline average value and fluctuation amplitude, and is constrained by the sample protection upper limit. The threshold remains unchanged within a single test run. The order of the time series model is uniquely determined by the minimum mean absolute error criterion. If the final selected order is 3, then three dynamic parameters are generated from the prediction results for the next 60 seconds output by that order. The deviation tolerance of 0.02 is selected based on the statistical results of the retest deviation of historical baseline samples. This value ensures the stability of the baseline parameters for subsequent simulations and evaluations.

[0037] The modeling and simulation module constructs a pore structure model using the finite element method (FEM) based on baseline parameters, incorporating ion migration rate and non-uniform distribution characteristics. It then determines a dynamic control scheme under an electric field. This includes meshing the pore structure model using FEM and obtaining initial electric field distribution data based on ion migration rate and non-uniform distribution characteristics. Based on this initial electric field distribution data, partial least squares regression (PLR) is used to analyze the correlation between ion migration rate and electric field strength, yielding an electric field control coefficient. If the electric field control coefficient exceeds a preset threshold, an iterative optimization method is used to adjust the electric field distribution in the pore structure model, obtaining optimized electric field distribution data. Based on the optimized electric field distribution data, time series analysis is used to extract the dynamic trend of ion migration rate, obtaining dynamic control parameters. Based on these dynamic control parameters, a support vector machine (SVM) algorithm is used to classify ion distribution characteristics and determine the applicability of the control parameters. If the applicability of the control parameters meets preset conditions, numerical simulation is used to update the pore structure model, resulting in the final dynamic control scheme. Finally, simulation verification is used to analyze the stability of ion migration rate and determine the control result under the electric field.

[0038] In one possible implementation, the modeling and simulation module takes baseline parameters as input and first establishes a three-dimensional cylindrical pore structure model. The sample length is 10 cm, the diameter is 5 cm, and the axis is aligned with the electric field direction. A uniform cubic mesh is used, with each element having a side length of 0.1 mm and a total number of elements of no less than 1,000,000. Two deterministic properties are assigned to each element: porosity and ion migration rate. The porosity is determined by mapping the initial ion concentration distribution data from the baseline parameters according to spatial location. Specifically, the sample axis is discretized in 0.1 mm increments. For each discrete location, the average concentration value at the same time is taken. The average concentration value at that location is subtracted from the minimum average concentration value of the entire sample, and then divided by the difference between the maximum and minimum average concentration values ​​of the entire sample to obtain a dimensionless value between 0 and 1. This dimensionless value is used as the porosity of each element at that location.

[0039] The initial ion migration rate is determined by the estimated ion migration rate in the baseline parameters and the spatial concentration gradient. Specifically, the average concentration difference between two adjacent discrete locations on the same time cross section is calculated and divided by 0.1 mm to obtain the gradient of that cross section. This gradient is then normalized by the difference between the minimum and maximum values ​​to obtain a dimensionless gradient value between 0 and 1. The initial ion migration rate of each cell is obtained by multiplying the estimated ion migration rate in the baseline parameters by this dimensionless gradient value. Boundary conditions and solution control parameters are then set: the inlet potential is fixed at 10 V, the outlet potential is fixed at 0 V, the cylindrical sidewall is set to insulation, the time step is fixed at 1 second, and the calculation duration is fixed at 300 seconds. The numerical solution uses successive over-relaxation iterations with a relaxation factor of 1.5 and a single-step residual convergence threshold of 0.0001. If the single-step residual does not reach the threshold within 1000 iterations, the relaxation factor is reduced to 1.2 and iteration continues until convergence.

[0040] After completing the transient solution, the electric field intensity and ion migration rate fields for each time step are obtained. The electric field intensity and ion migration rate at a time interval of 60 seconds are taken as the initial electric field distribution data and paired with the migration rate data for the correlation analysis stage. In this stage, the electric field control coefficient is calculated according to a fixed procedure of partial least squares regression. Specifically, the electric field intensity and ion migration rate of all units are first normalized to their minimum and maximum values, ensuring both are between 0 and 1. Then, the covariance direction between the two is calculated, and a score vector in one direction is extracted. This score vector is used to perform linear regression on the ion migration rate to obtain a set of regression coefficients. The absolute values ​​of the regression coefficients corresponding to the electric field intensity are then taken and the arithmetic mean is calculated over all units. The resulting average value is the electric field control coefficient. The threshold for the electric field control coefficient is determined using a pre-shipment statistical method. Thirty historical samples that have passed stability verification are selected, and the electric field control coefficient is calculated using the same steps, followed by an arithmetic mean. This average value is then multiplied by 1.20 to obtain the threshold value, which remains unchanged throughout the equipment's lifespan. If the electric field control coefficient obtained in this calculation is greater than the threshold, then proceed to the iterative optimization step; otherwise, skip this step and proceed directly to dynamic parameter extraction.

[0041] The iterative optimization step only adjusts the electric field distribution within the model without changing the boundary conditions. Specifically, it first calculates the arithmetic mean and standard deviation of the current electric field intensity across all elements. Elements with an electric field intensity higher than the average plus 0.5 times the standard deviation are classified as high-field elements, and elements with an electric field intensity lower than the average minus 0.5 times the standard deviation are classified as low-field elements. For high-field elements, the equivalent conduction weight is reduced by multiplying the current value by 0.90, and for low-field elements, the equivalent conduction weight is increased by multiplying the current value by 1.10. The remaining elements are left unchanged. After completing one weight adjustment, the electric field is solved again and the new standard deviation is calculated. If the new standard deviation is less than 10% lower than the previous one, the weight adjustment ratio remains unchanged. Otherwise, the weight adjustment ratio of both types of units is halved simultaneously, with a minimum of 0.02 and a maximum of 0.20. This process is repeated until the standard deviation of the electric field intensity is no greater than 50% of the initial standard deviation or the number of iterations reaches 20. Optimization stops when either condition is met, and the electric field intensity field at the time of stopping is the optimized electric field distribution data.

[0042] The dynamic trend of ion migration rate is extracted based on the optimized electric field distribution data. Specifically, a 300-second transient solution is run under fixed boundary conditions to obtain the ion migration rate sequence per second for each cell. Then, three dynamic indicators are calculated along the cell dimension. The first indicator is the average growth rate over the initial 60 seconds, calculated by subtracting the ion migration rate of the first second from the ion migration rate of the 60th second, dividing by 59 seconds to obtain the average cell growth rate, and then taking the arithmetic mean across all cells. The second indicator is the peak occurrence time, calculated by finding the time position of the maximum migration rate in each cell within the 300-second sequence and taking the arithmetic mean across all cells to obtain the average peak time. The third indicator is the steady-state fluctuation amplitude, calculated by subtracting the minimum migration rate from the maximum migration rate of each cell within the interval from the 241st second to the 300th second to obtain the cell fluctuation amplitude, and then taking the arithmetic mean across all cells. These three indicators together constitute the dynamic control parameters.

[0043] Ion distribution characteristics are classified and their applicability is determined based on dynamically adjusted parameters. The classification algorithm is a support vector machine algorithm with a radial basis function kernel, a penalty coefficient of 100, and a kernel width of 0.5. The training set consists of 200 historical model samples, including 100 applicable samples and 100 inapplicable samples. After training, the model parameters are fixed and not changed. During inference, the input features are the average growth rate, the average peak time, and the steady-state fluctuation amplitude. The output is applicable or inapplicable, along with a confidence score. The applicability determination condition is that the ion is applicable and the confidence score is not less than 0.90.

[0044] If the applicability criteria are met, the pore structure model is updated using numerical simulation to obtain the final dynamic control scheme. The update rule is to lower the ion migration rate of units with electric field strength higher than the optimized average value by multiplying the current value by 0.95, and to raise the ion migration rate of units with electric field strength lower than the optimized average value by multiplying the current value by 1.05. After one update, the transient solution is run again for 300 seconds, and the ion migration rate of all units per second is recorded. The final dynamic control scheme is the combination of the electric field strength field and the ion migration rate field obtained from this solution.

[0045] Finally, the stability of ion migration rate was analyzed using simulation verification methods, and the control results under the action of electric field were determined. The stability test criterion was fixed as follows: the ion migration rate of all units was spatially averaged at each time point in the interval from 241 seconds to 300 seconds to obtain a sequence, and then the arithmetic mean and standard deviation of the sequence in this interval were calculated. If the standard deviation was not greater than 0.10 of the arithmetic mean, it was considered stable; otherwise, the iterative optimization step was returned, and optimization and verification loops were executed no more than 10 times based on the current equivalent conduction weight. If the stability condition was still not met after 10 iterations, an inapplicable conclusion was output, and all intermediate data were retained.

[0046] The 0.1 mm side length of the grid cells is derived from the minimum identification requirement for the pore scale, ensuring at least 500 discrete points within a 5 mm diameter range. The 1-second time step and 300-second computation time are derived from the consistency requirement of the sampling and evaluation window in claim 2. The relaxation factor and convergence threshold of the successive over-relaxation method are determined by statistically analyzing the convergence speed and residuals of 10 sets of pre-test data. A relaxation factor of 1.5 results in an average of less than 800 iterations and a residual of less than 0.0001, and this is written as a fixed configuration. The electric field modulation coefficient threshold is obtained by multiplying the arithmetic mean of 30 historical samples by 1.20, ensuring that the false alarm rate in previous samples is no higher than 0.05. The high-field and low-field separation thresholds in iterative optimization use a fixed rule of the average value plus or minus 0.5 times the standard deviation. The upper and lower limits of the weight adjustment ratio, 0.02 and 0.20, are used to limit the amount of change in one iteration to avoid numerical divergence, and a 10% reduction in standard deviation is used to determine whether an adjustment is effective. The definitions and value ranges of the three dynamic control parameters are directly calculated from transient simulation data, without involving empirical coefficients. The kernel type, penalty coefficient, and kernel width of the support vector machine algorithm are determined and fixed before leaving the factory through gridded search and five-fold cross-validation, and remain unchanged during the inference stage. An applicability confidence threshold of 0.90 is used to ensure that the false positive rate is controlled within 0.10. The final stability judgment uses the last 60-second interval and uses a standard deviation no greater than 0.10 of the mean as the sole criterion.

[0047] The voltage optimization module acquires the voltage adjustment sequence from the dynamic control scheme, optimizes the local electric field intensity in the concentration change region, and if the optimized current change exceeds a preset threshold, it re-simulates the pore structure to obtain distribution data, including constructing a pore structure model using the finite element method, generating an initial voltage adjustment sequence, and acquiring electric field intensity distribution data in the concentration abrupt change region. If the electric field intensity distribution data exceeds the preset threshold, the gradient descent algorithm is used to adjust the voltage sequence to obtain the optimized voltage sequence. Based on the optimized voltage sequence, the current change data is calculated using numerical simulation to obtain the current change trend. If the current change trend exceeds the preset threshold, the pore structure model is updated using an iterative optimization method to obtain the adjusted structure model. Based on the adjusted structure model, the electric field intensity distribution data is recalculated to obtain the stable electric field distribution. The stable electric field distribution is processed using time series analysis to extract dynamic voltage adjustment parameters, obtaining a control parameter set. Based on the control parameter set, the support vector machine algorithm is used to classify the electric field distribution in the concentration abrupt change region, determine the stability of the control parameters, and obtain the final control scheme.

[0048] In one possible implementation, the voltage optimization module takes the dynamic control scheme as external input. First, it constructs a three-dimensional cylindrical porous structure model using the finite element method. The model has a sample length of 10 cm and a diameter of 5 cm. The mesh uses uniformly sized cubic elements with a side length of 0.1 mm, and the total number of elements is no less than 1,000,000. The inlet face is set as the terminal voltage boundary, the outlet face as the 0-volt boundary, and the sidewalls as the insulation boundary. The time step is 1 second, and the total simulation time is 300 seconds. The initial voltage adjustment sequence given by the dynamic control scheme is directly used as the boundary input. This sequence consists of 300 terminal voltage target values, corresponding to values ​​taken second by second from the 1st to the 300th second. The voltage optimization module calls the finite element solver second by second to obtain and store the full-field potential distribution and electric field intensity distribution per second.

[0049] Based on this, concentration abrupt change regions are identified. The identification method is to calculate the difference in ion concentration between the 60th and 61st seconds for each unit, divide it by 1 second to obtain the instantaneous growth rate, and calculate the arithmetic mean of the instantaneous growth rates of all units in the entire model to obtain the average growth rate of the entire model. Twice this average value is used as a threshold; units with instantaneous growth rates greater than this threshold are defined as concentration abrupt change units and classified into concentration abrupt change regions. Subsequently, the electric field intensity per second is extracted within this region, and the spatial mean and spatial standard deviation are calculated. The electric field intensity distribution threshold is 1.5 volts per centimeter. This threshold is determined before equipment delivery using a fixed method based on the electric field intensity data of concentration abrupt change regions from 20 sets of historical stable samples. The method involves calculating the spatial mean and spatial standard deviation for each set of data, adding 3 times the spatial standard deviation to the spatial mean, then calculating the arithmetic mean of the 20 sets of results and rounding it to 0.1 volts per centimeter. The calculated 1.5 volts per centimeter is then written into the system configuration and remains unchanged throughout its lifecycle.

[0050] If at any time point there is a cell with an electric field strength greater than 1.5 V / cm in the concentration abrupt change region, the gradient descent algorithm is initiated to adjust the voltage sequence. The goal is to bring the spatial average electric field strength in the concentration abrupt change region closer to 1.2 V / cm and stabilize it throughout the entire process. The objective function is the arithmetic mean over time of the square of the difference between the spatial average electric field strength in each second from the 1st to the 300th second and 1.2 V / cm. The learning rate is 0.05 V, the maximum number of iterations is 200, and the voltage value range is limited to 0 V to 10 V. Each iteration is implemented by numerically perturbing and calculating the gradient for each time step from second 1 to second 300. Specifically, the terminal voltage is increased by 0.1 volts at that time step, while keeping the voltages constant at other time steps. The single-step sensitivity simulation under that perturbation is quickly solved, and the new value of the objective function is calculated. Then, the terminal voltage is decreased by 0.1 volts at that time step, and the above calculation is repeated to obtain the upper and lower perturbation values ​​of the objective function at that time step. The upper perturbation value minus the lower perturbation value and divided by 0.2 volts yields the gradient estimate of the objective function at that time step. Subsequently, the terminal voltage at that time step is updated by subtracting the learning rate from the original value and multiplying by the gradient estimate. If the updated voltage exceeds 10 volts, it is truncated to 10 volts; if it is below 0 volts, it is truncated to 0 volts. After updating all time steps sequentially, a new voltage sequence is formed, and a complete transient solution is performed again to obtain a new objective function value. If the new objective function value is less than 0.001, the iteration stops immediately and the optimized voltage sequence is output; otherwise, the next iteration begins, until the stopping condition is met or the upper limit of 200 iterations is reached. After obtaining the optimized voltage sequence, a complete numerical simulation is performed to calculate the current change data. The calculation method is to extract the total current at the inlet face every second and form a time series of 300 values. The average rate of change is obtained by subtracting the current of the first second from the current of the 300th second and dividing by 299 seconds. At the same time, the maximum change is obtained by subtracting the minimum current from the maximum current in the sequence. The current change trend threshold adopts a two-condition parallel criterion. The average change rate threshold is 0.002 amps per second, and the maximum change threshold is 0.5 amps. These two thresholds were determined by statistical analysis of 30 historical samples before equipment delivery. The method is to calculate the average change rate and the maximum change for each sample, take their respective medians, and then multiply them by 1.25 to obtain fixed thresholds. Finally, they are rounded to three and one decimal places respectively to obtain 0.002 amps per second and 0.5 amps, and then fixed in the system.

[0051] If the average rate of change in this calculation is greater than 0.002 A / s or the maximum change is greater than 0.5 A, then the iterative optimization method is executed to update the pore structure model. The update rule only changes the equivalent conduction weight of the elements without changing the boundary conditions and mesh. Specifically, the average current density of each element in the concentration abrupt change region between 241 seconds and 300 seconds is used as the sorting criterion. The top 20% of elements are selected as high-current path elements, and their equivalent conduction weight is reduced by multiplying the current value by 0.95. The bottom 20% of elements are selected as low-current path elements, and their equivalent conduction weight is increased by multiplying the current value by 1.05. The weights of the remaining elements remain unchanged. After completing one weight update, the transient simulation is rerun, and the spatial standard deviation of the current density in the concentration abrupt change region is calculated. If the standard deviation is reduced by no less than 10% compared to the previous calculation result, the above weight adjustment ratio is kept unchanged and the next round is entered. Otherwise, the adjustment ratio of both types of elements is halved, but not lower than 0.02 and not higher than 0.20, and the iteration continues. Stop when the spatial standard deviation drops to 50% of the initial value or the number of iterations reaches 20, and the adjusted structural model is obtained.

[0052] The electric field intensity distribution data is recalculated on the adjusted structural model, and stability is determined. The determination method involves calculating the spatial mean and spatial standard deviation of the electric field intensity in the concentration abrupt change region second by second within the interval from 241 seconds to 300 seconds, forming two sequences at 60 time points. Then, the ratio of the spatial standard deviation to the spatial mean is calculated for each time point, and the maximum ratio is taken at the 60 time points. If this maximum ratio is not greater than 0.10, it is defined as stable, and the electric field intensity distribution obtained in this iteration is recorded as the stable electric field distribution. Otherwise, the weight update process is returned to the previous step, and the optimization and verification loop is executed for a maximum of 10 iterations. If stability is still not achieved after 10 iterations, the record that does not meet stability is output, and the process ends. After obtaining the stable electric field distribution, time series analysis is performed to extract dynamic voltage adjustment parameters and generate a control parameter set. The extraction method involves calculating the spatial average electric field intensity of the concentration abrupt change region second by second from the 1st second to the 300th second, obtaining a sequence of 300 values. Three parameters are directly calculated from this sequence and arranged in order to form a set of control parameters. The three parameters are the maximum spatial average electric field intensity throughout the process, the time point corresponding to the maximum value, and the average rise rate in the last 60 seconds. The average rise rate in the last 60 seconds is the spatial average electric field intensity of the 300th second minus the spatial average electric field intensity of the 241st second, divided by 59 seconds.

[0053] The set of control parameters is input into a support vector machine (SVM) algorithm for classification to determine the stability of the control parameters. The kernel function of the algorithm is a radial basis function with a penalty coefficient of 100 and a kernel width of 0.5. The classifier is trained and its parameters are frozen using 200 labeled historical samples before equipment delivery, with 100 stable samples and 100 unstable samples each accounting for half. In the inference phase, the output is either stable or unstable, along with a confidence level between 0 and 1. Stability is determined when the output is stable and the confidence level is not less than 0.90. The stable electric field distribution and the optimized voltage sequence are used together as the final control scheme output.

[0054] The permeation path calculation module extracts synchronous monitoring indicators from the distribution data and uses a diffusion equation algorithm to calculate the ion permeation path, obtaining a preliminary permeability coefficient estimate. This includes extracting ion concentration gradient features from the distribution data and generating concentration distribution features using data extraction methods; calculating the ion migration path using the diffusion equation algorithm based on the concentration distribution features to obtain migration path features; if the migration path features do not match a preset threshold, adjusting the path optimization parameters through an iterative optimization method to obtain an optimized migration path; calculating the ion diffusion rate based on the optimized migration path to obtain the diffusion rate distribution; extracting the permeability coefficient estimate from the diffusion rate distribution and classifying the permeability coefficient distribution using a support vector machine algorithm to determine distribution stability; if the distribution stability analysis result does not reach a preset threshold, recalculating the ion migration path through numerical simulation to obtain an updated permeability coefficient distribution; and extracting the final path optimization parameters based on the updated permeability coefficient distribution to obtain a stable permeability coefficient estimate.

[0055] In one possible implementation, the permeation path calculation module takes distribution data as input and executes in a fixed order without optional branches. First, it performs deterministic extraction of synchronous monitoring indicators from the distribution data. The distribution data includes the ion concentration value of each unit at each time point, with a fixed spatial step size of 0.1 mm and a fixed time step size of 1 second, for a total of 300 steps. At each time point, the absolute value of the concentration difference between each unit and its axially adjacent units is divided by 0.1 mm to obtain the concentration gradient value of that unit. Then, at that time point, the concentration gradient values ​​of all units are normalized based on the minimum and maximum values. Specifically, each gradient value is subtracted from the minimum gradient value at that time point, and then divided by the difference between the maximum and minimum gradient values, so that the normalization result is between 0 and 1.

[0056] To avoid the influence of sporadic noise, a median filter of length 5 was applied to the normalized gradient sequence of each unit on the time axis, followed by a moving average smoothing of length 7. The window length was determined based on the statistics of 20 sets of pre-test data, where a length of 5 achieved a peak removal rate of 90% without weakening the true edges, and a length of 7 yielded the best smoothness score. Subsequently, at each time point, the spatial arithmetic mean of the normalized gradients of all units was taken and recorded as a concentration distribution feature sequence. This sequence, with a length of 300 data points, served as the sole driving force for the diffusion equation algorithm. Next, the diffusion equation algorithm was used to calculate ion migration paths under a fixed grid and fixed time steps. The initial diffusion coefficient was set to 0.000000002 m² / s, a value fixed before equipment delivery based on the principle of minimizing fitting error from 30 sets of historical samples. The specific process of path calculation involved calculating the migration flux between adjacent units at each time step based on the concentration difference and spatial step size, and then continuously connecting adjacent units with the same flux direction to form candidate links. After removing discontinuous segments with continuous flux below 0.01 mmol / L / s from the candidate links, an effective migration path set is obtained. For each path, the starting position, ending position, total path length, average path gradient, and average path flux are calculated to form a migration path feature set.

[0057] The migration path characteristics are then matched against preset thresholds. Threshold 1 is that the difference between the median total path length and the target length is no greater than 0.5 cm. The target length is 5 cm, determined by using half the sample length from 50 historical stable samples as the effective penetration representative length, and this value yields the minimum bias in regression testing. Threshold 2 is that the median average flux of the path is no less than 10 mmol / L / s. This threshold is determined by multiplying the median of the 50 historical stable samples by 1.25 and rounding to 0.1. If the migration path characteristics at any time point do not meet any of the above thresholds, the iterative optimization step begins. The path optimization parameters include a time step adjustment factor and a flux weighting factor, both initially set to 1. The iterative optimization is specifically performed by enumerating all combinations of the two parameters in a grid scan within the range of 0.8 to 1.2 with a step size of 0.05. For each set of parameters, the migration path is recalculated and scored according to the objective function. The objective function is defined as follows: the closer the total path length is to 5 cm, the higher the score; the higher the average path throughput, the higher the score. The scoring interval is normalized to 0 to 1. The maximum number of iterations is 20. If the highest score of two adjacent iterations increases by less than 0.01, the iteration stops and the parameter set with the highest score is used as the optimized path optimization parameters and the corresponding migration path set.

[0058] After obtaining the optimized migration paths, the ion diffusion rate distribution is calculated. The calculation method involves dividing the average flux of each path by its total length to obtain the diffusion rate. Then, the diffusion rate of each path is mapped back to the cell level according to the cells it covers. For overlapping paths, the arithmetic mean is taken within the same cell to form the diffusion rate distribution field of the entire model. Subsequently, a permeability coefficient estimation sequence is calculated from the diffusion rate distribution. The calculation method involves taking the spatial arithmetic mean of the diffusion rates of all cells at each time point to obtain the estimated permeability coefficient value for that time point. This results in a permeability coefficient distribution sequence of length 300, which serves as the sole numerical basis for stability determination. The stability of the permeability coefficient distribution is first classified using a support vector machine (SVM) algorithm. The algorithm uses a radial basis function kernel with a fixed penalty coefficient of 100 and a fixed kernel width of 0.5. The classifier is trained with 200 historical distribution sequences and its parameters are frozen before device delivery, with 100 stable and 100 unstable sequences. During the inference phase, the output is either stable or unstable, along with a confidence level between 0 and 1. The stability threshold is a confidence level of not less than 0.9 and a classification result of stable. The confidence threshold is determined and fixed by the point corresponding to the receiver operating curve of the validation set when the false positive rate is controlled at 0.1.

[0059] If the stability determination at any time point fails to meet the aforementioned threshold, the numerical simulation method is invoked to recalculate the ion migration path. During recalculation, the spatial and time steps remain unchanged, and the path optimization parameters are set to the optimal values ​​from the previous round as new initial values. This recalculation is repeated only once to ensure process closure and obtain an updated permeability coefficient distribution sequence. After obtaining a permeability coefficient distribution that meets the stability threshold, the final path optimization parameters are extracted, and the stable permeability coefficient estimate is output. The final path optimization parameters are defined as the time step adjustment factor and flux weighting factor with the highest score in the last iteration. The stable permeability coefficient estimate is defined as the arithmetic mean of the permeability coefficient distribution sequence within the last 60-second interval. This is determined by averaging 60 values ​​from the 241st to the 300th second, and the standard deviation of this interval is recorded for future reference. If the standard deviation is greater than 0.1 of the mean, the diffusion rate and permeability coefficient calculations are repeated once without changing any thresholds or parameters, and the second result is used as the standard, without performing a third calculation.

[0060] The coupling analysis module analyzes the coupling between the initial permeability coefficient value and the real-time current change signal. If the coupling deviation is less than a preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution. This includes extracting dynamic features from the real-time current change signal, generating a current fluctuation sequence using Fourier transform, and obtaining frequency distribution features. Based on the frequency distribution features, the coupling deviation between the initial permeability coefficient estimate and the current fluctuation sequence is calculated to obtain the deviation distribution features. If the deviation distribution features are less than a preset threshold, the barrier performance evaluation model is determined using a logistic regression algorithm, and the model evaluation result is obtained. If the deviation distribution features are greater than a preset threshold, the concentration distribution parameters are adjusted using a gradient descent method to obtain an optimized concentration distribution. Based on the optimized concentration distribution, the dynamic features of the real-time current change signal are re-extracted to generate an updated current fluctuation sequence. Through coupling analysis between the updated current fluctuation sequence and the initial permeability coefficient estimate, it is determined whether the preset threshold is met, resulting in the final barrier performance evaluation model. Based on the final barrier performance evaluation model, key performance parameters are extracted to generate the corrected concentration distribution features.

[0061] In one possible implementation, the coupling analysis module takes a preliminary permeability coefficient estimate sequence and synchronously acquired real-time current change signals as input. First, it performs deterministic segmentation and frequency domain feature extraction on the real-time current change signals. The current signal sampling frequency is fixed at 100 Hz, the total duration is 300 seconds, and the total number of samples is 30,000. The time-domain signal is divided into 300 non-overlapping windows of 1 second length, each containing 100 samples. The signal within each window is first subtracted by the arithmetic mean of that window to remove the DC component, and then a fast Fourier transform is performed to obtain the signal from 0 Hz. The amplitude spectrum is obtained up to 50 Hz. Then, the amplitude spectrum is normalized by dividing the amplitude of each frequency component by the sum of the amplitudes of all frequency components within the window, so that the sum of all normalized amplitudes in the window equals 1. The energy proportion of three fixed frequency bands is calculated as the frequency distribution characteristics of the window, where the low frequency band is 0 to 5 Hz, the mid frequency band is 5 to 20 Hz, and the high frequency band is 20 to 50 Hz. At the same time, the dominant frequency, i.e. the frequency with the largest normalized amplitude, is determined within the window and the dominant frequency amplitude and the corresponding frequency band are recorded. Thus, a frequency distribution characteristic sequence and a dominant frequency amplitude sequence of length 300 are obtained.

[0062] Next, the preliminary permeability coefficient estimation sequence is aligned with the above sequence one value per second in time. Minimum and maximum values ​​are normalized for both sequences to ensure their values ​​fall between 0 and 1. Then, at each second, the absolute value of the difference between the normalized preliminary permeability coefficient and the dominant frequency amplitude is calculated and used as the coupling deviation for that time point. The arithmetic mean, median, 90th percentile, and maximum value of the coupling deviations at 300 time points are then calculated to form the deviation distribution characteristics. The benchmark deviation used for threshold determination is the arithmetic mean. The preset threshold is fixed at 0.05, which is determined by 50 historical samples before equipment delivery. The mean and standard deviation of the coupling deviation are calculated as follows: for each group of samples, the mean of the coupling deviation is calculated, the arithmetic mean of all sample means is taken, and three times the inter-sample standard deviation of that mean is added. This is then rounded to 0.01 to obtain 0.05, which is then written into the system configuration and not adjusted during a single run. If the baseline deviation in the deviation distribution characteristics is not greater than 0.05, the process directly proceeds to the barrier performance evaluation model confirmation step. The model type is logistic regression, and the input features are the arithmetic mean of the normalized permeability coefficient values ​​over the past 60 seconds, the arithmetic mean of the main frequency amplitude values ​​over the past 60 seconds, the low-frequency band energy proportion over the past 60 seconds, and the low-frequency band energy proportion over the past 60 seconds. The ratio of the frequency band energy percentage to the high-frequency band energy percentage, and the proportion of the dominant frequency occurring in the low-frequency band over the past 60 seconds, are used to determine whether the barrier performance meets the preset standards, along with the probability of this result. The model coefficients are trained and frozen using 200 sets of tagged historical data before equipment delivery. During runtime, only a single forward calculation is performed to obtain the model evaluation result, which is then used as the conclusion for this round of barrier performance evaluation. If the baseline deviation is greater than 0.05, gradient descent optimization of the concentration distribution parameters is performed to reduce coupling bias. The concentration distribution parameters are defined as the set of concentration values ​​at spatial and temporal discrete points, with a fixed spatial point spacing of 0.1 mm and a fixed time step... The time interval is fixed at 1 second and is completely consistent with the input data. The objective function is the arithmetic mean of the coupling deviations at 300 time points. The learning rate is fixed at 0.001, and the maximum number of iterations is fixed at 500. The iterative update is implemented by numerical difference gradient calculation without introducing other modules. That is, the parameter set is divided into groups of 1000 consecutive points in spatial order and processed one group at a time. For each discrete point being processed, the upper perturbation deviation is obtained by increasing the perturbation by 0.001 mmol / L and recalculating the baseline deviation. The lower perturbation deviation is obtained by decreasing the perturbation by 0.001 mmol / L and recalculating the baseline deviation. The difference between the upper and lower perturbation deviations is divided by 0.A gradient estimate of 0.002 mmol / L is obtained for each point. The updated value is obtained by subtracting the product of the learning rate and the gradient estimate from the current concentration value at that point. This updated value is then constrained within a physically feasible range: it must not be less than 0 mmol / L and must not exceed the maximum allowable value of the input distribution data at that point. After updating a set of 1000 points, the entire coupling bias baseline value is immediately recalculated using the updated concentration distribution as input for the next set. This process continues until all blocks are processed, marking one iteration as complete. At the end of each iteration, three stopping conditions are checked: whether the baseline bias is no greater than 0.05, whether the absolute change in the baseline bias over the last 20 iterations is less than 0.0001, and whether the number of iterations has reached 500. If any one of these conditions is met, optimization stops, and the latest concentration distribution is used as the optimized concentration distribution.

[0063] After obtaining the optimized concentration distribution, the initial frequency domain feature extraction process is strictly repeated. The same real-time current change signal is regenerated using the exact same sampling frequency, window length, and normalization rules to generate updated current fluctuation sequences and main frequency amplitude sequences. These are then realigned with the initial permeability coefficient estimation sequence, and the coupling deviation distribution features are recalculated using the same method. If the baseline deviation is no greater than 0.05, a one-time forward calculation is performed on the updated features using a logistic regression model with frozen parameters to obtain the final barrier performance evaluation model output. Otherwise, without changing any threshold, learning rate, or step size, a combination of gradient descent optimization, feature re-extraction, and deviation re-judgment is executed again, and this process is only allowed to be repeated once to ensure process closure. If the baseline deviation is still greater than 0.05 after re-judgment, records that do not meet the threshold are output, along with the reason for failure and the maximum deviation for the corresponding time period. Iteration is then discontinued. After obtaining the final barrier performance evaluation model, key performance parameters are extracted, and corrected concentration distribution features are generated. These key performance parameters include the barrier performance judgment result, judgment probability, and the arithmetic square of the permeability coefficient estimates over the past 300 seconds. The average, the 90th percentile of the estimated permeability coefficient over the past 300 seconds, the arithmetic mean of the low-frequency energy proportion over the past 300 seconds, the arithmetic mean of the coupling bias over the past 300 seconds, and the maximum value of these parameters are all rounded to three decimal places and used together with the optimized concentration distribution to form the corrected concentration distribution feature. The method for generating the corrected concentration distribution feature is to add three model-derived fields to the optimized concentration field at each second-level time point according to spatial location: the dominant frequency amplitude at that time point, the low-frequency energy proportion at that time point, and the coupling bias at that time point. This process is repeated for each... The linear correction coefficient for spatial points is calculated as the product of the model decision probability and the proportion of low-frequency energy at that time point. This coefficient is multiplied by the concentration change between the current time point and the adjacent previous time point as the correction amount, and added to the original concentration value to obtain the corrected concentration value constrained by the model. If the corrected concentration value exceeds the maximum allowable value for that point in the entire time period, it is truncated to the maximum allowable value; if it is lower than 0, it is truncated to 0. After processing all spatial points and all time points, the corrected concentration field at each time point is normalized to the minimum and maximum values ​​to form the final corrected concentration distribution feature sequence.

[0064] The iterative calculation module, based on the corrected concentration distribution and incorporating migration rate parameters, performs iterative calculations to determine the permeability coefficient under electric field assistance. This includes extracting migration rate parameters from the corrected concentration distribution and generating a preliminary permeability coefficient estimate using numerical integration; calculating the migration rate distribution under the electric field using the finite difference method based on the preliminary permeability coefficient estimate and the electric field-assisted parameters; adjusting the electric field-assisted parameters using gradient descent to obtain an optimized migration rate distribution; re-extracting dynamic features based on the optimized migration rate distribution to generate an updated permeability coefficient estimate; constructing a permeability coefficient model using logistic regression based on the updated permeability coefficient estimate to obtain the model output; obtaining the model output result if the deviation of the model output result is less than the preset threshold; acquiring key performance parameters using signal extraction methods to determine the final effective permeability coefficient; and generating corrected concentration distribution features based on the final effective permeability coefficient, outputting optimized distribution data.

[0065] In one possible implementation, the iterative calculation module takes the corrected concentration distribution output by claim 6 as its sole input. First, it discretizes the sample along the sample axis in space with a step size of 0.1 mm and in time in a step size of 1 second for the entire 300-second period. For each spatial node, it calculates the concentration change between two adjacent seconds and divides it by 1 second to obtain the concentration change rate of that node in that second. At the same time, it divides the concentration difference between two adjacent spatial nodes by 0.1 mm to obtain the spatial gradient of that node in that second. To avoid instability caused by an excessively small denominator, when the absolute value of the spatial gradient is less than 0.0001 mmol / L / mm, it is fixed at 0.0001 mmol / L / mm. Then, it divides the concentration change rate by the spatial gradient to obtain the local migration rate value under a unit electric field, and takes the arithmetic mean of this value for the same node within 300 seconds to obtain the migration rate parameter of that node.

[0066] After obtaining the migration rate parameters of all nodes, numerical integration is performed. The migration rate parameters of each node are summed in a trapezoidal pattern along the axis and multiplied by 0.1 mm to obtain the integral sum. The integral sum is then divided by the sample length of 10 cm to obtain a preliminary estimate of the permeability coefficient, which is recorded as the initial value for this round. Subsequently, the preliminary permeability coefficient estimate and the electric field auxiliary parameters jointly drive the finite difference calculation. The electric field auxiliary parameters are the electric field strength obtained by dividing the terminal voltage by the sample length. The terminal voltage is given by the previous module and remains unchanged in this round of calculation. In the difference calculation, a first-order forward propagation is used in the time direction and a second-order central difference is used in the spatial direction. The following update steps are performed second by second: At each node, the flux generated by diffusion is calculated. This flux equals the ratio of the concentration difference between adjacent nodes to the spatial step size multiplied by a preliminary estimate of the permeability coefficient. Simultaneously, the flux generated by electromigration is calculated. This flux equals the migration rate parameter of that node multiplied by the electric field strength multiplied by the arithmetic mean of the concentrations of adjacent nodes. The sum of these two fluxes is taken as the net flux of that node. The net flux is multiplied by the time step of 1 second to obtain the increment of the node's concentration, which is then added to the concentration value for the next second. This yields the migration rate distribution of all nodes over the entire time period. After completing one full-time advance, the deviation of the migration rate distribution is calculated. The difference is defined as follows: at each time point, the migration rate of all nodes is first calculated as a spatial arithmetic mean, then the absolute deviation of each node is calculated and its spatial arithmetic mean is taken. This value is then averaged again over 300 time points as the deviation metric for this round. The preset deviation threshold is fixed at 0.05. This threshold is calculated by adding three times the inter-sample standard deviation to the arithmetic mean of the above metrics for 50 historical samples, rounded to 0.01, and then fixed. If the deviation for this round is not greater than 0.05, parameter optimization is skipped and the next step is performed; otherwise, the gradient descent method is used to adjust only the electric field auxiliary parameters to reduce the deviation. Specifically, while keeping other quantities constant... Under the premise of keeping the electric field strength unchanged, a symmetrical perturbation is applied to increase the electric field strength by 0.1 V / cm and decrease it by 0.1 V / cm, and a complete finite difference propagation is performed once, and the deviation is calculated separately. The difference between the two deviations is divided by 0.2 V / cm to obtain the gradient estimate of the deviation with respect to the electric field strength. Then, the current electric field strength is subtracted from the learning rate and multiplied by the gradient estimate to obtain the updated value. The learning rate is fixed at 0.001. The updated electric field strength is limited to between 0.1 V / cm and 3.0 V / cm. After completing one update, a complete finite difference propagation is immediately performed again and the deviation is recalculated. This process is repeated until the deviation is no greater than 0.05. The iteration continues until 500 iterations are reached. If the threshold is not reached after 500 iterations, the electric field strength corresponding to the smallest deviation after 500 iterations is taken as the optimal electric field strength for this round, and the next step is initiated. After obtaining the optimized migration rate distribution, dynamic features are re-extracted and the updated permeability coefficient estimate is calculated. The dynamic features are fixed to three items: the average migration rate in the initial 60 seconds, the average time of the peak migration rate within a 300-second interval, and the steady-state fluctuation amplitude in the last 60 seconds. The steady-state fluctuation amplitude is the spatial arithmetic mean of the maximum and minimum migration rates within the last 60-second interval. The updated permeability coefficient estimate is calculated over the entire length using the same numerical integration method as the initial value. Subsequently, a permeability coefficient model is constructed based on the updated permeability coefficient estimate. The model type is logistic regression, and the input features are the average permeability coefficient of the past 60 seconds, the variance of the permeability coefficient of the past 60 seconds, and the permeability coefficient of the past 60 seconds. The algorithm retrieves the maximum and minimum penetration coefficient values ​​over 60 seconds, along with the values ​​of the three dynamic features mentioned above. The output is a binary determination of whether the model is valid or invalid, along with its probability. The model coefficients are trained and frozen with 200 labeled samples before equipment delivery and are not updated during runtime. The deviation of the model output is calculated, defined as the arithmetic mean of the absolute differences between the model probability and its corresponding binary label at all time points. The preset deviation threshold is fixed at 0.05, taken from the minimum probability interval on the validation set where the misclassification rate is controlled at 0.10, rounded to 0.01, and then fixed. If the deviation of the model output is not greater than 0.05, the algorithm proceeds to determine key performance parameters and output the final valid penetration coefficient. Otherwise, without changing any thresholds or learning rate, it performs a recalculation step starting from the optimal electric field strength of this round, involving only finite difference propagation and logistic regression evaluation, and uses the recalculated deviation as the standard. A third recalculation is not performed.

[0067] Key performance parameters were obtained through signal extraction methods. The following parameters were fixed: the overall arithmetic mean, overall standard deviation, 90th percentile, arithmetic mean and standard deviation of the last 60 seconds, and the percentage of time within the entire process that met the stability condition. The stability condition was that the standard deviation of the permeability coefficient within any consecutive 10-second window was no greater than 0.10 of the average value within that window. All values ​​were rounded to three decimal places. The final effective permeability coefficient was determined by taking the arithmetic mean of the estimated permeability coefficient values ​​for the last 60 seconds, provided that the model bias threshold was met. Furthermore, the effective permeability coefficient was determined when the standard deviation of the last 60 seconds was greater than the average value. At a value of 0.10, a robust average is performed using a weighted method, with the weight being the reciprocal of the standard deviation per second, after normalization. The result is the final effective permeability coefficient. After obtaining the final effective permeability coefficient, the corrected concentration distribution characteristics are generated and the optimized distribution data is output. The generation rule is to multiply the overall concentration field at each time point by the ratio of the final effective permeability coefficient to the updated permeability coefficient estimate at that time point to ensure consistency between the two. Subsequently, the corrected concentration field at each time point is normalized to the minimum and maximum values ​​to eliminate dimensional differences and form a time-ordered sequence of corrected concentration distribution characteristics.

[0068] The performance evaluation module acquires the permeability coefficient and quantitatively evaluates the overall barrier performance of recycled concrete. If the evaluation results meet the preset standards, a final monitoring report is output; otherwise, the system loops back to the dynamic control scheme to obtain optimized iterative data. This includes acquiring the barrier performance data of recycled concrete, analyzing the material composition distribution using X-ray fluorescence spectroscopy to obtain composition distribution characteristics; if the deviation of the composition distribution characteristics is greater than a preset threshold, adjusting the detection parameters of X-ray fluorescence spectroscopy using a particle swarm optimization algorithm to obtain optimized composition distribution characteristics; based on the optimized composition distribution characteristics, calculating the preliminary evaluation value of barrier performance using finite element analysis to obtain preliminary barrier performance data; if the deviation of the preliminary barrier performance data is greater than a preset threshold, classifying the preliminary barrier performance data using a support vector machine algorithm to obtain classified performance characteristics; based on the classified performance characteristics, processing the performance characteristics using a data standardization method to obtain standardized performance data; if the deviation of the standardized performance data is less than a preset threshold, extracting key performance parameters using signal processing methods to determine the final barrier performance data; and based on the final barrier performance data, adjusting the analysis parameters of the material composition distribution using a dynamic control scheme to obtain optimized material distribution data.

[0069] In one possible implementation, the performance evaluation module uses the permeability coefficient as an external input. In this implementation, the sample geometry is 10 cm in length and 5 cm in diameter; the evaluation time window is fixed at 300 seconds; the spatial step is fixed at 1 mm; and the time step is fixed at 1 second. First, the barrier performance data of the recycled concrete is acquired using fixed-energy X-ray fluorescence spectrometry, scanning the entire cross-section and length of the sample point by point. The X-ray source energy is fixed at 15 kEV, the initial beam diameter is 1 mm, the initial scan step is 1 mm, the detector energy resolution is better than 150 kEV, and the initial single-point integration time is 2 seconds. During acquisition, dark count subtraction and background subtraction are performed at each scan point. Then, the counts of each element are integrated within a pre-calibrated characteristic energy region. The integrated counts are converted into content percentages through two-point calibration. Calibration uses a zero-content standard sample and a known-content standard sample. The conversion coefficient is fixed before equipment delivery. The resulting element content arranged by spatial location is the material composition distribution data. The arithmetic mean of the content values ​​of each element at all scan points is used as the global mean for that element, and the absolute value of the difference between each scan point and the global mean is calculated. The arithmetic mean of the absolute differences of all scan points is used to obtain the composition distribution deviation of that element. The arithmetic mean of the composition distribution deviations of all elements is then used to obtain the deviation measure of composition distribution characteristics, expressed as a percentage. The composition distribution characteristic deviation threshold is fixed at 5.00. This threshold is calculated by adding three times the inter-sample standard deviation to the arithmetic mean of the composition distribution characteristic deviations of 50 historical samples that meet the design barrier performance requirements under the same process conditions, and rounding to 0.01 before being fixed. This threshold does not change during a single run. If the composition distribution characteristic deviation calculated in this run is not greater than 5.00, the process directly proceeds to the step of calculating the barrier performance using finite element analysis; if it is greater than 5.00, parameter optimization is performed, and the data is collected again and the calculation is repeated until the requirement is met or the iteration limit is reached.

[0070] When the component distribution characteristic deviation exceeds 5.00, the performance evaluation module initiates a particle swarm optimization algorithm to adjust the detection parameters of X-ray fluorescence spectrometry. The optimization objective is to ensure that the component distribution characteristic deviation does not exceed 5.00. The optimized parameters include the scan step size, integration time, and beam spot diameter, with values ​​ranging from 0.5 mm to 2.0 mm, 1 s to 5 s, and 0.5 mm to 2.0 mm, respectively. During initialization, 20 sets of parameters are uniformly and randomly generated within their respective ranges as initial values ​​for the population. The maximum number of iterations is fixed at 50. A fixed weight strategy is used for velocity and position updates, with an inertia weight of 0.7, an individual learning factor of 1.5, and a population learning factor of 1.5. The velocity upper limit is 20% of the length of each parameter's value range, and a truncation and write-back strategy is used when the position exceeds the limit. In each iteration, a rapid sampling is performed on each of the 20 sets of parameters. This is done by sampling one point every five points on the original scan grid for each candidate parameter set, and then filling in the unsampled points with linear interpolation. This reduces optimization overhead while maintaining trend consistency. Based on this rapid sampling data, the component distribution characteristic deviation is calculated using the aforementioned method, and its negative value is used as the fitness. After velocity and position updates are completed, the globally optimal parameter set is applied to the entire grid to complete a full acquisition, and the deviation is recalculated using the complete data as the recorded value for this iteration. Optimization terminates when the recorded value is no greater than 5.00 or when 50 iterations are reached, and the optimal detection parameters are output. If the deviation is still greater than 5.00 at termination, the parameter set corresponding to the smallest recorded value is used as the optimized detection parameters, and the process proceeds to the next step without further optimization, ensuring process closure and uniqueness.

[0071] After obtaining material composition distribution data that meets or approaches the threshold requirements, the performance evaluation module uses finite element analysis to calculate preliminary evaluation values ​​of barrier performance and obtain preliminary barrier performance data. The calculation rules are as follows: The three-dimensional space is divided into cubic units with 1 mm side lengths; the content of each element is mapped to the diffusion coefficient and permeability coefficient of each unit using a fixed lookup table mixing rule. The mapping table is established and solidified based on standard samples before equipment delivery; the boundary conditions are set as follows: the concentration at the inlet face is fixed at 1 mmol / L, the concentration at the outlet face is fixed at 0 mmol / L, and the sidewalls are flux-free boundaries; after solving the steady-state flux field, the spatial average of the flux across the entire cross-section is divided by the cross-sectional area and the concentration difference to obtain the overall permeability coefficient; the overall permeability coefficient is corrected according to the geometric dimensions in the barrier direction and the test environment temperature to obtain the preliminary evaluation value of barrier performance. This evaluation value is compared with the target barrier performance value, and the absolute value of the difference is calculated and divided by the target value to obtain the percentage deviation of the preliminary barrier performance data. The initial barrier performance deviation threshold is fixed at 10.00%. This threshold is obtained by adding twice the inter-sample standard deviation to the arithmetic mean of the deviations of 50 historical samples that meet the engineering durability requirements, and rounding to 0.01 before fixing. If the deviation is not greater than 10.00%, the classification step is skipped and standardization and key parameter extraction are performed; if it is greater than 10.00%, classification is performed to identify the problem type. When the deviation of the initial barrier performance data is greater than 10.00%, the performance evaluation module performs support vector machine classification on the initial barrier performance data to obtain the classified performance features. The input features are fixed at 5 items: overall permeability coefficient, maximum local permeability coefficient, standard deviation of permeability coefficient across the entire grid, maximum value of the distribution deviation of each element component, and standard deviation of the distribution deviation of each element component; the category labels are qualified and unqualified; the classifier kernel function type is radial basis function, the penalty parameter is fixed at 100, and the kernel width is fixed at 0.5; the training data scale is 200 sets, of which 100 sets are qualified and 100 sets are unqualified. The model is trained and frozen before equipment delivery. At runtime, a one-time forward inference is performed on the input features, outputting the category and its confidence level; the consistency index corresponding to the output category is recorded as the performance feature after classification, which is used for subsequent standardization and threshold determination.

[0072] Subsequently, the performance evaluation module standardizes the performance features and determines the standardization bias. Standardization employs the zero-mean unit variance rule, which involves subtracting the mean of each performance feature from the historical training set and dividing by the standard deviation of that feature; when the standard deviation is less than 0.0001, it is fixed at 0.0001 to avoid division by zero. The root mean square (RMS) value is calculated for the resulting standardized feature vector, and this RMS value is compared with the RMS value of the zero vector; the difference is used as a measure of the standardization bias. The standardization bias threshold is fixed at 0.50, determined by adding one time the inter-sample standard deviation to the mean of the standardization bias when the classification accuracy is not less than 95%, and rounded to 0.01 before being fixed. If the standardization bias measure is not greater than 0.50, the process proceeds to key performance parameter extraction; if it is greater than 0.50, the process directly enters the final dynamic control loop, and key parameters are no longer extracted, ensuring a deterministic correlation between the evaluation conclusion and the control action.

[0073] When extracting key performance parameters, the performance evaluation module obtains the final barrier performance data using a fixed signal processing method. The specific steps are as follows: a median filter of length 5 is applied to the 300-second permeability coefficient sequence to remove isolated spikes, followed by a moving average of length 7 to smooth short-term fluctuations; six parameters are extracted from the smoothed sequence: the overall arithmetic mean, the overall standard deviation, the 90th percentile, the arithmetic mean of the last 60 seconds, the standard deviation of the last 60 seconds, and the percentage of time that meets the steady-state condition; the steady-state condition is defined as the standard deviation of the permeability coefficient within any consecutive 10-second window not exceeding 0.10 of the average value within that window. If the standard deviation of the last 60 seconds is greater than 0.10 of the average value for that period, the permeability coefficient for each second within the last 60 seconds is weighted and normalized using the reciprocal of the standard deviation within that second, and then weighted and averaged to obtain a robust final barrier performance value; otherwise, the arithmetic mean of the last 60 seconds is directly used as the final barrier performance value. The final barrier performance value is compared with the target barrier performance value to calculate the final deviation percentage, which, along with the steady-state time percentage, serves as the evaluation conclusion. The preset criteria for passing the evaluation are fixed and must be met simultaneously: the final deviation percentage is no greater than 5.00%, and the steady-state time percentage is no less than 0.90%. These two thresholds are determined by the 95th percentile of the deviation distribution and steady-state percentage distribution in historical samples that meet the engineering acceptance requirements, respectively, and rounded to 0.01 before being fixed. If the preset criteria are met, the performance evaluation module generates a final monitoring report. The report content is fixed and includes detection parameters, grid and boundary settings, optimization record summary, material composition distribution summary, preliminary and final barrier performance results, key parameter table, and conclusions. When any threshold is determined not to meet the preset criteria, the performance evaluation module triggers a cyclical dynamic control scheme according to fixed rules to obtain optimization iteration data, and strictly limits the adjustment objects and magnitudes to ensure that modules outside the claims are not introduced.The execution sequence of the loop is as follows: First, determine the set of analysis parameters to be adjusted based on the specific unmet indicators, including only four items: scan step size, integration time, beam spot diameter, and data filter width; for each parameter, set a search interval of ±10% of the current value, with a step size of not less than 0.1 units; within the interval, use an exhaustive combination finite search, evaluating a maximum of 30 parameter combinations; after evaluating each combination, quickly collect and calculate the component distribution characteristic deviation and the preliminary barrier performance deviation given by the finite element method, and add the two with fixed weights to obtain a comprehensive score, with weights of 0.6 and 0.4 respectively; when the comprehensive score is better than the current... The parameter set is immediately recorded as a new candidate when it is the previous best. The search is terminated when the evaluation is completed or the score reaches the preset merit threshold, which is defined as an improvement of no less than 10% relative to the current score. The optimal parameter set is written into the dynamic control scheme of this cycle and returned to the first segment of this implementation method to re-execute data acquisition, finite element analysis, classification, standardization and key parameter extraction to form a new evaluation result. If the preset standard is not met after two consecutive cycles, the conclusion "the standard has not been met and two rounds of optimization have been completed" is output and all intermediate data is retained for traceability. The third round is not continued to ensure process closure and the uniqueness of the calculation result.

[0074] The intelligent monitoring module extracts key features from the optimized iterative data and uses a neural network algorithm to process the correlation between current changes and concentration abrupt changes, resulting in an enhanced synchronous monitoring framework. Based on this framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators. This includes extracting key features from the optimized iterative data, using principal component analysis to reduce the dimensionality of these features, obtaining dimensionality-reduced feature data. If the variance of the dimensionality-reduced feature data is less than a preset threshold, time series analysis is used to perform time-series modeling, obtaining a time-series feature model. Based on the time-series feature model, a convolutional neural network is used to analyze the correlation between current change data and concentration abrupt changes, obtaining a correlation feature vector. If the classification accuracy of the correlation feature vector is greater than a preset threshold, data standardization is used to process the correlation feature vector, obtaining standardized correlation data. Based on the standardized correlation data, signal processing methods are used to extract key parameters for synchronous monitoring, determining the synchronous monitoring parameter set. Time-series synchronous features are extracted from the synchronous monitoring parameter set and grouped using cluster analysis. A synchronous monitoring framework was obtained. Based on this framework, a data fusion method was used to integrate multi-source monitoring data streams to obtain enhanced synchronous monitoring results. Dynamic electric field signals were obtained from the multi-source electric field data, and time-series analysis was used to model the dynamic electric field signals to obtain a dynamic electric field model. If the fluctuation amplitude of the dynamic electric field model exceeded a preset threshold, feature extraction was performed on the dynamic electric field model using a signal decomposition method to obtain an ion permeation feature set. Based on the ion permeation feature set, a convolutional neural network was used to analyze the correlation between ion permeation behavior and electric field strength to obtain a permeation correlation vector. The permeation correlation vector was normalized using a data standardization method to obtain standardized permeation data. Cluster analysis was used to group the standardized permeation data to determine a preliminary barrier performance index set. If the classification consistency of the preliminary barrier performance index set exceeded a preset threshold, the multi-source electric field data and the preliminary barrier performance index set were integrated using a data fusion method to obtain enhanced barrier performance indices. Based on the enhanced barrier performance indices, a time-series prediction method was used to analyze the future trend of ion permeation behavior to determine the final barrier performance index set.

[0075] In one possible implementation, the intelligent monitoring module takes optimization iteration data as input and executes it in a fixed order without optional branches. First, it extracts key features from the optimization iteration data. This data contains a current change sequence, a concentration abrupt change marker sequence, and the electric field intensity at the corresponding time point per second. The sampling frequency is fixed at 100 Hz, the total duration is 300 seconds, and the total number of samples is 30,000. Key features are calculated according to defined rules within each second-level time window, including nine items: the mean and standard deviation of the current, the maximum rise slope, the energy proportion of the low, medium, and high frequency bands, the count of concentration abrupt change points, the shortest interval between abrupt change points, and the second-level average and standard deviation of the electric field intensity. These nine features are then processed window by window across the 300 time windows. The feature matrix is ​​dimensionality reduced by principal component analysis (PCA). The number of principal components retained is the smallest integer that ensures the cumulative variance percentage is not less than 0.85. If the second-level average of the cumulative variance percentages over 300 time windows is less than 0.85, the dimensionality-reduced feature data is considered to have insufficient variance and enters the time series modeling branch; otherwise, it directly enters the correlation analysis branch. The variance threshold of 0.85 is determined by multiplying the median of the cumulative variance percentages of the principal components of 30 stable samples by 1.00 and rounding to two decimal places before being fixed. When entering the time series modeling branch, time series analysis is performed on the dimensionality-reduced feature data. The modeling window is fixed at 60-second intervals with no overlap, and the candidate model order is from 1 to 5. The difference metric is the mean absolute error of one-step prediction. The array with the smallest error is selected as the time-series feature model for that window. The mean, standard deviation, and kurtosis of the prediction residuals obtained by this model within that window are used to construct a time-series feature vector. The vectors from each window are then concatenated in chronological order to form a time-series feature sequence. Subsequently, a convolutional neural network (CNN) is used to analyze the correlation between current changes and concentration abrupt changes, based on either dimensionality-reduced feature data or the time-series feature sequence (whichever is correct). The CNN structure is fixed as a one-dimensional three-layer convolutional network plus a fully connected layer. The kernel lengths are 5, 3, and 3, and the number of convolutional channels are 16, 32, and 64, respectively. The stride is 1 for all layers, and max pooling with a length of 2 is used. Each layer is followed by... The network employs a fixed random deactivation rate of 0.2. The output layer is a binary classification layer. The input consists of current change segments per second and corresponding labels indicating whether a concentration abrupt change occurs. The output is the category and probability of a strong correlation. Network parameters are trained and frozen with 200 labeled samples before device delivery. During operation, inference is performed every 60 seconds, and the classification accuracy is recorded. If the average classification accuracy of a segment is not less than 0.90, the data standardization step is initiated; otherwise, the network returns to the previous time-series modeling branch, adds a feature re-extraction, and performs another inference, with only one such addition allowed. The classification accuracy threshold of 0.90 is determined by rounding to two decimal places from the minimum accuracy point on the validation set that results in a false positive rate not exceeding 0.10.

[0076] Once the classification accuracy meets the threshold, data standardization is performed on the associated feature vectors output by the convolutional neural network. This is done by subtracting the arithmetic mean over 300 time windows from each vector component and then dividing by the corresponding standard deviation. When the standard deviation is less than 0.0001, it is fixed at 0.0001 to avoid division by zero, resulting in standardized associated data. Based on this standardized associated data, signal processing methods are used to extract key parameters for synchronous monitoring. Four parameters are extracted: the average absolute value of the time delay between the current peak and the concentration abrupt change point; the average synchronous energy obtained by summing the products of current energy and abrupt change counts over the entire second and dividing by 300; and the low-frequency band... The average difference between the energy percentage at the occurrence and non-occurrence of the abrupt change, and the proportion of the dominant frequency falling in the low-frequency band, constitute the synchronous monitoring parameter set. Temporal synchronization features are extracted from this parameter set by calculating the arithmetic mean of each of the four parameters within a consecutive 60-second window and concatenating them into a 4-dimensional window feature. Cluster analysis is then used to group the features. The number of clusters is fixed at 3, determined by the median of the cluster number corresponding to the largest silhouette coefficient in 50 historical samples before the product leaves the factory. Euclidean distance is used as the clustering distance metric, and the final grouping result is obtained by repeating clustering with 10 different initial centers and taking the minimum total squared error within the window. The resulting 3 groups... The central vectors and allocation rules together constitute the synchronous monitoring framework. Based on this framework, a data fusion method is used to integrate multi-source monitoring data streams to obtain enhanced synchronous monitoring results. The multi-source monitoring data streams are fixed to include three streams: current change sequence, concentration mutation marker sequence, and electric field intensity sequence. The fusion rule is to use the similarity of the group centers given by the synchronous monitoring framework as weights to perform a weighted average of the three streams per second. The similarity is defined as the reciprocal of the distance from the window feature to the corresponding center, after minimum-maximum normalization, with a normalization interval of 0 to 1. The weights and constraints are all set to 1. Subsequently, a dynamic electric field signal is obtained from the multi-source electric field data. The dynamic electric field signal is defined as... The time series of the spatial average electric field intensity per second over a 300-second period is used. Time series analysis is employed to model this series, with candidate orders ranging from 1 to 5. The evaluation metric is the mean absolute error of the one-step prediction, with the smallest value selected as the electric field dynamic model. The fluctuation amplitude of this model in the latter 60-second interval is calculated. The fluctuation amplitude is defined as the difference between the 95th percentile and the 5th percentile within this interval, divided by the median of that interval. The preset fluctuation amplitude threshold is fixed at 0.15, determined by adding twice the inter-sample standard deviation to the arithmetic mean of this metric for 30 stable electric field samples under the same calculation caliber, rounded to two decimal places. If the fluctuation amplitude is greater than 0.15...15. Then, a signal decomposition method is performed on the time series corresponding to the electric field dynamic model to extract the ion penetration feature set. The signal decomposition method is wavelet packet decomposition to 4 layers, and for each frequency band of each layer, four features are calculated: energy, energy ratio, dominant frequency, and bandwidth. The four features of all frequency bands are concatenated to form the ion penetration feature set. Based on the ion penetration feature set, a convolutional neural network with the aforementioned structure is used to analyze the correlation between ion penetration behavior and electric field strength. The input is the time-aligned ion penetration feature set and the electric field strength sequence, and the output is the penetration correlation vector with a fixed length of 16. The network parameters remain the same as before and are not updated during operation. Data standardization is performed on the penetration correlation vector by subtracting the average value of each component from its value over the last 60 seconds and then dividing by the corresponding standard deviation. If the standard deviation is less than 0.0001, it is fixed at 0.0001 to obtain standardized penetration data. Cluster analysis is then used to group the standardized penetration data, with a fixed number of clusters of 4. Euclidean distance is used as the distance metric. This initialization is repeated 10 times, and the result of the minimum total squared error is used to obtain a preliminary set of barrier performance indicators. The indicator set consists of a 16-dimensional vector from the 4 cluster centers and the corresponding cluster label group per second. Successfully complete the following step: Calculate the classification consistency of the initial barrier performance index set. Consistency is defined as follows: Within the next 60 seconds, perform three non-replacement 3 / 7 random partitioning training and validation on the data, and in each partition, repeat clustering once starting from the initial cluster center to obtain cluster labels per second. Calculate the average of the proportion of pairwise identical labels in the three partitions as the consistency metric. If the consistency metric is not less than 0.85, proceed to the enhanced barrier performance index generation step; otherwise, return to the wavelet packet decomposition step, add an upper limit of 1 to 5 to the decomposition level, and repeat the subsequent process once, allowing only one repetition to ensure the process is smooth. Closed; once the consistency reaches 0.85, the data fusion method is used to integrate the multi-source electric field data with the preliminary barrier performance index set. The fusion rule is to use the similarity between the cluster center per second and the vectorized description of the dynamic electric field signal in that second as the weighting coefficient to calculate the weighted sum of the four cluster centers to obtain the enhanced barrier performance index. The vectorized description is fixed by concatenating the mean, standard deviation, linear trend, and low-frequency and mid-frequency energy ratio of the dynamic electric field signal in the most recent 5-second window into a 5-dimensional vector. The similarity is the minimum-maximum normalization of the inverse distance, and the weight and constraint are 1. Finally, based on the enhanced barrier performance index, a time-series prediction method was used to analyze the future trend of ion permeation behavior. The time-series prediction method was exponential smoothing with a fixed smoothing coefficient of 0.3 and a prediction step size of 60 seconds. The enhanced barrier performance index sequence for the next 60 seconds was output, and the arithmetic mean, maximum value, minimum value, and last value were taken to form the final barrier performance index set. At the same time, the standard deviation of the sequence within 60 seconds was recorded as a stability index. If the standard deviation was greater than 0.10 of the average value of the sequence, the smoothing coefficient was adjusted between 0.2 and 0.4 with a step size of 0 without changing any parameters or thresholds.Each calculation is performed once, and the group with the smaller stability index is taken as the final result; a third calculation is not performed. The source, value range, calculation method, and stopping conditions of all parameters and thresholds are clearly stated in this section, including a sampling frequency of 100 Hz, a time window length of 60 seconds, a cumulative variance threshold of 0.85, a classification accuracy threshold of 0.90, a fluctuation amplitude threshold of 0.15, a classification consistency threshold of 0.85, 3 and 4 clusters, a fixed lower limit for data standardization of 0.0001, an exponential smoothing coefficient of 0.3 with alternative values ​​of 0.2 and 0.4, and a prediction step size of 60 seconds, etc., and remain constant within a single run.

[0077] like Figure 2 As shown, a method for detecting the permeability of recycled concrete is also provided. Using the aforementioned recycled concrete permeability detection system, the method includes: applying a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from the recycled concrete sample, thus obtaining baseline parameters for the permeation process; based on the baseline parameters, constructing a pore structure model using a finite element simulation method, incorporating ion migration rate and non-uniform distribution characteristics, and determining a dynamic control scheme under the action of the electric field; obtaining the voltage adjustment sequence in the dynamic control scheme, optimizing the local electric field strength for the concentration change region; if the current change exceeds a preset threshold after optimization, resimulating the pore structure to obtain distribution data; extracting synchronous monitoring indicators from the distribution data, calculating the ion permeation path using a diffusion equation algorithm, and obtaining a preliminary permeability coefficient estimate. The process involves: 1) Calculating the initial permeability coefficient value and the real-time current change signal. If the coupling deviation is less than a preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution. 2) Based on the corrected concentration distribution, migration rate parameters are incorporated for iterative calculation to determine the permeability coefficient under electric field assistance. 3) Obtaining the permeability coefficient, a quantitative evaluation of the overall barrier performance of recycled concrete is performed. If the evaluation result meets the preset standard, a final monitoring report is output; otherwise, the dynamic control scheme is looped back to obtain optimized iterative data. 4) Extracting key features from the optimized iterative data, a neural network algorithm is used to process the correlation between current change and concentration mutation, resulting in an enhanced synchronous monitoring framework. Based on this framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators.

[0078] First, a preset voltage gradient is applied to both ends of the recycled concrete sample using an electric field-assisted device. The magnitude of the voltage gradient is determined by dividing the terminal voltage by the sample length, with a fixed range between 0.1 V / cm and 3 V / cm. The specific value is determined through pre-experiments based on the sample's design strength grade and porosity. Under the voltage, a sensor array measures the ion concentration at each measuring point inside the sample, and a signal acquisition module synchronously acquires current change signals at a fixed sampling frequency of 100 Hz. The acquired ion concentration data is denoised and smoothed by a signal processing module and combined with the current change signals. The ion migration rate is calculated using a support vector machine algorithm, and this migration rate and concentration distribution curve are used as the baseline parameters for the infiltration process. Then, a three-dimensional pore structure model is established using the finite element method based on the baseline parameters. The mesh size is fixed at 0.1 mm cubic elements, and the material properties are mapped from the baseline parameters to the migration rate and non-uniform distribution characteristics of each element. The correlation coefficient between the ion migration rate and the electric field strength is calculated using a partial least squares regression algorithm. When this coefficient is greater than a preset threshold, the electric field distribution in the model is adjusted through iterative optimization until a dynamic control scheme that meets the conditions is obtained. After obtaining the dynamic control scheme, a voltage adjustment sequence is extracted from the scheme. Local electric field strength optimization is performed in areas where concentration changes exceed a set standard, using the gradient descent method to adjust the voltage amplitude. When the optimized current change exceeds a preset threshold, a new finite element simulation of the pore structure is performed based on the latest parameters to obtain updated distribution data. Subsequently, synchronous monitoring indicators are extracted from the distribution data, including ion concentration gradient, concentration change rate, and concentration distribution variance. The diffusion equation algorithm is used to calculate the ion permeation path and extract migration path characteristics. A preliminary permeability coefficient estimate is calculated based on the migration path characteristics. If the stability of the permeability coefficient distribution is insufficient, it is recalculated and corrected using numerical simulation. Next, the preliminary permeability coefficient value is coupled with the real-time current change signal. The frequency characteristics of the current signal are extracted using Fourier transform, and the coupling deviation is calculated. When the deviation is less than a preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted, and a corrected concentration distribution is obtained. For the corrected concentration distribution, the migration rate parameter is extracted, and iterative calculations using numerical integration and finite difference methods are performed to obtain the final permeability coefficient under electric field assistance. Based on this permeability coefficient, X-ray fluorescence spectroscopy is used to analyze the material composition distribution. Finite element analysis is then used to calculate the barrier performance, which is compared with preset performance standards. If the evaluation results meet the requirements, a final monitoring report is generated; otherwise, the process returns to the dynamic control step for optimization iteration. During the optimization iteration, key features of current changes and concentration abrupt changes are extracted from the data. Principal component analysis is used to reduce the dimensionality, and the results are then fed into a convolutional neural network to analyze the correlation between the two, constructing an enhanced synchronous monitoring framework. Finally, based on this framework, ion permeation behavior is predicted under various simulated electric field scenarios, and the final set of barrier performance indicators is calculated as the output of this detection method.

[0079] 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 system for testing the permeability of recycled concrete, characterized in that, include: The electric field-assisted module applies a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from the recycled concrete sample, thereby obtaining the baseline parameters of the infiltration process. The modeling and simulation module constructs a pore structure model using the finite element simulation method based on the benchmark parameters, incorporates ion migration rate and non-uniform distribution characteristics, and determines the dynamic control scheme under the action of electric field. The voltage optimization module acquires the voltage adjustment sequence in the dynamic control scheme, optimizes the local electric field intensity in the concentration change area, and if the current change after optimization exceeds the preset threshold, it re-simulates the pore structure to obtain distribution data. The permeation path calculation module extracts synchronous monitoring indicators from the distribution data, uses the diffusion equation algorithm to calculate the ion permeation path, and obtains a preliminary estimate of the permeability coefficient. The coupling analysis module analyzes the coupling between the initial permeability coefficient value and the real-time current change signal. If the coupling deviation is less than the preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution. The iterative calculation module performs iterative calculations based on the corrected concentration distribution and incorporates the migration rate parameter to determine the permeability coefficient under electric field assistance. The performance evaluation module obtains the permeability coefficient and performs a quantitative evaluation of the overall barrier performance of recycled concrete. If the evaluation results meet the preset standards, the final monitoring report is output; otherwise, the dynamic control scheme is circulated back to obtain optimized iterative data. The intelligent monitoring module extracts key features from the optimized iterative data and uses a neural network algorithm to process the correlation between current changes and concentration mutations, resulting in an enhanced synchronous monitoring framework. Based on this framework, the ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators.

2. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The electric field-assisted module applies a preset voltage gradient through an electric field-assisted device to obtain initial ion concentration distribution data and current change signals from the recycled concrete sample, thereby obtaining the baseline parameters of the infiltration process, including: A preset voltage gradient was applied to the recycled concrete sample using an electric field-assisted device, and initial ion concentration distribution data and current change signals were collected to determine the baseline parameters of the infiltration process. A sensor array was used to monitor the ion concentration distribution in recycled concrete samples in real time to obtain initial ion concentration distribution data. The signal processing module denoises the collected ion concentration distribution data to obtain a smooth ion concentration distribution curve. Based on the smooth ion concentration distribution curve and combined with the current change signal, the support vector machine algorithm is used to extract features and determine the ion migration rate. If the ion migration rate exceeds the preset threshold, the voltage gradient is adjusted by the control module to obtain a new current change signal; The new current change signal was processed by time series analysis to obtain the dynamic parameters of the permeation process; The baseline parameters for the permeation process are determined by comparing the dynamic parameters with the initial ion concentration distribution curve.

3. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The modeling and simulation module, based on benchmark parameters, uses the finite element method to construct a pore structure model, incorporating ion migration rate and non-uniform distribution characteristics, and determines the dynamic control scheme under the action of an electric field, including: The pore structure model was meshed using the finite element method, and the initial electric field distribution data was obtained by combining the characteristics of ion migration rate and non-uniform distribution. Based on the initial electric field distribution data, the correlation between ion migration rate and electric field strength is analyzed using the partial least squares regression algorithm to obtain the electric field modulation coefficient. If the electric field control coefficient exceeds the preset threshold, the electric field distribution in the pore structure model is adjusted by an iterative optimization method to obtain the optimized electric field distribution data. Based on the optimized electric field distribution data, the dynamic trend of ion migration rate is extracted by time series analysis to obtain dynamic control parameters; Based on the dynamic control parameters, the support vector machine algorithm is used to classify the ion distribution characteristics and determine the applicability of the control parameters; If the applicability of the control parameters meets the preset conditions, the pore structure model is updated by numerical simulation to obtain the final dynamic control scheme. Based on the final dynamic control scheme, the stability of ion migration rate was analyzed through simulation verification to determine the control results under the action of electric field.

4. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The voltage optimization module acquires the voltage adjustment sequence in the dynamic control scheme, optimizes the local electric field intensity in the concentration change region, and if the optimized current change exceeds a preset threshold, it re-simulates the pore structure to obtain distribution data, including: A pore structure model was constructed using the finite element method, an initial voltage adjustment sequence was generated, and electric field intensity distribution data in the concentration abrupt change region was obtained. If the electric field intensity distribution data exceeds the preset threshold, the gradient descent algorithm is used to adjust the voltage sequence to obtain the optimized voltage sequence. Based on the optimized voltage sequence, the current change data is calculated using numerical simulation methods to obtain the current change trend. If the current change trend exceeds the preset threshold, the pore structure model is updated by iterative optimization method to obtain the adjusted structure model; Based on the adjusted structural model, the electric field intensity distribution data is recalculated to obtain a stable electric field distribution. By processing the stable electric field distribution using time series analysis, dynamic voltage adjustment parameters are extracted to obtain the control parameter set; Based on the set of control parameters, the electric field distribution in the concentration abrupt change region is classified using the support vector machine algorithm to determine the stability of the control parameters and obtain the final control scheme.

5. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The permeation path calculation module extracts synchronous monitoring indicators from the distribution data, uses a diffusion equation algorithm to calculate the ion permeation path, and obtains preliminary permeability coefficient estimates, including: Ion concentration gradient features are extracted from the distribution data, and concentration distribution features are generated using data extraction methods. Based on the concentration distribution characteristics, the diffusion equation algorithm is used to calculate the ion migration path and obtain the migration path characteristics; If the migration path characteristics do not match the preset threshold, the path optimization parameters are adjusted through iterative optimization to obtain the optimized migration path. Based on the optimized migration path, the ion diffusion rate is calculated, and the diffusion rate distribution is obtained. Permeability coefficient estimates are extracted from the diffusion rate distribution, and the permeability coefficient distribution is classified using the support vector machine algorithm to determine the stability of the distribution; If the distribution stability analysis results do not reach the preset threshold, the ion migration path is recalculated using numerical simulation to obtain the updated permeability coefficient distribution. Based on the updated permeability coefficient distribution, the final path optimization parameters are extracted to obtain a stable permeability coefficient estimate.

6. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The coupling analysis module, through coupling analysis of the preliminary permeability coefficient value and the real-time current change signal, confirms the barrier performance evaluation model if the coupling deviation is less than a preset threshold; otherwise, it adjusts the dynamic control scheme to obtain the corrected concentration distribution, including: Dynamic features are extracted from real-time current change signals, and current fluctuation sequences are generated using the Fourier transform method to obtain frequency distribution characteristics; Based on the frequency distribution characteristics, the coupling deviation between the preliminary permeability coefficient estimate and the current fluctuation sequence is calculated, and the deviation distribution characteristics are obtained. If the deviation distribution characteristics are less than the preset threshold, the barrier performance evaluation model is determined by the logistic regression algorithm, and the model evaluation result is obtained. If the deviation distribution characteristics are greater than the preset threshold, the gradient descent method is used to adjust the concentration distribution parameters to obtain the optimized concentration distribution. Based on the optimized concentration distribution, the dynamic features of the real-time current change signal are re-extracted to generate an updated current fluctuation sequence. By coupling analysis of the updated current fluctuation sequence and the preliminary permeability coefficient estimate, it is determined whether the preset threshold is met, and the final barrier performance evaluation model is obtained. Based on the final barrier performance evaluation model, key performance parameters are extracted, and the corrected concentration distribution characteristics are generated.

7. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The iterative calculation module, based on the corrected concentration distribution and incorporating migration rate parameters, performs iterative calculations to determine the permeability coefficient under electric field assistance, including: Based on the corrected concentration distribution, the migration rate parameter is extracted, and a preliminary estimate of the permeability coefficient is generated using a numerical integration method. Using a preliminary estimate of the permeability coefficient, combined with auxiliary electric field parameters, the migration rate distribution under the influence of the electric field is calculated using the finite difference method. If the deviation of the migration rate distribution is greater than the preset threshold, the electric field auxiliary parameters are adjusted by the gradient descent method to obtain the optimized migration rate distribution. Based on the optimized migration rate distribution, dynamic features are re-extracted to generate updated penetration coefficient estimates. Using the updated permeability coefficient estimate, a permeability coefficient model is constructed using a logistic regression algorithm, and the model output results are obtained. If the deviation of the model output is less than the preset threshold, the key performance parameters are obtained by signal extraction method to determine the final effective penetration coefficient. Based on the final effective permeability coefficient, the corrected concentration distribution characteristics are generated, and the optimized distribution data is output.

8. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The performance evaluation module obtains the permeability coefficient and quantitatively evaluates the overall barrier performance of recycled concrete. If the evaluation result meets the preset standard, a final monitoring report is output; otherwise, the dynamic control scheme is cyclically re-implemented to obtain optimized iterative data, including: Barrier performance data of recycled concrete were obtained, and the material composition distribution was analyzed by X-ray fluorescence spectroscopy to obtain composition distribution characteristics. If the deviation of the component distribution characteristics is greater than the preset threshold, the detection parameters of X-ray fluorescence spectroscopy are adjusted by the particle swarm optimization algorithm to obtain the optimized component distribution characteristics. Based on the optimized component distribution characteristics, the finite element analysis method is used to calculate the preliminary evaluation value of the barrier performance, and the preliminary barrier performance data is obtained. If the deviation of the initial barrier performance data is greater than the preset threshold, the initial barrier performance data is classified using the support vector machine algorithm to obtain the classified performance characteristics. Based on the classified performance characteristics, the performance characteristics are processed using data standardization methods to obtain standardized performance data. If the deviation of the standardized performance data is less than the preset threshold, key performance parameters are extracted by signal processing methods to determine the final blocking performance data. Based on the final barrier performance data, a dynamic control scheme was used to adjust the analytical parameters of the material composition distribution to obtain optimized material distribution data.

9. The system for testing the permeability of recycled concrete according to claim 1, characterized in that: The intelligent monitoring module extracts key features from the optimized iterative data, uses a neural network algorithm to process the correlation between current changes and concentration abrupt changes, and obtains an enhanced synchronous monitoring framework. Based on the enhanced synchronous monitoring framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance indicators, including: Key features are extracted from the optimization iteration data, and principal component analysis is used to reduce the dimensionality of the key features to obtain dimensionality-reduced feature data. If the variance of the dimensionality-reduced feature data is less than a preset threshold, then time series analysis is used to perform time series modeling on the dimensionality-reduced feature data to obtain a time series feature model. Based on the time-series feature model, a convolutional neural network is used to analyze the correlation between current change data and concentration abrupt change points to obtain the correlation feature vector; If the classification accuracy of the associated feature vector is greater than a preset threshold, the associated feature vector is processed by a data standardization method to obtain standardized associated data. Based on standardized correlation data, signal processing methods are used to extract key parameters for synchronous monitoring and determine the synchronous monitoring parameter set. Temporal synchronization features are extracted from the set of synchronous monitoring parameters, and cluster analysis is used to group the temporal synchronization features to obtain the synchronous monitoring framework; Based on the synchronous monitoring framework, a data fusion method is used to integrate multi-source monitoring data streams to obtain enhanced synchronous monitoring results; Dynamic electric field signals are obtained from multi-source electric field data, and time series analysis is used to model the dynamic electric field signals to obtain a dynamic electric field model. If the fluctuation amplitude of the electric field dynamic model is greater than the preset threshold, the electric field dynamic model is subjected to feature extraction by signal decomposition method to obtain the ion permeation feature set. Based on the ion permeation feature set, a convolutional neural network is used to analyze the correlation between ion permeation behavior and electric field strength, and a permeation correlation vector is obtained. Standardized penetration data is obtained by normalizing the penetration correlation vector using data standardization methods. Cluster analysis was used to group the standardized penetration data to determine a preliminary set of barrier performance indicators; If the classification consistency of the preliminary barrier performance index set is greater than the preset threshold, the multi-source electric field data and the preliminary barrier performance index set are integrated by the data fusion method to obtain the enhanced barrier performance index. Based on the enhanced barrier performance indicators, a time-series prediction method is used to analyze the future trend of ion permeation behavior and determine the final set of barrier performance indicators.

10. A method for testing the permeability of recycled concrete, using the recycled concrete permeability testing system according to any one of claims 1-9, characterized in that, The method includes: By applying a preset voltage gradient through an electric field-assisted device, initial ion concentration distribution data and current change signals are obtained from recycled concrete samples to obtain the baseline parameters of the infiltration process. Based on the baseline parameters, a pore structure model is constructed using the finite element simulation method, incorporating ion migration rate and non-uniform distribution characteristics, and a dynamic control scheme under the action of an electric field is determined. Obtain the voltage adjustment sequence in the dynamic control scheme, optimize the local electric field intensity in the concentration change region, and if the current change after optimization exceeds the preset threshold, re-simulate the pore structure to obtain distribution data. Synchronous monitoring indicators were extracted from the distribution data, and the ion permeation path was calculated using the diffusion equation algorithm to obtain a preliminary estimate of the permeability coefficient. By coupling analysis of the preliminary permeability coefficient value and the real-time current change signal, if the coupling deviation is less than the preset threshold, the barrier performance evaluation model is confirmed; otherwise, the dynamic control scheme is adjusted to obtain the corrected concentration distribution. Based on the corrected concentration distribution, the migration rate parameter is incorporated into iterative calculations to determine the permeability coefficient under electric field assistance. The permeability coefficient is obtained to quantitatively evaluate the overall barrier performance of recycled concrete. If the evaluation results meet the preset standards, the final monitoring report is output; otherwise, the dynamic control scheme is circulated back to obtain optimized iterative data. Key features are extracted from the optimized iterative data, and a neural network algorithm is used to process the correlation between current changes and concentration abrupt changes to obtain an enhanced synchronous monitoring framework. Based on the enhanced synchronous monitoring framework, ion permeation behavior under various electric field scenarios is simulated to determine the final barrier performance index.

Images