Foundation pit slope protection pile concrete strength evaluation and intelligent reinforcement integrated method and system

Abstract

The invention relates to the technical field of civil engineering, and discloses a foundation pit slope protection pile concrete strength evaluation and intelligent reinforcement integrated method which comprises the following steps: S1, data acquisition and transmission: acquiring a slope protection pile structure state and environmental parameters in real time through a multi-source sensor network, the sensor network comprises a mechanical sensor, an environment sensor and a hydrological sensor which are deployed in a spatial distribution manner; s2, data preprocessing and feature extraction: performing space-time fusion processing, including noise suppression, space-time reference unification and multi-dimensional feature extraction, on the acquired multi-source sensing data, and generating a structured feature matrix; according to the technical scheme, the multi-source sensor network and the deep learning model are combined, the strength of the foundation pit slope protection pile concrete is monitored and evaluated in real time, the accurate and lossless strength prediction effect is achieved, the problems of destructive testing and precision limitation are avoided, and the test efficiency is improved. The defects of low precision and poor real-time performance under complex geological conditions in the prior art are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of civil engineering, and specifically to an integrated method and system for concrete strength evaluation and intelligent reinforcement of foundation pit slope protection piles. Background Technique

[0002] With the acceleration of the urbanization process, foundation pit support piles, as one of the key structures in urban infrastructure construction, their safety and stability directly affect the overall quality of buildings. Foundation pit slope protection piles play a supporting role during the excavation of deep foundation pits, ensuring the stability of the surrounding soil mass and preventing ground settlement and soil sliding. To ensure the structural safety, it is necessary to regularly evaluate the concrete strength and timely detect potential risks. However, traditional concrete strength detection methods rely on manual operation and are mostly destructive tests, which not only increase the cost but also cannot monitor the foundation pit state in real time.

[0003] Currently, the evaluation methods of concrete strength are roughly divided into two categories: traditional sampling detection and non-destructive detection. Although the traditional core drilling method is accurate, it will damage the structure of the slope protection pile and cannot provide real-time data; non-destructive testing methods, such as the rebound hammer method and ultrasonic testing method, although they do not damage the concrete structure, their accuracy is limited, especially in complex geological conditions, it is difficult to accurately reflect the actual situation of concrete strength. In addition, these detection methods usually rely on manual judgment and cannot meet the requirements of real-time data and rapid response during foundation pit construction.

[0004] The main problem faced by the existing technology is the lack of effective means to evaluate concrete strength in real time and accurately, especially in complex geological environments and dynamic construction conditions. Traditional detection methods not only have accuracy problems but also have a slow response speed, unable to timely identify the change trend of concrete strength, resulting in a lag in reinforcement measures. This lag in reinforcement response not only affects the safety of the foundation pit but also may lead to unnecessary increase in project costs. Therefore, how to achieve real-time monitoring and intelligent reinforcement of the concrete strength of foundation pit slope protection piles has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of the deficiencies of the existing technology, the present invention provides an integrated method and system for concrete strength evaluation and intelligent reinforcement of foundation pit slope protection piles, which solves the problem of lag in real-time evaluation and intelligent reinforcement response of the concrete strength of foundation pit slope protection piles.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: An integrated method for concrete strength evaluation and intelligent reinforcement of foundation pit slope protection piles, including the following steps: S1. Data collection and transmission: Real-time collect the structural state and environmental parameters of the slope protection pile through a multi-source sensor network, and the sensor network includes mechanically sensors, environmental sensors and hydrological sensors deployed in a spatially distributed manner; S2. Data preprocessing and feature extraction: Perform spatio-temporal fusion processing on the collected multi-source sensing data, including noise suppression, spatio-temporal benchmark unification, and multi-dimensional feature extraction, to generate a structured feature matrix; S3. Strength evaluation and prediction: Input the structured feature matrix into a spatio-temporal coupled deep learning model, dynamically predict the evolution trend of concrete strength by parallel processing of time series features and spatial topological features, and output the strength prediction value and the confidence interval evaluation result; S4. Intelligent decision-making and reinforcement execution: Based on the evaluation result of the strength prediction value and the preset safety threshold, autonomously decide the reinforcement strategy and execute the reinforcement operation, and at the same time feedback the reinforcement execution status to the cloud analysis platform through the industrial Internet of Things protocol.

[0007] Preferably, in the step S1: The mechanical sensors include distributed fiber optic strain sensors and piezoelectric ceramic arrays buried inside the concrete; The environmental sensors include temperature and humidity sensors and embedded temperature chains; The hydrographic sensors include pore water pressure gauges and TDR time domain reflectometers.

[0008] Preferably, the step S2 specifically includes: Adopt an improved empirical wavelet transform to perform adaptive frequency band segmentation on the multi-source sensing data; Based on the dynamic time warping algorithm, perform non-linear alignment on the multi-source sensing data; Extract time domain statistics, frequency domain energy entropy, and time-frequency domain wavelet packet coefficients to generate a structured feature matrix.

[0009] Preferably, in the step S3: Process the time series signal through a one-dimensional depthwise separable convolutional network, with a convolution kernel size of 3-4 and a stride of 1-2, to extract local dependence features; Model the sensor spatial topological relationship through a graph attention network, and the edge weights are determined by the sensor spatial distance and the signal coherence coefficient to capture spatial correlation; Adopt a gated recurrent unit to fuse spatio-temporal features, and output the strength prediction value and the confidence interval evaluation result through a Bayesian neural network.

[0010] Preferably, in the step S4: When the strength prediction value is lower than the safety threshold, trigger the dynamic adjustment of the grouting pressure and the epoxy resin ratio; Based on the strain feedback, calibrate the inclination angle and prestress value of the support member in real time.

[0011] The integrated system for evaluating the concrete strength of the foundation pit slope protection pile and intelligent reinforcement includes: The sensing and acquisition module includes a distributed fiber optic strain sensor, a piezoelectric ceramic array, a temperature and humidity sensor, and a TDR time domain reflectometer; The data processing module is integrated into an FPGA-accelerated embedded unit and is configured to perform improved empirical wavelet transform, dynamic time warping algorithm, and time-frequency domain feature extraction; The strength evaluation module has a spatio-temporal coupled deep learning model built-in, including a one-dimensional depthwise separable convolutional network and a graph attention network; The intelligent reinforcement module includes a grouting pressure controller and a hydraulic support adjustment mechanism.

[0012] Preferably, in the sensing and acquisition module: The distributed fiber optic strain sensor is spirally wound around the pile body, and the node spacing decreases exponentially along the pile depth; The piezoelectric ceramic array is embedded inside the pile body concrete in a matrix form to monitor the stress wave propagation characteristics.

[0013] Preferably, the data processing module is configured to: Suppress power frequency interference through improved empirical wavelet transform and retain the signals in the effective frequency band; Use the dynamic time warping algorithm to achieve spatio-temporal alignment of multi-sensor data and eliminate the sampling frequency difference.

[0014] Preferably, in the strength evaluation module: The edge weights of the graph attention network are jointly determined by the sensor spatial distance and the signal coherence coefficient; The Bayesian neural network outputs the strength prediction value and the confidence interval evaluation result, and the confidence interval is 95 - 97%.

[0015] Preferably, the intelligent reinforcement module includes: The grouting robot is equipped with an ultrasonic guided wave imaging device for real-time detection of the slurry diffusion front and generation of a three-dimensional filling distribution map; The hydraulic support system feedback-controls and adjusts the inclination angle of the support rod through a strain sensor, and the adjustment accuracy is ±0.1°.

[0016] The present invention provides an integrated method and system for evaluating the concrete strength of foundation pit retaining piles and intelligent reinforcement. It has the following Beneficial effects: 1. The present invention adopts a technical solution combining a multi-source sensor network and a deep learning model to monitor and evaluate the strength of the concrete of the foundation pit retaining pile in real time, achieving an accurate and non-destructive strength prediction effect. Compared with the traditional core drilling sampling method and ultrasonic testing method, it avoids the problems of destructive testing and limited accuracy, and solves the deficiencies of the existing technology in low accuracy and poor real-time performance under complex geological conditions.

[0017] 2. The present invention adopts an intelligent reinforcement decision-making and execution mechanism. Based on the comparison between the strength prediction value and the safety threshold, the reinforcement process is automatically started. Compared with the traditional solution that relies on manual judgment of the reinforcement timing, the present invention eliminates human errors, reduces the lag of reinforcement decision-making, realizes efficient and automated reinforcement response, and greatly improves the safety of the foundation pit and the construction efficiency of the project.

[0018] 3. The present invention combines time-frequency domain feature extraction technology with a graph neural network model, effectively improving the accuracy and robustness of concrete strength evaluation. Compared with the existing methods based on a single sensor or traditional neural network models, the present invention can better process complex spatio-temporal features and solve the problem of insufficient ability of the existing technology to accurately capture and evaluate signals in a dynamic environment.

[0019] 4. The present invention realizes intelligent monitoring and reinforcement of the foundation pit slope protection piles. The system monitors real-time data and optimizes the reinforcement strategy through feedback, achieving higher reinforcement accuracy and efficiency. Compared with the traditional manual monitoring and manual reinforcement operations, the present invention solves the deficiencies of unstable reinforcement effect and frequent manual intervention in the existing technology through automatic control and optimized adjustment, improves the construction quality and reduces the construction cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is one of the system flow diagrams of the present invention; Figure 2 is another system flow diagram of the present invention; Figure 3 is the third system flow diagram of the present invention; Figure 4 is the method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0022] Please refer to the attached Figures 1-4 , the embodiments of the present invention provide an integrated method and system for evaluating the concrete strength of foundation pit slope protection piles and intelligent reinforcement, including: In the present invention, the first key step of the concrete strength assessment and intelligent reinforcement system for foundation pit retaining piles is data collection and transmission. This step involves obtaining the structural status and environmental change data of the retaining piles in real time through a multi-source sensor network, and transmitting this data to the central computing platform in an efficient and low-latency manner. Effective data collection provides a reliable basis for subsequent strength assessment and reinforcement decision-making. To achieve this goal, the reasonable layout of sensors and accurate data transmission are the keys to the successful implementation of this technical solution.

[0023] In this embodiment, data collection is completed by a variety of sensors deployed at key positions of the retaining piles. The sensor network includes mechanical sensors, environmental sensors, and hydrological sensors, which can comprehensively monitor the dynamic changes around the foundation pit. The data of the sensors are preliminarily processed by edge computing nodes, and then uploaded to the central database in real time using industrial Internet of Things (IIoT) or 5G technology to achieve efficient data transmission and storage.

[0024] In the monitoring of concrete structures, the accuracy and response speed of mechanical sensors are crucial. In this embodiment, mechanical sensors mainly include two types: distributed optical fiber strain sensors (DOFS) and piezoelectric ceramic arrays. Distributed optical fiber strain sensors have extremely high precision and can be arranged longitudinally along the retaining piles to monitor the strain changes of concrete in real time. This sensor can sensitively detect tiny strain changes with an accuracy of ±1μm. Its working principle is to measure the tiny displacement of the optical fiber under stress by reflecting the changes in the optical signal propagating in the optical fiber, and then obtain the strain information. In this way, the stress distribution and the expansion of microcracks in concrete under load can be efficiently captured.

[0025] Piezoelectric ceramic array sensors monitor the early damage of concrete through acoustic emission signals. These sensors respond to changes in external stress by generating internal electrical signals and can highly sensitively detect the occurrence and expansion of microcracks inside the concrete. The layout of the piezoelectric ceramic array inside the retaining piles can provide real-time feedback on the damage status of the concrete and provide an important basis for subsequent assessment.

[0026] Environmental and hydrological sensors are used to monitor the external environmental factors around the foundation pit in real time, especially temperature, humidity, and hydrological conditions. These factors have an important impact on the concrete strength. For example, the change of the groundwater level may cause the concrete to be affected by the humid environment, while the change of temperature and humidity directly affects the cement hydration process of the concrete. Environmental sensors include temperature and humidity composite sensors and embedded temperature chains, which cover key positions in the foundation pit area.

[0027] TDR time domain reflectometer and pore water pressure gauge belong to the core equipment of hydrological sensors. The TDR time domain reflectometer is based on time domain reflection technology and can accurately measure the change of groundwater level with an error accuracy of ±2 cm. The pore water pressure gauge is used to measure the water pressure in the foundation pit soil, and through real-time data analysis, it provides influencing factors of the hydrological environment for strength assessment.

[0028] Once the sensor data is collected, the stability and real-time performance of data transmission become particularly important. In this invention, low-latency industrial Internet of Things protocol (IIoT) or 5G network is adopted for data transmission to ensure that a large amount of real-time data can be quickly transmitted to the cloud database for storage and processing. During the transmission process, the edge computing node is responsible for preliminary filtering and compression of the data to reduce the data volume and improve the data transmission efficiency.

[0029] The key to transmission lies in real-time performance and stability. Especially in a complex foundation pit environment, the interruption or delay of data transmission may lead to a lag in reinforcement decisions and affect the stability of the retaining piles. Therefore, the system adopts redundant design to ensure the reliability of data transmission, and at the same time, real-time monitoring of key data is carried out to avoid packet loss or data missing.

[0030] To further ensure the accuracy and consistency of data, the multi-source sensor data collected needs to be aligned in time and space during the transmission process. Suppose the strain data collected by the sensor is ε(t,x), where t represents time; x represents the spatial position.

[0031] For different types of sensors (such as fiber optic strain sensors, temperature and humidity sensors, etc.), the data collected by each sensor will have different spatio-temporal characteristics. For example, the data S piezo (t,x) of the piezoelectric ceramic sensor may not be completely consistent with the data S DOFS (t,x) of the distributed fiber optic sensor, and its time resolution and spatial distribution will also be different.

[0032] Therefore, in the data preprocessing stage, these data need to be aligned through a time synchronization algorithm, and possible sensor blind areas need to be filled through spatial interpolation methods. The spatio-temporal synchronization process can be modeled by the following formula: Among them, S aligned (t,x) is the sensor data after spatio-temporal synchronization processing, S sensor (t,x) is the original sensor data, and Weight(t,x) is the time and space weighting coefficient, which is used to ensure that the synchronized data accurately reflects the actual measured value.

[0033] Through the above processing, the obtained data structure will be further fed into the data preprocessing module to provide structured data input for subsequent strength prediction and decision-making.

[0034] Through the implementation of the above steps, the data acquisition and transmission link can efficiently and stably obtain the real-time status information of the foundation pit retaining piles. Different types of sensor networks cooperate with each other to comprehensively monitor the changes in the foundation pit environment from different angles and provide important basis for concrete strength evaluation. The low latency and high stability of data transmission ensure the real-time response ability of the system and lay a solid foundation for subsequent strength prediction and intelligent reinforcement decision-making.

[0035] After the data collection of the sensor network is completed in the data acquisition stage, the next key task is to effectively preprocess and extract features from these data. The purpose of this link is to convert multi-source and multi-dimensional raw data into a structured feature matrix suitable for subsequent analysis. This process not only includes noise suppression and data spatio-temporal alignment, but also forms a data input with sufficient representativeness through multi-dimensional feature extraction for use by subsequent strength evaluation models. Through effective preprocessing, the uncertainty in the data can be eliminated, and the prediction accuracy and reliability of subsequent models can be improved.

[0036] In this embodiment, first, the multi-source sensing data collected is subjected to spatio-temporal fusion processing, which mainly includes two parts: noise suppression and spatio-temporal alignment. Wavelet transform algorithm is used for noise suppression, while dynamic time warping algorithm is used for spatio-temporal alignment. Subsequently, feature extraction is carried out, combining the features in the time domain, frequency domain and time-frequency domain to form a structured multi-dimensional feature matrix. These features include but are not limited to time-domain statistics (such as mean, variance), frequency-domain energy entropy, and time-frequency domain wavelet packet coefficients. The original data often contains unnecessary interference information caused by environmental factors, equipment noise or communication interference, etc. These noises will affect the accuracy of subsequent analysis. To improve the reliability of the data, the present invention performs noise filtering through wavelet transform (Wavelet Transform, WT). In the wavelet transform process, Daubechies5 basis function is used to decompose and reconstruct the original data, and an appropriate frequency band is selected for denoising processing. Through wavelet transform, low-frequency or high-frequency noise signals can be effectively filtered out, and only the main effective signals are retained.

[0037] Let the original data be X(t), and the result of wavelet transform is: Among them, ψ j (t) is the Daubechies wavelet basis function, α j is the transformation coefficient, and j is the scale parameter. In this way, we can remove environmental noise from the original data and obtain a smoother and more reliable data input.

[0038] Due to factors such as sensor distribution, different sampling frequencies, and clock synchronization, the data in the sensor network often has asynchrony on the time axis. To ensure the consistency of data in space and time, this embodiment uses the Dynamic Time Warping (DTW) algorithm to perform non-linear alignment on multi-source sensing data. DTW finds the optimal matching path by minimizing the distance between two time series, aligning the data on the time axis, and thus eliminating the timing errors caused by sampling frequency differences or signal transmission delays.

[0039] Suppose there are two sets of time series X = (x1, x2, …, x m ) and Y = (y1, y2, …, y n ). DTW calculates the shortest distance path between them, which is defined as: where D(i, j) represents the cumulative distance when matching the i-th and j-th data points, and d(x i , y j ) is the Euclidean distance between x i and y j .

[0040] Through dynamic time warping, the time series of all sensors can be aligned to ensure their consistency in space and time.

[0041] After noise suppression and spatio-temporal alignment, the data will enter the feature extraction stage. The purpose of feature extraction is to extract key features from the time series data that are helpful for subsequent intensity prediction. The present invention adopts a multi-dimensional feature extraction method in the time domain, frequency domain, and time-frequency domain.

[0042] Time-domain feature extraction: Time-domain features usually include statistics such as the mean, variance, kurtosis, and skewness of the data. The mean and variance respectively represent the central position and fluctuation amplitude of the data, and are important indicators for analyzing signal characteristics.

[0043] The mean is a description of the overall level of a set of data, representing the average value of all data points. In the formula, x i represents the i-th data point, and N is the total number of data points. By summing all data points and dividing by the total number of data points, the average value of the data can be calculated. Mean(x) represents the mean of the data, and Variance(x) represents the variance.

[0044] Variance is used to measure the degree of dispersion of data points, that is, the degree of deviation of data points from the mean. When calculating, it is necessary to first calculate the mean, and then for each data point, find the difference between it and the mean, square the difference, and finally find the average of all these squared differences. The larger the variance, the higher the degree of dispersion of the data.

[0045] Frequency domain feature extraction: Frequency domain features are based on the Fast Fourier Transform to extract the spectral information of the signal. By calculating the spectrum of the signal, the energy distribution of each frequency component in the signal can be obtained. Frequency domain features are often used to analyze periodic or regular signals.

[0046] This formula is the standard expression of the Fast Fourier Transform, used to convert a time-domain signal into a frequency-domain signal. By performing the Fourier transform on the time-domain signal x n the spectral information of the signal at different frequencies can be obtained. X(f) is the frequency-domain signal, representing the signal component at frequency f.

[0047] x n is the nth data point in the original time-domain signal; N is the total number of sampling points of the signal; f is the frequency; j is the imaginary unit.

[0048] By calculating the Fourier transform, the amplitude and phase information of each frequency component in the signal can be obtained. This is very useful for analyzing the periodic or regular components in the signal and is often used to detect periodic vibrations, noises, and the frequency characteristics of signals.

[0049] This formula represents the wavelet packet transform, which combines the characteristics of the time domain and the frequency domain and is suitable for processing non-stationary signals. The wavelet packet transform extracts multi-scale features by decomposing the time-domain and frequency-domain information of the signal. The α in the formula k is the wavelet packet coefficient, ψ k (t) is the wavelet packet basis function, and x(t) is the original signal.

[0050] α k is the wavelet packet coefficient, representing the energy distribution of the signal at different scales and frequencies.

[0051] ψ k (t) is the wavelet packet basis function, used to decompose the signal.

[0052] t represents time.

[0053] The wavelet packet transform is particularly suitable for analyzing signals with time-varying characteristics and can capture the details of the signal in both time and frequency simultaneously. This makes the wavelet packet transform widely used in the processing of non-stationary signals (such as transient signals and non-periodic signals).

[0054] Through the processes of noise suppression, spatio-temporal alignment, and multi-dimensional feature extraction, this embodiment successfully converts multi-source and heterogeneous sensor data into a structured feature matrix. These matrices contain key information about the evolution of concrete strength, providing accurate data support for subsequent strength assessment models. This preprocessing process ensures the efficiency and accuracy of the system, laying a solid foundation for intelligent reinforcement decisions for foundation pit slope protection piles.

[0055] After the data preprocessing and feature extraction stages are completed, the next step is strength assessment and prediction. The goal of this step is to predict the concrete strength based on the feature matrix extracted in the previous steps, combined with a deep learning model. By establishing a spatio-temporal coupled deep learning model, it is able to simultaneously process time series features and spatial topological features, dynamically predict the evolution trend of concrete strength, and output the predicted value of strength and the corresponding confidence interval assessment results. This link is closely connected to the feature extraction in the previous step S2 and relies on data features from different sensors to generate predictions.

[0056] In this embodiment, the strength assessment uses a hybrid model based on a deep neural network (DNN) and a graph neural network (GNN), combines the local dependence features of time series signals and the topological relationship of sensor spatial positions, and uses a gated recurrent unit (GRU) and a Bayesian neural network to predict the strength of concrete. This model can adaptively adjust the prediction strategy, process features in different time and space dimensions, and thus provide a more accurate and stable assessment.

[0057] The core of strength assessment is to model the feature data based on a deep neural network (DNN) and a graph neural network (GNN) to form a strength prediction model. In this process, the deep neural network first processes the time series data through a one-dimensional convolutional neural network (1D-CNN) to extract the local dependence features in the time series signal; the graph neural network is responsible for modeling the spatial topological relationship of the sensors and capturing the correlation of spatial data. The two are combined through a fusion layer, and the attention mechanism is used to weight the outputs of the time series features and spatial features, thereby obtaining the final strength prediction value.

[0058] A one-dimensional convolutional neural network (1D-CNN) is used to process the time series data from the sensors. During the convolution process, assuming the time series of the signal is x n , the convolution operation can be expressed as: where y n is the convolution result, w kis the convolution kernel, k is the size of the convolution kernel, and n is the current time step. The convolution operation helps the model extract local temporal features from the input temporal data, especially the short-term fluctuations and trends of the signal.

[0059] The graph neural network is used to model the spatial topological relationship between sensors. Assume the spatial position of the sensor is x i , and the graph neural network learns the spatial dependence relationship through the adjacency matrix A between sensors. The graph convolution operation can be expressed as: where, represents the feature representation of the i-th node in the l-th layer, N(i) is the set of nodes adjacent to node i, A ij is the element of the adjacency matrix, W (l) and b (l) are the weights and biases of the l-th layer. Through the graph neural network, the model can effectively capture the spatial dependence and structural features between sensors.

[0060] The processing of temporal signals needs to consider long-term dependencies. To enhance the memory ability of the model for temporal data, the present invention adopts a gated recurrent unit (GRU) to process temporal features. GRU is a variant of the recurrent neural network (RNN), which can effectively capture long-term dependencies in sequential data while maintaining computational efficiency.

[0061] Assume h t is the hidden state at time t, and the update rule of GRU is as follows: r t = σ(W r · [h t-1 , x t + b r ); z t = σ(W z · [h t-1 , x t + b z ); where, r t is the reset gate, z t is the update gate, is the candidate hidden state, x t is the input data, and W and b are the learned weights and biases. GRU adjusts the memory ability of the network through the gating mechanism, thereby effectively processing the temporal changes of concrete strength.

[0062] To improve the credibility and robustness of the prediction results, in this embodiment, a Bayesian neural network is used to estimate the confidence interval of the strength evaluation results. The Bayesian neural network not only outputs the predicted value of the concrete strength but also can calculate the confidence interval of the strength prediction. This enables the system to quantify the uncertainty and further provides a basis for the reinforcement decision-making.

[0063] In the Bayesian neural network, assume the prediction result is Through the variational inference method, the predicted value and the estimation of its confidence interval can be obtained.

[0064] Among them, E[f(x)] is the expectation of the strength prediction value, and Var[f(x)] is the variance of the strength prediction value, which reflects the reliability of the prediction result.

[0065] By combining deep neural networks, graph neural networks, gated recurrent units (GRUs), and Bayesian neural networks, this embodiment provides a strength evaluation and prediction model that can accurately predict the strength evolution trend of the foundation pit slope protection pile concrete and output the strength prediction value and its confidence interval. Through the deep fusion of spatio-temporal features, the model effectively improves the accuracy and robustness of the strength prediction and provides reliable data support for subsequent intelligent reinforcement decision-making.

[0066] Based on the aforementioned data collection, feature extraction, and strength evaluation, the intelligent decision-making and reinforcement execution steps are crucial. This step automatically makes decisions and executes concrete strength reinforcement operations based on the strength prediction results of the previous stage. During the reinforcement execution process, not only the difference between the predicted strength value and the preset safety threshold needs to be considered, but also the reinforcement execution status needs to be fed back in real time to ensure the effectiveness and safety of the system. The accuracy of the intelligent decision-making, the precision of the reinforcement operation, and the stability of the feedback mechanism directly affect the safety and economy of the entire system.

[0067] In this embodiment, when the strength evaluation is completed by the model combining deep neural networks and graph neural networks, the system will automatically make a reinforcement decision. The decision-making basis is the comparison result between the set safety threshold and the predicted strength value. Once the predicted value is lower than the safety threshold, the system will trigger an automated reinforcement process. This process includes adjusting the grouting pressure and the proportion of the reinforcement material, calibrating and adjusting the support members, and monitoring and optimizing the reinforcement effect through a real-time feedback mechanism.

[0068] The intelligent decision-making mechanism is the core of this embodiment. Based on the comparison between the strength prediction value and the safety threshold, it automatically activates the reinforcement response. When the system detects that the concrete strength is lower than the preset threshold, the intelligent decision-making mechanism will dynamically calculate the operation parameters required for reinforcement according to the predicted strength value and generate a reinforcement plan. Assume the current strength prediction value of the system is The safety threshold is The decision logic is as follows: Yellow warning: If Initiate manual review and prepare additional contingency plans; Orange warning: If Initiate preparatory grouting operation.

[0069] Red warning: If Immediately initiate full-automatic reinforcement, including grouting and support structure adjustment.

[0070] This process is automatically completed by the algorithm, greatly reducing the error of manual judgment and improving the response speed.

[0071] The reinforcement execution mainly consists of two parts: grouting and support adjustment. For different predicted strength values, the system will automatically adjust the grouting pressure, the ratio of epoxy resin, and the inclination angle and prestress of the support rods.

[0072] The grouting operation injects the reinforcement material by controlling the pressure of the grouting equipment. When the automatic reinforcement process is initiated, the system will dynamically adjust the pressure and proportion of the grouting material according to the difference between the predicted strength value and the safety threshold. Assuming that the grouting pressure P and the ratio of epoxy resin R are two key parameters for the reinforcement operation, the adjustment formula is: where P max is the maximum grouting pressure, R max is the maximum ratio of epoxy resin, is the predicted strength, is the safety threshold. The grouting pressure and the ratio of epoxy resin increase as the predicted strength value decreases, thus ensuring the accuracy of the reinforcement effect.

[0073] In addition to grouting, the adjustment of the support structure is equally important. After detecting a decrease in concrete strength and triggering the reinforcement response, the system will adjust the angle and prestress of the support structure through the hydraulic servo system. Through the real-time feedback of the strain sensor, the system will calculate the required support force F support , and adjust the inclination angle θ and prestress P support of the support rods in real time. The adjustment formula is: F support = k·Δε; θ = θ initial +Δθ; where k is the stiffness of the support rod, Δε is the strain change detected by the sensor, θ initial is the initial support angle, Δθ is the angle adjustment amount, P max is the maximum support force, and They are the predicted strength and the safety threshold respectively.

[0074] The reinforcement operation is not only executed, but also requires real-time monitoring and feedback during the execution process. In this embodiment, the execution status of all reinforcement operations is synchronously updated in real time through Internet of Things (IoT) devices and a central database. After each reinforcement operation, the system will adjust the reinforcement strategy in real time according to the feedback information. Through real-time data monitoring, the system can obtain the feedback on the reinforcement effect and further optimize the subsequent reinforcement decisions.

[0075] After the reinforcement operation is completed, the system calculates the evaluation value E of the reinforcement effect effect , and conducts an uncertainty assessment of the reinforcement effect through a Bayesian network to further adjust the reinforcement strategy. For example, if the reinforcement effect does not meet the expectation (E effect <0.9), the system will recalculate the parameters of grouting and support adjustment and re-execute the reinforcement operation.

[0076] Through intelligent decision-making and automated reinforcement execution, this embodiment realizes the intelligent reinforcement process of the concrete strength of the foundation pit retaining piles. Based on the real-time strength prediction value, the system automatically makes decisions and accurately executes the reinforcement operation, which not only improves the response speed of the reinforcement but also ensures the accuracy and effectiveness of the reinforcement. The automated reinforcement execution mechanism significantly reduces the risk of manual intervention and enhances the safety and reliability of the overall project.

[0077] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An integrated method for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles, characterized in that: The following steps are involved: S1. Data collection and transmission: Real-time data collection of slope protection pile structure status and environmental parameters through a multi-source sensor network, wherein the sensor network includes mechanical sensors, environmental sensors and hydrological sensors deployed in a spatially distributed manner; S2, data preprocessing and feature extraction: perform spatiotemporal fusion processing on the collected multi-source sensor data, including noise suppression, spatiotemporal benchmark unification and multi-dimensional feature extraction, and generate a structured feature matrix; S3, strength assessment and prediction: input the structured feature matrix into the spatiotemporal coupling deep learning model, dynamically predict the evolution trend of concrete strength by processing the temporal features and spatial topological features in parallel, and output the strength prediction value and confidence interval assessment results; S4. Intelligent decision-making and reinforcement execution: Based on the evaluation results of the strength prediction value and the preset safety threshold, the reinforcement strategy is independently decided and the reinforcement operation is executed. At the same time, the reinforcement execution status is fed back to the cloud analysis platform through the industrial Internet of Things protocol.

2. The integrated method for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 1 is characterized in that: In the step S1: The mechanical sensor includes a distributed optical fiber strain sensor and a piezoelectric ceramic array buried inside the concrete; The environmental sensor includes a temperature and humidity sensor and an embedded temperature chain; The hydrological sensor includes a pore water pressure meter and a TDR time domain reflectometer.

3. The integrated method for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 1 is characterized in that: The S2 step specifically includes: Adopting improved empirical wavelet transform to adaptively segment the multi-source sensor data into frequency bands; Nonlinear alignment of multi-source sensor data based on dynamic time warping algorithm; The time domain statistics, frequency domain energy entropy and time-frequency domain wavelet packet coefficients are extracted to generate a structured feature matrix.

4. The method for integrating concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 1 is characterized in that: In step S3: The time series signal is processed by a one-dimensional depthwise separable convolutional network with a convolution kernel size of 3-4 and a step size of 1-2 to extract local dependency features; The spatial topological relationship of sensors is modeled through the graph attention network, and the edge weight is determined by the sensor spatial distance and signal coherence coefficient; The gated recurrent unit is used to fuse the spatiotemporal features, and the intensity prediction value and confidence interval evaluation results are output through a Bayesian neural network.

5. The method for integrating concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 1 is characterized in that: In step S4: When the predicted strength value is lower than the safety threshold, the dynamic adjustment of the grouting pressure and epoxy resin ratio is triggered; The inclination angle and prestress value of the support rod are calibrated in real time based on strain feedback.

6. An integrated system for evaluating the concrete strength of foundation pit slope protection piles and intelligent reinforcement, according to the integrated method for evaluating the concrete strength of foundation pit slope protection piles and intelligent reinforcement as claimed in any one of claims 1 to 5, characterized in that: include: The sensing acquisition module includes a distributed optical fiber strain sensor, a piezoelectric ceramic array, a temperature and humidity sensor, and a TDR time domain reflectometer; A data processing module, integrated in an embedded unit accelerated by an FPGA, configured to perform improved empirical wavelet transform, dynamic time warping algorithm, and time-frequency domain feature extraction; Strength assessment module with built-in spatiotemporal coupling deep learning model, including one-dimensional deep separable convolutional network and graph attention network; The intelligent reinforcement module includes a grouting pressure controller and a hydraulic support adjustment mechanism.

7. The integrated system for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 6 is characterized in that: In the sensor acquisition module: Distributed optical fiber strain sensors are laid in the pile in a spiral winding manner, and the node spacing decreases exponentially along the pile depth; The piezoelectric ceramic array is embedded in the pile concrete in a matrix form to monitor the stress wave propagation characteristics.

8. The integrated system for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 6 is characterized in that: The data processing module is configured as follows: By improving the empirical wavelet transform, power frequency interference is suppressed and the effective frequency band signal is retained; The dynamic time warping algorithm is used to achieve spatiotemporal alignment of multi-sensor data and eliminate sampling frequency differences.

9. The integrated system for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 6 is characterized in that: In the strength assessment module: The edge weights of the graph attention network are determined by the sensor space distance and the signal coherence coefficient; The Bayesian neural network output intensity prediction value and confidence interval evaluation results, with a confidence interval of 95-97%.

10. The integrated system for concrete strength assessment and intelligent reinforcement of foundation pit slope protection piles according to claim 6, characterized in that: The intelligent reinforcement module comprises: The grouting robot is equipped with an ultrasonic guided wave imaging device to detect the slurry diffusion front in real time and generate a three-dimensional filling distribution map; The hydraulic support system adjusts the inclination angle of the support rod through a strain sensor feedback closed loop.

Images