Iot-driven underground pipe gallery grouting material leakage monitoring method and system

Abstract

The application discloses an Internet of Things driven underground pipe gallery grouting material leakage monitoring method and system, relates to the technical field of data processing and analysis, and comprises the following steps: obtaining a dynamic parameter data set; generating a grouting leakage feature vector; positioning potential leakage coordinates according to the leakage risk level; and triggering a hierarchical early warning instruction through the Internet of Things platform. The application solves the technical problems of isolated multi-source data, low positioning accuracy and early warning lag in the traditional monitoring method in underground pipe gallery grouting engineering, achieves multi-source data fusion analysis, improves positioning accuracy, can timely perform leakage early warning, and further guarantees the safe and stable operation of underground pipe gallery grouting engineering.

Description

Technical Field

[0001] This invention relates to the field of data processing and analysis technology, and in particular to an Internet of Things-driven method and system for monitoring leakage of grouting materials in underground utility tunnels. Background Technology

[0002] In the construction and operation of underground utility tunnels, the leakage of grouting materials is crucial to the structural safety and service life of the tunnels. Currently, traditional methods for monitoring grouting material leakage in underground utility tunnels mainly rely on regular manual inspections or simple single-point sensor monitoring. Manual inspections are not only inefficient and time-consuming, but also greatly affected by subjective factors, making it difficult to detect hidden leakage points; single-point sensor monitoring can only obtain local information and cannot comprehensively grasp the overall leakage situation of the tunnel. With the development of technologies such as the Internet of Things and big data, the requirements for the accuracy and real-time performance of underground utility tunnel leakage monitoring are increasing. The limitations of these traditional monitoring methods are becoming increasingly apparent, failing to meet the needs for precise location, real-time monitoring, and effective early warning of grouting material leakage in underground utility tunnels. More advanced monitoring technologies are urgently needed to ensure the safe and stable operation of underground utility tunnels. Summary of the Invention

[0003] This application addresses the technical problems of isolated multi-source data, low positioning accuracy, and delayed early warning in traditional monitoring methods for underground utility tunnel grouting projects. This application collects dynamic parameter data by deploying a multi-source sensor network in the grouting section of the underground utility tunnel. Multimodal feature extraction generates grouting leakage feature vectors, and a leakage risk model determines the leakage risk level and locates potential leakage coordinates. Combining the risk level and coordinates, a leakage monitoring report is generated and a tiered early warning system is triggered. This allows for accurate monitoring of grouting material leakage in underground utility tunnels, making grouting leakage monitoring more efficient, accurate, and reliable.

[0004] To address the aforementioned technical issues, this application proposes a technical solution for a method and system for monitoring leakage of grouting materials in underground utility tunnels driven by the Internet of Things.

[0005] In a first aspect, this application provides an IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels. The method includes: real-time data acquisition of the underground utility tunnel via an IoT sensor network to obtain a dynamic parameter dataset; multimodal feature extraction of the dynamic parameter data to generate a grouting leakage feature vector; inputting the grouting leakage feature vector into a leakage risk model for real-time analysis to determine the leakage risk level; locating potential leakage coordinates based on the leakage risk level; generating a leakage monitoring report based on the leakage risk level and the potential leakage coordinates; sending the leakage monitoring report to an IoT platform; and triggering a tiered early warning command through the IoT platform.

[0006] Secondly, this application provides an IoT-driven underground utility tunnel grouting material leakage monitoring system, wherein the system includes: a dataset acquisition module, used to collect data on the underground utility tunnel in real time through an IoT sensor network to obtain a dynamic parameter dataset; a vector generation module, used to extract multimodal features from the dynamic parameter data to generate a grouting leakage feature vector; a coordinate positioning module, used to input the grouting leakage feature vector into a leakage risk model for real-time analysis to determine the leakage risk level and locate potential leakage coordinates based on the leakage risk level; and an instruction triggering module, used to generate a leakage monitoring report based on the leakage risk level and the potential leakage coordinates, send the leakage monitoring report to an IoT platform, and trigger a graded early warning instruction through the IoT platform.

[0007] This application proposes one or more technical solutions, which have at least the following technical effects:

[0008] This application utilizes an IoT sensor network deployed in the grouting area of ​​underground utility tunnels to collect dynamic parameter datasets in real time. Multimodal feature extraction is performed on the data to generate grouting leakage feature vectors. These vectors are then input into a leakage risk model to determine the leakage risk level and potential leakage coordinates. Based on the risk level and coordinates, a leakage monitoring report is generated, triggering tiered early warnings. In this process, multi-source sensor collaborative acquisition achieves multi-source data fusion, and 3D modeling and sound source localization technologies improve positioning accuracy. Real-time analysis and a tiered early warning mechanism address the issue of early warning lag. This achieves the technical effects of realizing multi-source data fusion analysis, improving positioning accuracy, and enabling timely leakage early warnings, thereby ensuring the safe and stable operation of underground utility tunnel grouting projects.

[0009] The above outlines the method and system for monitoring leakage of grouting materials in underground utility tunnels driven by the Internet of Things. The following detailed embodiments will describe the steps of the technical solution in detail to enable those skilled in the art to have a clear and complete understanding of this application. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic flowchart of the IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels provided in this application embodiment.

[0012] Figure 2 This is a schematic diagram of the structure of the IoT-driven underground utility tunnel grouting material leakage monitoring system provided in this application embodiment.

[0013] Figure labeling: Dataset acquisition module 1, vector generation module 2, coordinate positioning module 3, command triggering module 4. Detailed Implementation

[0014] This application utilizes a multi-source sensor network deployed in the grouting section of underground utility tunnels to collect data. After preprocessing, multimodal feature extraction is performed to generate grouting leakage feature vectors. These vectors are input into a leakage risk model to determine the risk level and potential leakage coordinates, thereby generating a leakage monitoring report and triggering tiered early warnings. If multiple sensor data anomalies are found, a comprehensive analysis is conducted to determine key monitoring areas. Simultaneously, the monitoring frequency is adjusted according to the risk level. Finally, the monitoring data is categorized and stored to provide a basis for subsequent maintenance and management. This achieves the technical effect of realizing multi-source data fusion analysis, improving positioning accuracy, and enabling timely leakage early warning, thereby ensuring the safe and stable operation of underground utility tunnel grouting projects.

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0017] Example 1, as Figure 1 As shown, an IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels includes:

[0018] Step A100: Collect data in real time from the underground utility tunnel using an IoT sensor network to obtain a dynamic parameter dataset.

[0019] In this embodiment, the IoT sensor network is a multi-source sensor network deployed in the grouting section of the underground utility tunnel. The underground utility tunnel is a public tunnel space in an urban area used for the centralized laying of municipal pipelines. The dynamic parameter dataset is the original dynamic parameter dataset generated in real time by the multi-source sensor network from the grouting area of ​​the underground utility tunnel.

[0020] Specifically, a multi-source sensor network, such as distributed fiber optic pressure sensors and acoustic emission sensors, is deployed in the grouting section of the underground utility tunnel based on the Internet of Things (IoT). These sensors collect data in real time from the grouting area, generating a raw dynamic parameter dataset containing various information. Subsequently, this raw data undergoes multi-level preprocessing to generate a dynamic parameter dataset for subsequent analysis, laying a solid foundation for accurate leakage monitoring. The specific steps are detailed in A110-A120.

[0021] Step A200: Perform multimodal feature extraction on the dynamic parameter data to generate a grouting leakage feature vector.

[0022] In this embodiment, multimodal feature extraction is the process of parsing multiple data from a dynamic parameter dataset, extracting corresponding features using specific methods, and then normalizing them. The grouting leakage feature vector is a vector generated by normalizing the pressure fluctuation feature matrix, flow velocity chaos feature vector, and frequency band energy distribution feature obtained during the multimodal feature extraction process.

[0023] Optionally, the dynamic parameter dataset is parsed to obtain grouting pressure time-series data, grout flow velocity time-series data, and acoustic waveform data. Time-frequency domain joint analysis, nonlinear dynamic analysis, and multi-scale decomposition are then performed on these data to extract pressure fluctuation feature matrix, flow velocity chaotic feature vector, and frequency band energy distribution features. These features are then normalized to generate grouting leakage feature vector. Specific steps are detailed in A210-A250.

[0024] By generating this feature vector, the leakage status of the grouting material can be reflected more accurately, providing key data support for subsequent leakage risk assessment.

[0025] Step A300: Input the grouting leakage feature vector into the leakage risk model for real-time analysis, determine the leakage risk level, and locate the potential leakage coordinates based on the leakage risk level.

[0026] In this embodiment, the leakage risk model is a model constructed by integrating convolutional neural networks and random forest algorithms. The leakage risk level is obtained by real-time analysis of the grouting leakage feature vector input into the leakage risk model. The potential leakage coordinates are the spatial coordinates of possible leakage locations determined based on the leakage risk level.

[0027] In one embodiment of this application, the grouting leakage feature vector is synchronized to a pre-trained leakage risk model. The specific steps of the model generation process are detailed in A351-A353. A leakage event probability distribution map is generated through the model's multi-scale spatiotemporal feature extraction layer. Dynamic mode decomposition is performed to extract dominant mode data. Based on this, historical leakage similarity calculation is performed to determine the similarity matrix, thereby determining the leakage risk level. The specific steps are detailed in A310-A340. Simultaneously, a strain dataset is obtained by extracting a distributed fiber optic pressure sensor array through a multi-source sensor network for strain analysis. Based on this, a three-dimensional pressure distribution model of the underground utility tunnel is constructed. Then, acoustic emission sensors are extracted to locate the sound source and determine the sound source location data. Finally, the coordinates are calculated by combining the three-dimensional pressure distribution model and the sound source location data to determine the potential leakage coordinates. The specific steps are detailed in A360-A390.

[0028] By determining the leakage risk level and locating the coordinates of potential leaks, effective assessment of leakage risks and precise location of potential leak points can be achieved, ensuring the safe and stable operation of underground utility tunnels.

[0029] Step A400: Generate a leakage monitoring report based on the leakage risk level and the potential leakage coordinates, send the leakage monitoring report to the Internet of Things (IoT) platform, and trigger a graded early warning command through the IoT platform.

[0030] In this embodiment, the leakage monitoring report is a comprehensive analysis report generated based on the leakage risk level and potential leakage coordinates. The tiered early warning command is a warning command issued by the IoT platform based on the leakage monitoring report, indicating different risk levels.

[0031] Specifically, based on the leakage risk level and potential leakage coordinates, a grouting risk analysis is first performed on the underground utility tunnel to generate a risk level label. Then, a grouting leakage analysis is performed to obtain the three-dimensional coordinates of the leakage point. Next, the grouting prediction is performed by combining the risk level label and the three-dimensional coordinates of the leakage point to obtain the leakage diffusion prediction range. Finally, based on the leakage diffusion prediction range and according to the environmental parameter impact factors, the risk level label and the three-dimensional coordinates of the leakage point are linked and integrated to generate a leakage monitoring report. The specific steps are explained in detail in A410-A440.

[0032] The generated leakage monitoring report is transmitted to the cloud server via the communication interface of the IoT platform. The platform system first analyzes and verifies key data in the report, such as the risk level label, the three-dimensional coordinates of the leakage point, and the predicted range of leakage diffusion. After successful verification, the platform triggers the corresponding level of warning instruction according to the preset graded warning rules (such as low, medium, and high risk thresholds). For example, when a high-risk leakage is detected, the platform will automatically send an immediate notification containing the leakage coordinates and diffusion range to the maintenance personnel, and simultaneously activate the emergency equipment in the utility tunnel to initiate a local sealing procedure.

[0033] Through the above steps, the risk of leakage during grouting in underground utility tunnels can be effectively controlled.

[0034] Furthermore, step A100 in the method provided in this application embodiment includes:

[0035] A110: Based on the Internet of Things, a multi-source sensor network is deployed in the grouting section of the underground utility tunnel. The multi-source sensor network is used to collect data in real time on the grouting area of ​​the underground utility tunnel and generate a raw dynamic parameter dataset.

[0036] A120: Perform multi-level preprocessing on the original dynamic parameter dataset to generate a dynamic parameter dataset.

[0037] In this embodiment, the multi-source sensor network includes a distributed fiber optic pressure sensor array, a high-precision flow sensor, and an acoustic emission sensor network. The original dynamic parameter dataset is a data set generated based on real-time data acquisition of the underground utility tunnel grouting area using the multi-source sensor network. Multi-level preprocessing includes noise filtering, anomaly verification, and timestamp alignment. The dynamic parameter dataset is the dataset obtained after multi-level preprocessing of the original dynamic parameter dataset.

[0038] Specifically, a multi-source sensor network is deployed in the grouting section of the underground utility tunnel. A distributed fiber optic pressure sensor array accurately collects grouting pressure data; sensors are placed at regular intervals (e.g., 10 meters) to acquire real-time pressure changes at different locations with an accuracy of up to 0.1 MPa. High-precision flow sensors collect grout velocity data with an accuracy of ±1%, ensuring accurate velocity information. An acoustic emission sensor network collects acoustic signature data, sensitively capturing minute sound changes within the tunnel, providing multi-dimensional data support for leak detection. This data is then integrated and aggregated—that is, a unified data format is used to ensure consistency, and the data is arranged chronologically, linking data collected by different sensors at the same time to form a complete dataset. This generates a raw dynamic parameter dataset containing grouting pressure data, grout velocity data, and acoustic signature data.

[0039] Next, the original dynamic parameter dataset undergoes multi-level preprocessing. First, noise filtering is performed to remove noise caused by sensor errors and environmental electromagnetic interference, making the data more stable and reliable. Then, abnormal data verification is performed by setting reasonable data thresholds and logical rules. For grouting pressure data, these thresholds can be set to 1MPa-8MPa, and for grout flow rate data, they can be set to 0.5-2m. 3 / h, the voiceprint signal data can be set between 50-500Hz. Data outside this range can be judged as abnormal. Identify and remove abnormal data that are obviously inconsistent with the actual situation. Finally, timestamp alignment is performed to ensure that the data collected by different types of sensors are synchronized in time, and a dynamic parameter dataset is generated to facilitate subsequent comprehensive analysis.

[0040] By acquiring and preprocessing the original dynamic parameter dataset, a dynamic parameter dataset that can be used for subsequent analysis is generated, laying a solid data foundation for accurately monitoring the leakage of grouting materials in underground utility tunnels.

[0041] Furthermore, step A200 in the method provided in this application embodiment includes:

[0042] A210: Analyze the dynamic parameter dataset to obtain grouting pressure time-series data, grout flow rate time-series data, and acoustic waveform data.

[0043] A220: Perform time-frequency domain joint analysis on the grouting pressure time series data to extract the pressure fluctuation feature matrix.

[0044] A230: Perform nonlinear dynamic analysis on the time-series data of the slurry flow velocity to generate a chaotic feature vector of the flow velocity.

[0045] A240: Perform multi-scale decomposition on the acoustic waveform data to extract frequency band energy distribution features.

[0046] A250: Normalize the pressure fluctuation feature matrix, the flow velocity chaotic feature vector, and the frequency band energy distribution feature to generate the grouting leakage feature vector.

[0047] Optionally, the dynamic parameter dataset is first parsed to separate the grouting pressure time series data, grout flow rate time series data, and acoustic waveform data.

[0048] Next, a joint time-frequency domain analysis was performed on the grouting pressure time-series data. First, Fourier transform was used to decompose the time-varying grouting pressure signal into a superposition of sine and cosine waves of different frequencies to obtain global frequency information. Then, wavelet transform was used to reflect the frequency changes of the signal at different time points by convolving the signal with a series of wavelet functions obtained by scaling and shifting the mother wavelet function. Combining the results of Fourier transform and wavelet transform, and comprehensively considering the information of the pressure signal at different times and frequencies, the pressure fluctuation characteristics at different times and frequencies were organized into a matrix form, thus generating the pressure fluctuation characteristic matrix.

[0049] Then, nonlinear dynamic analysis is performed on the slurry flow velocity time series data, because the flow velocity change of slurry during leakage often exhibits nonlinear chaotic characteristics. The delay time and embedding dimension are determined first by phase space reconstruction and calculation of the Lyapunov exponent. The delay time is generally determined by the autocorrelation function (calculating the correlation between the time series of slurry flow velocity data and its delay series; the correlation changes as the delay time increases; generally, the delay time corresponding to the first decrease of the autocorrelation function value to a certain proportion of the initial value (e.g., 1 / e) is selected). The embedding dimension is commonly calculated using the spurious nearest neighbor method (mapping data points in the low-dimensional space to the high-dimensional space, calculating the nearest neighbors of each data point, and determining whether these nearest neighbors are still nearest neighbors in the higher-dimensional space; as the embedding dimension increases, the proportion of spurious nearest neighbors changes; when the proportion of spurious nearest neighbors is below a certain threshold (e.g., 5%), the embedding dimension is considered appropriate). After determining the parameters, phase space reconstruction is applied to one-dimensional slurry velocity time-series data, extending it to a higher-dimensional space to reveal the nonlinear characteristics inherent in the data. In phase space, phase points and their neighboring points are selected, and the distance changes between these neighboring points during the evolution process are calculated. Over time, the average exponential divergence rate of these distances, i.e., the Lyapunov exponent, is calculated. If the Lyapunov exponent is greater than zero, it indicates that the system exhibits chaotic characteristics; if it is less than zero, it indicates that the system is stable, ultimately generating a chaotic eigenvector of the flow velocity.

[0050] For the voiceprint signal waveform data, multi-scale decomposition is performed. The Empirical Mode Decomposition (EMD) algorithm is used to determine all maxima and minima in the voiceprint signal. Cubic spline interpolation is used to fit the upper and lower envelopes, and the mean of the upper and lower envelopes is calculated to obtain an average envelope. The average envelope is subtracted from the original voiceprint signal to obtain a preliminary Intrinsic Mode Function (IMF). Next, it is determined whether the IMF meets the conditions for an IMF. If not, the above steps are repeated for filtering. IMFs that meet the conditions are separated, and the remaining signals continue the decomposition process until the remaining signals are monotonic functions or constants. Through this processing, the complex voiceprint signal is decomposed into multiple IMFs of different frequencies. By analyzing the energy distribution of these IMFs in different frequency bands, the frequency band energy distribution characteristics are extracted.

[0051] Finally, the extracted pressure fluctuation feature matrix, flow velocity chaos feature vector, and frequency band energy distribution feature are normalized using a weighted feature fusion algorithm. First, the weights of the pressure fluctuation feature matrix, flow velocity chaos feature vector, and frequency band energy distribution feature are determined, based on the importance of each feature in reflecting the leakage of grouting materials. Next, each feature is normalized separately, mapping its value to a specific interval, such as [0, 1], to eliminate differences in dimensions and numerical ranges. For example, a minimum-maximum normalization formula can be used. Where x is the original feature value, x max and x min These are the minimum and maximum values ​​of the feature. Then, based on the determined weights, the normalized features are weighted and calculated, that is, each normalized feature is multiplied by its corresponding weight and then summed to obtain a fused and normalized grouting leakage feature vector, which is used for subsequent leakage risk analysis.

[0052] The above steps provide comprehensive and accurate data support for the leakage risk model, improving the accuracy and reliability of leakage monitoring.

[0053] Furthermore, step A300 in the method provided in this application embodiment includes:

[0054] A310: Input the grouting leakage feature vector into the leakage risk model for real-time analysis to determine the leakage risk level.

[0055] A320: Synchronize the grouting leakage feature vector to the pre-trained leakage risk model, and generate a leakage event probability distribution map through the multi-scale spatiotemporal feature extraction layer of the leakage risk model.

[0056] A330: Perform dynamic mode decomposition on the probability distribution map of the leakage event and extract the dominant mode data.

[0057] A340: Based on the dominant mode, perform historical leakage similarity calculations to determine the similarity matrix, and set the leakage risk level according to the similarity matrix.

[0058] In this embodiment, the multi-scale spatiotemporal feature extraction layer is a key component of the leakage risk model, and this layer processes vectors. The leakage event probability distribution map is a graph generated by the model's multi-scale spatiotemporal feature extraction layer. The dominant modal data is the data extracted after performing dynamic modal decomposition on the leakage event probability distribution map.

[0059] Specifically, firstly, the grouting leakage feature vector is input into the leakage risk model for real-time analysis. This vector integrates feature information from multiple aspects such as grouting pressure, grout flow rate, and acoustic signature, and can comprehensively reflect the state of the grouting process.

[0060] Next, the grouting leakage feature vectors are synchronized to a pre-trained leakage risk model. This model's multi-scale spatiotemporal feature extraction layer uses multiple convolutional and pooling layers to deeply analyze the vectors. In the convolutional layers, kernels of different sizes slide across the vectors. Small kernels capture local details such as short-term fluctuations in grouting pressure and small-area changes in grout flow velocity, while large kernels acquire macroscopic features such as long-term trends in grouting pressure and the distribution characteristics of grout flow velocity in different areas of the pipe gallery, transforming these into feature maps. The pooling layers then filter and reduce the dimensionality of the feature maps output by the convolutional layers; for example, max pooling selects the maximum value in a small area to highlight significant features and filters out unimportant details. Through alternating processing by multiple convolutional and pooling layers, the vectors are comprehensively and deeply analyzed from different scales and spatiotemporal dimensions. Finally, the processed features are integrated to generate a leakage event probability distribution map, which shows the probability of leakage events occurring in different areas.

[0061] Then, a cascaded classifier is used to classify the risk level of the leakage event probability distribution map. The cascaded classifier sequentially filters and classifies the data through multiple classifiers. The first classifier performs preliminary filtering of the data in the leakage event probability distribution map based on preset rules and features, identifying some obvious features and trends, and classifying the data into preliminary categories. Subsequent classifiers then further analyze and classify the data based on the results of the previous classifier. In this step-by-step filtering and classification process, various factors are comprehensively considered (factors related to grouting leakage feature vectors, leakage event probability distribution map factors, dominant modal data factors, and historical leakage similarity calculation factors) to accurately determine low-risk, medium-risk, and high-risk levels. Simultaneously, dynamic modal decomposition technology is used to process the leakage event probability distribution map and extract dominant modal data. During the decomposition process, the amplitude and frequency parameters of the dominant mode are obtained. The amplitude reflects the intensity of leakage-related features, while the frequency parameter reflects the rate of feature change. These parameters reflect the main characteristics of the leakage event and are crucial for judging the degree of risk.

[0062] Finally, historical leakage similarity calculations are performed based on the dominant mode, using extracted dominant mode data, which includes the amplitude and frequency parameters of the dominant mode. Next, historically known leakage data is collected, covering leakage characteristics at different times and under different conditions. Then, the current dominant mode data is compared point-by-point with the historical leakage data, and the cosine similarity algorithm is used to calculate the degree of similarity between the two.

[0063] First, the current dominant mode and historical leakage data are converted into vector form, with elements representing characteristic values ​​such as amplitude and frequency. The product of corresponding elements of two vectors is multiplied and then summed. The square root of the sum of the squares of the elements of each vector is then calculated; this is similar to measuring the "length" of the vector. The summation result is divided by the product of the two "lengths" to obtain the cosine similarity value. The closer the value is to 1, the more similar the two vectors are; the closer it is to -1, the greater the difference; and 0 indicates that they are independent and unrelated. The degree of similarity is determined by calculating this for each set of data. Leakage risk levels are then set according to a three-tiered risk threshold: similarity < 0.3 is low risk, 0.3 ≤ similarity < 0.7 is medium risk, and similarity ≥ 0.7 is high risk.

[0064] By following the steps above, leakage risks can be assessed more accurately, improving the accuracy and reliability of leakage monitoring for grouting materials in underground utility tunnels.

[0065] Furthermore, step A350 in the method provided in this application embodiment includes:

[0066] A351: A convolutional neural network is used to extract spatiotemporal features from the grouting leakage feature vector to generate a multi-scale feature map, which is located in the multi-scale spatiotemporal feature extraction layer.

[0067] A352: Perform time series modeling on the multi-scale feature map to construct a leakage probability prediction function.

[0068] A353: Combine the random forest algorithm to optimize the decision boundary of the leakage probability prediction function and generate the leakage risk model.

[0069] In this embodiment, the multi-scale feature map is generated by extracting spatiotemporal features from the grouting leakage feature vector using a convolutional neural network. The leakage probability prediction function is a function constructed by time-series modeling of the multi-scale feature map. The random forest algorithm is used to optimize the decision boundary of the leakage probability prediction function.

[0070] Specifically, firstly, convolutional kernels of different sizes are prepared. These kernels act like "feature detectors," sliding along the vector when processing the grouting leakage feature vector. When a small kernel slides, it focuses on local areas within the vector, capturing detailed features such as minute fluctuations in grouting pressure over a short period and changes in grout flow rate within a small area. When a large kernel slides, it focuses on a more macroscopic range, such as the trend of grouting pressure over a longer period and the distribution characteristics of grout flow rate in different areas of the entire pipe gallery.

[0071] As the convolution kernel slides, each element in the kernel is multiplied by the corresponding element of the grouting leakage feature vector region it covers. These products are then summed to obtain a convolution result. As the kernel slides point-by-point along the vector, a series of such convolution results are generated, forming a new feature map. Multiple convolution kernels of different sizes generate multiple different feature maps, reflecting the characteristics of the grouting leakage feature vector at different scales. Combining these feature maps at different scales transforms them into multi-scale feature maps. These maps reside in a multi-scale spatiotemporal feature extraction layer, presenting the grouting leakage features from different scales, temporal dimensions, and spatial dimensions.

[0072] Next, time series modeling is performed on the multi-scale feature maps, and a leakage probability prediction function is constructed using a long short-term memory network.

[0073] Step A: Arrange the multi-scale feature maps in chronological order and use them as input data for a Long Short-Term Memory (LSTM) network. The LSTM network includes an input gate, a forget gate, and an output gate. The input gate controls the input of new information from the multi-scale feature maps, the forget gate determines whether to retain or discard past information, and the output gate determines the output content. This process handles and learns multi-scale feature maps from different time points.

[0074] Step B: When processing input data, the LSTM network learns multi-scale feature maps at different times based on its own structure. Through a gating mechanism, it can effectively capture long-term dependencies in time series data, preventing the loss of key information due to time span. For example, when dealing with features that change over time, such as grouting pressure and grout flow rate, LSTM can remember the impact of earlier features on the current moment.

[0075] Step C: As time progresses, the LSTM network continuously updates its internal state, gradually uncovering the intrinsic relationship between multi-scale feature maps and leakage probability during the learning process (the causal relationship and mutual influence between the changing trends and fluctuations of features such as grouting pressure, grout flow rate, and acoustic signature signals in the multi-scale feature maps and the probability of leakage in underground pipe corridors).

[0076] Step D: After multiple iterations of training, the backpropagation algorithm first calculates the difference between the predicted leakage probability and the actual leakage probability to obtain the error. Then, starting from the output layer, it calculates the derivative of the activation function with respect to the error and, combined with the network weights, propagates the error forward step by step to calculate the gradient of each layer. This determines how to adjust the weights and parameters to reduce the error and make the network output closer to the true value. Finally, the trained LSTM network is used as the core computing unit. Multi-scale feature map data is input, and after internal network processing, the predicted leakage probability is output, thus constructing a function model with multi-scale feature maps as input and leakage probability as output.

[0077] Finally, the decision boundary of the leakage probability prediction function is optimized by combining the random forest algorithm to generate a leakage risk model.

[0078] The Random Forest algorithm first randomly selects multiple subsets of samples with replacement from the training dataset. For each subset, a decision tree is constructed. During the construction of the decision tree, for each node split, a subset of features is randomly selected from numerous features, and the optimal split point is found among these features. After the numerous decision trees form a forest, each decision tree makes a prediction on the input multi-scale feature map data. The prediction results of all decision trees are then integrated, typically using a majority voting method (for classification problems). The predictions of each decision tree are counted, and the prediction with the most frequent occurrence is taken as the final prediction result. By continuously adjusting the parameters of the random forest, such as the number of decision trees and the number of features randomly selected for each node, the prediction results are made closer to the actual leakage probability, thereby optimizing the decision boundary of the leakage probability prediction function.

[0079] After optimizing the decision boundary of the leakage probability prediction function using the random forest algorithm, the optimized function is combined with the leakage probability prediction function previously constructed using multi-scale feature maps extracted by convolutional neural networks and time series modeling. The grouting leakage feature vector obtained through multi-modal feature extraction is used as the input to the entire model. First, a multi-scale feature map is generated through spatiotemporal feature extraction using a convolutional neural network. Then, the multi-scale feature map is input into the optimized leakage probability prediction function. Based on the function calculation and the judgment of the decision boundary, the corresponding leakage risk level is output, thus generating a leakage risk model that can accurately assess the leakage risk of underground utility tunnel grouting materials.

[0080] Through the above steps, the generated leakage risk model can more accurately assess the leakage risk of underground utility tunnel grouting materials, effectively improving the accuracy and reliability of leakage monitoring and providing strong support for ensuring the safe operation of underground utility tunnels.

[0081] Furthermore, step A300 in the method provided in this application embodiment includes:

[0082] A360: Extract the distributed fiber optic pressure sensor array through the multi-source sensor network, perform strain analysis based on the distributed fiber optic pressure sensor array, and obtain a strain dataset.

[0083] A370: Perform three-dimensional modeling of the underground utility tunnel according to the strain dataset to construct a three-dimensional pressure distribution model.

[0084] A380: Extract acoustic emission sensors through the multi-source sensor network, locate the sound source based on the acoustic emission sensors, and determine the sound source location data.

[0085] A390: Based on the three-dimensional pressure distribution model and the sound source location data, coordinate calculation is performed to determine the coordinates of the potential leakage.

[0086] In this embodiment, the three-dimensional pressure distribution model is a model constructed by three-dimensional modeling of the underground utility tunnel based on the strain dataset. The sound source localization data is determined by using the propagation characteristics of sound signals, such as the time difference of arrival at different sensors, to locate the sound source.

[0087] Specifically, the process begins by extracting data from a distributed fiber optic pressure sensor array using a multi-source sensor network. Strain analysis is then performed on this pressure data. The collected pressure data is preprocessed to remove noise and outliers. Next, based on the principles of materials mechanics and the characteristics of fiber optic sensing, a mathematical model of the relationship between pressure and strain is established. Those skilled in the art first determine parameters such as the elastic modulus of commonly used materials in underground utility tunnels (like concrete of specific grades) using theories such as Hooke's Law in materials mechanics, thus clarifying the basic ratio of stress to strain. Regarding fiber optic sensing, the photoelastic effect of optical fibers is studied. For example, using the fiber Bragg grating sensing principle, experimental data on the Bragg wavelength drift of a specific fiber under different pressures is obtained to understand the correlation between pressure and changes in the fiber's refractive index. Combining these two aspects, an equation is established encompassing utility tunnel material parameters, fiber optic parameters, and the relationship between pressure and strain. Furthermore, considering the influence of interference factors such as temperature on the fiber optics and utility tunnel materials, correction coefficients are introduced through experimental or theoretical analysis, and temperature correction terms are added, ultimately forming a mathematical model that accurately describes the relationship between pressure and strain and is used to calculate strain data from pressure data.

[0088] Using this model, pressure data is converted into strain data. Considering the differences in pressure distribution at different locations within the utility tunnel, data for different areas are converted and calculated separately. During the calculation process, factors such as the material properties and geometric structure of the utility tunnel are also considered for correction to more accurately reflect the actual strain of the tunnel structure. Finally, these calculated and corrected strain data are organized and summarized to form a strain dataset reflecting the deformation of the utility tunnel structure.

[0089] Next, a 3D model of the underground utility tunnel is constructed based on the obtained strain dataset, creating a 3D pressure distribution model. The collected strain data is organized and preprocessed (data cleaning, normalization, and format conversion). Then, using professional 3D modeling software (SolidWorks), the geometric structure of the underground utility tunnel is accurately constructed, clearly defining its shape, dimensions, and the relative positions of its various parts. The strain data is then mapped onto the 3D model of the tunnel, associating the strain data at each location with the corresponding point in the model. Based on the principles of mechanics of materials, the stress and pressure values ​​at each point in the tunnel are calculated using the strain data. To more intuitively display the pressure distribution, visualization methods such as color and contour lines are used to render the pressure values. Finally, the constructed model is validated and optimized. The calculation results of the model are compared with actual monitoring data or theoretical analysis results. Based on the comparison results, the model parameters are adjusted to improve the accuracy and reliability of the model, thus completing the construction of the 3D pressure distribution model.

[0090] Then, acoustic emission sensor data is extracted through a multi-source sensor network. Acoustic emission sensors can capture minute acoustic signals generated inside the pipe gallery due to leaks or other reasons. Based on the propagation characteristics of these acoustic signals and the time difference of their arrival at the sensor, sound source localization is performed, thereby determining the sound source localization data.

[0091] Finally, the three-dimensional pressure distribution model was combined with the sound source location data for coordinate calculation. The three-dimensional pressure distribution model provided background information on the spatial structure and pressure distribution of the utility tunnel, while the sound source location data determined the approximate area where leakage might occur. By combining these two sets of data, the coordinates of potential leaks could be accurately calculated.

[0092] The above steps improve the accuracy and reliability of monitoring leakage of grouting materials in underground utility tunnels, providing strong support for timely remedial measures.

[0093] Furthermore, step A400 in the method provided in this application embodiment includes:

[0094] A410: Based on the leakage risk level and the potential leakage coordinates, perform grouting risk analysis on the underground utility tunnel and generate risk level labels.

[0095] A420: Based on the leakage risk level and the potential leakage coordinates, perform grouting leakage analysis on the underground utility tunnel to generate three-dimensional coordinates of the leakage point.

[0096] A430: Based on the risk level label and the three-dimensional coordinates of the leakage point, grouting prediction is performed on the underground utility tunnel to obtain the predicted leakage diffusion range.

[0097] A440: Based on the predicted leakage diffusion range and according to the environmental parameter impact factors, the risk level label and the three-dimensional coordinates of the leakage point are correlated and integrated to generate the leakage monitoring report.

[0098] In this embodiment, grouting risk analysis is conducted by constructing a grouting risk analysis checklist to analyze leakage risk levels and potential leakage coordinates. Grouting leakage analysis utilizes Geographic Information System (GIS) and Computer-Aided Design (CAD) technologies to integrate leakage risk levels and potential leakage coordinates into a three-dimensional model of the underground utility tunnel. Grouting prediction simulates and predicts the diffusion trend of grouting material leakage, providing a predicted range for leakage diffusion. Environmental parameter influencing factors refer to external environmental factors affecting grouting leakage in the utility tunnel, including surrounding soil moisture, changes in groundwater level, ambient temperature, and air pressure.

[0099] In one embodiment, after the system obtains the leakage risk level and potential leakage coordinates, the first step is to conduct a grouting risk analysis. A detailed grouting risk analysis checklist is developed, listing various factors related to the leakage risk level and potential leakage coordinates, along with their corresponding risk descriptions. When developing the grouting risk analysis checklist, extensive data on past underground utility tunnel grouting projects are collected, including different types of leakage accidents and their corresponding grouting treatments, while also referencing relevant industry standards and specifications. Based on this, factors closely related to the leakage risk level and potential leakage coordinates are identified, such as the utility tunnel structure type (rectangular, circular), the distribution of pipelines around the leakage point, and the basis for classifying the leakage risk level.

[0100] Based on these factors, risk levels are categorized as low, medium, and high. For each risk level, a detailed description of the corresponding risk situation is provided, along with the coordinates of potential leaks. For example, when the leakage risk level is low, and the potential leak coordinates are located in a non-critical area of ​​the utility tunnel with no important pipelines nearby, it is described as "due to the relatively independent location of the leak point, the simple surrounding environment, and the low difficulty of grouting repair, it is judged as low risk." All identified factors, risk levels, and corresponding risk descriptions are arranged in a logical order to create a grouting risk analysis checklist, which is continuously optimized and improved based on new feedback during practical application.

[0101] Next, grouting leakage analysis was conducted. Using the three-dimensional pressure distribution model of the underground utility tunnel constructed in step A370, the leakage risk level and potential leakage coordinates were integrated. Simultaneously, the system collected multi-source sensor information, including strain data from distributed fiber optic sensors within the tunnel and internal pressure data monitored by pressure sensors. Based on this data, the leakage process of the grouting material from the potential leakage point was simulated in the three-dimensional model. The possible flow direction was inferred based on the pressure difference, material properties, and the internal spatial structure of the tunnel. Fluid dynamics algorithms were used to continuously correct and optimize the simulation path according to the flow laws of fluids in complex spaces. The simulated path was compared and verified with actual data such as the approximate leakage area located by acoustic emission sensors. After multiple adjustments and calculations, the precise location of the leakage point in three-dimensional space was calculated.

[0102] Next, grouting prediction is conducted. A large number of historical cases with similar characteristics to the current underground utility tunnel are collected. These characteristics include the tunnel's material, structural type, geological environment, and leakage situation. For example, past cases with similar soil conditions, similar tunnel structures, and similar leakage risk levels and leakage point locations are identified. The leakage and diffusion of grouting materials in these cases are analyzed, such as diffusion rate, diffusion direction, and final diffusion range. Based on the current utility tunnel's risk level label and the three-dimensional coordinates of the leakage point, and referring to the leakage and diffusion data under similar conditions in historical cases, the predicted leakage and diffusion range of the current utility tunnel's grouting material is inferred by analogy. For example, if in a historical case, under similar conditions, the grouting material diffused 2 meters axially and 0.5 meters radially within 24 hours, this can be used as a reference for the current utility tunnel's leakage and diffusion range.

[0103] Finally, a leakage monitoring report is generated. Environmental parameters such as surrounding soil moisture and groundwater level changes are taken into account. Risk level labels, three-dimensional coordinates of leakage points, and predicted leakage spread range are linked and integrated. Using professional report generation software, a clear and detailed leakage monitoring report is automatically generated. The report covers basic information about the utility tunnel, leakage risk status, precise location of leakage points, and estimated future leakage spread range, providing comprehensive and reliable decision-making basis for the operation and maintenance management of the utility tunnel.

[0104] In summary, the IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels provided in this application has the following technical advantages:

[0105] This application establishes a data transmission link between an IoT sensor network and a leakage risk model, processes the data using multimodal feature extraction technology, and obtains grouting leakage feature vectors through parsing and analysis. These vectors are then analyzed in real-time within the leakage risk model. Through historical leakage similarity calculations, dynamic modal decomposition, and mechanisms such as risk level classification and coordinate positioning, potential leakage coordinates are determined based on a three-dimensional pressure distribution model and sound source location data, generating a leakage monitoring report. This ensures effective monitoring of leakage in underground utility tunnel grouting materials, achieving the technical effect of multi-source data fusion analysis, improving positioning accuracy, and enabling timely leakage early warning, thereby ensuring the safe and stable operation of underground utility tunnel grouting projects.

[0106] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides an Internet of Things-driven underground utility tunnel grouting material leakage monitoring system, the system comprising:

[0107] Data set acquisition module 1 is used to collect dynamic parameter datasets in real time from underground utility tunnels through an Internet of Things sensor network.

[0108] Vector generation module 2 is used to extract multimodal features from the dynamic parameter data and generate grouting leakage feature vectors.

[0109] The coordinate positioning module 3 is used to input the grouting leakage feature vector into the leakage risk model for real-time analysis, determine the leakage risk level, and locate the potential leakage coordinates according to the leakage risk level.

[0110] The instruction triggering module 4 is used to generate a leakage monitoring report based on the leakage risk level and the potential leakage coordinates, send the leakage monitoring report to the Internet of Things platform, and trigger a graded early warning instruction through the Internet of Things platform.

[0111] Furthermore, the dataset acquisition module 1 is used to perform the following steps:

[0112] Based on the Internet of Things, a multi-source sensor network is deployed in the grouting section of the underground utility tunnel. The multi-source sensor network collects data in real time on the grouting area of ​​the underground utility tunnel and generates a raw dynamic parameter dataset.

[0113] The original dynamic parameter dataset is preprocessed in multiple stages to generate a new dynamic parameter dataset.

[0114] Furthermore, the vector generation module 2 is used to perform the following steps:

[0115] The dynamic parameter dataset is analyzed to obtain grouting pressure time-series data, grout flow rate time-series data, and acoustic waveform data.

[0116] A joint time-frequency domain analysis was performed on the grouting pressure time series data to extract the pressure fluctuation feature matrix.

[0117] Nonlinear dynamic analysis is performed on the time-series data of the slurry flow velocity to generate a chaotic feature vector of the flow velocity.

[0118] The waveform data of the voiceprint signal is decomposed into multiple scales to extract the frequency band energy distribution characteristics.

[0119] The pressure fluctuation feature matrix, the flow velocity chaotic feature vector, and the frequency band energy distribution feature are normalized to generate the grouting leakage feature vector.

[0120] Furthermore, the coordinate positioning module 3 is used to perform the following steps:

[0121] The grouting leakage feature vector is input into the leakage risk model for real-time analysis to determine the leakage risk level.

[0122] The grouting leakage feature vector is synchronized to the pre-trained leakage risk model, and a leakage event probability distribution map is generated through the multi-scale spatiotemporal feature extraction layer of the leakage risk model.

[0123] Dynamic mode decomposition is performed on the probability distribution map of the leakage event to extract the dominant mode data.

[0124] Based on the dominant mode, historical leakage similarity calculations are performed to determine the similarity matrix, and the leakage risk level is set according to the similarity matrix.

[0125] Furthermore, the coordinate positioning module 3 is used to perform the following steps:

[0126] A convolutional neural network is used to extract spatiotemporal features from the grouting leakage feature vector to generate a multi-scale feature map, which is located in the multi-scale spatiotemporal feature extraction layer.

[0127] Time series modeling is performed on the multi-scale feature map to construct a leakage probability prediction function.

[0128] The decision boundary of the leakage probability prediction function is optimized by combining the random forest algorithm to generate the leakage risk model.

[0129] Furthermore, the coordinate positioning module 3 is used to perform the following steps:

[0130] The distributed fiber optic pressure sensor array is extracted through the multi-source sensor network, and strain analysis is performed based on the distributed fiber optic pressure sensor array to obtain a strain dataset.

[0131] Based on the strain dataset, a three-dimensional model of the underground utility tunnel was created to construct a three-dimensional pressure distribution model.

[0132] The acoustic emission sensors are extracted through the multi-source sensor network, and the sound source is located based on the acoustic emission sensors to determine the sound source location data.

[0133] The coordinates of the potential leakage are determined by calculating the coordinates based on the three-dimensional pressure distribution model and the sound source location data.

[0134] Furthermore, the instruction triggering module 4 is used to perform the following steps:

[0135] Based on the leakage risk level and the potential leakage coordinates, a grouting risk analysis is performed on the underground utility tunnel to generate a risk level label.

[0136] Based on the leakage risk level and the potential leakage coordinates, grouting leakage analysis is performed on the underground utility tunnel to generate three-dimensional coordinates of the leakage points.

[0137] Based on the risk level label and the three-dimensional coordinates of the leakage point, grouting prediction is performed on the underground utility tunnel to obtain the predicted leakage diffusion range.

[0138] Based on the predicted leakage diffusion range, the risk level label and the three-dimensional coordinates of the leakage point are linked and integrated according to the environmental parameter impact factors to generate the leakage monitoring report.

[0139] The IoT-driven underground utility tunnel grouting material leakage monitoring system provided in this embodiment of the invention can execute the IoT-driven underground utility tunnel grouting material leakage monitoring method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0140] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0141] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for monitoring leakage of grouting materials in underground utility tunnels driven by the Internet of Things, characterized in that, The method includes: The underground utility tunnel is monitored in real time using an Internet of Things (IoT) sensor network to obtain a dynamic parameter dataset. Multimodal feature extraction is performed on the dynamic parameter data to generate a grouting leakage feature vector; The grouting leakage feature vector is input into the leakage risk model for real-time analysis to determine the leakage risk level, and the potential leakage coordinates are located based on the leakage risk level. A leakage monitoring report is generated based on the leakage risk level and the potential leakage coordinates. The leakage monitoring report is sent to the Internet of Things (IoT) platform, and a graded early warning command is triggered through the IoT platform. The method for extracting multimodal features from the dynamic parameter data to generate grouting leakage feature vectors includes: The dynamic parameter dataset is parsed to obtain grouting pressure time-series data, grout flow rate time-series data, and acoustic waveform data; The time-frequency domain joint analysis of the grouting pressure time series data is performed to extract the pressure fluctuation feature matrix; Nonlinear dynamic analysis is performed on the time-series data of the slurry flow velocity to generate a chaotic flow velocity feature vector; The waveform data of the acoustic signature signal is decomposed into multiple scales to extract the frequency band energy distribution characteristics; The pressure fluctuation feature matrix, the flow velocity chaotic feature vector, and the frequency band energy distribution feature are normalized to generate the grouting leakage feature vector. The grouting leakage feature vector is input into a leakage risk model for real-time analysis to determine the leakage risk level. The method includes: The grouting leakage feature vector is input into the leakage risk model for real-time analysis to determine the leakage risk level. The grouting leakage feature vector is synchronized to the pre-trained leakage risk model, and a leakage event probability distribution map is generated through the multi-scale spatiotemporal feature extraction layer of the leakage risk model. Dynamic mode decomposition was performed on the probability distribution map of the leakage event to extract the dominant mode data; Based on the dominant mode, historical leakage similarity calculations are performed to determine the similarity matrix, and the leakage risk level is set according to the similarity matrix. The method for locating potential leak coordinates based on the aforementioned leak risk level includes: A distributed fiber optic pressure sensor array is extracted through a multi-source sensor network, and strain analysis is performed based on the distributed fiber optic pressure sensor array to obtain a strain dataset. Based on the strain dataset, a three-dimensional model of the underground utility tunnel was created to construct a three-dimensional pressure distribution model. The acoustic emission sensors are extracted through the multi-source sensor network, and the sound source is located based on the acoustic emission sensors to determine the sound source location data. The coordinates of the potential leakage are determined by calculating the coordinates based on the three-dimensional pressure distribution model and the sound source location data.

2. The IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels as described in claim 1, characterized in that, Real-time data collection of dynamic parameters from underground utility tunnels is achieved through IoT sensor networks. Methods include: Based on the Internet of Things, a multi-source sensor network is deployed in the grouting section of the underground utility tunnel. The multi-source sensor network collects data in real time on the grouting area of ​​the underground utility tunnel and generates a raw dynamic parameter dataset. The original dynamic parameter dataset is preprocessed in multiple stages to generate a new dynamic parameter dataset.

3. The IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels as described in claim 1, characterized in that, The process of generating the leakage risk model includes: A convolutional neural network is used to extract spatiotemporal features from the grouting leakage feature vector to generate a multi-scale feature map, which is located in the multi-scale spatiotemporal feature extraction layer. Time series modeling is performed on the multi-scale feature map to construct a leakage probability prediction function; The decision boundary of the leakage probability prediction function is optimized by combining the random forest algorithm to generate the leakage risk model.

4. The IoT-driven method for monitoring leakage of grouting materials in underground utility tunnels as described in claim 1, characterized in that, A leakage monitoring report is generated based on the leakage risk level and the potential leakage coordinates, the method including: Based on the leakage risk level and the potential leakage coordinates, a grouting risk analysis is performed on the underground utility tunnel to generate a risk level label. Based on the leakage risk level and the potential leakage coordinates, grouting leakage analysis is performed on the underground utility tunnel to generate three-dimensional coordinates of the leakage points. Based on the risk level label and the three-dimensional coordinates of the leakage point, grouting prediction is performed on the underground utility tunnel to obtain the predicted leakage diffusion range. Based on the predicted leakage diffusion range, the risk level label and the three-dimensional coordinates of the leakage point are linked and integrated according to the environmental parameter impact factors to generate the leakage monitoring report.

5. An Internet of Things-driven underground utility tunnel grouting material leakage monitoring system, characterized in that, For implementing the Internet of Things-driven method for monitoring leakage of grouting materials in underground utility tunnels as described in any one of claims 1-4, the system comprises: The dataset acquisition module is used to collect dynamic parameter datasets in real time from underground utility tunnels through an IoT sensor network; The vector generation module is used to extract multimodal features from the dynamic parameter data and generate grouting leakage feature vectors. The coordinate positioning module is used to input the grouting leakage feature vector into the leakage risk model for real-time analysis, determine the leakage risk level, and locate the potential leakage coordinates based on the leakage risk level. The instruction triggering module is used to generate a leakage monitoring report based on the leakage risk level and the potential leakage coordinates, send the leakage monitoring report to the Internet of Things platform, and trigger a graded early warning instruction through the Internet of Things platform.

Images