Tunnel grouting material application effect prediction method and system based on multi-scale layering

Abstract

The disclosure provides a tunnel grouting material application effect prediction method and system based on multi-scale layering, relating to the technical field of grouting material prediction, obtaining multi-source data; the multi-source data is weighted and integrated to form a key data table; the data is divided into levels to form geological global structure layer data, geological local structure layer data, and semantic layer data of grouting material and hydrological data; the geological global structure layer data, the geological local structure layer data and the semantic layer data are input into a multi-task prediction model, the geological global structure layer data and the geological local structure layer data are input into a GNN network structure to extract geological topological structure features, the semantic layer data is input into a DNN network structure to extract high-dimensional semantic features, the geological topological structure features and the high-dimensional semantic features are extracted in a shared feature layer to obtain a shared feature map, and different target prediction results are output based on the shared feature map through a multi-source output regression model.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of grouting treatment material prediction, in particular to a tunnel grouting material application effect prediction method and system based on multi-scale layering. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] Tunnel gushing water is a common construction risk, especially when passing through complex geology or high water pressure areas. Water gushing into the tunnel excavation face not only increases the construction difficulty and prolongs the construction period, but also may even cause safety accidents and threaten the safety of construction personnel. Therefore, the reasonable application and effect prediction of grouting materials have become the key in gushing water treatment.

[0004] However, the existing technology still has deficiencies in predicting the grouting effect, and it is difficult to accurately predict the key performance indicators such as impermeability, diffusion range and consolidation strength of the material. Mainly because the commonly used grouting treatment methods at this stage rely on experience design and static models, lack the ability of dynamic prediction and feedback optimization, and are difficult to respond to complex geological conditions and sudden hydrological environment in construction in real time. This leads to the inability to adjust the proportioning and grouting parameters of the grouting material in real time, affecting the actual treatment effect of the grouting material. In addition, the grouting effect is affected by multiple factors such as geology, hydrology, material and construction parameters, and there is a complex nonlinear relationship between these factors. Traditional methods are difficult to capture these relationships and effectively predict and optimize in the treatment process. SUMMARY

[0005] In order to solve the above problems, the present disclosure proposes a tunnel grouting material application effect prediction method and system based on multi-scale layering. Through multi-level data layering and scale division, independent multi-output layers are constructed, fine modeling is performed for multi-task targets such as impermeability, diffusion range and consolidation strength, and each output layer optimizes parameters independently according to different task requirements, so that each prediction has high-precision independence. At the same time, in order to realize the coordinated optimization of multi-target tasks, a weighted loss function and an adaptive weight distribution strategy are adopted to ensure the balance of the model between multiple task targets, significantly improving the comprehensiveness and stability of the prediction model.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions:

[0007] The tunnel grouting material application effect prediction method based on multi-scale layering comprises:

[0008] The multi-source data of tunnel geological data, sudden gushing water hydrological data, grouting material data and construction parameters are acquired and preprocessed; the preprocessed multi-source data is weighted and integrated according to geological structure, hydrological characteristics, material characteristics and construction characteristics to form a key data table;

[0009] The data of the key data table is divided into levels according to multi-scale hierarchical dimensions to form geological global structure layer data, geological local structure layer data and semantic layer data of grouting material and hydrological data;

[0010] The geological global structure layer data, the geological local structure layer data and the semantic layer data are input into a multi-task prediction model, in which the geological global structure layer data and the geological local structure layer data are input into a GNN network structure to extract geological topological structure features, the semantic layer data are input into a DNN network structure to extract high-dimensional semantic features, the geological topological structure features and the high-dimensional semantic features are extracted into a shared feature map in a shared feature layer, and a target prediction result is output based on the shared feature map through multi-source output regression.

[0011] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0012] The tunnel grouting material application effect prediction system based on multi-scale hierarchical layering comprises:

[0013] The data acquisition module is configured to acquire multi-source data of tunnel geological data, sudden gushing water hydrological data, grouting material data and construction parameters, and pre-process the multi-source data; the pre-processed multi-source data is weighted and integrated according to geological structure, hydrological characteristics, material characteristics and construction characteristics to form a key data table;

[0014] The hierarchical module is configured to divide the data of the key data table into levels according to multi-scale hierarchical dimensions to form geological global structure layer data, geological local structure layer data and semantic layer data of grouting material and hydrological data;

[0015] The prediction module is configured to input the geological global structure layer data, the geological local structure layer data and the semantic layer data into a multi-task prediction model, in which the geological global structure layer data and the geological local structure layer data are input into a GNN network structure to extract geological topological structure features, the semantic layer data are input into a DNN network structure to extract high-dimensional semantic features, the geological topological structure features and the high-dimensional semantic features are extracted into a shared feature map in a shared feature layer, and a target prediction result is output based on the shared feature map through multi-source output regression.

[0016] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0017] A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the method for predicting the application effect of tunnel grouting materials based on multi-scale layering.

[0018] According to some embodiments, the present disclosure adopts the following technical solutions:

[0019] An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the method for predicting the application effect of tunnel grouting materials based on multi-scale layering.

[0020] Compared with the prior art, the beneficial effects of this disclosure are as follows:

[0021] This disclosed method for predicting the application effect of tunnel grouting materials based on multi-scale layering achieves cross-scale feature extraction from geological, hydrological, material, and construction data through macro-, meso-, and micro-level data stratification and scale division. The macro-level identifies the stability of the large-scale global geological structure, the meso-level assesses local surrounding rock characteristics, and the micro-level analyzes the diffusion and solidification characteristics of the grouting material in a sudden water inrush environment. This multi-scale feature extraction method significantly enhances the model's feature representation capability, laying a solid foundation for accurate prediction under complex tunnel conditions.

[0022] This disclosed method for predicting the application effect of tunnel grouting materials based on multi-scale hierarchical structure combines DNN and GNN to construct a multi-task prediction model. GNN captures the spatial correlation of geological and hydrological features, while DNN processes high-dimensional grouting material and construction parameter data, forming a more refined multi-objective prediction model. The GNN component is responsible for transforming geological structures such as faults and water flow paths into topological structures, while the DNN learns the physicochemical properties of the grouting materials. This joint model fully leverages the advantages of different algorithms to accurately describe the multi-dimensional data feature relationships in complex tunnel environments.

[0023] This disclosed method for predicting the application effect of tunnel grouting materials based on multi-scale hierarchical layering introduces a shared feature layer, which fuses features extracted by GNN and DNN in the same feature space to improve information sharing efficiency. A weighted aggregation strategy is employed to ensure the effectiveness of different data features, and residual connections (ResNet) are used to mitigate the gradient vanishing problem caused by excessively deep layers. The shared layer integrates multi-source information, enhancing the model's ability to extract common features and improving prediction accuracy and data sharing effectiveness.

[0024] This disclosed method for predicting the application effect of tunnel grouting materials based on multi-scale layering constructs a multi-source output regression model to perform detailed modeling of the task objectives of impermeability, diffusion range, and consolidation strength. The model independently optimizes parameters according to different task requirements, ensuring high-precision independence of each prediction. Furthermore, to achieve coordinated optimization of multiple tasks, a weighted loss function and adaptive weight allocation strategy are adopted to ensure the model's balance among multiple task objectives, significantly improving the comprehensiveness and stability of the prediction model.

[0025] This disclosed method for predicting the application effect of tunnel grouting materials based on multi-scale layering receives the latest data during construction (such as the grouting material mix ratio and rheological properties) in real time and performs dynamic predictions based on current geological and hydrological conditions. If the predicted effect deviates from the target standard, the system will automatically trigger a feedback optimization mechanism to adjust the component ratio of the grouting material and the configuration of grouting parameters to ensure the stability of the treatment effect. This real-time feedback and optimization function ensures that various parameters during construction can be dynamically adjusted according to environmental changes, enhancing the system's responsiveness and treatment reliability in complex environments. Attached Figure Description

[0026] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0027] Figure 1 This is a diagram illustrating the implementation architecture of the method application process in this embodiment of the disclosure. Detailed Implementation

[0028] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0029] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0030] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0031] Example 1

[0032] One embodiment of this disclosure provides a method for predicting the application effect of tunnel grouting materials based on multi-scale layering, including the following steps:

[0033] Step 1: Acquire multi-source data including tunnel geological data, sudden water inrush hydrological data, grouting material data, and construction parameters, and preprocess the data; then weight and integrate the preprocessed multi-source data according to geological structure, hydrological characteristics, material characteristics, and construction characteristics to form a key data table.

[0034] Step 2: Divide the data in the key data table into layers according to multi-scale hierarchical dimensions to form semantic layer data of global geological structure, local geological structure, grouting material, and hydrological data;

[0035] Step 3: Input the global geological structure layer data, local geological structure layer data, and semantic layer data into the multi-task prediction model. In the multi-task prediction model, the global geological structure layer data and local geological structure layer data are input into the GNN network structure to extract geological topological features, and the semantic layer data is input into the DNN network structure to extract high-dimensional semantic features. The geological topological features and high-dimensional semantic features are used to extract a shared feature map in the shared feature layer. Based on the shared feature map, the target prediction result is output through multi-source output regression.

[0036] As one embodiment, the specific implementation process of the method for predicting the application effect of tunnel grouting materials based on multi-scale layering disclosed herein is as follows:

[0037] Step 1: Acquire multi-source data including tunnel geological data, sudden water inrush hydrological data, grouting material data, and construction parameters, and preprocess the data;

[0038] Specifically, (1) the source and acquisition of data

[0039] First, geological exploration technology and on-site monitoring systems are used to collect geological structure information of the tunnel area. Specifically, the system obtains the distribution of rock strata through drilling, ultrasonic imaging, and other methods, while identifying and recording special geological structures such as fault structures and karst cave structures. A thorough understanding of the regional geological conditions provides a reliable geological basis for subsequent data analysis and model building. Hydrological data is collected in real-time by deploying flow meters, pressure gauges, and temperature sensors at the tunnel construction site to collect key hydrological parameters of the sudden inrush water, including water pressure, water level, flow direction, and temperature. The monitoring equipment transmits this data via a network, continuously updating the dynamic changes of the sudden inrush water and sending this real-time data to the data center. Through dynamic monitoring of hydrological data, the system can capture the changing trends of the sudden inrush water in the tunnel area over different time periods, providing crucial hydrological input for intelligent prediction models. Finally, grouting material data is established: to accurately analyze the adaptability of different grouting materials, a database of the physicochemical properties of grouting materials is built through laboratory tests and historical project data. Specifically, through laboratory experiments, parameters such as construction density, viscosity, setting time, and impermeability of commonly used grouting materials (such as cement grout, chemical grout, etc.) are measured. The composition ratio and types of admixtures of different materials are systematically recorded. At the same time, by sorting out and summarizing the grouting treatment of historical projects, a reference is provided for the application effect of different grouting materials under various geological conditions.

[0040] Furthermore, all collected data is effectively stored and managed through the system's database module, establishing a unified data structure and access interface. Geological data, hydrological data, grouting material characteristics, and construction parameters are all stored in the database for subsequent retrieval and analysis. Simultaneously, the database possesses data backup and recovery mechanisms to ensure data security and integrity, providing long-term support for continuous model updates and optimization.

[0041] Step 2: The preprocessed multi-source data is weighted and integrated according to geological structure, hydrological characteristics, material characteristics, and construction characteristics to form a key data table;

[0042] Specifically, all acquired data undergoes cleaning, completion, and noise and outlier removal, followed by standardization and normalization to complete the preprocessing process, ensuring the integrity, validity, and consistency of the model input data. This includes:

[0043] Missing values ​​are filled using interpolation or mean substitution methods. For time-series data, such as hydrological parameters like water pressure and water level, linear interpolation is preferred to maintain the temporal continuity and rationality of the data. For non-time-series data, such as material property parameters, mean substitution is used to fill missing items to ensure the stability of the data distribution.

[0044] Anomaly detection algorithms are employed to identify and remove noisy data and outliers. For geological parameter data, clustering algorithms are used to identify outliers that deviate significantly from the mainstream distribution; for time-series data, a sliding window method is used to detect abrupt changes that do not conform to the trend. Through outlier removal, the system can eliminate invalid information, providing more realistic and reliable data support for model training.

[0045] Geological, hydrological, material properties, and construction parameters were standardized and normalized. Geological parameters and material property data, due to differences in units and dimensions, required standardization, transforming each data feature into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate dimensional differences. Hydrological data and construction parameters were normalized, compressing the data to the [0,1] interval to improve the model's convergence speed.

[0046] Filtering algorithms are used to denoise time-series data such as hydrological and construction parameters. Specifically, for the fluctuations in hydrological parameters such as water pressure and flow velocity, low-pass filtering or Kalman filtering algorithms are used to smooth data curves to remove high-frequency noise interference. For instantaneous abnormal peaks in construction parameters, mean filtering is used to eliminate abnormal fluctuations while retaining the main trend information of the data.

[0047] The standardized formula is as follows:

[0048]

[0049] Where X represents the original data, μ represents the mean, and σ represents the standard deviation.

[0050] Step 3: The preprocessed multi-source data is weighted and integrated according to geological structure, hydrological characteristics, material characteristics, and construction characteristics to form a key data table;

[0051] Specifically, the preprocessed multi-source data is aligned temporally and spatially. Then, the aligned multi-source data is processed hierarchically along the time dimension, and a comprehensive data table is constructed based on the stability of the geological structure, the dynamic changes of hydrological characteristics, the physicochemical properties of materials, and the operational conditions of construction parameters. Specifically, the structural parameters and hydrological characteristics of the tunnel geological environment are fused hierarchically, and the chemical composition of the grouting material is correlated with the construction parameters, forming a sequence in a network input format. Finally, the data is preliminarily weighted and integrated based on the correlations between various data features to obtain a key data table that significantly influences the prediction of grouting effects.

[0052] To ensure temporal and spatial consistency of geological, hydrological, material property, and construction data, the first step is to align all data according to their spatiotemporal attributes. This alignment process matches data from different sources by timestamp and spatial location, ensuring that each data record corresponds to the same time period and location. For geological data, this involves matching spatial information such as fault and stratum distribution with collected hydrological parameters (such as water pressure and water level). For time-series data, such as hydrological monitoring data like flow velocity and temperature, and construction parameters (such as grouting speed and pressure), the alignment step precisely matches these data according to timestamps, ensuring dynamic consistency.

[0053] Furthermore, after data alignment, this disclosure performs layered fusion processing on multi-source data based on data source and content. Specifically, standardized and spatiotemporally aligned data on geological structure, hydrological characteristics, material properties, and construction parameters are processed layered along the time dimension. A comprehensive data table is constructed according to the stability of the geological structure, the dynamic changes of hydrological characteristics, the physicochemical properties of material properties, and the operational conditions of construction parameters. In this process, the system integrates macroscopic parameters of the tunnel geological environment (such as fault location and karst cave distribution) with hydrological conditions (such as groundwater flow direction and water level fluctuations) layer by layer. At the same time, it correlates the chemical composition of grouting materials with construction process parameters to form the basic structure (sequence) of the network input format, providing consistent and comprehensive data input for the model.

[0054] Furthermore, after multi-source data fusion, the data is initially weighted and integrated based on the correlation between various data features. By comparing the correlation between historical project data such as geological conditions, hydrological environment, and construction effects, the system automatically identifies key features affecting the effectiveness of grouting control and assigns weights to these features. For example, in areas with significant water pressure changes, the system assigns higher weights to hydrological parameters; in geological environments with dense fault distribution, it assigns higher weights to fault distribution and rock strata type characteristics, ensuring that the data fusion process can reasonably present data features that have a significant impact on the prediction of grouting effects. Furthermore, the high-quality data that has completed hierarchical fusion and weighted integration is stored as the system's multidimensional dataset to support subsequent model training and analysis.

[0055] The weighted average formula is as follows:

[0056]

[0057] Among them, f i For the i-th scale feature, w i Let m be the weight of the scale feature.

[0058] Step 4: Divide the data in the key data table into layers according to multi-scale hierarchical dimensions to form semantic layer data of global geological structure, local geological structure, grouting material, and hydrological data;

[0059] Specifically, firstly, based on the spatial scope and semantic information of the data, the data is divided into three levels: macroscopic, mesoscopic, and microscopic. The specific division principle is as follows: the macroscopic level includes the global geological structural characteristics of the tunnel area, used to identify the stability of the large-scale geological environment; the mesoscopic level focuses on the local geological structural characteristics of the tunnel surrounding rock, assessing the impact of local geological conditions on the grouting effect; the microscopic level contains the semantic information of the refined physicochemical properties of the grouting material and the hydrological parameters of the inrush water, used to analyze the diffusion and coagulation behavior of the material in the micro-environment.

[0060] The macro-level data primarily encompasses the large-scale, global geological structural characteristics of the tunnel construction area. It collects and integrates key geological information such as faults, folds, karst caves, and rock strata boundaries to form the basis for a stability assessment of the entire tunnel area. At the macro-level, data is mainly used to identify large-scale geological condition trends and predict possible paths for sudden water inrushes. Macro-level data also includes dynamic information on regional geological environment changes, such as regional water level variations and fault activity, to assess the overall safety of the tunnel.

[0061] The meso-level data focuses on the local geological characteristics of the tunnel surrounding rock, encompassing information such as porosity, density, rock type, and water content during tunnel excavation. The primary objective of meso-level data is to assess the direct impact of local geological conditions on grouting effectiveness, facilitating the analysis of the diffusion path and setting behavior of grouting materials within the local environment.

[0062] The microscopic level mainly covers the physicochemical properties of the grouting material and the hydrological parameters of the inrush water, in order to accurately analyze the diffusion, permeation, and coagulation characteristics of the material in the microscopic environment. The data includes the particle distribution, chemical composition, coagulation rate, and impermeability of the grouting material, as well as parameters such as water pressure, water temperature, and flow velocity of the inrush water.

[0063] Finally, after the data hierarchy is defined, the system stores the macro, meso, and micro data in a structured manner, forming a multi-level dataset. By labeling and indexing the data at different levels, the system can flexibly access data from each level, facilitating collaborative analysis of macro trends, meso details, and micro characteristics across different analysis modules.

[0064] Step 5: Input the global geological structure layer data, local geological structure layer data, and semantic layer data into the multi-task prediction model. In the multi-task prediction model, the global geological structure layer data and local geological structure layer data are input into the GNN network structure to extract geological topological features, and the semantic layer data are input into the DNN network structure to extract high-dimensional semantic features. The geological topological features and high-dimensional semantic features are used to extract a shared feature map in the shared feature layer. Based on the shared feature map, the target prediction result is output through multi-source output regression.

[0065] Specifically, S1: Construction and pre-training of the multi-task prediction model

[0066] First, the multi-task prediction model includes a dual-channel network structure. It adopts a multi-task learning framework and combines the advantages of deep neural networks (DNN) and graph neural networks (GNN). The dual-channel network structures are CNN network structure and GNN network structure, respectively, to build a multi-objective prediction model suitable for complex tunnel environments.

[0067] In the model structure, the GNN network is mainly used to capture the topological structure and nonlinear spatial relationships of geological and underground structural features, while the DNN network is used to handle high-dimensional, continuous hydrological parameters, grouting materials, and construction parameters. The specific setup steps are as follows:

[0068] 1) Constructing the GNN network structure: First, the features of the geological structure, such as faults, fissures, and water flow paths, are transformed into a graph structure, where nodes represent spatial locations and edges represent spatial relationships between nodes (such as water flow paths and fissure extension directions). Next, an initial feature vector is assigned to each node, containing geological features (water content, porosity, density, etc.) and hydrological characteristics (such as flow velocity and pressure) of the region corresponding to the node.

[0069] Then, through a message-passing mechanism, the GNN transmits information between each node and its neighboring nodes to update the node's feature representation. Finally, the GNN gradually captures the spatial correlation between nodes through multiple iterations, integrating the information of surrounding nodes into the feature representation of each node, thereby learning the nonlinear spatial relationships of geological structural features.

[0070] The formula for the message passing mechanism is as follows:

[0071]

[0072] in, Let N(v) represent the message aggregation of node v at layer k, and N(v) be the set of neighboring nodes of node v. Let e ​​be the feature vector of neighbor node u. uv Represents the edge characteristics between node u and node v; Let v be the feature vector of node v.

[0073] Furthermore, the node update formula is:

[0074]

[0075] in, W represents the feature representation of node v at the (k+1)th layer. (k) Let σ be the weight matrix for learning, and b be the activation function. (k) Let be the bias vector of the k-th layer.

[0076] Furthermore, the input to the GNN model is the topological graph structure formed by tunnel geology and hydrological flow, and the output is the embedding vector of geological and hydrological dynamics, i.e., geological topological structure features. The update formula is:

[0077]

[0078] in, W is the feature vector of node v in the (k+1)th layer. (k) Let N(v) be the weight matrix for learning, and c be the neighboring nodes. vu This is the normalization constant.

[0079] 2) Construct a DNN network structure. The input to the DNN model consists of standardized hydrological parameters, material properties, and construction parameters, while the output is high-dimensional semantic features. The calculation formula is as follows:

[0080]

[0081] Where y represents the high-dimensional semantic features of the output, and x i For input data, w i denoted as weight, b as bias constant, and f as activation function.

[0082] Furthermore, the formula for forward propagation of the hidden layers in a DNN model is:

[0083] h (l) =f(W (l-1) h (l-1) +b (l-1) )

[0084] Among them, h (l) W is the output of the neuron in layer l. (l-1) and b (l-1) Let f be the weight matrix and bias from layer (l-1) to layer l, and f be the activation function.

[0085] 3) Construction of a shared feature layer: A shared feature layer is set after the CNN and GNN network structures to extract common information from geological, hydrological, material, and construction parameters, thereby improving the efficiency of information sharing among multiple tasks. The shared feature layer extracts the common information from the feature outputs h of the GNN and DNN. combined This system integrates the topological structure of geological and hydrological information with the high-dimensional features of materials and construction parameters into a single feature space. The feature fusion strategy of the shared layer employs a weighted aggregation method to ensure the effectiveness of various data features during the fusion process. Simultaneously, the system uses ResNet (Residual Connections) to mitigate the gradient vanishing problem caused by excessively deep layers. After weighted aggregation and residual connection processing, the final shared feature h... shared It contains spatial topological features of geological and hydrological information, as well as high-dimensional features of grouting materials and construction parameters. The output of this shared feature layer will serve as the input for multi-task prediction, providing common feature support for various tasks (such as impermeability prediction, diffusion range prediction, and consolidation strength prediction).

[0086] Assume the feature vector output by the GNN is h GNN The feature vector output by the DNN is h DNN Comprehensive feature vector h combined for:

[0087] h combined =[h GNN ,h DNN ]

[0088] Furthermore, the weighted aggregation formula is:

[0089] h shared =α GNN ·h GNN +α DNN ·h DNN

[0090] Where, α GNN With α DNN The weight coefficients of the features of the GNN and DNN are respectively, satisfying α GNN +α DNN =1.

[0091] Furthermore, the residual connection formula is:

[0092] h residual =h combined +F(h combined )

[0093] Among them, F(h) combined ) represents a nonlinear transformation performed on the fused features, such as a linear transformation and activation performed through one or more layers of neural networks.

[0094] 4) Construction of multi-source output regression model: In order to achieve multi-objective prediction, a multi-source output regression model is set up after the shared feature layer. This model combines dynamic weighting mechanism and task-specific input adjustment strategy to design independent output modules for each task that focus on different input data characteristics. Each output module corresponds to a different prediction task, including impermeability, diffusion range and consolidation strength.

[0095] Among them, the anti-seepage performance prediction mainly focuses on the anti-seepage effect of the material, using tunnel water pressure and the permeability parameters of the grouting material as inputs; the diffusion range prediction predicts the diffusion area of ​​the material in the sudden water inrush based on data such as grouting pressure, viscosity, and water flow rate; and the consolidation strength prediction predicts the consolidation effect based on information such as the setting time of the grouting material and construction temperature. Each output layer performs independent parameter optimization and error feedback for its respective task objectives.

[0096] Furthermore, before entering each output layer, the model assigns different weights to the input features according to task requirements, generating a task-specific subset of features. The weighted features for task t are:

[0097] h t =Weighting(h shared W t )

[0098] Among them, W t It is the feature weighting matrix for task t, which dynamically adjusts the contribution of each feature.

[0099] Furthermore, the regression model for each task is designed independently with different model structures and hyperparameters tailored to the specific characteristics of the task, achieving nonlinear modeling of specific features. Specifically:

[0100] The anti-permeability prediction uses a multilayer perceptron (MLP) to extract the nonlinear relationship of features layer by layer:

[0101] y 抗渗 =f 抗渗 (W 抗渗 ·h 抗渗 +b 抗渗 )

[0102] Among them, f 抗渗 It is a nonlinear activation function designed based on anti-permeability characteristic prediction, h 抗渗 W is the weighted feature set for anti-seepage prediction. 抗渗 and b 抗渗 These are the weights and bias parameters of the prediction model, respectively.

[0103] The diffusion range prediction uses a convolutional neural network (CNN) to capture the local correlation between grouting material properties and hydrological parameters:

[0104] y 扩散 =CNN(h 扩散 ;θ 扩散 )

[0105] Among them, h 扩散 For the weighted diffusion performance prediction feature set, θ 扩散 These are the parameters of the diffusion performance prediction task model.

[0106] Consolidation strength prediction uses a recurrent neural network (RNN) combined with time series data analysis to examine the dynamic changes in the consolidation process:

[0107] y 固结 =RNN(h 固结 ;θ 固结 )

[0108] Among them, h 固结 For the weighted consolidation strength performance prediction feature set, θ 固结 These are the parameters of the model for the task of predicting the consolidation strength performance.

[0109] 5) Model optimization

[0110] To balance the losses across different task objectives, the model employs a weighted loss function, allocating loss weights based on the importance of each task and the convergence performance during training. Initial weight settings are based on task priority and the model's initial performance on each task. During training, the model dynamically adjusts the weights to achieve coordinated optimization across tasks. An adaptive weight allocation strategy uses a weight adjustment function to adaptively adjust based on the loss feedback from each task, ensuring the model dynamically responds to the convergence of each task during multi-task learning. Specifically, when the error of a certain task is large, the system increases the weight of that task to enhance its focus during training, enabling the model to achieve better coordination among different objectives.

[0111] Furthermore, the formula for calculating the multi-task weighted loss function is as follows:

[0112]

[0113] in, For the loss of the t-th task, α t Assign weights to them.

[0114] Furthermore, during model training, the stochastic gradient descent (SGD) algorithm is used to optimize parameters. To improve the efficiency and accuracy of model training, the system employs a mini-batch training strategy, updating model parameters with each batch of data to gradually reduce the loss function and approach the global optimum. To further improve the model's convergence speed and prediction accuracy, the model uses a Bayesian optimization algorithm to automatically select the optimal hyperparameter combination. Bayesian optimization intelligently selects new hyperparameters based on the training results of historical hyperparameter combinations, avoiding redundant parameter tuning processes and enabling the model to quickly reach the best training effect.

[0115] Furthermore, the formula for the stochastic gradient descent (SGD) optimization algorithm is as follows:

[0116]

[0117] Where θ is the model parameter and η is the learning rate.

[0118] Furthermore, after training, the model's multi-objective prediction performance is evaluated using a validation set. Mean absolute error (MAE) is used to quantify the prediction results for anti-permeability, diffusion range, and consolidation strength. The system fine-tunes the model based on the validation set performance to ensure its stability and accuracy across different task objectives. Finally, the system saves the optimal model parameters for each task for subsequent model loading and rapid deployment in applications.

[0119] The model performance is evaluated by calculating the mean squared error (MSE) using the validation set. The formula is as follows:

[0120]

[0121] Among them, y i This is the actual value. is the predicted value, and n is the number of samples in the validation set.

[0122] As one embodiment, the method disclosed herein is used for real-time prediction and feedback optimization. By establishing a real-time prediction and feedback optimization mechanism and combining real-time data during the grouting process, dynamic material ratio adjustment is achieved. The specific operation is as follows:

[0123] First, during the grouting process, the system collects key data in real time from sensors and monitoring equipment, including: the proportion parameters of the grouting material (such as the ratio of cement, water, additives, etc.), the rheological properties of the material (such as viscosity, fluidity, etc.), and geological and hydrological conditions (the distribution of faults and fissures, the direction of water flow and the velocity of flow in the current construction section, etc.). All data is transmitted to the system's central database through the network and used as real-time input to the model.

[0124] Next, after receiving the latest data, the system performs predictions based on a multi-task learning model and a multi-objective optimization algorithm. The model uses a DNN component to analyze the high-dimensional properties of the material (such as viscosity and rheology). The GNN component is responsible for modeling the topological relationships of geological structures and hydrological features, and generating geological and hydrological feature vectors. The outputs of the DNN and GNN are fused in a shared feature layer, concatenated to generate a comprehensive feature vector, and then used a multi-source output regression model to predict the following three objectives: anti-seepage effect prediction, diffusion range prediction, and consolidation strength prediction. The prediction results are weighted and synthesized using a multi-objective optimization algorithm to determine the applicability of the current material mix, and the predicted values ​​are output for subsequent feedback optimization.

[0125] Finally, if the real-time predicted effect deviates from the design requirements or standard value, the system triggers a feedback optimization mechanism. First, a difference analysis and deviation feedback are performed: the system compares the predicted value with the target treatment standard, calculating the deviations in impermeability, diffusion range, and consolidation strength. While ensuring a balance between impermeability, diffusion, and consolidation targets, a multi-objective optimization algorithm is used to adjust the material proportioning parameters. The optimization algorithm dynamically adjusts the material component ratios based on the current actual deviations to achieve the optimization objectives. After the model derives the new material proportioning parameters, the system feeds the optimization results back to the construction site, updates the grouting material proportions, and continues to monitor the construction progress. If the deviation is effectively corrected, the current proportions are maintained; if the deviation persists, the feedback optimization process is repeated until the treatment standard is met.

[0126] Furthermore, the core of multi-objective optimization is to construct an objective function that transforms the deviations of multiple prediction indicators (permeability resistance, diffusion range, consolidation strength) into an optimizable mathematical form:

[0127] First, let the mixing ratio parameter of the grouting material be P = [p1, p2, ..., p]. n ], where p i This indicates the proportion of a certain component (such as cement, water, or additives), and the proportioning parameter must satisfy the condition that the total proportion is 1: p i ≥0. Next, minimize the weighted sum of the biases of each forecast indicator: Where α t The weight of task t is dynamically adjusted based on task priority to ensure goal balance. It is the deviation loss function for task t (impermeability, diffusion, consolidation): in It is a model prediction value based on the current material proportion parameter p. It is the target value of task t.

[0128] Then, a weighted summation method is used to combine multiple objectives through weights α. t We optimize the weighted averages into a single objective, resulting in the objective function: Select the initial ratio parameter P 0 Update the matching parameters using gradient descent: Where η is the learning rate. The gradient of the objective function.

[0129] Finally, the globally minimum P obtained after optimization using the weighted summation method is chosen. * The optimal mix proportions are submitted to the construction system for the preparation of grouting materials. The model predictions are updated based on the actual construction results. If deviations still exist, the optimization process is repeated until the target requirements are met.

[0130] Example 2

[0131] One embodiment of this disclosure provides a multi-scale layered tunnel grouting material application effect prediction system, including:

[0132] The data acquisition module is used to acquire multi-source data such as tunnel geological data, sudden water inrush hydrological data, grouting material data, and construction parameters, and preprocess the data. The preprocessed multi-source data is then weighted and integrated according to geological structure, hydrological characteristics, material characteristics, and construction characteristics to form a key data table.

[0133] The layered module is used to divide the data in the key data table into layers according to multi-scale layered dimensions, forming semantic layer data of global geological structure layer data, local geological structure layer data, and grouting material and hydrological data.

[0134] The prediction module is used to input global geological structure layer data, local geological structure layer data, and semantic layer data into a multi-task prediction model. In the multi-task prediction model, the global geological structure layer data and local geological structure layer data are input into a GNN network structure to extract geological topological features, and the semantic layer data are input into a DNN network structure to extract high-dimensional semantic features. The geological topological features and high-dimensional semantic features are used to extract a shared feature map in a shared feature layer. Based on the shared feature map, the target prediction result is output through a multi-source output regression model.

[0135] Example 3

[0136] One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, they implement the method for predicting the application effect of tunnel grouting materials based on multi-scale layering.

[0137] Example 4

[0138] One embodiment of this disclosure provides an electronic device, including a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the method for predicting the application effect of tunnel grouting materials based on multi-scale layering.

[0139] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0140] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0141] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A method for predicting the application effect of tunnel grouting material based on multi-scale layering, characterized in that, The method comprises the following steps: Obtain multi-source data of tunnel geological data, sudden gushing hydrological data, grouting material data and construction parameters, and preprocess them; Integrate the preprocessed multi-source data according to the geological structure, hydrological characteristics, material characteristics and construction characteristics, and form a key data table; According to the multi-scale hierarchical dimension, divide the data in the key data table into levels to form geological global structure layer data, geological local structure layer data and semantic layer data of grouting material and hydrological data, including: according to the spatial range and semantic information of the data, divide the data into three levels of macro, meso and micro, and the specific division principle is: the macro level contains the global geological structure characteristics of the tunnel area, which is used to identify the stability of the large-scale geological environment; the meso level is the local geological structure characteristics of the tunnel surrounding rock, which is used to evaluate the influence of local geological conditions on the grouting effect; the micro level is the semantic information of the physical and chemical properties of the grouting material and the hydrological parameters of the sudden gushing water, which is used to analyze the diffusion and condensation behavior of the material in the micro environment; Input the geological global structure layer data, the geological local structure layer data and the semantic layer data into the multi-task prediction model, in which the geological global structure layer data and the geological local structure layer data are input into the GNN network structure to extract the geological topological structure characteristics, the semantic layer data is input into the DNN network structure to extract the high-dimensional semantic characteristics, the geological topological structure characteristics and the high-dimensional semantic characteristics are extracted into a shared feature map in a shared feature layer, and the target prediction result is output based on the shared feature map through multi-source output regression. 2.The method of claim 1, wherein, Through drilling and ultrasonic imaging, obtain geological data of rock layer distribution, fault structure and karst cave structure, hydrological data including water pressure, water level, flow direction and temperature data, grouting material data including grouting material, proportioning and physical and chemical properties, and construction parameters including density, viscosity, setting time and various performance indicators, respectively clean, complete and remove noise and outliers of all obtained data, and complete the preprocessing process. 3.The method of claim 1, wherein, Integrate the preprocessed multi-source data according to the geological structure, hydrological characteristics, material characteristics and construction characteristics to form a key data table, including: After preprocessing, align the multi-source data in time and space, then process the aligned multi-source data in time dimension, construct a comprehensive data table according to the stability of the geological structure, the dynamic change of the hydrological characteristics, the physical and chemical properties of the material characteristics and the operation conditions of the construction parameters; wherein, the structure parameters of the tunnel geological environment and the hydrological characteristics are fused by level, and the chemical composition of the grouting material and the construction parameters are associated to form a network input format sequence; then, according to the correlation between the data characteristics, the data is preliminarily weighted and integrated to obtain a key data table which has an important influence on the prediction of grouting effect. 4.The method of claim 1, wherein, The macro, meso and micro three-layer data are stored in a structured manner to form multi-level data, and the multi-level data is input into a multi-task prediction model. The multi-task prediction model includes a double-channel network structure, and the double-channel network structure is respectively a CNN network structure and a GNN network structure. After the multi-level data is input into the multi-task prediction model, the macro and meso level data enter the GNN network structure to capture the topological structure and nonlinear spatial relationship of the geology and extract the geological topological structure features; The micro level data enters the CNN network structure to process high-dimensional and continuous hydrological data, grouting materials and construction parameters, and extract high-dimensional semantic features. 5.The method of claim 1, wherein, A shared feature layer is arranged after the CNN network structure and the GNN network structure, and the geological topological structure features and the high-dimensional semantic features are spliced and combined on the shared feature layer to extract shared feature maps. A multi-source output regression model is arranged on the shared feature layer of the model, a dynamic weighting mechanism and a task-specific input adjustment strategy are combined, and an independent output module focusing on different input data characteristics is designed for each task. Each output module corresponds to a different prediction task. 6.The method of claim 5, wherein, Each output module corresponds to a different prediction task, including impermeability, diffusion range and consolidation strength.

7. A tunnel grouting material application effect prediction system based on multi-scale layering, characterized by, It comprises: A data acquisition module is used to acquire multi-source data of tunnel geological data, gushing water hydrological data, grouting material data and construction parameters, and to preprocess the data; The preprocessed multi-source data is weighted and integrated according to the geological structure, hydrological characteristics, material characteristics and construction characteristics to form a key data table; A hierarchical module is used to divide the data of the key data table into levels according to the multi-scale hierarchical dimensions to form geological global structure layer data, geological local structure layer data and semantic layer data of grouting materials and hydrological data. The data is divided into macro, meso and micro three levels according to the spatial range and semantic information of the data. The specific division principle is that the macro level contains the geological global structure features of the tunnel area, which is used to identify the stability of the large-scale geological environment; the meso level is the geological local structure features of the tunnel surrounding rock, which is used to evaluate the influence of local geological conditions on the grouting effect; the micro level is the semantic information of the physical and chemical properties of the grouting material and the hydrological parameters of the gushing water, which is used to analyze the diffusion and condensation behavior of the material in the micro environment; A prediction module is used to input the geological global structure layer data, the geological local structure layer data and the semantic layer data into a multi-task prediction model. In the multi-task prediction model, the geological global structure layer data and the geological local structure layer data are input into a GNN network structure to extract geological topological structure features, the semantic layer data is input into a DNN network structure to extract high-dimensional semantic features, the geological topological structure features and the high-dimensional semantic features are extracted into shared feature maps in a shared feature layer, and target prediction results are output based on the shared feature maps through a multi-source output regression.

8. A non-transitory computer-readable storage medium, comprising: The non-transitory computer readable storage medium is used to store computer instructions, which are executed by a processor to implement the multi-scale hierarchical tunnel grouting material application effect prediction method according to any one of claims 1-6.

9. An electronic device, comprising: It comprises: A processor, a memory and a computer program; wherein the processor is connected with the memory, the computer program is stored in the memory, when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the method for predicting the application effect of the tunnel grouting material based on the multi-scale hierarchy as claimed in any one of claims 1-6.

Images