Tunnel surrounding rock pressure arch calculation system

Abstract

The invention, which belongs to the technical field of tunnel engineering, discloses a tunnel surrounding rock pressure arch calculation system comprising a multi-source data fusion acquisition module, a surrounding rock grade intelligent identification module, a self-adaptive pressure arch parameter calculation module and a dynamic load prediction and optimization module. The multi-source data fusion acquisition module acquires geological parameters, drilling monitoring data and case data and generates fusion feature vectors; the surrounding rock grade intelligent identification module outputs a surrounding rock classification result based on a deep residual network and an attention mechanism; the self-adaptive pressure arch parameter calculation module calculates pressure arch parameters by adopting an improved Prscherski theory and a stress release time-varying function; the dynamic load prediction and optimization module predicts the load through the space-time collaborative network and feeds the deviation back to the calculation module to form closed-loop optimization, precise dynamic calculation of the surrounding rock pressure arch parameters is achieved, and the calculation precision is improved by 20% or above compared with a traditional method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel engineering, and in particular to a tunnel surrounding rock pressure arch calculation system. BACKGROUND

[0002] With the rapid development of infrastructure construction in China, the scale of tunnel engineering construction is continuously expanding, and the safety problem of tunnel construction is increasingly prominent. In the process of tunnel excavation, the stress redistribution of surrounding rock forms a pressure arch structure, and accurate calculation of the pressure arch parameters is of great significance for reasonable design of support structure and ensuring construction safety.

[0003] In the prior art, such as the deep Shandong West Water Source and Water Supply Project water tunnel construction scheme, water tunnel construction involves multiple surrounding rock types, including Class II surrounding rock, Class III surrounding rock, Class IV surrounding rock and Class V surrounding rock, and different surrounding rock types have different physical and mechanical properties and stability characteristics. The traditional surrounding rock pressure arch calculation method mainly uses the Prouse theory, the Terzaghi theory and the empirical formula recommended by the Chinese Highway Tunnel Design Specification, and these methods have the following technical defects:

[0004] First, the traditional calculation method mainly relies on static geological parameters and fails to fully consider the time-varying characteristics of stress release of surrounding rock in the process of tunnel excavation, resulting in deviations between the calculation results and the actual working conditions. When the tunnel passes through complex geological conditions, such as fault fracture zones, water-rich sections or weak surrounding rock sections, the traditional method is difficult to accurately reflect the dynamic evolution law of the surrounding rock pressure arch.

[0005] Second, the existing method is relatively rough in processing the classification of surrounding rock, usually only relying on a single index for classification, and failing to comprehensively utilize drilling parameters, geological exploration data and construction monitoring information, resulting in insufficient accuracy of surrounding rock type identification, and further affecting the reliability of pressure arch calculation.

[0006] Third, the traditional method lacks deep correlation analysis capability between surrounding rock parameters and pressure arch characteristics, and is difficult to establish the internal relationship between multi-source heterogeneous data, and cannot realize self-adaptive and accurate calculation of pressure arch parameters.

[0007] Fourth, the existing technology fails to establish a closed-loop feedback mechanism between pressure arch calculation and support design, and cannot dynamically adjust the calculation model parameters according to construction monitoring data, limiting the engineering applicability and prediction accuracy of the calculation system.

[0008] Therefore, there is an urgent need for a tunnel surrounding rock pressure arch calculation system that can integrate multi-source data, intelligently identify surrounding rock types, adaptively calculate pressure arch parameters and have closed-loop feedback optimization capability. SUMMARY

[0009] To solve the above problems in the prior art, the tunnel surrounding rock pressure arch calculation system is provided, which realizes accurate calculation of tunnel surrounding rock pressure arch parameters through the synergistic effect of multi-source data deep fusion, surrounding rock intelligent grading, self-adaptive pressure arch calculation and dynamic closed-loop optimization.

[0010] The tunnel surrounding rock pressure arch calculation system provided by the application comprises:

[0011] The multi-source data fusion acquisition module is used for acquiring tunnel surrounding rock geological parameter data, construction drilling monitoring data and historical engineering case data, performing space-time alignment and feature standardization processing on the acquired data, and generating a fusion feature vector in a unified format.

[0012] The surrounding rock grade intelligent identification module is used for receiving the fusion feature vector, extracting deep features of surrounding rock geological features based on a deep residual network, weighting and enhancing key features in combination with an attention mechanism, and outputting surrounding rock grade classification results and feature importance weights.

[0013] The self-adaptive pressure arch parameter calculation module is used for calculating a pressure arch span reference value by using an improved Prandtl theory model according to the surrounding rock grade classification results and the feature importance weights, calculating a pressure arch height dynamic value based on a stress release time-varying function, and generating a pressure arch parameter set comprising a pressure arch span, a height and a lateral pressure coefficient.

[0014] The dynamic load prediction and optimization module is used for predicting a surrounding rock load evolution trend by using a space-time collaborative prediction network according to the pressure arch parameter set and real-time monitoring data, calculating a deviation feedback signal of a predicted load and a measured load, and transmitting the deviation feedback signal to the self-adaptive pressure arch parameter calculation module for parameter correction to form a closed-loop optimization control.

[0015] In an embodiment of the application, the multi-source data fusion acquisition module comprises:

[0016] The geological parameter acquisition unit is used for acquiring surrounding rock lithology, joint fissure development degree, underground water level and surrounding rock integrity index.

[0017] The drilling monitoring acquisition unit is used for acquiring drilling speed, impact pressure, rotary pressure and pushing pressure in real time.

[0018] The case data acquisition unit is used for retrieving historical pressure arch calculation case data under similar geological conditions from an engineering case database.

[0019] In an embodiment of the application, the multi-source data fusion acquisition module further comprises a feature fusion unit, which is used for performing space-time alignment on geological parameters, drilling monitoring data and case data, filling missing data points by using least squares interpolation, and splicing the normalized data of various types to generate a fusion feature vector.

[0020] In an embodiment of the present application, the surrounding rock grade intelligent identification module comprises:

[0021] The feature extraction unit is configured to perform multi-layer convolution operation on the fusion feature vector by using a deep residual network to extract hierarchical representation of the surrounding rock geological features.

[0022] The attention enhancement unit is configured to calculate a correlation matrix between feature dimensions based on a self-attention mechanism and dynamically weight the key geological features.

[0023] The classification output unit is configured to input the weighted feature representation into a fully connected classification layer to output classification probability distribution of Class II to Class V surrounding rocks and importance weight of each feature.

[0024] In an embodiment of the present application, the adaptive pressure arch parameter calculation module comprises:

[0025] The span calculation unit is configured to calculate a pressure arch span reference value according to the tunnel excavation width, the surrounding rock grade and the feature importance weight.

[0026] The height calculation unit is configured to calculate a pressure arch height dynamic value based on a stress release time-varying function and a surrounding rock Prandtl coefficient.

[0027] The lateral pressure calculation unit is configured to calculate a lateral pressure value according to the burial depth, the surrounding rock lateral pressure coefficient and the horizontal stress.

[0028] In an embodiment of the present application, the stress release time-varying function adopted by the height calculation unit is a Weibull time function, and the function parameters are determined by fitting the field monitoring data to realize dynamic tracking of the evolution of the pressure arch height over time.

[0029] In an embodiment of the present application, the dynamic load prediction and optimization module comprises:

[0030] The space-time feature extraction unit is configured to perform one-dimensional convolution operation on the construction monitoring time series data to extract spatial features and use a gated recurrent unit to extract time series features.

[0031] The load prediction unit is configured to input the space-time features into a prediction network to output surrounding rock load prediction values in a future preset time period.

[0032] The bias calculation unit is configured to calculate the root mean square bias between the predicted load and the measured load.

[0033] The feedback adjustment unit is configured to generate a parameter correction signal according to the bias value to adjust the calculation coefficients in the adaptive pressure arch parameter calculation module.

[0034] In one embodiment of the present invention, the system further includes a visualization output module for generating a three-dimensional spatial distribution map of the pressure arch, a load time history curve, and a surrounding rock stability assessment report.

[0035] In one embodiment of the present invention, a parameter-level deep coupling relationship is established between the multi-source data fusion acquisition module, the surrounding rock grade intelligent identification module, the adaptive pressure arch parameter calculation module, and the dynamic load prediction and optimization module, with the output of the former module directly serving as the key input parameter of the latter module.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention achieves deep integration of geological parameters, borehole monitoring data, and historical case data through a multi-source data fusion acquisition module, overcoming the limitations of traditional methods that rely on a single data source and improving the completeness and reliability of the input data.

[0038] This invention uses a deep residual network and attention mechanism to classify surrounding rocks through an intelligent identification module for surrounding rock grade. Compared with traditional empirical discrimination methods, the classification accuracy is improved by more than 15%, providing more accurate surrounding rock parameter input for subsequent pressure arch calculation.

[0039] This invention introduces a stress release time-varying function through an adaptive pressure arch parameter calculation module, which can accurately reflect the dynamic evolution law of pressure arch parameters during tunnel excavation, and improves the calculation accuracy by more than 20% compared with the traditional static method.

[0040] This invention establishes a closed-loop feedback control mechanism through a dynamic load prediction and optimization module, enabling online adaptive adjustment of calculation model parameters. This allows the system to adapt to complex geological conditions and effectively reduces prediction errors.

[0041] The four core modules of this invention form a deeply coupled synergistic relationship, realizing integrated processing of data acquisition, feature recognition, parameter calculation and dynamic optimization. The overall technical solution has the synergistic effect of 1+1>2. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the overall structure of the tunnel surrounding rock pressure arch calculation system of the present invention;

[0043] Figure 2 This is a schematic diagram of the structure of the multi-source data fusion acquisition module of the present invention;

[0044] Figure 3 This is a schematic diagram of the intelligent identification module for surrounding rock grade of the present invention;

[0045] Figure 4 This is a schematic diagram of the adaptive pressure arch parameter calculation module of the present invention;

[0046] Figure 5 This is a structural schematic diagram of the dynamic load prediction and optimization module of the present invention;

[0047] Figure 6 This is a schematic diagram of the data flow of the system of the present invention. Detailed Implementation

[0048] Please refer to the attached document. Figures 1-6 The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0049] See Figure 1 The tunnel surrounding rock pressure arch calculation system provided by this invention includes a multi-source data fusion acquisition module 1, a surrounding rock grade intelligent identification module 2, an adaptive pressure arch parameter calculation module 3, and a dynamic load prediction and optimization module 4. These four modules are deeply coupled at the parameter level, forming a complete closed-loop collaborative system. In the technical solution of this invention, the multi-source data fusion acquisition module 1 is responsible for data acquisition and preprocessing, and its output fused feature vector serves as the input to the surrounding rock grade intelligent identification module 2; the surrounding rock classification results and feature weights output by the surrounding rock grade intelligent identification module 2 serve as the key input to the adaptive pressure arch parameter calculation module 3; the pressure arch parameter set generated by the adaptive pressure arch parameter calculation module 3 is transmitted to the dynamic load prediction and optimization module 4 for load prediction; the deviation feedback signal calculated by the dynamic load prediction and optimization module 4 is transmitted back to the adaptive pressure arch parameter calculation module 3 for parameter correction, forming a complete closed-loop control.

[0050] See Figure 2 The multi-source data fusion acquisition module 1 includes a geological parameter acquisition unit 11, a borehole monitoring acquisition unit 12, a case data acquisition unit 13, and a feature fusion unit 14. In one embodiment of the present invention, the design of the multi-source data fusion acquisition module 1 fully considers the multi-source heterogeneous characteristics of tunnel engineering data, and solves the problem of spatiotemporal inconsistency between different types of data by establishing a unified data acquisition and fusion framework.

[0051] The geological parameter acquisition unit 11 is used to collect basic geological parameters of the tunnel surrounding rock, specifically including lithological parameters, joint and fracture development degree, groundwater level, and surrounding rock integrity index. In one embodiment of the invention, the surrounding rock lithological parameters include uniaxial compressive strength, elastic modulus, Poisson's ratio, and internal friction angle, which are obtained through field sampling and laboratory tests. The joint and fracture development degree is quantitatively characterized by fracture spacing, fracture width, and fracture penetration rate. The groundwater level is obtained through real-time measurement via a water level monitoring well, with a sampling period of 1 hour. The surrounding rock integrity index is evaluated using a combination of RQD and BQ values, where the RQD value is obtained through statistical calculation of borehole cores, and the BQ value is calculated using the method specified in my country's engineering rock mass classification standard. In a specific embodiment, for the water conveyance tunnel project of the Shenzhen-Shantou West Water Source and Water Supply Project, the surrounding rock types collected by the system include medium- and coarse-grained granite and tuff, and the surrounding rock grades cover Class II to Class V. Among them, the length of Class II surrounding rock is about 4008.6m, the length of Class III surrounding rock is about 1216.7m, the length of Class IV surrounding rock is about 287.3m, and the length of Class V surrounding rock is about 688.8m.

[0052] The borehole monitoring and acquisition unit 12 is used to collect the working parameters of the intelligent drilling rig in real time during the drilling process, mainly including drilling speed, impact pressure, rotational pressure, and propulsion pressure. It is worth noting that these borehole parameters are significantly correlated with the quality of the surrounding rock; analyzing the variation patterns of these parameters can indirectly reflect the mechanical properties of the surrounding rock. In one embodiment of the invention, the borehole monitoring and acquisition unit 12 adopts a high-frequency data acquisition mode, with a sampling frequency set to 100Hz, which can capture subtle parameter fluctuations during the drilling process. The collected raw data is filtered to remove noise interference, and then aggregated at 0.1m drilling depths to calculate the statistical characteristic values ​​of each parameter, including the mean, standard deviation, maximum value, and minimum value. According to relevant research from the International Society for Rock Mechanics, drilling speed is negatively correlated with surrounding rock strength, while pressure-related parameters are positively correlated with the degree of surrounding rock fragmentation. This pattern provides an important basis for subsequent surrounding rock classification.

[0053] The case data acquisition unit 13 is used to retrieve historical engineering case data with similar geological conditions to the current project from the engineering case database. In one embodiment of the present invention, the case database covers calculation cases of surrounding rock pressure arches for typical tunnel engineering projects at home and abroad, including basic project information, geological condition parameters, calculation methods used, calculation results, and measured verification data. The case retrieval adopts a similarity-based calculation method. The system calculates the comprehensive similarity with each case in the database based on the surrounding rock type, burial depth, cross-sectional dimensions, and geostress conditions of the current project, and selects the top 5 cases with the highest similarity as references. The introduction of case data provides engineering experience support for pressure arch calculation, effectively making up for the insufficient applicability of purely theoretical calculation methods under complex geological conditions.

[0054] The feature fusion unit 14 is a core component of the multi-source data fusion acquisition module 1, responsible for deep fusion processing of geological parameters, borehole monitoring data, and case data. In one embodiment of the invention, the feature fusion unit 14 first performs spatiotemporal alignment processing on the multi-source data, unifying data with different sampling frequencies and acquisition times under the same spatiotemporal reference framework. Specifically, using the tunnel face location as the spatial reference benchmark and the excavation time as the temporal reference benchmark, missing data points are filled using the least squares interpolation method. After completing the spatiotemporal alignment, the feature fusion unit 14 normalizes various types of data, mapping parameters of different dimensions to the 0-1 interval, eliminating the influence of dimensional differences on subsequent calculations. The normalization process uses the maximum-minimum normalization method, which can maintain the distribution characteristics of the data. Finally, the feature fusion unit 14 concatenates the normalized features according to a preset feature order to generate a unified format fused feature vector. In one embodiment of the invention, the fused feature vector has 64 dimensions, of which geological parameter features account for 20 dimensions, borehole monitoring features account for 32 dimensions, and case similarity features account for 12 dimensions.

[0055] See Figure 3 The intelligent rock grade identification module 2 includes a feature extraction unit 21, an attention enhancement unit 22, and a classification output unit 23. The design goal of the intelligent rock grade identification module 2 is to achieve high-precision automatic classification of rock types, providing accurate rock grade input for subsequent pressure arch parameter calculations. When working in conjunction with the multi-source data fusion acquisition module 1, the intelligent rock grade identification module 2 receives the fused feature vector as input, extracts deep representations of the geological features of the surrounding rock through deep learning methods, and utilizes an attention mechanism to identify the most critical feature dimensions for rock classification.

[0056] The feature extraction unit 21 is used to perform multi-layer feature extraction on the fused feature vector using a deep residual network to obtain a hierarchical representation of the geological features of the surrounding rock. In one embodiment of the present invention, the feature extraction unit 21 adopts a one-dimensional residual network structure, which contains four residual blocks. Each residual block contains two layers of one-dimensional convolution and one skip connection. The convolution kernel size is set to 3, the stride is set to 1, and the padding method is the same to keep the feature map size unchanged. The number of channels of the network is set to an increasing pattern of 64-128-256-512, and the hierarchical representation from low-level features to high-level features is achieved by extracting layers one by one. The introduction of residual connections solves the gradient vanishing problem in the training process of deep networks, enabling the network to learn richer feature representations. In a specific embodiment, the network uses batch normalization and ReLU activation function. Batch normalization helps to accelerate network convergence and improve generalization ability, while ReLU activation function introduces nonlinear transformation capability. After processing by the feature extraction unit 21, the output feature representation dimension is 512-dimensional.

[0057] Attention enhancement unit 22 is used to dynamically weight the extracted feature representations based on a self-attention mechanism, identifying and enhancing feature dimensions that significantly contribute to rock classification. Self-attention is a significant technological breakthrough in deep learning in recent years; its core idea is to dynamically adjust feature weights by calculating the correlation between features. In one embodiment of the invention, attention enhancement unit 22 employs a multi-head self-attention mechanism with 8 attention heads. In one possible implementation, the self-attention calculation process first maps the input features to a query vector, a key vector, and a value vector through three linear projection layers. Then, the dot product of the query vector and the key vector is calculated to obtain the attention score matrix. After scaling and normalization, this matrix is ​​multiplied by the value vector to obtain the weighted feature output. The multi-head attention mechanism executes the above calculation process in parallel multiple times and concatenates the outputs of each head before performing a linear transformation, enabling it to capture the correlation information between features from different representation subspaces. In the technical solution of the invention, attention enhancement unit 22 also outputs the attention weights for each feature dimension. These weights reflect the importance of each feature to rock classification and will be used as weight inputs for subsequent pressure arch parameter calculations.

[0058] The classification output unit 23 is used to map the attention-enhanced feature representation to the surrounding rock grade classification result. In one embodiment of the present invention, the classification output unit 23 includes two fully connected layers and one output layer, wherein the first fully connected layer has 256 neurons, the second fully connected layer has 128 neurons, and the output layer has 4 neurons, corresponding to Class II, Class III, Class IV, and Class V surrounding rocks, respectively. The output layer uses the Softmax activation function to convert the network output into a probability distribution form, with the sum of the probabilities of each class being 1. The classification output unit 23 also obtains feature importance weights from the attention enhancement unit 22 and outputs them together with the classification result to the adaptive pressure arch parameter calculation module 3. In a specific embodiment, based on the test data of the Shenzhen-Shantou West Water Source and Water Supply Project, the classification accuracy of the surrounding rock grade intelligent identification module 2 reaches 92.5%, which is about 16 percentage points higher than the traditional discrimination method based on empirical formulas.

[0059] See Figure 4 The adaptive pressure arch parameter calculation module 3 includes a span calculation unit 31, a height calculation unit 32, and a lateral pressure calculation unit 33. This module is the core innovative module of the invention, and its main function is to calculate the key geometric and load parameters of the pressure arch based on the surrounding rock grade classification results and feature importance weights. This module adopts an improved Protodyakonov theoretical model and introduces a stress release time-varying function, which can accurately reflect the dynamic evolution of pressure arch parameters during tunnel excavation.

[0060] The span calculation unit 31 is used to calculate the reference value of the pressure arch span based on the tunnel excavation width, surrounding rock grade, and characteristic importance weight. In the technical solution of this invention, the pressure arch span refers to the horizontal distribution range of the balanced arch structure formed by the self-supporting capacity of the surrounding rock, and its size directly affects the design parameters of the tunnel support structure. In one embodiment of this invention, the span calculation unit 31 first determines the initial value of the excavation width based on the tunnel cross-section dimensions, then calculates the influence width coefficient based on the surrounding rock grade and burial depth conditions, and finally calculates the reference value of the pressure arch span.

[0061] This invention proposes an innovative adaptive algorithm for calculating the span of a pressure arch. This algorithm integrates the weights of surrounding rock classification features to achieve refined calculation of the span parameters. The formula for calculating the baseline value of the pressure arch span is as follows:

[0062] ,

[0063] in, This is the reference value for the span of the pressure arch (in meters). This refers to the tunnel excavation width (in meters). The tunnel excavation height (in meters). Calculate the friction angle (in radians) for the surrounding rock. The number of feature dimensions, Let the importance weights of the i-th feature output by the attention mechanism be such that the sum of all weights equals 1. The normalized influence factor for the i-th feature ranges from 0.8 to 1.2. The innovation of this algorithm lies in the introduction of a feature importance weighting mechanism, enabling the calculation of the pressure arch span to adaptively adjust to different geological conditions.

[0064] The height calculation unit 32 is used to calculate the dynamic value of the pressure arch height based on the stress release time-varying function and the Protodyakonov coefficient of the surrounding rock. The pressure arch height is a key parameter affecting the magnitude of the surrounding rock load at the top of the tunnel, and its accurate calculation is of great significance for support design. In one embodiment of the present invention, the height calculation unit 32 uses the Weibull time function to describe the time-varying law of stress release, which can accurately reflect the process of the surrounding rock stress gradually being released and tending to stabilize after tunnel excavation.

[0065] This invention proposes an innovative time-varying pressure arch height calculation algorithm. This algorithm comprehensively considers surrounding rock strength parameters, stress release rate, and time effects to achieve dynamic and accurate calculation of the pressure arch height. The calculation formula for the dynamic value of the pressure arch height is as follows:

[0066] ,

[0067] in, The dynamic value of the pressure arch height at time t (in meters). The span reference value (in meters) of the pressure arch output by the span calculation unit. The Protodextrin coefficient of the surrounding rock. The time after tunnel excavation (in days). denoted as the scale parameter (in d) of the Weibull distribution. is the shape parameter (dimensionless) of the Weibull distribution. This is a correction function for the stress release rate. In one possible implementation, the Protodyakonov coefficient... Based on the uniaxial compressive strength of the surrounding rock Calculation, the specific calculation method is as follows ,in The unit is MPa. Weibull distribution parameters. and In one embodiment of the present invention, the water source and water supply project in the western part of the Shenzhen-Shantou region are fitted with on-site monitoring data to determine the optimal method. The value is 15d. The value is 1.5.

[0068] Stress relief rate correction function The formula used to describe the change in stress relief rate over time is as follows:

[0069] ,

[0070] in, The initial stress release rate, ranging from 0.3 to 0.5, The stress release attenuation coefficient (unit: d^-1) ranges from 0.05 to 0.15. In one embodiment of the present invention, for Class II and Class III surrounding rock, the initial stress release rate is... The value is 0.5, which is the stress relief attenuation coefficient. The value is set to 0.1; for Class IV and Class V surrounding rock, the initial stress release rate is... The value is 0.3, which is the stress release attenuation coefficient. The value is 0.08.

[0071] The lateral pressure calculation unit 33 is used to calculate the lateral pressure of the surrounding rock acting on the tunnel sidewall based on the tunnel burial depth, the lateral pressure coefficient of the surrounding rock, and the horizontal ground stress. In one embodiment of the present invention, the formula for calculating the lateral pressure value is as follows:

[0072] ,

[0073] in, This represents the lateral pressure value of the surrounding rock (in kPa). The lateral pressure coefficient of the surrounding rock is dimensionless. Unit weight of surrounding rock (in kN / m³). The depth of the tunnel (in meters). Groundwater pressure (in kPa). Lateral pressure coefficient. Based on the surrounding rock grade and tectonic stress conditions, in one embodiment of the present invention, for Class II surrounding rock... The value ranges from 0.3 to 0.5 for Class III surrounding rock. The value ranges from 0.4 to 0.6 for Class IV and Class V surrounding rock. The value ranges from 0.5 to 0.8.

[0074] See Figure 5 The dynamic load prediction and optimization module 4 includes a spatiotemporal feature extraction unit 41, a load prediction unit 42, a deviation calculation unit 43, and a feedback adjustment unit 44. The dynamic load prediction and optimization module 4 is a key module for achieving closed-loop control in this invention. Its main function is to predict the surrounding rock load based on the pressure arch parameters and real-time monitoring data, and to dynamically adjust the calculation model parameters through a deviation feedback mechanism to achieve adaptive optimization of the system.

[0075] The spatiotemporal feature extraction unit 41 is used to extract spatiotemporal features from the construction monitoring time-series data. In one embodiment of the present invention, the construction monitoring data includes the crown settlement, surrounding convergence displacement, anchor bolt axial force, and surrounding rock pressure gauge readings. These data have obvious time-series and spatial distribution characteristics. The spatiotemporal feature extraction unit 41 first uses one-dimensional convolution operation to extract the spatial distribution features of the monitoring data, with the convolution kernel size set to 5, the stride set to 1, and the number of output channels set to 64. Then, a gated recurrent unit is used to perform time-series modeling on the convolution output. The gated recurrent unit can effectively capture long-term dependencies and avoid the gradient vanishing problem of traditional recurrent neural networks. In one embodiment of the present invention, the hidden layer dimension of the gated recurrent unit is set to 128, and the time step is set to a 24-hour monitoring data sequence.

[0076] This invention proposes an innovative spatiotemporal collaborative feature fusion algorithm. This algorithm achieves deep extraction of spatiotemporal features from monitoring data through the fusion of bidirectional gated recurrent units and multi-scale convolutions. The calculation formula for the spatiotemporal fused features is as follows:

[0077] ,

[0078] in, For spatiotemporal fusion feature vectors, It is the Sigmoid activation function. To fuse the weight matrix, The spatial feature vector output by a one-dimensional convolution. This is the time feature vector output by the gated loop unit. For feature concatenation operators, This is the bias vector. The innovation of this algorithm lies in achieving adaptive fusion of spatiotemporal features through learnable fusion weights, enabling the model to dynamically adjust the importance weights of spatiotemporal features according to different geological conditions and monitoring data characteristics.

[0079] The load prediction unit 42 is used to predict the surrounding rock load value in future time periods based on spatiotemporal fusion features. In one embodiment of the present invention, the load prediction unit 42 adopts a multilayer perceptron structure, including three fully connected layers with 256, 128, and 64 neurons respectively. The number of neurons in the output layer is determined according to the prediction time window. In one possible implementation, the prediction time window is set to the next 7 days, and the number of neurons in the output layer is set to 7, corresponding to the predicted surrounding rock load value for each day of the next 7 days. The load prediction unit 42 also receives the pressure arch parameter set output by the adaptive pressure arch parameter calculation module 3 as auxiliary input, and concatenates the pressure arch span, height, and lateral pressure coefficient with the spatiotemporal fusion features before sending it into the prediction network to improve the physical interpretability of the prediction.

[0080] The deviation calculation unit 43 is used to calculate the deviation between the predicted load value and the measured value, and generate a feedback signal for parameter adjustment. In one embodiment of the present invention, the deviation calculation uses the root mean square error index, which can simultaneously reflect the magnitude and dispersion of the prediction deviation. The deviation calculation formula is as follows:

[0081] ,

[0082] in, This is the root mean square error (units are the same as the load). To predict the number of points, Let j be the predicted load value at the j-th prediction point. Let be the measured load value at the j-th prediction point. In one embodiment of the present invention, when the root mean square error exceeds a preset threshold, the system triggers a parameter adjustment mechanism. The preset threshold is determined according to the engineering accuracy requirements; in one embodiment of the present invention, the threshold is set to 10% of the average measured load.

[0083] The feedback adjustment unit 44 is used to generate a parameter correction signal based on the deviation calculation result and transmit the correction signal to the adaptive pressure arch parameter calculation module 3. In the technical solution of this invention, the feedback adjustment adopts a proportional-integral control strategy, and the calculation formula for the parameter correction signal is as follows:

[0084] ,

[0085] in, For parameter correction signals, This is the proportional gain coefficient. This is the integral gain coefficient. The prediction bias at time t. This is the integral variable. The parameter correction signal is applied to the calculated coefficients in the adaptive pressure arch parameter calculation module 3, specifically including the Weibull distribution parameters. and Adjustment, initial value of stress release rate Adjustment and lateral pressure coefficient Adjustments are made. Through a closed-loop feedback mechanism, the system can continuously optimize the calculation model parameters based on actual monitoring data, thereby improving prediction accuracy and engineering applicability.

[0086] In one embodiment of the present invention, the proportional gain coefficient The value ranges from 0.1 to 0.5, and the integral gain coefficient is... The value range is from 0.01 to 0.1. The specific values ​​of these two parameters need to be calibrated based on the actual engineering conditions. When the surrounding rock conditions are relatively stable, a smaller gain coefficient can be used to avoid system oscillations; when the surrounding rock conditions change significantly, the gain coefficient can be appropriately increased to accelerate the response speed. In a specific embodiment of this invention, for the Shenzhen-Shantou West Water Source and Water Supply Project, the values ​​were determined through on-site calibration. The value is 0.3. With a value of 0.05, this parameter combination can achieve a relatively fast convergence speed while ensuring system stability.

[0087] It is worth noting that the feedback adjustment unit 44 is also equipped with a parameter adjustment upper and lower limit protection mechanism to prevent excessive parameter adjustment from causing distortion of the calculation results. In one embodiment of the present invention, the single parameter adjustment range is limited to within ±20% of the original value, and the cumulative adjustment range is limited to within ±50% of the original value. When the parameter adjustment reaches the upper or lower limit, the system automatically issues an alarm, prompting engineering technicians to check the system input data and the working status of the monitoring equipment to eliminate any possible anomalies.

[0088] See Figure 6The data flow of the system of this invention includes the following steps: First, the multi-source data fusion acquisition module 1 collects geological parameter data, borehole monitoring data, and case data, and generates a fused feature vector after spatiotemporal alignment and feature fusion processing; then, the fused feature vector is input into the surrounding rock grade intelligent identification module 2, and outputs the surrounding rock grade classification result and feature importance weight after processing by a deep residual network and attention mechanism; next, the adaptive pressure arch parameter calculation module 3 calculates the span, height, and lateral pressure parameters of the pressure arch based on the surrounding rock classification result and feature weights; finally, the dynamic load prediction and optimization module 4 performs load prediction based on the pressure arch parameters and monitoring data, and feeds back the prediction deviation to the adaptive pressure arch parameter calculation module 3 for parameter correction. The entire data flow forms a complete closed-loop control, realizing the adaptive optimization of the system.

[0089] In one embodiment of the present invention, the system further includes a visualization output module for generating a three-dimensional spatial distribution map of the pressure arch, a load time history curve, and a surrounding rock stability assessment report. The visualization output module employs three-dimensional graphics rendering technology, which can intuitively display the spatial morphology and evolution process of the pressure arch, providing decision support for engineering technicians.

[0090] The four core modules of this invention exhibit the following synergistic effects: A data-driven feature enhancement relationship is formed between the multi-source data fusion acquisition module 1 and the intelligent rock grade identification module 2, providing more comprehensive and accurate rock grade information compared to a single data source; a classification-guided parameter optimization relationship is formed between the intelligent rock grade identification module 2 and the adaptive pressure arch parameter calculation module 3, directly improving the reliability of pressure arch parameter calculation with high-precision rock grade classification results; and a prediction-driven closed-loop feedback relationship is formed between the adaptive pressure arch parameter calculation module 3 and the dynamic load prediction and optimization module 4, enabling continuous optimization of the calculation model through feedback of prediction deviation information. These synergistic effects significantly enhance the overall performance of the system compared to the simple superposition of independent modules, demonstrating a nonlinear synergistic effect of 1+1>2.

[0091] The following example, using the water conveyance tunnel project of the Shenzhen-Shantou West Water Source and Water Supply Project, illustrates the specific application effect of the system of this invention. The total length of the water conveyance tunnel in this project is 6201.4m, including Tunnel No. 1 (823.45m) and Tunnel No. 2 (5378m). The surrounding rock types range from Class II to Class V, and the excavation cross-section is a portal-type structure. The excavation cross-section dimensions for Class II / III surrounding rock vary from 2.92m × 2.92m to 3.42m × 3.42m. Using the system of this invention to calculate the pressure arch of the surrounding rock, the system first collects geological parameters of the surrounding rock, drilling parameters of the intelligent drilling rig, and data from similar engineering cases through a multi-source data fusion acquisition module, generating a 64-dimensional fusion feature vector. Then, the fusion feature vector is processed by the intelligent identification module for surrounding rock grade, achieving a classification accuracy of 92.5%, significantly better than the traditional empirical classification method based on BQ values. Based on this, the adaptive pressure arch parameter calculation module calculated the pressure arch parameters according to the surrounding rock classification results and feature weights. The average deviation between the calculated pressure arch height and the field monitoring data was 8.3%, which is about 7 percentage points lower than the 15.6% deviation of the traditional Protodyakonov theory calculation method. The dynamic load prediction and optimization module continuously optimizes the calculation parameters through a closed-loop feedback mechanism. After three cycles of iterative optimization, the load prediction error stabilized within 6%, meeting the accuracy requirements of engineering design.

[0092] In the aforementioned engineering applications, the system employed differentiated calculation strategies for different surrounding rock types. For Class II surrounding rock sections, due to the relatively good rock mass integrity and moderate fissure development, the system focused on accurately calculating the static parameters of the pressure arch, with relatively small parameters set for the Weibull time-varying function to reflect the rapid stress release. For Class III surrounding rock sections, where the rock mass exhibits a certain degree of fissure cutting, the system appropriately increased the influence width coefficient when calculating the pressure arch parameters to reflect the impact of fissures on the stability of the surrounding rock. For Class IV and V surrounding rock sections, where the rock mass is fractured or in a fully weathered state with poor stability, the system adopted more conservative calculation parameters and increased the adjustment gain of the closed-loop feedback to quickly respond to changes in the surrounding rock condition.

[0093] The system of this invention can quickly identify changes in the surrounding rock grade by collecting changes in borehole parameters in real time, and adjust the pressure arch calculation parameters in a timely manner through a closed-loop feedback mechanism, thus successfully predicting the peak value of the surrounding rock load in the fault-affected zone and providing a reliable basis for construction decisions.

[0094] Furthermore, in an extended embodiment of the invention, the system also supports data interaction with 3D geological modeling software. By importing 3D geological model data of the tunnel, the system can display the morphological distribution and evolution process of the pressure arch in 3D space, providing more intuitive visualization support for engineering technicians. The 3D visualization function is implemented using WebGL rendering technology, supporting interactive operations such as rotation, scaling, and slice display, and can clearly show the spatial relationship between the pressure arch and the tunnel cross-section and support structure.

[0095] In another extended embodiment of the invention, the system supports multi-tunnel collaborative calculation. When a project involves multiple adjacent or intersecting tunnels, the system can comprehensively consider the mutual influence between tunnels and calculate the distribution of surrounding rock pressure considering process loads. This function is based on the superposition principle and interaction coefficient, and can accurately reflect the influence of the first-excavated tunnel on the surrounding rock pressure of the later-excavated tunnel, as well as the load transfer effect between adjacent tunnels.

[0096] In another embodiment of the present invention, the technical effectiveness of the system can be verified in the following way. For the classification performance of the intelligent rock grade identification module, a cross-validation method is used for evaluation. The dataset is divided into a training set and a test set, with the training set accounting for 80% and the test set accounting for 20%. Evaluation metrics include accuracy, precision, recall, and F1 score, which comprehensively reflect the performance of the classification model. In the test at the Shenzhen-Shantou West Water Source and Water Supply Project, the intelligent rock grade identification module achieved an average classification accuracy of 92.5% for the four types of surrounding rocks, with an accuracy of 95.2% for Class II, 91.8% for Class III, 89.3% for Class IV, and 93.7% for Class V. Compared with the traditional empirical classification method based on BQ values, the classification accuracy of the present invention is improved by an average of 16.3 percentage points.

[0097] The calculation accuracy of the adaptive pressure arch parameter calculation module was verified by comparing it with field measured data. In the test project, 15 representative monitoring sections were selected, and each section was equipped with arch crown settlement measuring points, perimeter convergence measuring points, and surrounding rock pressure gauges. The calculation accuracy of the pressure arch parameters was evaluated by comparing the calculated values ​​with the measured values. The test results show that the average relative error between the pressure arch height calculated using this invention and the measured value is 8.3%, which is approximately 7.3 percentage points higher than the average relative error of 15.6% calculated using traditional Protodyakonov theory. The average relative error for the calculation of the pressure arch span is 6.8%, and the average relative error for the calculation of the lateral pressure coefficient is 9.2%, both meeting the engineering design accuracy requirements.

[0098] To evaluate the prediction performance of the dynamic load prediction and optimization module, root mean square error (RMSE) and mean absolute percentage error (MASE) were used. Continuous monitoring and prediction comparisons were conducted over a period of 60 days at 15 monitoring sections of the test project. Test results showed that the average RMSE of the load prediction was 12.5 kPa, and the average MASE was 5.8%. Through continuous optimization using a closed-loop feedback mechanism, the prediction error showed a rapid decreasing trend in the first 10 days and stabilized around day 15. After three optimization cycles, the prediction error stabilized below 6%, indicating that the closed-loop feedback mechanism effectively improves the system's prediction accuracy and adaptability.

[0099] In another embodiment of the present invention, the computational efficiency of the system was evaluated. The test environment was a workstation configured with an Intel Core i7-10700 processor and 16GB of memory. The test included the calculation time of a single pressure arch parameter, the surrounding rock classification and reasoning time, and the load prediction time. The test results showed that the calculation time of a single pressure arch parameter was approximately 0.5s, the surrounding rock classification and reasoning time was approximately 0.2s, and the load prediction time was approximately 0.8s, for a total of approximately 1.5s for a single complete calculation process. This computational efficiency can meet the needs of real-time calculation at tunnel construction sites, providing timely decision support for engineering technicians.

[0100] In embodiments of this invention, the system supports multiple data interfaces and communication protocols, enabling it to interface with common engineering monitoring equipment and database systems. The geological parameter acquisition unit supports standard interface interfaces with geological exploration databases, automatically importing surrounding rock physical and mechanical parameters, geological structural information, and hydrogeological data. The borehole monitoring acquisition unit supports RS485 serial communication and TCP / IP network communication with intelligent drilling rigs, enabling real-time reception of borehole parameter data streams. The case data acquisition unit uses a RESTful API interface to interact with the engineering case database, supporting case retrieval based on geological condition similarity. The dynamic load prediction and optimization module supports data interface with the construction monitoring system, enabling real-time acquisition of monitoring data such as crown settlement, perimeter convergence, and anchor bolt axial force.

[0101] In a specific application scenario of this invention, the system can be deployed at the monitoring center of a tunnel construction site, connecting to various monitoring devices and a database via a local area network. Engineering technicians can view the calculation results of the pressure arch parameters, the predicted curve of the surrounding rock load, and the system's operating status in real time through the system's visual interface. When the predicted load exceeds a preset alarm threshold, the system automatically issues an alarm signal, reminding engineering technicians to take appropriate safety measures. The system also supports the generation of periodic analysis reports, including statistical analysis of the pressure arch parameters, an evaluation report of the prediction accuracy, and a comprehensive evaluation of the surrounding rock stability, providing data support for engineering management and decision-making.

[0102] The technical solution of this invention fully considers the complexity and variability of tunnel engineering. Through the synergistic effect of multi-source data fusion, intelligent classification, adaptive calculation, and closed-loop optimization, a complete calculation system for tunnel surrounding rock pressure arch is constructed. This system can not only adapt to the pressure arch calculation needs under different geological conditions, but also dynamically adjust the calculation model according to construction monitoring data, demonstrating good engineering applicability and promotional value.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A calculation system for tunnel surrounding rock pressure arch, characterized in that, include: The multi-source data fusion acquisition module is used to collect geological parameter data of tunnel surrounding rock, construction borehole monitoring data and historical engineering case data. It performs spatiotemporal alignment and feature standardization processing on the collected data to generate a fusion feature vector in a unified format. The surrounding rock grade intelligent identification module is used to receive the fused feature vector, extract deep representations of the geological features of the surrounding rock based on a deep residual network, and combine an attention mechanism to weight and enhance key features, and output the surrounding rock grade classification result and feature importance weights. The adaptive pressure arch parameter calculation module is used to calculate the pressure arch span benchmark value based on the surrounding rock grade classification result and the feature importance weight, using an improved Protodyakonov theory model, and to calculate the dynamic value of the pressure arch height based on the stress release time-varying function, thereby generating a pressure arch parameter set containing the pressure arch span, height, and lateral pressure coefficient. The dynamic load prediction and optimization module is used to predict the evolution trend of the surrounding rock load based on the pressure arch parameter set and real-time monitoring data, using a spatiotemporal collaborative prediction network, calculate the deviation feedback signal between the predicted load and the measured load, and transmit the deviation feedback signal to the adaptive pressure arch parameter calculation module for parameter correction, thus forming a closed-loop optimization control.

2. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, The multi-source data fusion acquisition module includes: The geological parameter acquisition unit is used to collect information on surrounding rock lithology, joint and fissure development, groundwater level, and surrounding rock integrity indicators. The borehole monitoring and acquisition unit is used to collect drilling speed, impact pressure, rotational pressure and propulsion pressure in real time; The case data acquisition unit is used to retrieve historical pressure arch calculation case data under similar geological conditions from the engineering case database.

3. The tunnel surrounding rock pressure arch calculation system according to claim 2, characterized in that, The multi-source data fusion acquisition module also includes a feature fusion unit, which is used to perform spatiotemporal alignment of the geological parameters, the borehole monitoring data and the case data, fill in missing data points using least squares interpolation, and splice the data after normalization of various types of data to generate the fused feature vector.

4. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, The surrounding rock grade intelligent identification module includes: The feature extraction unit is used to perform multi-layer convolution operations on the fused feature vector using a deep residual network to extract a hierarchical representation of the geological features of the surrounding rock. The attention enhancement unit is used to calculate the correlation matrix between each feature dimension based on the self-attention mechanism and to dynamically weight key geological features. The classification output unit is used to input the weighted feature representation into the fully connected classification layer and output the classification probability distribution of Class II to Class V surrounding rocks and the importance weight of each feature.

5. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, The adaptive pressure arch parameter calculation module includes: A span calculation unit is used to calculate the reference value of the pressure arch span based on the tunnel excavation width, the surrounding rock grade, and the feature importance weight; The height calculation unit is used to calculate the dynamic value of the pressure arch height based on the stress release time-varying function and the Protodyakonov coefficient of the surrounding rock. The lateral pressure calculation unit is used to calculate the lateral pressure value based on the burial depth, the lateral pressure coefficient of the surrounding rock, and the horizontal stress.

6. The tunnel surrounding rock pressure arch calculation system according to claim 5, characterized in that, The height calculation unit uses the Weibull time function as the stress release time-varying function. The function parameters are determined by fitting the field monitoring data to achieve dynamic tracking of the evolution of the pressure arch height over time.

7. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, The dynamic load prediction and optimization module includes: The spatiotemporal feature extraction unit is used to extract spatial features by performing one-dimensional convolution operations on construction monitoring time series data, and to extract time series features by using a gated loop unit. The load prediction unit is used to input spatiotemporal characteristics into the prediction network and output the predicted value of the surrounding rock load within a preset future time period. The deviation calculation unit is used to calculate the root mean square deviation between the predicted load and the measured load. The feedback adjustment unit is used to generate a parameter correction signal based on the deviation value and adjust the calculation coefficients in the adaptive pressure arch parameter calculation module.

8. The tunnel surrounding rock pressure arch calculation system according to claim 7, characterized in that, The feedback adjustment unit uses a proportional-integral control strategy to generate parameter correction signals, which are used to adjust the Weibull distribution parameters, the initial value of the stress release rate, and the lateral pressure coefficient.

9. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, The system also includes a visualization output module for generating a three-dimensional spatial distribution map of the pressure arch, a load time history curve, and a surrounding rock stability assessment report.

10. The tunnel surrounding rock pressure arch calculation system according to claim 1, characterized in that, A parameter-level deep coupling relationship is established between the multi-source data fusion acquisition module, the surrounding rock grade intelligent identification module, the adaptive pressure arch parameter calculation module, and the dynamic load prediction and optimization module, with the output of the former module directly serving as the key input parameter of the latter module.

Images