Response test method and device for soil layer deformation and tunnel stress in foundation pit excavation

Abstract

The invention discloses a response test method and device for soil layer deformation and tunnel stress in foundation pit excavation, and relates to the technical field of underground engineering tests. The method comprises the following steps: acquiring engineering geology and tunnel structure parameters, and constructing a geological tunnel coupling similarity model embedded with a layered sensing array; step-by-step excavation is adopted to simulate foundation pit construction, and soil layer displacement strain, tunnel stress contact pressure and excavation disturbance parameters are synchronously collected through multiple channels; constructing a deformation stress coupling prediction model and an excavation disturbance quantification model; and constructing a dynamic coupling response model based on the two, and outputting a dynamic coupling curve of soil layer deformation and tunnel stress. The device comprises a model box body, a data processing host and a display terminal, a layered sensing array, a dynamic loading device, a servo excavation mechanism and a multi-channel data acquisition instrument are arranged in the box body, and the whole test process integrated operation can be realized. According to the method, the dynamic coupling rule can be accurately captured, and technical support is provided for safety assessment and parameter optimization of underground engineering construction.

Description

Technical Field

[0001] This invention relates to the field of underground engineering testing technology. Specifically, it relates to a method and apparatus for real-time testing of soil deformation and tunnel stress during foundation pit excavation. Background Technology

[0002] With the increasing intensity of urban underground space development, the spatial intersection of foundation pit projects and existing tunnels is becoming more frequent. During foundation pit excavation, soil unloading triggers stress redistribution, causing deformation of the surrounding soil and consequently influencing the adjacent tunnel structure with additional stress and displacement. This coupling effect can easily lead to tunnel structural cracking, water leakage, and other defects, and in severe cases, threaten the operational safety of the tunnel. Therefore, accurately capturing the dynamic response relationship between foundation pit excavation and tunnel stress has become a key technical problem that urgently needs to be solved in the field of underground engineering. In current engineering practice, traditional monitoring methods mostly rely on single-point sensors, which have shortcomings such as limited monitoring range and poor data synchronization, making it difficult to fully reflect the coupling mechanism between the soil layer and the tunnel, thus posing a challenge to engineering safety risk assessment.

[0003] Existing research on foundation pit excavation and tunnel response mainly falls into three technical approaches: field measurement, numerical simulation, and model testing. Field measurement is limited by factors such as the complexity of geological conditions and construction interference, resulting in high testing costs and difficulty in repeated verification. While numerical simulation methods can achieve multi-condition analysis, the selection of model parameters relies on empirical assumptions, leading to insufficient adaptability to complex geological conditions and difficulty in guaranteeing the accuracy of prediction results. Model testing, as an important means of bridging theoretical analysis and engineering practice, can recreate actual engineering scenarios through the principle of similarity. However, traditional model testing suffers from problems such as outdated sensing technology, limited loading methods, and difficulty in accurately simulating the excavation process. It cannot achieve real-time synchronous monitoring of soil deformation and tunnel stress, thus hindering in-depth research into the dynamic coupling response law.

[0004] At the technical level, existing model testing methods have not yet formed a complete system of parameter acquisition, similarity modeling, real-time monitoring, and coupled prediction technologies. On the one hand, the design of similar material mixes lacks precise control methods for the physical and mechanical properties of different soil layers, resulting in insufficient consistency between the mechanical response of the model and the prototype. On the other hand, monitoring systems mostly use single-type sensors, making it difficult to simultaneously capture multi-dimensional deformation of soil layers and the complex stress state of tunnels. Furthermore, data processing is mostly based on static analysis, lacking quantitative assessment of the dynamic impact of excavation disturbance. In addition, existing prediction models mostly focus on single-dimensional deformation or stress prediction, failing to fully consider the dynamic correlation between excavation disturbance and coupling effects, and thus unable to provide real-time and accurate risk warnings for engineering construction.

[0005] Therefore, there is an urgent need for a real-time response testing method and device that integrates monitoring, simulation and prediction to accurately capture the dynamic coupling law between soil deformation and tunnel stress, thereby improving the safety level of underground engineering construction and optimizing engineering design schemes, which has important engineering value and practical significance. Summary of the Invention

[0006] Based on the aforementioned technical problems, this application discloses a test method and apparatus for the response of soil deformation and tunnel stress during foundation pit excavation. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation includes:

[0007] Obtain engineering geological parameters and tunnel structural parameters. The engineering geological parameters include soil physical and mechanical properties, groundwater distribution, and initial in-situ stress. The tunnel structural parameters include cross-sectional dimensions, material strength, and burial depth.

[0008] Based on the engineering geological parameters and tunnel structural parameters, a geological tunnel coupled similarity model is constructed. The geological tunnel coupled similarity model is embedded with a layered sensor array and a dynamic loading interface. The layered sensor array is arranged according to the soil layer depth and the tunnel circumferential axis.

[0009] The excavation of the foundation pit was simulated by a step-by-step excavation method. The excavation was carried out in layers and sections according to the preset excavation thickness and speed. Real-time monitoring data of the layered sensor array was acquired through multi-channel synchronous acquisition. The real-time monitoring data included soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters.

[0010] A first model and a second model are constructed. The first model is a deformation-stress coupling prediction model, which outputs a multi-dimensional deformation-stress prediction vector. The second model is an excavation disturbance quantification model, which outputs the excavation disturbance intensity coefficient.

[0011] Based on the multi-dimensional deformation and stress prediction vector and the excavation disturbance intensity coefficient, a dynamic coupled response model is constructed to output the dynamic coupling curve of soil deformation and tunnel stress during the excavation process.

[0012] Preferably, the construction of the coupled similarity model of the geological tunnel includes:

[0013] Based on similarity theory, the geometric similarity ratio, material similarity ratio and stress similarity ratio are determined. Soil layer similarity material is prepared according to a preset ratio. The soil layer similarity material is made by mixing prototype soil sand particles and binder according to a preset ratio. The compaction degree is controlled by compaction test.

[0014] The tunnel model is scaled down proportionally according to the tunnel structural parameters, and reserved holes for installing the sensor array are provided. The positions of the holes correspond to the positions of the layered sensor arrays.

[0015] Initial ground stress is applied through a dynamic loading interface to simulate a natural stress field. The loading process adopts a graded loading method, and the next level of loading is carried out only after a preset stabilization time is reached after each level of loading.

[0016] Preferably, the first model outputs a multi-dimensional deformation and stress prediction vector, converting the soil displacement and strain, and tunnel stress and contact pressure in the real-time monitoring data into a standardized feature vector; the vector is input into the first model for coupling calculation, and the first model outputs a multi-dimensional vector containing predicted values ​​of vertical soil deformation, horizontal displacement, tunnel circumferential stress, and axial strain, as shown in the formula: ,in, This is a multi-dimensional deformation force prediction vector. To standardize the feature vector, , These are the model weight matrices, , These are the bias vectors, Activation function It is the hyperbolic tangent activation function.

[0017] Preferably, the first model deformation-stress coupling prediction model adopts a dual-branch network architecture. The first branch processes soil deformation-related features, and the second branch processes tunnel stress-related features. A cross-branch attention mechanism is used to capture the coupling relationship between the two types of features. The attention weight formula is as follows: ,in, For the first The characteristics of the soil layer and the first Coupling weights for each tunnel feature, This is a query vector for soil layer features. The tunnel feature key vector. The total number of tunnel features is represented by the value of 1; coupled features and branch features are fused together, and a multi-dimensional deformation and stress prediction vector is output through a fully connected layer.

[0018] Preferably, the excavation disturbance intensity coefficient is output through the second model, and the excavation disturbance parameters, including excavation speed, excavation thickness, and disturbance duration, are extracted from the real-time monitoring data; the disturbance parameters are quantified and calculated using the second model to output the excavation disturbance intensity coefficient. ,in, This is the excavation disturbance intensity coefficient. For a moment The excavation speed For a moment The excavation thickness, For a moment The depth of the disturbance's impact. For tunnel burial depth, The attenuation coefficient is... , These represent the start and end times of the excavation period.

[0019] Preferably, the second model excavation disturbance quantification model is based on historical excavation test data, establishing a sample library of mapping relationships between disturbance parameters and disturbance intensity; the model is trained using the gradient boosting tree algorithm, and the model's loss function formula is: ,in, This represents the model loss value. For the first The actual perturbation intensity of each sample For predicted values, The regularization coefficient is . For the first The complexity of a decision tree, The total number of samples, The number of decision trees.

[0020] Preferably, the construction of the dynamic coupled response model is achieved by fusing the multi-dimensional deformation-stress prediction vector with the excavation disturbance intensity coefficient, as shown in the following formula: ,in, To fuse feature vectors, This is a multi-dimensional deformation force prediction vector. The excavation disturbance intensity coefficient is used to construct a dynamic coupling response model based on a long short-term memory network. The input is a fused feature vector and a time series of real-time monitoring data, and the output is the feature parameters of the dynamic coupling curve.

[0021] Preferably, the construction of the dynamic coupling response model involves introducing a time decay factor to correct the coupling weights at different excavation stages. The time decay factor is: ,in, For a moment Time decay factor, The decay rate coefficient, The excavation start time is used; a gating mechanism controls the fusion ratio of historical and current data, and the calculation of update and reset gates satisfies the following: ,in, To update the door, To reset the door, For a moment The fused feature vector, This is the hidden state from the previous moment. , These are the weight matrices, , These are the bias vectors.

[0022] Preferably, the output of the dynamic coupling curve includes: constructing a piecewise function to describe the coupling relationship based on the feature parameters output by the dynamic coupling response model, including curve peak value, rate of change, and stability threshold; and converting the feature parameters into a continuous curve according to the time series, with the curve equation being:

[0023]

[0024] in, For the dynamic coupling curve at time 10:00 The value of , , , , , These are the curve fitting coefficients, The attenuation coefficient is... This refers to the time interval during the excavation phase. For the stable phase time interval, through data smoothing, a continuous dynamic coupling curve of soil deformation and tunnel stress is output.

[0025] The real-time response test device for soil deformation and tunnel stress during the excavation of the foundation pit includes a model box, a data processing host and a display terminal; the model box is equipped with a layered sensor array, a dynamic loading device, a servo excavation mechanism and a multi-channel data acquisition instrument.

[0026] The model box is filled with soil-like materials and a tunnel model is laid out inside to form a geological tunnel coupling similar model. The side walls of the model box are equipped with transparent observation windows, and the top is equipped with a loading interface adapted to the dynamic loading device.

[0027] The layered sensing array includes distributed optical fiber sensors and miniature piezoelectric sensors. The distributed optical fiber sensors are buried in similar materials in the soil layer according to the soil layer depth. The miniature piezoelectric sensors are pasted and fixed along the circumferential and axial directions of the tunnel model. The signal output terminals of all sensors are connected to the signal input terminals of the multi-channel data acquisition instrument through shielded cables.

[0028] The dynamic loading device includes a hydraulic loader and a pressure controller. The output end of the hydraulic loader is attached to the bottom and side wall of the model box through a loading interface. The pressure controller is connected to the hydraulic loader through a control line and is used to adjust the loading pressure to simulate the initial ground stress.

[0029] The servo excavation mechanism includes a servo motor, a mechanical excavator arm, and a displacement sensor. The servo motor is fixed to the top of the model box by a bracket. The mechanical excavator arm is connected to the output shaft of the servo motor. The displacement sensor is installed on the mechanical excavator arm. Both the servo motor and the displacement sensor are connected to the data processing host via data cables. The data processing host adjusts the excavation thickness and speed of the mechanical excavator arm through control commands.

[0030] The signal output terminal of the multi-channel data acquisition instrument is connected to the data processing host via a data cable, and is used to transmit the acquired soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters to the data processing host.

[0031] The data processing host has a built-in deformation-force coupling prediction model, excavation disturbance quantification model and dynamic coupling response model. The signal output terminal of the data processing host is connected to the display terminal to output the dynamic coupling curve and real-time monitoring data to the display terminal.

[0032] Compared with the prior art, the technical solution of this application has the following technical effects:

[0033] This invention constructs a geological tunnel coupled similarity model with an embedded layered sensor array and combines it with multi-channel synchronous acquisition technology to achieve comprehensive real-time monitoring of soil displacement and strain, tunnel stress and contact pressure, and excavation disturbance parameters. By deploying the layered sensor array according to the soil depth and the tunnel circumferential and axial directions, it can cover a wider monitoring range, ensure data synchronization and integrity, and solve the problems of incomplete information and poor synchronization of traditional monitoring methods.

[0034] The first model of this invention adopts a dual-branch network architecture and a cross-branch attention mechanism, which can accurately output multi-dimensional deformation and stress prediction vectors and fully capture the coupling characteristics of soil deformation and tunnel stress. The second model is based on the gradient boosting tree algorithm, which can scientifically quantify the excavation disturbance intensity, break through the limitations of single-dimensional prediction, make deformation and stress prediction and disturbance quantification more scientific, and greatly improve the prediction accuracy of the dynamic interaction between soil and tunnel.

[0035] The dynamic coupling response model of this invention, through feature fusion and long short-term memory network, introduces time decay factor and gating mechanism, and can output continuous dynamic coupling curves of soil deformation and tunnel stress. It can clearly present the changing law of coupling relationship at different excavation stages, intuitively reflect the dynamic impact of excavation disturbance on coupling effect, and provide a visual basis for engineers to grasp the response status of soil and tunnel in real time during construction, helping to adjust construction parameters in a timely manner.

[0036] The experimental device of this invention integrates dynamic loading, servo excavation, and multi-channel acquisition functions, realizing integrated operation from similar model construction to data processing and result display. The dynamic loading device can accurately simulate the initial ground stress, the servo excavation mechanism can perform layered and segmented excavation according to preset parameters, and the data processing host has built-in multiple models to realize real-time analysis, which improves the overall experimental efficiency and accuracy, and provides reliable technical support for underground engineering safety risk assessment and construction optimization.

[0037] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0038] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0040] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0041] Figure 1 A schematic diagram of the test method for real-time response of soil deformation and tunnel stress during foundation pit excavation;

[0042] Figure 2 This is a schematic diagram of the dual-branch network architecture of the deformation-force coupling prediction model (first model);

[0043] Figure 3 A schematic diagram of the architecture process for the excavation disturbance quantification model (second model);

[0044] Figure 4 A schematic diagram of the test device for real-time response of soil deformation and tunnel stress during foundation pit excavation;

[0045] Figure 5 A schematic diagram of the working process of the test device for real-time response of soil deformation and tunnel stress during foundation pit excavation;

[0046] Figure 6 Comparison of dynamic coupling curves under conditions of sandy silt and moderate excavation intensity;

[0047] Figure 7 The diagram shows the dynamic coupling curve characteristics of this application under three typical soil layer conditions. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0049] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0050] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0051] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0052] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0053] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0054] Example 1

[0055] This embodiment mainly describes the real-time response test method for soil deformation and tunnel stress during foundation pit excavation, such as... Figure 1 As shown, it specifically includes:

[0056] Obtain engineering geological parameters and tunnel structural parameters. The engineering geological parameters include soil physical and mechanical properties, groundwater distribution and initial ground stress. The tunnel structural parameters include cross-sectional dimensions, material strength and burial depth.

[0057] Based on the engineering geological parameters and tunnel structural parameters, a geological tunnel coupled similarity model is constructed. The geological tunnel coupled similarity model is embedded with a layered sensor array and a dynamic loading interface. The layered sensor array is arranged according to the soil layer depth and the tunnel circumferential and axial directions.

[0058] The excavation of the foundation pit was simulated by a step-by-step excavation method. The excavation was carried out in layers and sections according to the preset excavation thickness and speed. Real-time monitoring data of the layered sensor array was acquired through multi-channel synchronous acquisition. The real-time monitoring data included soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters.

[0059] A first model and a second model are constructed. The first model is a deformation-stress coupling prediction model, which outputs a multi-dimensional deformation-stress prediction vector. The second model is an excavation disturbance quantification model, which outputs the excavation disturbance intensity coefficient.

[0060] Based on the multi-dimensional deformation and stress prediction vector and the excavation disturbance intensity coefficient, a dynamic coupled response model is constructed to output the dynamic coupling curve of soil deformation and tunnel stress during the excavation process.

[0061] Furthermore, the engineering geological parameters and tunnel structural parameters need to be obtained. Specifically, when obtaining engineering geological parameters, the physical and mechanical properties of the soil layer should include the natural density, water content, void ratio, compression modulus, cohesion, and internal friction angle of the soil layer. These parameters should be determined through indoor geotechnical tests, such as using the ring cutter method to measure natural density, the oven-drying method to measure water content, the consolidation test to measure compression modulus, and the direct shear test or triaxial compression test to measure cohesion and internal friction angle. The groundwater distribution should clearly define the groundwater level depth, aquifer thickness, and permeability coefficient, which can be obtained through borehole water level observation, pumping tests, or pressure tests. The initial in-situ stress should determine the vertical and horizontal initial stress values, which can be measured using on-site testing methods such as hydraulic fracturing and stress relief methods.

[0062] When obtaining tunnel structural parameters, the tunnel cross-sectional dimensions need to be accurately measured, including the outer diameter, inner diameter, cross-sectional height, and width. The material strength needs to be obtained, including the cubic compressive strength, axial compressive strength, and tensile strength of the tunnel segment concrete, determined through standard concrete test blocks. The tunnel burial depth needs to be measured, including the vertical distance from the tunnel top to the ground surface, and determined by combining engineering survey data with on-site measurement data. This ensures the accuracy and completeness of the parameters and provides precise basic data for the subsequent construction of a geological tunnel coupling similarity model.

[0063] A coupled similarity model of a geological tunnel is constructed, and the similarity ratio and similar materials are determined. Based on similarity theory, the geometric similarity ratio, material similarity ratio, and stress similarity ratio are first determined. The three must satisfy the constraint relationship of the similarity principle to ensure that the model and the prototype have the same mechanical response. The geometric similarity ratio is determined according to the space of the test site, the size of the model box and the scale of the prototype project. The material similarity ratio needs to match the geometric similarity ratio and the stress similarity ratio. The stress similarity ratio is set by combining the initial ground stress of the prototype and the loading capacity of the model.

[0064] The soil-similar material is made by mixing prototype soil, sand, and binder in a preset ratio. The prototype soil must be a soil sample with properties similar to the actual soil layer in the project. The sand must be selected from a specific particle size range to match the particle size distribution of the prototype soil. The binder can be gypsum, cement, or other materials that meet the mechanical performance requirements. By adjusting the ratio of the three components, the physical and mechanical properties (such as density, compression modulus, cohesion, etc.) of the similar material are made to meet the material similarity ratio requirements. After mixing, the compaction degree of the similar material is controlled by a compaction test. The compaction test must be carried out according to a preset compaction energy to ensure that the compaction degree of the similar material is consistent with the compaction degree of the prototype soil layer, simulating the dense state of the real soil layer.

[0065] The tunnel model is made in scale according to the determined geometric similarity ratio and the tunnel structural parameters. The material of the tunnel model must be selected to be similar to the mechanical properties of the prototype tunnel segments, such as micro concrete and resin composite materials, to ensure that the elastic modulus, compressive strength and other parameters of the model tunnel meet the material similarity ratio requirements.

[0066] During the tunnel model fabrication process, it is necessary to pre-reserve mounting channels for the sensor array. The location, number, and size of the channels must strictly correspond to the layout scheme of the layered sensor array. The channels along the tunnel circumference must be evenly distributed to cover different angular positions along the tunnel circumference, while the channels along the tunnel axis must be laid out at a set interval to ensure that the miniature piezoelectric sensors can be accurately attached and fixed in the channels, enabling effective monitoring of the tunnel's circumferential and axial stress and strain. The inner walls of the channels must be smoothed to avoid damage to the sensors during installation, while ensuring that the sensors are in close contact with the surface of the tunnel model to ensure the accuracy of the monitoring data.

[0067] The initial ground stress is simulated by applying it to the natural stress field through a dynamic loading interface of a geological tunnel-coupled similar model. This dynamic loading interface must be precisely connected to a dynamic loading device (such as a hydraulic loader) to ensure that the loading force is uniformly transmitted to the model. The loading process employs a staged loading method. First, the pressure value and loading time for each stage are set. The loading pressure for each stage is determined based on the stress similarity ratio and the initial ground stress of the prototype. During loading, the output pressure of the hydraulic loader is precisely adjusted by a pressure controller, allowing the stress on the model to gradually reach the set value for each stage.

[0068] After each loading stage is completed, a preset stabilization time is required. The stabilization time needs to be determined based on the creep characteristics of similar materials to ensure that the deformation of similar materials is stable under the stress of that stage, and to avoid affecting the accuracy of subsequent loading due to unstable deformation. After the deformation is stable, the deformation of the model is monitored by displacement sensors in the layered sensing array. The next loading stage is carried out only after the deformation is confirmed to be stable, until the initial ground stress on the model reaches the set value, so as to accurately simulate the natural stress field of the engineering site.

[0069] When simulating foundation pit construction using a step-by-step excavation method, the total excavation depth and excavation range of the model foundation pit are first determined based on the excavation depth, excavation range, and construction plan of the prototype foundation pit, combined with the geometric similarity ratio. Then, the preset excavation thickness and excavation speed are set. The excavation thickness needs to be determined according to the layout spacing of the layered sensor array to ensure that the soil deformation can be monitored by the corresponding depth sensor after each layer is excavated. The excavation speed needs to be referenced to the actual excavation speed of the prototype foundation pit and converted into the excavation speed of the model based on the similarity ratio to ensure that the excavation process is consistent with the prototype construction conditions.

[0070] During the excavation process, the servo excavation mechanism (including servo motors, mechanical excavators, and displacement sensors) performs layered and segmented excavation according to the set excavation thickness and speed. The servo motor needs to precisely adjust its speed according to the control commands to drive the mechanical excavator. The displacement sensors monitor the displacement of the mechanical excavator in real time to ensure that the thickness of each excavation layer is precisely controlled within the preset range. The segmented excavation needs to be carried out sequentially according to the set excavation zones. After the excavation of each zone is completed, the excavation is paused and stabilized for a preset time. After the soil deformation stabilizes, the excavation of the next zone is carried out, simulating the layered and segmented excavation construction process of the prototype foundation pit.

[0071] Multi-channel synchronous acquisition of the layered sensor array is achieved through a multi-channel data acquisition instrument. The multi-channel data acquisition instrument needs to be connected to all sensors (distributed fiber optic sensors, miniature piezoelectric sensors, etc.) in the layered sensor array through shielded cables. The shielded cables need to have anti-interference capabilities to avoid external electromagnetic interference affecting the transmission of monitoring signals.

[0072] Before data acquisition, the acquisition frequency and duration need to be set. The acquisition frequency should be determined based on the excavation speed and the response characteristics of the soil layer and tunnel to ensure that the dynamic changes in soil deformation and tunnel stress during the excavation process can be captured. The acquisition duration should cover the entire excavation process and the stable stage after excavation. During the acquisition process, the multi-channel data acquisition instrument simultaneously acquires soil displacement and strain data monitored by distributed fiber optic sensors and tunnel stress contact pressure data monitored by micro piezoelectric sensors. At the same time, the displacement and velocity sensors of the servo excavation mechanism acquire excavation disturbance parameters (such as excavation speed, excavation thickness, and disturbance duration). All monitoring data are stored in real time to the storage module of the data acquisition instrument to ensure data integrity and synchronization, providing continuous and accurate raw data for subsequent model calculations.

[0073] Furthermore, such as Figure 2 As shown, a first model deformation-stress coupling prediction model is constructed, outputting a multi-dimensional deformation-stress prediction vector. Specifically, this includes preprocessing the soil displacement-strain and tunnel stress-contact pressure data from real-time monitoring, removing outliers (such as those obtained through...). The criteria identify and remove outlier data. Then, the preprocessed data is standardized using the Z-score standardization method, transforming the data into a standardized feature vector X with a mean of 0 and a standard deviation of 1. The formula is: ,in This is the raw monitoring data. The mean of the data. The standard deviation of the data is used to ensure that the monitoring data of different dimensions have the same order of magnitude, so as to avoid the model calculation accuracy being affected by the difference in data volume.

[0074] The first model employs a dual-branch network architecture. The first branch (soil deformation branch) processes soil deformation-related features. The input is the feature components related to soil displacement and strain in the standardized feature vector. This branch contains multiple convolutional and pooling layers. The convolutional layers perform convolution operations on the input features using a set kernel size and stride to extract local features of soil deformation. The pooling layers use max pooling or average pooling to downsample the output features of the convolutional layers, reducing feature dimensionality while retaining key features. The second branch (tunnel stress branch) processes tunnel stress-related features. The input is the feature components related to tunnel stress contact pressure in the standardized feature vector. The network structure is consistent with the first branch, extracting local features of tunnel stress through the same convolution and pooling operations.

[0075] By employing a cross-branch attention mechanism to capture the coupling relationship between soil deformation features and tunnel stress features, the soil features output from the first branch are first converted into soil feature query vectors. ( (The number of soil layer features) converts the tunnel features output from the second branch into a tunnel feature key vector. ( (This refers to the number of tunnel features); then, the attention weight formula is used to calculate the first... The characteristics of the soil layer and the first Coupling weights of tunnel features The formula is: ,in The formula normalizes the inner product of the soil layer feature query vector and the tunnel feature key vector using the Softmax function to obtain the coupling weight. The larger the weight value, the stronger the coupling relationship between the corresponding soil layer feature and the tunnel feature.

[0076] The calculated coupling weights are multiplied by the tunnel features to obtain the coupling features. These coupling features are then fused with the soil layer features of the first branch and the tunnel features of the second branch to form a fused feature vector. This fused feature vector is input into a fully connected layer, which contains an input layer, hidden layers, and an output layer. The hidden layers use the tanh hyperbolic tangent activation function, and the output layer uses... Activation functions (such as the Sigmoid function), through formulas Perform coupled operations, where , These are the model weight matrices for the hidden layer and the output layer, respectively. , These are the bias vectors for the hidden layer and the output layer, respectively. The output is a multi-dimensional deformation and stress prediction vector containing predicted values ​​for vertical soil deformation, horizontal displacement, tunnel circumferential stress, and axial strain. .

[0077] Furthermore, such as Figure 3 As shown, the construction of the second model and the output of the excavation disturbance intensity coefficient specifically involves: extracting excavation disturbance parameters from real-time monitoring data, including time... Excavation speed (Acquired by the speed sensor of the servo excavation mechanism), time Excavation thickness (Calculated by displacement sensor monitoring of mechanical excavator arm displacement), Time Depth of disturbance impact (Using distributed fiber optic sensors to monitor soil deformation at different depths to determine the extent of disturbance impact), tunnel depth (Determined based on tunnel structural parameters), start time of excavation period With end time (Determined based on excavation execution records).

[0078] Based on historical excavation test data, excavation disturbance parameters were collected under different engineering scenarios. , , , , , And the corresponding actual excavation disturbance intensity (determined through field measurement or numerical simulation), establish a sample library of mapping relationship between disturbance parameters and disturbance intensity. The sample library should contain a sufficient number of samples (such as thousands of samples) to cover different geological conditions, tunnel parameters and excavation conditions, to ensure the diversity and representativeness of the samples and provide sufficient data support for model training.

[0079] The second model is trained using the gradient boosting tree algorithm, which consists of multiple decision trees (M trees), each acting as a weak learner. Model performance is gradually improved through iterative training. During training, the first decision tree is initialized using the mean of the actual perturbation strengths of all samples in the sample library as the initial predicted value. Then, the residual for each sample (the difference between the actual perturbation strength and the current predicted value) is calculated, and the residual is used as the target variable to train the next decision tree, enabling it to fit the residual. This process is repeated until training is complete. A decision tree;

[0080] The loss function for model training uses the formula ,in This represents the model loss value. For the first The actual perturbation intensity of each sample For the first The predicted perturbation intensity for each sample. The total number of samples, This is the regularization coefficient (used to control model complexity and avoid overfitting). For the first The complexity of each decision tree (measured by parameters such as the number of nodes and depth) is adjusted by minimizing the loss function to ensure that the model has high prediction accuracy and generalization ability.

[0081] The real-time extracted excavation disturbance parameters are input into the trained second model, and the model outputs the excavation disturbance intensity coefficient through quantization calculation. The calculation formula is: ,in This is the integral of the product of excavation speed and excavation thickness during the excavation period, reflecting the total disturbance energy during the excavation process. The average perturbation energy; For disturbance attenuation term, The attenuation coefficient (determined based on fitting historical sample data, reflecting the attenuation pattern of disturbance with depth). As the relative depth of disturbance influence, the formula comprehensively considers the impact of excavation speed, excavation thickness, depth of disturbance influence, and tunnel burial depth on the intensity of disturbance, and accurately quantifies the degree of effect of excavation disturbance on soil layers and tunnels;

[0082] Furthermore, a dynamic coupling response model is constructed and a dynamic coupling curve is output, specifically: based on the multi-dimensional deformation force prediction vector output by the first model. The excavation disturbance intensity coefficient output by the second model To construct a fused feature vector Z, feature fusion is performed. The fusion formula is as follows: ,in Multidimensional deformation force prediction vector transpose, Excavation disturbance intensity coefficient and The element-wise product combines the excavation disturbance intensity with the deformation and stress prediction features, so that the fused feature vector contains both the prediction information of soil deformation and tunnel stress, and reflects the influence of excavation disturbance on both, providing a comprehensive feature input for dynamic coupling analysis.

[0083] The dynamic coupling response model is built on a long short-term memory network (LSTM). The LSTM network includes an input gate, a forget gate, an update gate, and a reset gate, which can effectively handle long-term dependencies in time series data and is suitable for dynamic analysis of soil deformation and tunnel stress during excavation.

[0084] By introducing a time decay factor The coupling weights for different excavation stages are adjusted, and the time decay factor formula is as follows: ,in For the current moment, The start time of excavation. The attenuation rate coefficient (determined by fitting the attenuation characteristics of the coupling relationship during historical excavation) makes the model assign different weights to the coupling features at different excavation stages. The attenuation coefficient is small and the coupling weight is high in the early stage of excavation, while the attenuation coefficient is large and the coupling weight is low in the later stage of excavation, which is consistent with the law of the change of disturbance influence over time in the actual excavation process.

[0085] The gating mechanism controls the fusion ratio of historical and current data, and updates the gate. With Reset Door The calculation satisfies: ,in For a moment The fused feature vector, For the previous moment ( The network's hidden state at any given time. , These are the weight matrices for the update gate and the reset gate, respectively. , These are the bias vectors for the update gate and the reset gate, respectively. The sigmoid activation function is used; the update gate controls the influence of the historical hidden state on the current hidden state, and the reset gate controls the dependence of the current input on the historical hidden state. Through the synergistic effect of the two gates, the model can adaptively adjust the fusion ratio of historical and current data, accurately capturing changes in dynamic coupling relationships.

[0086] The dynamic coupling curve is output by extracting feature parameters. Based on the output of the dynamic coupling response model, the feature parameters of the dynamic coupling curve are extracted, including the curve peak value (maximum soil deformation value, maximum tunnel stress value and corresponding occurrence time), rate of change (the slope of the change of soil deformation and tunnel stress at different excavation stages, reflecting how fast the deformation and stress change), and stability threshold (the numerical range of soil deformation and tunnel stress when they tend to stabilize after excavation and the corresponding stability time). These feature parameters need to be calculated and obtained through the time series prediction data output by the model to ensure that the parameters can accurately reflect the key features of the coupling curve.

[0087] A piecewise function is constructed based on the extracted feature parameters to describe the coupling relationship between soil deformation and tunnel stress. The piecewise function is divided into an excavation stage and a steady-state stage, and the formula is as follows:

[0088]

[0089] in For the dynamic coupling curve at time 10:00 The value of can represent soil displacement, strain, or tunnel stress. The start time of excavation. This marks the end of the excavation. This is the moment when deformation and stress reach stability; , , The fitting coefficients of the quadratic function during the excavation stage are obtained through the characteristic parameters of the excavation stage (such as...). Initial value at time, The values ​​at each time point and the peak value of the curve are determined by fitting. , , These are the fitting coefficients of the exponential function during the steady-state phase. The attenuation coefficient is determined by characteristic parameters of the steady-state phase (such as...). The values ​​at time points, the stability threshold, and the stable time are determined by fitting.

[0090] Data smoothing and curve output: The curves generated by the piecewise function are smoothed using moving average or Gaussian filtering to remove noise interference and make the curves continuous and smooth, ensuring that the curves accurately reflect the dynamic changes in soil deformation and tunnel stress. After processing, a continuous dynamic coupling curve of soil deformation and tunnel stress is output. The curves must include a time axis and the corresponding deformation / stress axis, clearly showing the changes in the coupling relationship between soil deformation and tunnel stress at different excavation stages, providing intuitive visualization results for engineering analysis.

[0091] This implementation accurately acquires engineering geological and tunnel structural parameters, constructs a geological tunnel coupling similarity model embedded with a layered sensor array, and combines step-by-step excavation simulation with multi-channel synchronous acquisition to achieve comprehensive real-time capture of soil deformation, tunnel stress, and excavation disturbance parameters. Utilizing a deformation-stress coupling prediction model with a dual-branch network architecture, an excavation disturbance quantification model using a gradient boosting tree algorithm, and a dynamic coupling response model incorporating time decay factors and gating mechanisms, it accurately quantifies the intensity of excavation disturbance and clearly presents the dynamic coupling relationship between soil layers and the tunnel at different stages. This breakthrough overcomes the limitations of traditional monitoring and prediction methods and provides reliable technical support for underground engineering construction safety assessment and parameter optimization.

[0092] Example 2

[0093] This embodiment describes in detail the real-time response test device for soil deformation and tunnel stress during foundation pit excavation, such as... Figure 4 As shown, it includes a model box, a data processing host, and a display terminal; the model box is equipped with a layered sensor array, a dynamic loading device, a servo excavation mechanism, and a multi-channel data acquisition instrument.

[0094] The interior of the model box is used to fill soil-like materials and lay out tunnel models to form a geological tunnel coupled similar model. The side walls of the model box are equipped with transparent observation windows, and the top is equipped with a loading interface adapted to the dynamic loading device.

[0095] The model box has a rectangular closed structure with sufficient rigidity and load-bearing capacity to withstand the maximum pressure applied by the dynamic loading device without deformation. Its external dimensions need to be designed according to the test scale and the scale of similar models. The internal part of the box is divided into a soil filling area, a tunnel model placement area, and an equipment installation area. The soil filling area is the core area inside the box, used to fill soil similar materials prepared according to the similar ratio. The filling process requires layered compaction, and the compaction degree must be consistent with the prototype soil layer. The tunnel model placement area is located in the middle of the soil filling area or at a preset position, which needs to be determined according to the tunnel burial depth parameters. The placement area needs to reserve space matching the size of the tunnel model. After the tunnel model is placed, the surrounding soil similar materials are filled and compacted to ensure that the tunnel model is in close contact with the soil layer. The equipment installation area is distributed on the top, bottom, and side walls of the box. It is used to install the loading end of the dynamic loading device, the fixing structure of the servo excavation mechanism, and the wiring terminals of the layered sensor array. The installation area is equipped with standardized installation interfaces with interface sizes adapted to the corresponding equipment, which facilitates equipment disassembly, assembly, and debugging.

[0096] Transparent observation windows are installed on one or two adjacent side walls of the test chamber, facilitating observation of soil deformation and tunnel model status via high-speed camera or naked eye during the experiment. The connection between the observation windows and the side walls of the test chamber is sealed with rubber sealing rings to ensure that similar materials to the soil inside the chamber do not leak from gaps.

[0097] The loading interfaces are divided into top loading interfaces, bottom loading interfaces, and side wall loading interfaces. The top loading interfaces are circular or square openings, evenly distributed on the top of the housing. The opening size is adapted to the output end of the hydraulic loader. The interface edge is equipped with a flange structure and is fixedly connected to the hydraulic loader by bolts. The bottom loading interfaces are symmetrical to the top loading interfaces and have the same number. They are located at the bottom of the housing and are used to apply vertical loading force from the bottom. The side wall loading interfaces are elongated openings with a guide structure to ensure that the output end of the hydraulic loader can apply horizontal loading force in a direction perpendicular to the side wall. The loading interfaces are equipped with dustproof sealing gaskets to prevent soil-like material particles from entering the interface and affecting the operation of the loading device.

[0098] The layered sensing array includes distributed optical fiber sensors and miniature piezoelectric sensors. The distributed optical fiber sensors are buried in similar materials in the soil layer according to the soil depth. The miniature piezoelectric sensors are pasted and fixed along the circumferential and axial directions of the tunnel model. The signal output terminals of all sensors are connected to the signal input terminals of the multi-channel data acquisition instrument through shielded cables.

[0099] The distributed optical fiber sensor is made of single-mode or multi-mode optical fiber. During deployment, it is buried in layers according to the soil depth, and the laying length is consistent with the length of the model box. The two ends of the optical fiber are led out to the junction box outside the box through optical fiber connectors to ensure compatibility with the optical fiber interface of the multi-channel data acquisition instrument. During the burial process, the optical fiber must be in close contact with similar materials to the soil layer to avoid gaps between the optical fiber and the soil layer, which would affect the accuracy of displacement and strain monitoring.

[0100] The miniature piezoelectric sensor has a patch structure and accurately monitors the stress and contact pressure on the surface of the tunnel model. The sensor is attached and fixed along the circumferential and axial directions of the tunnel model. The circumferential attachment is evenly distributed on the outer periphery of the tunnel model, and the distribution along the length of the tunnel model ensures that the sensor can cover the entire surface of the tunnel model.

[0101] The signal output terminals of the sensors are all connected to the multi-channel data acquisition instrument via shielded cables. The cables are arranged in the box along the pre-set cable grooves on the side wall of the box and distributed along the height of the box. Cable fixing clips are installed in the grooves to fix the cables and prevent them from being pulled and damaged during soil filling or excavation. A waterproof sealing connector is provided at the cable exit point of the box. The connector is made of brass and is threaded to the side wall of the box. A rubber sealing sleeve is installed inside the connector to ensure that the inside of the box remains sealed after the cable is exited.

[0102] The dynamic loading device includes a hydraulic loader and a pressure controller. The output end of the hydraulic loader is attached to the bottom and side wall of the model box through a loading interface. The pressure controller is connected to the hydraulic loader through a control line and is used to adjust the loading pressure to simulate the initial ground stress.

[0103] Hydraulic loaders are divided into vertical hydraulic loaders and horizontal hydraulic loaders. Vertical hydraulic loaders are used to apply vertical initial ground stress. The number of vertical loaders matches the number of top and bottom loading interfaces. The output end of the loader is equipped with a pressure sensor, which can monitor the output pressure in real time. The vertical loaders are connected to the loading interfaces at the top and bottom of the housing through flanges.

[0104] The horizontal hydraulic loader is used to apply the initial horizontal ground stress. The number of loading ports installed on the side wall of the box matches the number of side wall loading ports. The output end is also equipped with a pressure sensor with parameters consistent with the vertical loader.

[0105] The servo excavation mechanism includes a servo motor, a mechanical excavator arm, and a displacement sensor. The servo motor is fixed to the top of the model box by a bracket. The mechanical excavator arm is connected to the output shaft of the servo motor. The displacement sensor is installed on the mechanical excavator arm. Both the servo motor and the displacement sensor are connected to the data processing host via data cables. The data processing host adjusts the excavation thickness and speed of the mechanical excavator arm through control commands.

[0106] The mechanical excavator arm consists of a boom, a bucket, and connecting joints. It can swing up and down and left and right through the joint structure. The maximum digging depth of the excavator arm is adapted to the height of the model box, and the maximum digging speed can be adjusted by the servo motor speed.

[0107] Both the servo motor and the displacement sensor are connected to the data processing host via data cables. The data processing host sends control commands to the servo motor, which include parameters such as the target excavation position and excavation speed. The servo motor drives the mechanical excavator arm to move according to the commands. The displacement sensor collects the displacement data of the excavator arm in real time and transmits it to the data processing host. The host performs closed-loop control of the servo motor by comparing the actual displacement with the target displacement.

[0108] The signal output terminal of the multi-channel data acquisition instrument is connected to the data processing host via a data cable, and is used to transmit the acquired soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters to the data processing host.

[0109] The multi-channel data acquisition instrument is equipped with an operation panel featuring a power switch, a reset button, and status indicator lights. These lights display the acquisition status of each channel (normal, fault, not connected), allowing operators to monitor the equipment's operation in real time. Furthermore, the instrument supports remote control, enabling the setting of acquisition parameters (sampling frequency, acquisition duration, channel gain) remotely via the data processing host, eliminating the need for on-site operation and improving experimental convenience.

[0110] The data processing host has a built-in deformation-force coupling prediction model, excavation disturbance quantification model and dynamic coupling response model. The signal output terminal of the data processing host is connected to the display terminal to output the dynamic coupling curve and real-time monitoring data to the display terminal.

[0111] Furthermore, such as Figure 5 As shown, the working process of the real-time response test device for soil deformation and tunnel stress during the foundation pit excavation is as follows:

[0112] Test preparation phase: The data processing host sends instructions to the pressure controller through the equipment control module. The pressure controller adjusts the hydraulic loader to apply initial ground stress to the model box. At the same time, the displacement sensor monitors the soil deformation during loading. The data is transmitted to the host through the multi-channel data acquisition instrument. After the host confirms that the initial ground stress has reached the preset value, loading stops. The servo excavation mechanism is reset to the initial position under the control of the host, with the bucket above the soil surface. The layered sensor array completes preheating and calibration to ensure the accuracy of the monitoring data.

[0113] Excavation and monitoring phase: The host sends excavation commands to the servo motor, which drives the mechanical excavator arm to perform layered and segmented excavation according to the preset excavation thickness and speed. Displacement sensors collect excavator arm displacement data in real time and feed it back to the host to achieve closed-loop control. During the excavation process, the layered sensor array continuously monitors soil layer displacement and strain, tunnel stress and contact pressure. The multi-channel data acquisition instrument collects these data synchronously and transmits them to the host. The host's data processing module runs three built-in models in real time to calculate multi-dimensional deformation and stress prediction vectors, excavation disturbance intensity coefficients, and dynamic coupling curve parameters.

[0114] Results Output Stage: The host computer transmits the dynamic coupling curve and real-time monitoring data (displacement, strain, stress, disturbance parameters) to the display terminal, which displays this information in real time. The test personnel can observe the test process through the display terminal. If abnormal data is found, the host computer can send instructions to adjust the excavation parameters or stop the test. After the test, the host computer automatically stores all test data, which can be exported to an external storage device for subsequent analysis and report writing.

[0115] This embodiment details the integrated testing device with dynamic loading, servo excavation, multi-channel acquisition, and intelligent analysis functions, enabling integrated operation from similar model construction to data processing and result display. Dynamic loading accurately simulates initial ground stress, servo excavation precisely controls layered and segmented excavation, and multi-model real-time data analysis improves testing efficiency and accuracy, providing reliable support for underground engineering safety risk assessment and construction optimization.

[0116] Based on Embodiment 1 or 2, this embodiment details the implementation and verification of the real-time response test method for soil deformation and tunnel stress during foundation pit excavation. The technical effectiveness of the method is quantitatively verified using a 1:20 scale similar model test. Taking a deep foundation pit project adjacent to a subway tunnel as a prototype, 120 sets of working condition samples covering different excavation speeds, excavation thicknesses, and geological conditions are selected. All data are collected through the test device to ensure the reliability and generalization ability of the verification results. The core verification indicators focus on the fitting accuracy of the dynamic coupling curve, the accuracy of deformation and stress prediction, and the effectiveness of excavation disturbance quantification. The method is benchmarked against the current mainstream multiphysics numerical simulation method and distributed sensing and static coupling model method in the industry.

[0117] The experimental dataset is divided into three working conditions based on the combination of geological conditions and excavation intensity. Each working condition has 40 samples, covering three typical soil layers: silty clay, sandy silt, and clayey silt, with different combinations of excavation speed of 1-5 mm / s and excavation thickness of 5-15 cm. The comparison results of various core indicators are shown in the table below. The dynamic coupling curve fitting accuracy (curve similarity) of the method in this application reaches 96.8%, with a peak error of only 2.1% and a stable stage error of 1.5%. This is 17.5 percentage points higher than the mainstream multiphysics numerical simulation method (curve similarity 82.4%, peak error 8.7%) and 9.2 percentage points higher than the distributed sensing + static coupling model method (curve similarity 88.6%, peak error 5.3%). In terms of deformation and stress prediction, the average absolute error of the method in this application is only 0.12 mm (displacement) and 0.03 MPa (stress), with an average relative error of 3.2%. The average relative error of the multiphysics numerical simulation method is 11.5%, and that of the distributed sensing + static coupling model method is 7.8%. The relative error of the excavation disturbance quantification in this application is only 3.8%, which is much lower than the 16.3% of the multiphysics numerical simulation method and the 9.4% of the distributed sensing + static coupling model method. All core indicators are at the leading level in the industry.

[0118]

[0119] The table above clearly shows the differences in core performance indicators between our method and two mainstream methods in the industry. The fitting accuracy of the dynamic coupling curve is our core advantage; the curve similarity is 9.2 percentage points higher than the closest mainstream method, and the peak error is reduced by 60.4%. This difference stems from our innovative deep integration of the dynamic coupling response model with real-time monitoring data and excavation disturbance quantification results, rather than the limitations of traditional methods that rely on static models or pure numerical simulations. The advantage in error reduction for deformation stress prediction and excavation disturbance quantification is due to the collaborative design of the dual models (deformation stress coupling prediction model and excavation disturbance quantification model), especially the application of cross-branch attention mechanism and gradient boosting tree algorithm. This makes the prediction and quantification process more closely aligned with the dynamic changes in engineering practice, avoiding the problem of model decoupling from actual working conditions found in mainstream methods.

[0120] like Figure 6As shown in the figure, the black solid line is the measured curve of the layered sensing array, the red solid line is the dynamic coupling curve output by the method of this application, the blue dashed line is the output curve of the distributed sensing + static coupling model method, and the green dotted line is the output curve of the multiphysics numerical simulation method. The measured curve shows a displacement peak of 7.1 mm and a stress peak of 1.7 MPa at 180 min (end of the 3rd layer excavation). After the excavation is completed (360 min), it stabilizes at a displacement of 4.8 mm and a stress of 1.2 MPa. The output curve of this application almost completely overlaps with the measured curve, not only accurately capturing the peak position and value, but also completely restoring the dynamic process of "rapid growth during the excavation stage - slow adjustment during the excavation interval - and gradual flattening during the stable stage". Especially during the excavation interval, it can capture the hysteresis response of the soil layer and the tunnel. In contrast, the distributed sensing + static coupling model method lacks a dynamic adjustment mechanism, and the curve is flat during the interval, with the deviation from the measured value gradually increasing. The multiphysics numerical simulation method is limited by parameter assumptions, with the peak position lagging by 15 minutes and the numerical deviation reaching 8.7%. The data in the stable stage differs significantly from the measured value.

[0121] like Figure 7 As shown, the figure includes dynamic coupling curves for three working conditions: silty clay, sandy silt, and clayey silt, along with corresponding peak values, rates of change, and stability thresholds. Under the silty clay condition, the peak displacement is 5.8 mm, the peak stress is 1.4 MPa, the rate of change is gradual (0.023 mm / min, 0.006 MPa / min), and the stability threshold is 4.1 mm / 1.0 MPa. Under the sandy silt condition, the peak displacement is 7.1 mm, the peak stress is 1.7 MPa, the rate of change increases to 0.038 mm / min and 0.009 MPa / min, and the stability threshold is 4.8 mm / 1.2 MPa. Under the clayey silt condition, the peak displacement is 6.5 mm, the peak stress is 1.5 MPa, the rate of change is between the two (0.031 mm / min, 0.007 MPa / min), and the stability threshold is 4.5 mm / 1.1 MPa. The three curves clearly quantify the influence of different soil layer physical and mechanical properties on the coupled response. In particular, sandy silt, due to its high permeability and low shear strength, has larger deformation and stress peaks and faster change rates. The method of this application can accurately adapt to the response differences of different soil layers and output continuous and experimentally accurate dynamic coupling curves. This capability is difficult to achieve by mainstream methods. The similarity of curves in multiphysics numerical simulation methods under different soil layer conditions fluctuates by more than 10%, while the fluctuation of the method of this application is only 2.3%, which fully verifies its stability and adaptability under complex geological conditions.

[0122] This embodiment fully verifies that through the all-round synchronous monitoring of the hierarchical sensor array, the accurate prediction and quantification of the dual models, and the deep integration of the dynamic coupling response model, it comprehensively surpasses the mainstream methods in the industry in terms of core performance indicators. In particular, the core output dynamic coupling curve can accurately and continuously capture the coupling response law of soil deformation and tunnel stress under different geological conditions and different excavation intensities. It provides a visualized and quantitative technical basis for safety risk assessment in the engineering construction process, and fully verifies the technical advancement and practicality of the method in engineering applications.

[0123] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A real-time test method for soil deformation and tunnel stress response during foundation pit excavation, characterized in that, include: Obtain engineering geological parameters and tunnel structural parameters. The engineering geological parameters include soil physical and mechanical properties, groundwater distribution, and initial in-situ stress. The tunnel structural parameters include cross-sectional dimensions, material strength, and burial depth. Based on the engineering geological parameters and tunnel structural parameters, a geological tunnel coupled similarity model is constructed. The geological tunnel coupled similarity model is embedded with a layered sensor array and a dynamic loading interface. The layered sensor array is arranged according to the soil layer depth and the tunnel circumferential axis. The excavation of the foundation pit was simulated by a step-by-step excavation method. The excavation was carried out in layers and sections according to the preset excavation thickness and speed. Real-time monitoring data of the layered sensor array was acquired through multi-channel synchronous acquisition. The real-time monitoring data included soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters. A first model and a second model are constructed. The first model is a deformation-stress coupling prediction model, which outputs a multi-dimensional deformation-stress prediction vector. The second model is an excavation disturbance quantification model, which outputs the excavation disturbance intensity coefficient. Based on the multi-dimensional deformation and stress prediction vector and the excavation disturbance intensity coefficient, a dynamic coupled response model is constructed to output the dynamic coupling curve of soil deformation and tunnel stress during the excavation process.

2. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1, characterized in that, The construction of the coupled similarity model of the geological tunnel includes: Based on similarity theory, the geometric similarity ratio, material similarity ratio and stress similarity ratio are determined. Soil layer similarity material is prepared according to a preset ratio. The soil layer similarity material is made by mixing prototype soil sand particles and binder according to a preset ratio. The compaction degree is controlled by compaction test. The tunnel model is scaled down proportionally according to the tunnel structural parameters, and reserved holes for installing the sensor array are provided. The positions of the holes correspond to the positions of the layered sensor arrays. Initial ground stress is applied through a dynamic loading interface to simulate a natural stress field. The loading process adopts a graded loading method, and the next level of loading is carried out only after a preset stabilization time is reached after each level of loading.

3. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1, characterized in that, The first model outputs a multi-dimensional deformation and stress prediction vector, which converts the soil displacement and strain and tunnel stress and contact pressure in the real-time monitoring data into a standardized feature vector. The first model is input for coupling calculation, and the first model outputs a multi-dimensional vector containing predicted values ​​of vertical soil deformation, horizontal displacement, tunnel circumferential stress, and axial strain. The formula is as follows: ,in, This is a multi-dimensional deformation force prediction vector. To standardize the feature vector, , These are the model weight matrices, , These are the bias vectors, Activation function It is the hyperbolic tangent activation function.

4. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1 or 3, characterized in that, The first model deformation-stress coupling prediction model adopts a dual-branch network architecture, with the first branch processing soil deformation-related features and the second branch processing tunnel stress-related features. The attention weight formula is as follows: A cross-branch attention mechanism is used to capture the coupled feature relationship between two types of features. ,in, For the first The characteristics of the soil layer and the first Coupling weights for each tunnel feature, This is a query vector for soil layer features. The tunnel feature key vector. This represents the total number of tunnel features. By fusing coupling features and branching features, a multi-dimensional deformation force prediction vector is output through a fully connected layer.

5. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1, characterized in that, The second model outputs the excavation disturbance intensity coefficient and extracts the excavation disturbance parameters from the real-time monitoring data, including excavation speed, excavation thickness, and disturbance duration. The second model is used to quantify the disturbance parameters and output the excavation disturbance intensity coefficient: ,in, This is the excavation disturbance intensity coefficient. For a moment The excavation speed For a moment The excavation thickness, For a moment The depth of the disturbance's impact. For tunnel burial depth, The attenuation coefficient is... , These represent the start and end times of the excavation period.

6. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1 or 5, characterized in that, The second model, the excavation disturbance quantification model, is based on historical excavation test data, establishing a sample database of mapping relationships between disturbance parameters and disturbance intensity. The model is trained using the gradient boosting tree algorithm, and the model's loss function formula is as follows: ,in, This represents the model loss value. For the first The actual perturbation intensity of each sample For predicted values, The regularization coefficient is . For the first The complexity of a decision tree, The total number of samples, The number of decision trees.

7. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1, characterized in that, The construction of the dynamic coupled response model involves feature fusion of the multi-dimensional deformation force prediction vector and the excavation disturbance intensity coefficient, as shown in the formula: ,in, To fuse feature vectors, This is a multi-dimensional deformation force prediction vector. The excavation disturbance intensity coefficient is used to construct a dynamic coupling response model based on a long short-term memory network. The input is a fused feature vector and a time series of real-time monitoring data, and the output is the feature parameters of the dynamic coupling curve.

8. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 7, characterized in that, The constructed dynamic coupling response model, by introducing a time decay factor, corrects the coupling weights at different excavation stages. The time decay factor is: ,in, For a moment Time decay factor, The decay rate coefficient, The excavation start time is used; a gating mechanism controls the fusion ratio of historical and current data, and the calculation of update and reset gates satisfies the following: ,in, To update the door, To reset the door, For a moment The fused feature vector, This is the hidden state from the previous moment. , These are the weight matrices, , These are the bias vectors.

9. The real-time response test method for soil deformation and tunnel stress during foundation pit excavation as described in claim 1, characterized in that, The output of the dynamic coupling curve includes: feature parameters based on the output of the dynamic coupling response model, including curve peak value, rate of change, and stability threshold; constructing a piecewise function to describe the coupling relationship; and converting the feature parameters into a continuous curve according to the time series, with the curve equation as follows: ； in, For the dynamic coupling curve at time 10:00 The value of , , , , , These are the curve fitting coefficients, The attenuation coefficient is... This refers to the time interval during the excavation phase. For the stable phase time interval, through data smoothing, a continuous dynamic coupling curve of soil deformation and tunnel stress is output.

10. A real-time test device for soil deformation and tunnel stress during foundation pit excavation, characterized in that, It includes a model box, a data processing host, and a display terminal; the model box is equipped with a layered sensor array, a dynamic loading device, a servo excavation mechanism, and a multi-channel data acquisition instrument. The model box is filled with soil-like materials and a tunnel model is laid out inside to form a geological tunnel coupling similar model. The side walls of the model box are equipped with transparent observation windows, and the top is equipped with a loading interface adapted to the dynamic loading device. The layered sensing array includes distributed optical fiber sensors and miniature piezoelectric sensors. The distributed optical fiber sensors are buried in similar materials in the soil layer according to the soil layer depth. The miniature piezoelectric sensors are pasted and fixed along the circumferential and axial directions of the tunnel model. The signal output terminals of all sensors are connected to the signal input terminals of the multi-channel data acquisition instrument through shielded cables. The dynamic loading device includes a hydraulic loader and a pressure controller. The output end of the hydraulic loader is attached to the bottom and side wall of the model box through a loading interface. The pressure controller is connected to the hydraulic loader through a control line and is used to adjust the loading pressure to simulate the initial ground stress. The servo excavation mechanism includes a servo motor, a mechanical excavator arm, and a displacement sensor. The servo motor is fixed to the top of the model box by a bracket. The mechanical excavator arm is connected to the output shaft of the servo motor. The displacement sensor is installed on the mechanical excavator arm. Both the servo motor and the displacement sensor are connected to the data processing host via data cables. The data processing host adjusts the excavation thickness and speed of the mechanical excavator arm through control commands. The signal output terminal of the multi-channel data acquisition instrument is connected to the data processing host via a data cable, and is used to transmit the acquired soil displacement and strain, tunnel stress contact pressure and excavation disturbance parameters to the data processing host. The data processing host has a built-in deformation-force coupling prediction model, excavation disturbance quantification model and dynamic coupling response model. The signal output terminal of the data processing host is connected to the display terminal to output the dynamic coupling curve and real-time monitoring data to the display terminal.

Images