Model experiment tunnel longitudinal and transverse full-section deformation monitoring device

Abstract

A tunnel longitudinal and transverse full-section deformation monitoring device in a model experiment comprises a model box, a horizontal positioning support, a vertical adjustable positioning support, a longitudinal transmission mechanism, a displacement meter rotating mechanism, a speed controller and a data processing module, and an electric push rod of the longitudinal transmission mechanism controls a laser displacement meter of the displacement meter rotating mechanism to translate along the axis of a tunnel. A hollow shaft stepping motor of a displacement meter rotating mechanism is used for controlling a laser displacement meter to rotate in the circumferential direction of a tunnel, the laser displacement meter covers measurement of any point on the inner side of the tunnel, and therefore all tunnel deformation response data are obtained, intelligent data noise reduction filtering processing is conducted on the basis of a novel unsupervised generative adversarial network (GAN) model architecture, and the tunnel deformation response accuracy is improved. And the data quality and the deformation mode capturing capability are improved. The method is suitable for exploring the longitudinal and transverse global deformation conditions of the tunnel structure under various disturbance effects, can be flexibly adjusted according to the burial depth, size, horizontal position, monitoring density and the like of the tunnel required by the test, and has wide applicability.

Description

Technical Field

[0001] The present application belongs to the technical field of tunnel physics similarity model experiments. Background Art

[0002] The monitoring method of tunnel deformation in tunnel engineering is an important means of evaluating the safety and health of tunnel structures, and has attracted extensive attention from researchers and engineers. Tunnel deformation is mainly divided into circumferential and longitudinal deformations: in the circumferential direction, the tunnel ring has a complex deformation pattern; in the longitudinal direction, the tunnel may also produce complex uneven settlements under various disturbances, and mechanical behaviors such as misalignment and opening may also occur between the rings. Capturing the longitudinal and lateral deformations of the tunnel is of great value to research and engineering practice.

[0003] With the rapid development of underground space in my country, various new tunnels and existing tunnels are intertwined, and tunnel engineering has the characteristic of poor visibility. Therefore, in order to have a deeper understanding of the influence mechanism of mechanical action on tunnels, how to obtain better quality and more comprehensive deformation monitoring data in physical similarity model tests is one of the problems that need to be solved urgently. The existing monitoring equipment of physical similarity model tests can usually only simply measure a small number of measuring points on a few rings, and cannot monitor the deformation data at each angle on each ring. For this reason, a large number of traditional research methods currently use numerical simulation or analytical solutions to obtain the continuous deformation of the entire section of the tunnel, most of which are based on finite element analysis or finite difference analysis and elastic foundation beam method. However, the calculation accuracy of these methods cannot be well verified.

[0004] At present, physical similarity model test monitoring usually uses LVDT or dial indicator for static monitoring, which provides single-point deformation data and cannot effectively capture the complex multi-dimensional deformation mode of the tunnel, especially the dynamic deformation of the entire longitudinal and transverse section. In addition, the physical similarity model test space is limited, and how to achieve dynamic monitoring of the entire tunnel section in a limited space is also a major challenge. The existing methods of deploying monitoring inside the tunnel are severely restricted by space, which may cause certain errors in the monitoring data. In summary, there is no better way in the current tunnel model test monitoring method to meet the dynamic monitoring needs of the longitudinal and transverse deformation of the entire tunnel section in a small space. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present application provides a model experimental tunnel longitudinal and transverse full-section deformation monitoring device, which controls the laser displacement meter to move horizontally along the tunnel axis at a specified speed through an electric push rod, and controls the laser displacement meter to rotate along the tunnel circumference at a specified speed through a hollow shaft stepper motor, so that the laser displacement meter covers the measurement of any point inside the tunnel, thereby obtaining all the tunnel deformation response data. In addition, the present application performs intelligent data noise reduction and filtering based on a new generative adversarial network (GAN) model architecture to improve data quality and deformation pattern capture capabilities.

[0006] The technical solution of this application is as follows:

[0007] The model test tunnel longitudinal and transverse full-section deformation monitoring device comprises a model box 16, a horizontal positioning bracket, a vertical adjustable positioning bracket, a longitudinal transmission mechanism, a displacement meter rotation mechanism, a speed controller 15 and a data processing module, wherein:

[0008] The model box 16 is arranged on a horizontal positioning bracket and is used to fill soil and simulate the tunnel space;

[0009] There are multiple vertically adjustable positioning brackets, which are arranged on the horizontal positioning bracket. The horizontal positioning bracket and the multiple vertically adjustable positioning brackets together constitute a bracket system for supporting the longitudinal transmission mechanism and the displacement meter rotation mechanism;

[0010] The longitudinal transmission mechanism is supported on a vertically adjustable positioning bracket, and the displacement meter rotation mechanism is arranged on the longitudinal transmission mechanism, and is used to obtain longitudinal and transverse deformation monitoring data of the entire cross section in the tunnel space;

[0011] The speed controller is connected to the longitudinal transmission mechanism and the displacement meter rotation mechanism, and is used to control the translation speed of the longitudinal transmission mechanism and the rotation speed of the displacement meter rotation mechanism;

[0012] The data processing module is used to process the full-section longitudinal and transverse deformation monitoring data.

[0013] Furthermore, the horizontal positioning bracket includes two brackets 1 and a base 2 connected between the two brackets 1; the model box 16 is arranged on the base 2, and the two brackets 1 are respectively located at two ends of the model box 16;

[0014] The vertically adjustable positioning device includes a push rod positioning slot 3, a vertical bracket 4, and also includes a vertical positioning adjustment bolt 5, an anchor bolt 6, and a positioning bolt 7; a bolt hole is provided on the upper part of the bracket 1 of the horizontal positioning bracket, and the bottom of the vertical bracket 4 is connected to the bracket 1 of the horizontal positioning bracket by the anchor bolt 6 to achieve fixation; a threaded hole is provided on the top of the vertical bracket 4, and the vertical positioning adjustment bolt 5 is threadedly connected to the top of the vertical bracket 4, and the lower part of the vertical positioning adjustment bolt 5 is connected to the push rod positioning slot 3 by welding, and the push rod positioning slot 3 is driven to move vertically by rotating the adjustment bolt 5, thereby changing the vertical position of the push rod positioning slot 3; at the same time, the push rod positioning slot 3 and the vertical bracket 4 are fixed together by the positioning bolt 7 to limit the jitter of the overall system during the measurement process.

[0015] Furthermore, the longitudinal transmission mechanism includes a DC electric push rod 13 and an extension rod 14, wherein the extension rod 14 is connected to the DC electric push rod 13, and the extension rod 14 performs linear reciprocating motion under the action of the DC electric push rod 13; the DC electric push rod 13 and the extension rod 14 are arranged in the push rod positioning groove 3 of the vertically adjustable positioning device;

[0016] The displacement meter rotating mechanism is fixed on the extension rod 14 of the longitudinal transmission mechanism, and includes a slewing bearing 8, a hollow shaft stepper motor 9, a rotating mechanism positioner 10, a displacement meter positioning disk 11, and a laser displacement meter 12; wherein, the hollow shaft stepper motor 9 is arranged on the extension rod 14 of the longitudinal transmission mechanism; the laser displacement meter 12 is fixed on the displacement meter positioning disk 11, and the laser measurement direction is consistent with the radial direction of the positioning meter positioning disk; the displacement meter positioning disk 11 is arranged on the extension rod 14 of the longitudinal transmission mechanism through the slewing bearing 8, and the other end of the displacement meter positioning disk 11 is connected to the hollow shaft stepper motor 9, and the hollow shaft stepper motor 9 drives the displacement meter positioning disk 11 to rotate and then drives the laser displacement meter 12 to rotate; the rotating mechanism positioner 10 has two, which are used to fix the slewing bearing 8 and the hollow shaft stepper motor 9 on the extension rod 14 of the longitudinal transmission mechanism.

[0017] Furthermore, the DC electric push rod 13 is arranged in the push rod positioning groove 3 of the two vertically adjustable positioning devices at the left end of the model box 16, and the extension rod 14 is arranged in the push rod positioning groove 3 of the two vertically adjustable positioning devices at the right end of the model box 16, and gaskets are placed in the push rod positioning grooves to increase friction so as to control jitter and rotation during the measurement process.

[0018] Furthermore, the longitudinal transmission mechanism: its DC electric push rod 13 integrates a DC motor, a reducer, a screw, and a guide rail component. The DC motor uses a permanent magnet DC motor to generate electromagnetic force to drive the rotor to rotate. The reducer is used to reduce the motor rotation speed to prevent a large disturbance to the monitoring system. The screw is used to convert the rotational motion of the DC motor into linear motion. The guide rail is used to maintain the direction of the screw and keep its movement stable; the extension rod 14 is connected to the screw of the DC electric push rod 13.

[0019] Furthermore, the size, layout position and measuring range of the laser displacement meter 12 are selected according to the internal space of the tunnel model.

[0020] Furthermore, the speed controller 15 integrates multiple components, including a power supply, a DC motor speed regulator, a pulse generator, and a stepper motor driver; the power supply is used to supply power to other components; the DC motor speed regulator is used to adjust the input current of the DC electric push rod 13 of the longitudinal transmission mechanism to control the propulsion speed of the DC electric push rod 13; the stepper motor driver receives and analyzes the pulse signal of the pulse generator to operate the hollow shaft stepper motor 9 of the displacement meter rotating mechanism to achieve precise control of the rotation speed of the displacement meter rotating mechanism.

[0021] Furthermore, the data processing module includes data preprocessing and post-processing, wherein: the data preprocessing includes data cleaning and normalization; the post-processing processes the measurement data of the laser displacement meter based on a new generative adversarial network (GAN) model architecture based on artificial intelligence, filters out outliers and oscillations in the data, and retains the core deformation features in the originally highly complex and irregular data.

[0022] The above-mentioned model test tunnel longitudinal and transverse full-section deformation monitoring device is applied to the relevant model tests of tunnel transverse and longitudinal structure stress deformation and its influence analysis, and its implementation steps include:

[0023] Step 1: Lay out the device of the present application, confirm that the initial state is correct, control the hollow shaft stepper motor 9 and the DC electric push rod 13 through the speed controller 15, realize the precise control of the translation and rotation speed of the laser displacement meter 12, and collect the monitoring data of the laser displacement meter as the benchmark data;

[0024] Step 2: Simulate tunnel deformation, and during the tunnel deformation process, control the hollow shaft stepper motor 9 and the DC electric push rod 13 through the speed controller 15 to achieve precise control of the translation and rotation speeds of the laser displacement meter 12, dynamically collect monitoring data, and compare them with the benchmark data obtained in step 1, so as to obtain the longitudinal and transverse deformation data of the entire section of the tunnel structure;

[0025] Step 3: Preprocess the full-section longitudinal and transverse deformation monitoring data obtained in step 2, and divide the preprocessed data into a training set and a test set;

[0026] Step 4: In order to post-process the noisy data, design and initialize the GAN model architecture, wherein the GAN model includes a generator and a discriminator;

[0027] Step 5: Iteratively update the generator and discriminator to implement cyclic training of the GAN model and save the trained bidirectional optimal model;

[0028] Step 6: Based on the bidirectional optimal model trained in step 5, the noisy raw data is input into the generator to generate high-quality data for analysis and prediction of the tunnel deformation state.

[0029] Compared with the prior art, this application has the following beneficial effects:

[0030] The present application provides a physical similarity model test and measurement device that can simultaneously measure the deformation of the entire longitudinal and transverse sections of a tunnel, providing an alternative solution for existing structural response data measurement. The present application fills the gap in the current experimental research that the monitoring data is relatively discrete, resulting in insufficient grasp of the overall deformation mode of long linear structures; the new generative adversarial network (GAN) model architecture based on artificial intelligence realizes intelligent data noise reduction and filtering processing, and proposes to integrate the idea of ​​two-dimensional image denoising into one-dimensional sequences. At the same time, through the design of the discriminator, the conventional need for known "clean data" labels in generative model training is abandoned, and unsupervised self-learning is realized to process the monitored original oscillation data into high-quality data with both natural and smooth characteristics. The present application can be used for relevant model tests on the force deformation and influence analysis of the transverse and longitudinal structures of tunnels under any circumstances. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the overall device of this application.

[0032] Figure 2 Schematic diagram of the structure of the horizontal positioning bracket and the vertical adjustable positioning bracket of the device of the present application.

[0033] Figure 3 A partially enlarged schematic diagram of the details of the vertically adjustable positioning bracket of the device of the present application.

[0034] Figure 4 Schematic diagram of the structure of the displacement meter rotation mechanism and longitudinal transmission mechanism of the device of the present application.

[0035] Figure 5 A partially enlarged schematic diagram of the details of the displacement meter rotation mechanism of the device of the present application.

[0036] Figure 6 Schematic diagram of the actual application flow chart of the device of this application.

[0037] Figure 7 Schematic diagram of different working modes of the device of this application.

[0038] Figure 8 Schematic diagram of the GAN model used in this application for data post-processing.

[0039] Reference numerals:

[0040] 1- bracket, 2- base, 3- push rod positioning slot, 4- vertical bracket, 5- vertical positioning adjustment bolt, 6- anchor bolt, 7- positioning bolt, 8- slewing bearing, 9- hollow shaft stepper motor, 10- rotating mechanism positioner, 11- displacement meter positioning plate, 12- laser displacement meter, 13- DC electric push rod, 14- extension rod, 15- control box, 16- model box. DETAILED DESCRIPTION

[0041] The technical solution provided by the present application will be further described below in conjunction with specific embodiments and accompanying drawings. The advantages and features of the present application will become more apparent with the following description.

[0042] like Figure 1 As shown, a model test tunnel longitudinal and transverse full-section deformation monitoring device is used for monitoring the longitudinal and transverse full-section continuous deformation of long linear structures such as tunnels or pipelines in the physical model test in the field of geotechnical engineering, including a model box 16, a horizontal positioning bracket, a vertically adjustable positioning bracket, a displacement meter rotation mechanism, a longitudinal transmission mechanism, a speed controller 15 and a data processing module, wherein:

[0043] The model box 16 is arranged on a horizontal positioning bracket and is used to fill soil and simulate the tunnel space;

[0044] There are multiple vertically adjustable positioning brackets, which are arranged on the horizontal positioning bracket. The horizontal positioning bracket and the multiple vertically adjustable positioning brackets together constitute a bracket system for supporting the longitudinal transmission mechanism and the displacement meter rotation mechanism;

[0045] The longitudinal transmission mechanism is supported on a vertically adjustable positioning bracket, and the displacement meter rotation mechanism is arranged on the longitudinal transmission mechanism, and is used to obtain longitudinal and transverse deformation monitoring data of the entire cross section in the tunnel space;

[0046] The speed controller is connected to the longitudinal transmission mechanism and the displacement meter rotation mechanism, and is used to control the translation speed of the longitudinal transmission mechanism and the rotation speed of the displacement meter rotation mechanism;

[0047] The data processing module is used to process the full-section longitudinal and transverse deformation monitoring data.

[0048] like Figure 1 , Figure 2 As shown, the horizontal positioning bracket includes two brackets 1 and a base 2 connected between the two brackets 1; a model box 16 is arranged on the base 2; and the two brackets 1 are at both ends of the model box 16.

[0049] Preferably, the two brackets 1 are fixed on both sides of the base 2 by welding.

[0050] Furthermore, the base 2 is provided with lateral constraints for limiting the horizontal displacement of the model box 16 .

[0051] like Figure 2 , Figure 3 As shown, the vertically adjustable positioning device includes a push rod positioning slot 3, a vertical bracket 4, and also includes a vertical positioning adjustment bolt 5, an anchor bolt 6, and a positioning bolt 7; a bolt hole is provided on the upper part of the bracket 1 of the horizontal positioning bracket, and the bottom of the vertical bracket 4 is connected to the bracket 1 of the horizontal positioning bracket by the anchor bolt 6 to achieve fixation; a threaded hole is provided on the top of the vertical bracket 4, and the vertical positioning adjustment bolt 5 is threadedly connected to the top of the vertical bracket 4, and the lower part of the vertical positioning adjustment bolt 5 is connected to the push rod positioning slot 3 by welding, and the push rod positioning slot 3 can be driven to move vertically by rotating the adjustment bolt 5, thereby changing the vertical position of the push rod positioning slot 3; at the same time, the push rod positioning slot 3 and the vertical bracket 4 can be fixed together by the positioning bolt 7 to limit the jitter of the overall system during the measurement process.

[0052] In the embodiment, there are four vertically adjustable positioning devices, two of which are arranged on the bracket 1 of the horizontal positioning bracket at the left end of the model box 16, and another two vertically adjustable positioning devices are arranged on the bracket 1 of the horizontal positioning bracket at the right end of the model box 16. The four vertically adjustable positioning devices can be leveled by a horizontal laser line measurement, and the height of the monitoring line can be adjusted by bolts.

[0053] like Figure 4 As shown, the longitudinal transmission mechanism includes a DC electric push rod 13 and an extension rod 14. The extension rod 14 is connected to the DC electric push rod 13. The extension rod 14 performs linear reciprocating motion under the action of the DC electric push rod 13. The DC electric push rod 13 and the extension rod 14 are arranged in the push rod positioning groove 3 of the vertically adjustable positioning device.

[0054] Specifically, the DC electric push rod 13 integrates components such as a DC motor, a reducer, a screw, and a guide rail. The DC motor uses a permanent magnet DC motor to generate electromagnetic force to drive the rotor to rotate. The reducer can reduce the motor rotation speed to prevent a large disturbance to the monitoring system. The screw can convert the rotational motion of the DC motor into linear motion. The guide rail is used to maintain the direction of the screw and keep its movement stable. The extension rod 14 is connected to the screw of the DC electric push rod 13 by bolts.

[0055] Specifically, the DC electric push rod 13 is arranged in the push rod positioning slots 3 of the two vertically adjustable positioning devices at the left end of the model box 16, and the extension rod 14 is arranged in the push rod positioning slots 3 of the two vertically adjustable positioning devices at the right end of the model box 16. The vertical movement of the push rod positioning slot 3 can be driven by rotating the vertical positioning adjustment bolt 5 as needed to achieve the measurement of tunnel deformation at different buried depths. The push rod positioning slot 3 of the vertically adjustable positioning device needs to be preset with positioning slot sizes according to the electric push rod 13 and the extension rod 14 of the longitudinal transmission mechanism, and a gasket is placed in the push rod positioning slot to increase friction, so as to control the jitter and rotation during the measurement process.

[0056] like Figure 4 , Figure 5 As shown, the displacement meter rotating mechanism is fixed on the extension rod 14 of the longitudinal transmission mechanism, including a slewing bearing 8, a hollow shaft stepping motor 9, a rotating mechanism positioner 10, a displacement meter positioning plate 11, and a laser displacement meter 12; wherein the hollow shaft stepping motor 9 is arranged on the extension rod 14 of the longitudinal transmission mechanism; the laser displacement meter 12 is fixed on the displacement meter positioning plate 11, and the laser measurement direction is consistent with the radial direction of the positioning plate of the positioning meter; the displacement meter positioning plate 11 is arranged on the extension rod 14 of the longitudinal transmission mechanism through the slewing bearing 8 , and the other end of the displacement meter positioning plate 11 is connected to the hollow shaft stepper motor 9, which drives the displacement meter positioning plate 11 to rotate and then drives the laser displacement meter 12 to rotate. At the same time, the displacement meter positioning plate 11 is a part of the displacement meter rotation mechanism, and its translational freedom is fixed on the longitudinal transmission mechanism, and moves longitudinally under the drive of the DC electric push rod 13 of the longitudinal transmission mechanism; there are two rotation mechanism positioners 10, which are used to fix the slewing bearing 8 and the hollow shaft stepper motor 9 on the extension rod 14 of the longitudinal transmission mechanism.

[0057] Specifically, the laser displacement meter 12 is fixed to the displacement meter positioning plate 11 by bolts or adhesive, the inner ring of the slewing bearing 8 is connected to the displacement meter positioning plate 11 by bolts, the other end of the displacement meter positioning plate 11 is bolted to the hollow shaft stepper motor 9, the rotating mechanism positioner 10 farther away from the DC electric push rod 13 is bolted to the hollow shaft stepper motor 9, and the rotating mechanism positioner 10 closer to the DC electric push rod 13 is bolted to the outer ring of the slewing bearing 8, thereby achieving the fixation of the overall horizontal degree of freedom of the displacement meter rotating mechanism.

[0058] Furthermore, the size, layout position and measuring range of the laser displacement meter 12 can be selected according to the internal space of the tunnel model.

[0059] The speed controller 15 integrates multiple components, including a power supply, a DC motor speed regulator, a pulse generator, and a stepper motor driver; the power supply is used to supply power to other components; the DC motor speed regulator can adjust the input current of the DC electric push rod 13 of the longitudinal transmission mechanism to control the propulsion speed of the DC electric push rod 13; the stepper motor driver receives and analyzes the pulse signal of the pulse generator to operate the hollow shaft stepper motor 9 of the displacement meter rotating mechanism to achieve precise control of the rotation speed of the displacement meter rotating mechanism. The speed controller can adopt different propulsion speeds of the DC electric push rod 13 and the rotation speeds of the hollow shaft stepper motor 9 to achieve different measurement accuracy and resolution requirements.

[0060] The data processing module includes data preprocessing and post-processing, wherein: the data preprocessing includes data cleaning, normalization processing, etc.; the post-processing processes the measurement data of the laser displacement meter based on a new unsupervised generative adversarial network (GAN) model architecture based on artificial intelligence, filters out outliers and oscillations in the data, and retains the core deformation features in the originally highly complex and irregular data.

[0061] Based on the above model experimental tunnel longitudinal and transverse full-section deformation monitoring device, the longitudinal and transverse deformation data of the tunnel structure can be dynamically obtained in a small space, and intelligent data noise reduction and filtering processing is realized through a new unsupervised generative adversarial network (GAN) model architecture based on artificial intelligence, which improves data quality, better reveals the deformation mode of the tunnel structure, and meets the dynamic monitoring needs of the longitudinal and transverse deformation of the tunnel. Figure 6 As shown, the implementation steps include:

[0062] Step 1: Lay out the device of the present application, confirm that the initial state is correct, control the hollow shaft stepper motor 9 and the DC electric push rod 13 through the speed controller 15, realize the precise control of the translation and rotation speed of the laser displacement meter 12, and collect the monitoring data of the laser displacement meter as the benchmark data;

[0063] Step 2: Simulate tunnel deformation, and during the tunnel deformation process, control the hollow shaft stepper motor 9 and the DC electric push rod 13 through the speed controller 15 to achieve precise control of the translation and rotation speeds of the laser displacement meter 12, dynamically collect monitoring data, and compare them with the benchmark data obtained in step 1, so as to obtain the longitudinal and transverse deformation data of the entire section of the tunnel structure;

[0064] Step 3: Preprocess the full-section longitudinal and transverse deformation monitoring data obtained in step 2, and divide the preprocessed data into a training set and a test set;

[0065] Step 4: In order to post-process the noisy data, design and initialize the GAN model architecture, wherein the GAN model includes a generator and a discriminator;

[0066] Step 5: Iteratively update the generator and discriminator to implement cyclic training of the GAN model and save the trained bidirectional optimal model;

[0067] Step 6: Based on the bidirectional optimal model trained in step 5, the noisy raw data is input into the generator to generate high-quality data for analysis and prediction of the tunnel deformation state.

[0068] like Figure 7 As shown, the working modes of the longitudinal and transverse full-section deformation monitoring device of the model test tunnel include linear and spiral types; one embodiment of the linear type is: the laser displacement meter 12 is installed at the arch top, arch bottom, left waist and right waist positions of the positioning plate 11, and only the DC electric push rod 13 is used to drive the laser displacement meter 12 to move horizontally at a uniform speed along the longitudinal direction of the tunnel to obtain deformation data on four continuous measuring lines at four positions; one embodiment of the spiral type is: the laser displacement meter 12 is installed at the arch top of the positioning plate 11, and the DC electric push rod 13 and the hollow shaft stepping motor 9 are used to drive the laser displacement meter 11 to move and rotate at a uniform speed to obtain deformation data on a spiral line.

[0069] Specifically, the movement of the laser displacement meter 12 from one end to the other end is regarded as completing a full-section measurement; before the formal experiment begins, the translational speed and rotational speed of the laser displacement meter 12 need to be calibrated in advance; in addition, a measurement is required to determine the reference state of the tunnel; in subsequent tests, the difference between the measured data and the measured data under the reference state is calculated to obtain the deformation of the tunnel at each point on the survey line.

[0070] Given that the inner diameter of the tunnel is R, the translational velocity v and the angular velocity ω of the laser displacement meter are measured by calibration, then the position of the laser measuring point in three-dimensional space at any time t can be obtained by the formula:

[0071]

[0072] Furthermore, after each cycle of T = 2π / ω, the laser measuring point reaches the same position in the annular direction after traveling a distance of δx = vT, so that the longitudinal measuring line of any point on the entire section of the tunnel can be obtained. Furthermore, in the actual monitoring application of the spiral line, the ratio of the rotation speed to the translation speed can be increased, or the number of laser displacement meters can be increased to obtain denser measuring lines, so as to achieve full-section longitudinal and transverse deformation measurement that approximately covers any point in the tunnel. After obtaining the measuring line, the data is post-processed based on the GAN model architecture to capture the nonlinear characteristics in the core deformation mode of the tunnel.

[0073] In step 3 above, the process of preprocessing the full-section longitudinal and transverse deformation monitoring data is as follows: First, the acquired spiral longitudinal and transverse full-section monitoring data is converted into multiple longitudinal measurement lines at different angles to facilitate the understanding of the physical meaning of the data. Due to the complex internal environment during the experiment, there are multiple influencing factors such as foreign body obstruction, unstable signal transmission, etc., which may cause various problems such as invalid laser sensor data and abnormal data mutation. Therefore, first of all, outliers and invalid time periods in the data must be removed, and the missing data must be processed by methods such as average value replacement and 3sigma rule. In the next step, the data is normalized, and the tunnel deformation value is mapped to between 0 and 1 using the normalization method, so as to provide better GAN model generation performance.

[0074]

[0075] Where x is the original data, min(X) and max(X) are the minimum and maximum values ​​in the data respectively. Finally, the preprocessed data is divided into a training set and a test set.

[0076] In step 4 above, the specific process of initializing the GAN model architecture is as follows: First, define the GAN model. The model structure mainly includes the generator G and the discriminator D, such as Figure 8 shown.

[0077] The information received by the generator G is the pre-processed original tunnel deformation survey line data. The sequence features (Conv1D) are extracted through multiple one-dimensional convolutional layers, and ReLU activation is added after each layer to increase the nonlinear modeling capability. Then the deconvolution layer is used to map the features back to the same dimension as the input. Finally, the Tanh activation function is used to generate the output. The ideal output is the denoised deformation data, which can gradually capture and restore the dynamic changes and trends of the sequence data. The output of the generator can be expressed as follows:

[0078] G(x)=Tanh(Deconv(BatchNorm(ReLU(Conv(x)))))

[0079] The input of the discriminator D can be the output G(x) of the generator or the original noisy sequence data x, and the input data is classified to maximize the difference in distribution between noisy data and denoised data. The discriminator includes multiple layers of one-dimensional convolutional layers and Leaky ReLU activation layers to extract features, and finally a fully connected layer + Sigmoid activation layer is added to output a value indicating the probability that the input data is noisy data. The output of the discriminator can be expressed as follows:

[0080] D(x)=Sigmoid(Dense(LeakyReLU(Conv(x))))

[0081] The core idea of ​​step 5 above is to treat the training of the generator and the discriminator as a two-way optimal solution problem, so that the two can reach Nash equilibrium during the training process. The generator and the discriminator are updated alternately in an adversarial game, which is specifically manifested as follows: the discriminator tries to distinguish whether the data is original noise data or forged smooth data, while the generator tries to generate high-quality data from the noise data that removes most of the noise but retains the main features, so that the discriminator cannot easily distinguish the denoised data from the original data, that is, to find the best balance between "smoothness" and "naturalness".

[0082] For this, we first need to set up the discriminator. Define that when the input data is the original noisy data, D(x) = 1; when the input data is the smoothed data G(x) generated by the generator, let D(G(x)) = 0. Then the loss function of the discriminator can be defined as:

[0083]

[0084] Where E x～pdata is the expectation sampled in the real measured tunnel deformation time series data set, D(x) is the discrimination result of the original data, D(G(x)) is the discrimination result of the generated smoothed data, and L D is the discriminator loss function. This loss can encourage the discriminator to maximize the difference between the original noisy data and the generated smooth data, thereby encouraging the generator to more effectively generate smooth data that can capture the main features.

[0085] The loss function of the generator is:

[0086]

[0087] Where L G is the generator loss function. This loss will guide the generated smooth data to be closer to the distribution mode of the noisy data, and gradually learn to generate trend data with random noise removed.

[0088] In step 5 above, the specific training model optimization process is as follows: First, select the initial learning rate and the learning rate decay strategy. Perform batch sampling in the training set containing deformation time series data on several survey lines, select a batch of data x, and then use the generator to generate G(x) based on the input x. Then, calculate the discriminator loss function, optimize the discriminator parameters according to the gradient descent algorithm, and update the discriminator. Next, keep the discriminator parameters unchanged, calculate the generator loss function, and optimize the generator parameters according to the gradient descent algorithm. In this way, repeat the above iterative training steps, alternately optimize the generator and the discriminator, until the denoised data generated by the generator is smooth and close to the main trend of the original noise data.

[0089] If the model performs poorly, adjust the model hyperparameters and retrain until the model can perform well in the test set, that is, it can generate high-quality deformation sequence data that meets the requirements based on the original high-noise data. In one embodiment of the present application, the model hardware is an Intel Core i9-14700k CPU and an NvidiaRTX4060 GPU with 8GB of memory. The software environment is configured with the PyTorch framework in the Windows system. The convolutional layers for encoding and decoding are set to three layers, and the other fully connected layers and activation layers are all one layer. The initial learning rate is 1e-4, and the Adam learning rate decay strategy is adopted.

[0090] In the above step 6, after a GAN model has been trained, the generator parameters are saved to build the final denoising model to generate high-quality data from the original noisy data. In practical applications, it is only necessary to input the deformation time series data on a certain survey line into the generator G to obtain the denoised deformation data that retains the main trend, such as y = G (x).

[0091] This application is mainly used to measure the deformation of the entire tunnel section in a physical similarity model test device, and is suitable for exploring the global deformation of the tunnel structure in the longitudinal and transverse directions under various disturbances. It can also be flexibly adjusted according to the test requirements of the tunnel's burial depth, size, horizontal position and monitoring density, and has a wide range of applicability. On this basis, the data post-processing proposed in this application ensures the quality of the acquired data.

[0092] The above description is only a description of the preferred embodiments of the present application, and is not intended to limit the scope of the present application. Any changes or modifications made by any person skilled in the art based on the above disclosed technical contents shall be deemed as equivalent effective embodiments and shall fall within the scope of protection of the technical solution of the present application.

Claims

1. The model test tunnel longitudinal and transverse full-section deformation monitoring device is characterized by: It comprises a model box (16), a horizontal positioning bracket, a vertically adjustable positioning bracket, a longitudinal transmission mechanism, a displacement meter rotation mechanism, a speed controller (15) and a data processing module, wherein: The model box (16) is arranged on a horizontal positioning bracket and is used for filling soil and simulating tunnel space; There are multiple vertically adjustable positioning brackets, which are arranged on the horizontal positioning bracket. The horizontal positioning bracket and the multiple vertically adjustable positioning brackets together constitute a bracket system for supporting the longitudinal transmission mechanism and the displacement meter rotation mechanism; The longitudinal transmission mechanism is supported on a vertically adjustable positioning bracket, and the displacement meter rotation mechanism is arranged on the longitudinal transmission mechanism, and is used to obtain longitudinal and transverse deformation monitoring data of the entire cross section in the tunnel space; The speed controller (15) is connected to the longitudinal transmission mechanism and the displacement meter rotation mechanism, and is used to control the translation speed of the longitudinal transmission mechanism and the rotation speed of the displacement meter rotation mechanism; The data processing module is used to process the full-section longitudinal and transverse deformation monitoring data.

2. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 1 is characterized in that: The horizontal positioning bracket comprises two brackets (1) and a base (2) connected between the two brackets (1); the model box (16) is arranged on the base (2), and the two brackets (1) are respectively located at two ends of the model box (16); The vertically adjustable positioning device comprises a push rod positioning groove (3), a vertical bracket (4), and also comprises a vertical positioning adjustment bolt (5), an anchor bolt (6), and a positioning bolt (7); a bolt hole is arranged on the upper part of the bracket (1) of the horizontal positioning bracket, and the bottom of the vertical bracket (4) is connected to the bracket (1) of the horizontal positioning bracket by the anchor bolt (6) to achieve fixation; a threaded hole is arranged on the top of the vertical bracket (4), and the vertical positioning adjustment bolt (5) is threadedly connected to the top of the vertical bracket (4), and the lower part of the vertical positioning adjustment bolt (5) is connected to the push rod positioning groove (3) by welding, and the push rod positioning groove (3) is driven to move vertically by rotating the adjustment bolt (5), thereby changing the vertical position of the push rod positioning groove (3); at the same time, the push rod positioning groove (3) and the vertical bracket (4) are fixed together by the positioning bolt (7) to limit the vibration of the overall system during the measurement process.

3. The longitudinal and transverse full-section deformation monitoring device for a model test tunnel according to claim 2 is characterized in that: The longitudinal transmission mechanism comprises a DC electric push rod (13) and an extension rod (14); the extension rod (14) is connected to the DC electric push rod (13); the extension rod (14) performs linear reciprocating motion under the action of the DC electric push rod (13); the DC electric push rod (13) and the extension rod (14) are arranged in a push rod positioning groove (3) of a vertically adjustable positioning device; The displacement meter rotating mechanism is fixed on the extension rod (14) of the longitudinal transmission mechanism, and comprises a slewing bearing (8), a hollow shaft stepping motor (9), a rotating mechanism positioner (10), a displacement meter positioning disk (11), and a laser displacement meter (12); wherein the hollow shaft stepping motor (9) is arranged on the extension rod (14) of the longitudinal transmission mechanism; the laser displacement meter (12) is fixed on the displacement meter positioning disk (11), and the laser measurement direction is consistent with the radial direction of the positioning disk of the positioning meter; the displacement meter positioning disk (11) is arranged on the extension rod (14) of the longitudinal transmission mechanism through the slewing bearing (8), and the other end of the displacement meter positioning disk (11) is connected to the hollow shaft stepping motor (9), and the hollow shaft stepping motor (9) drives the displacement meter positioning disk (11) to rotate, thereby driving the laser displacement meter (12) to rotate; the rotating mechanism positioner (10) is provided with two, and is used to fix the slewing bearing (8) and the hollow shaft stepping motor (9) on the extension rod (14) of the longitudinal transmission mechanism.

4. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 3 is characterized in that: The longitudinal transmission mechanism has a DC electric push rod (13) arranged in push rod positioning grooves (3) of two vertically adjustable positioning devices at the left end of the model box (16), and an extension rod (14) arranged in push rod positioning grooves (3) of two vertically adjustable positioning devices at the right end of the model box (16), and a gasket is placed in the push rod positioning groove to increase friction so as to control jitter and rotation during the measurement process.

5. The longitudinal and transverse full-section deformation monitoring device for a model test tunnel according to claim 3 is characterized in that: The longitudinal transmission mechanism: its DC electric push rod (13) integrates a DC motor, a reducer, a screw, and a guide rail component; the DC motor uses a permanent magnet DC motor to generate electromagnetic force to drive the rotor to rotate; the reducer is used to reduce the motor rotation speed to prevent a large disturbance to the monitoring system; the screw is used to convert the rotational motion of the DC motor into linear motion; the guide rail is used to maintain the direction of the screw and keep its motion stable; the extension rod (14) is connected to the screw of the DC electric push rod (13).

6. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 3 is characterized in that: The size, layout position and measuring range of the laser displacement meter (12) are selected according to the internal space of the tunnel model.

7. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 3 is characterized in that: The speed controller integrates multiple components, including a power supply, a DC motor speed regulator, a pulse generator, and a stepper motor driver; the power supply is used to supply power to other components; the DC motor speed regulator is used to adjust the input current of the DC electric push rod (13) of the longitudinal transmission mechanism to control the propulsion speed of the DC electric push rod (13); the stepper motor driver receives and analyzes the pulse signal of the pulse generator to operate the hollow shaft stepper motor (9) of the displacement meter rotating mechanism to achieve precise control of the rotation speed of the displacement meter rotating mechanism.

8. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 1 is characterized in that: The data processing module includes data preprocessing and post-processing, wherein: the data preprocessing includes data cleaning and normalization processing; the post-processing processes the measurement data of the laser displacement meter based on a new unsupervised generative adversarial network (GAN) model architecture based on artificial intelligence, filters out outliers and oscillations in the data, and retains the core deformation features in the originally highly complex and irregular data.

9. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 3, characterized in that: The relevant model test applied to the analysis of the stress and deformation of the tunnel's transverse and longitudinal structures and their influence includes the following steps: Step 1: Arrange the device of the present application, confirm that the initial state is correct, control the hollow shaft stepper motor (9) and the DC electric push rod (13) through the speed controller (15), realize the precise control of the translation and rotation speed of the laser displacement meter (12), and collect the monitoring data of the laser displacement meter as the reference data; Step 2: Simulate tunnel deformation, and during the tunnel deformation process, control the hollow shaft stepper motor (9) and the DC electric push rod (13) through the speed controller (15) to achieve accurate control of the translation and rotation speeds of the laser displacement meter (12), dynamically collect monitoring data, and compare them with the benchmark data obtained in step 1, so as to obtain the longitudinal and transverse deformation data of the entire cross-section of the tunnel structure; Step 3: Preprocess the full-section longitudinal and transverse deformation monitoring data obtained in step 2, and divide the preprocessed data into a training set and a test set; Step 4: In order to post-process the noisy data, design and initialize the GAN model architecture, wherein the GAN model includes a generator and a discriminator; Step 5: Iteratively update the generator and discriminator to implement cyclic training of the GAN model and save the trained bidirectional optimal model; Step 6: Based on the bidirectional optimal model trained in step 5, the noisy raw data is input into the generator to generate high-quality data for analysis and prediction of the tunnel deformation state.

10. The model test tunnel longitudinal and transverse full-section deformation monitoring device according to claim 9, characterized in that: The GAN model consists of a generator G and a discriminator D, where: The information received by the generator G is the preprocessed original tunnel deformation survey line data. The sequence features are extracted through multiple one-dimensional convolutional layers (Conv1D), and ReLU activation is added after each layer to increase the nonlinear modeling capability. Then, the deconvolution layer is used to map the features back to the same dimension as the input. Finally, the Tanh activation function is used to generate the output. The ideal output is the denoised deformation data, which can gradually capture and restore the dynamic changes and trends of the sequence data. The output of the generator is expressed as follows: G(x)=Tanh(Deconv(BatchNorm(ReLU(Conv(x))))) The input of the discriminator D is the output G(x) of the generator or the original noisy sequence data x, and it classifies the input data to maximize the difference in distribution between noisy data and denoised data. The discriminator includes multiple layers of one-dimensional convolutional layers and Leaky ReLU activation layers to extract features, and finally adds a fully connected layer + Sigmoid activation layer to output a value indicating the probability that the input data is noisy data. The output of the discriminator is expressed as follows: D(x)=Sigmoid(Dense(LeakyReLU(Conv(x)))) Define that when the input data is the original noisy data, D(x) = 1; when the input data is the smoothed data G(x) generated by the generator, let D(G(x)) = 0. Then define the loss function of the discriminator as: Among them, E x～pdata is the expectation sampled in the real measured tunnel deformation time series data set, D(x) is the discrimination result of the original data, D(G(x)) is the discrimination result of the generated smoothed data, and L D is the discriminator loss function; The loss function of the generator is: Where L G is the generator loss function; The training model optimization process is as follows: First, select the initial learning rate and learning rate decay strategy; Secondly, batch sampling is performed in the training set containing deformation time series data on several survey lines, a batch of data x is selected, and then the generator is used to generate G(x) according to the input x; Then, the discriminator loss function is calculated, and the discriminator is optimized and updated according to the gradient descent algorithm; Next, keep the discriminator parameters unchanged, calculate the generator loss function, and optimize the generator according to the gradient descent algorithm; In this way, the above iterative training steps are repeated, alternately optimizing the generator and the discriminator until the denoised data generated by the generator is smooth and close to the main trend of the original noisy data.

Images