Permeation column test device and method for mud membrane identification of slurry balance shield

Abstract

The application discloses a kind of mud balance shield mud film identification permeameter test device and method, it is related to geotechnical engineering test field, it includes organic glass cylinder, air compressor, pressure gauge, pore water pressure gauge and ground penetrating radar detection system.The organic glass cylinder forms closed test cavity, air compressor is controlled air pressure to cavity interior by gas pressure regulating valve to drive mud penetration soil sample to generate mud film, and pressure gauge and pore water pressure gauge are used to monitor gas pressure in cavity and internal pore water pressure of medium.Radar antenna of ground penetrating radar detection system is installed in the way of emission surface no gap adhering to the top surface of organic glass cylinder, and the center normal direction is perpendicular to the layered interface of medium in cylinder, electromagnetic wave is vertically incident to obtain the best reflection echo signal.The application integrates ground penetrating radar and permeameter device structure, realizes the real-time nondestructive detection of mud film formation process.

Description

Technical Field

[0001] This application relates to the field of geotechnical engineering testing technology, and more specifically, to a permeability column test device and method for identifying mud film in slurry balance shield tunneling machines. Background Technology

[0002] In cross-river and cross-sea tunnel projects, slurry-balanced shield tunneling is widely used due to its easily controllable support pressure. During excavation, pressurized slurry infiltrates the strata at the excavation face. Bentonite particles in the slurry gradually deposit in the soil pores, forming a dense, low-permeability mud film on the surface of the excavation face. The formation of this mud film allows the pressure of the slurry chamber to be effectively transmitted to the soil skeleton, establishing a support system balanced with the soil and water pressure of the strata. The quality and thickness of this mud film directly affect the efficiency of support pressure transmission and the overall stability of the excavation face. However, current engineering practice relies mainly on indirect inference from slurry parameters and tunneling data for assessment of the mud film's condition, or requires manual inspection by stopping the machine and opening the chamber. The former makes it difficult to obtain direct structural information about the mud film, while the latter seriously affects construction progress and poses safety risks. Neither approach can meet the need for real-time continuous monitoring of the mud film's condition.

[0003] The permeation column test, a classic indoor simulation method for studying the formation mechanism of mud films, reproduces the mud film formation process by pressurizing a soil sample within a sealed columnar container to allow mud to permeate the sample. However, existing devices still rely on destructive sampling after the test for mud film observation, failing to capture real-time information on the dynamic formation of mud films. Ground-penetrating radar (GPR) technology offers the advantages of being non-destructive and efficient. It obtains information about the internal layering of the medium by emitting electromagnetic waves and receiving reflected echoes at interfaces with different dielectric constants, theoretically enabling the detection of local dielectric constant changes caused by mud films. However, mud films are typically only millimeters to centimeters thick, and the multi-layered water-bearing environment causes complex interference to the signal. Furthermore, there is currently a lack of technical solutions for effectively integrating GPR with permeation column devices, making it difficult to support the necessary experimental platform for research on real-time non-destructive monitoring of mud films using radar technology. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this application provides a permeability column test device and method for identifying mud film in slurry balance shield tunnels.

[0005] According to a first aspect of this application, a permeable column test device for identifying mud film in a slurry-water balance shield tunnel is provided, comprising: an plexiglass cylinder, an air compressor, a pressure gauge, a pore water pressure gauge, and a ground-penetrating radar detection system; wherein, the plexiglass cylinder forms a closed test chamber; the output port of the air compressor is connected to the upper space of the inner cavity of the plexiglass cylinder through an air pipeline and an air pressure regulating valve; the pressure-sensing end of the pressure gauge extends into the inner cavity of the plexiglass cylinder; the sensing probe of the pore water pressure gauge is disposed through the side wall of the plexiglass cylinder and extends into the interior of the test medium, and is in direct contact with the pore water inside the plexiglass cylinder; wherein, the ground-penetrating radar detection system is installed at a preset position, and the ground-penetrating radar detection system includes a radar antenna, the emitting surface of the radar antenna is in close contact with the top surface of the plexiglass cylinder without gaps, and the central normal direction of the radar antenna is perpendicular to the medium layering interface inside the plexiglass cylinder.

[0006] According to a second aspect of this application, a method for operating a permeable column test device for identifying mud film in a slurry-water balance shield tunnel is provided, comprising: sequentially filling a gravel filter layer and a layered compacted sand layer into the interior of an acrylic cylinder; then closing the drain valve and slowly injecting water in reverse through the drainage channel of the bottom glass sealing plate to fully saturate the sand layer; after saturation, injecting a quantitative amount of mud slurry onto the top surface of the sand layer; and then sealing and installing the top glass sealing plate on the top of the acrylic cylinder; starting an air compressor to output compressed gas, adjusting the compressed gas to the target pressure value through a pressure regulating valve, and then entering the acrylic cylinder through an air path to apply uniform downward air pressure to the surface of the mud slurry; then opening the drain valve, and under the action of pressure difference, the mud slurry permeates the sand layer from top to bottom, and the bentonite particles in the mud slurry gradually deposit on the surface of the sand layer to form a mud film; wherein, a pore water pressure gauge continuously monitors the pore water pressure change in the medium inside the acrylic cylinder; and simultaneously starting the pressurization, activating a ground-penetrating radar detection system, wherein the ground-penetrating radar detection system receives and processes the echo signal to obtain the actual thickness of the mud film.

[0007] Compared with existing technologies, this application provides a permeable column test device and method for mud film identification in slurry-water balance shield tunneling. This device enables continuous radar signal acquisition throughout the entire process of dynamic mud film formation under pressure, achieving direct coupling between the physical process of mud film formation and the radar detection process. Unlike existing technologies that require destructive sampling and observation only after the test is completed, this device continuously acquires echo data reflecting the stratification changes of the medium throughout the entire mud film formation process, fully recording the electromagnetic response characteristics of the mud film from initial deposition to stable film formation. This provides continuous raw detection data in the time dimension for in-depth research on the evolution of mud film. Simultaneously, the coordinated configuration of the pneumatic loading system and the pore water pressure monitoring system allows for precise control and synchronous recording of the mechanical boundary conditions during the test, ensuring the temporal correspondence between radar detection data and mechanical parameters. This lays a data foundation for establishing a quantitative correlation between the physical properties of the mud film and radar signal characteristics. Attached Figure Description

[0008] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 Side view of the permeable column test apparatus for identifying mud film in slurry balance shield tunneling according to this application. Figure 1 A schematic diagram showing the radar antenna mounted on top of the test setup.

[0010] Figure 2 Side view of the permeable column test apparatus for identifying mud film in slurry balance shield tunneling according to this application. Figure 2 A schematic diagram showing the radar antenna installed at the bottom of the test device.

[0011] Figure 3 This is a flowchart illustrating the working method of the permeable column test device for mud film identification in slurry balance shield tunneling according to this application.

[0012] Figure 4 This is a flowchart illustrating step S3 of the working method of the permeable column test device for identifying mud film in a slurry balance shield tunneling machine according to this application.

[0013] Figure 5 This is a flowchart illustrating the second embodiment of step S33 in the working method of the permeable column test device for identifying mud film in a slurry balance shield tunneling machine according to this application.

[0014] In the diagram: 1. Acrylic cylinder; 2. Air compressor; 3. Pressure gauge; 4. Pore pressure gauge; 5. Measuring cylinder; 6. Support and protective components; 7. Transmission cable; 8. A ground-penetrating radar detection system with replaceable antenna. Detailed Implementation

[0015] The embodiments of this application will now be described in more detail with reference to the accompanying drawings. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0016] like Figure 1 and Figure 2 As shown, this application provides a permeable column test device for mud film identification in slurry-water balance shield tunneling. The device includes an acrylic cylinder 1, an air compressor 2, a pressure gauge 3, a pore water pressure gauge 4, and a ground-penetrating radar detection system 8. The acrylic cylinder 1 serves as the main container of the permeable column test device, containing sand layers and mud, forming a simulated stratified geological structure. The air compressor 2 provides an adjustable compressed gas source to pressurize the air above the mud inside the acrylic cylinder 1. The pressure gauge 3 measures the air pressure inside the test device. The pore water pressure gauge 4 monitors the pore water pressure within the medium during the test. The ground-penetrating radar detection system 8 includes a radar antenna, a radar host, and a transceiver control unit, used for real-time electromagnetic wave detection of the medium stratification interfaces within the acrylic cylinder 1 during the test. The entire device is primarily constructed of acrylic glass to minimize the use of metal materials and avoid strong interference with the radar signal.

[0017] In this embodiment, the acrylic cylinder 1 includes an acrylic cylinder body, a top glass sealing plate, and a bottom glass sealing plate. The top glass sealing plate is fixedly connected to the top opening of the acrylic cylinder 1 in a sealed manner, and the bottom glass sealing plate is fixedly connected to the bottom opening of the acrylic cylinder 1 in a sealed manner. The acrylic cylinder body, the top glass sealing plate, and the bottom glass sealing plate together form a closed test chamber. The acrylic cylinder 1 has good sealing performance and can maintain the airtightness of the chamber under pressure, providing a stable sealed environment for the mud to penetrate the sand layer from top to bottom under the action of pressure difference and gradually deposit on the surface of the sand layer to form a mud film. Figure 1 and Figure 2As shown, the inner cavity of the acrylic cylinder 1 vertically contains, from bottom to top, a gravel filter layer, a sand layer, and a mud layer. The gravel filter layer, with a thickness S3 of approximately 30 mm, is laid at the bottom of the acrylic cylinder 1 as a filter bed. The sand layer, with a thickness S2 of approximately 120 mm, is prepared using a layered compaction method and its relative density is controlled to approximately 80% through hammering. The mud layer is located above the sand layer, and the space between the mud layer and the top glass sealing plate is the air pressure loading space. The total height S1 of the inner cavity of the acrylic cylinder 1 is approximately 250 mm, thus providing sufficient space for the upper mud layer and air pressure loading.

[0018] In this embodiment, the output port of the air compressor 2 is threadedly connected to the inlet end of the air pressure regulating valve via an air passage pipe. The outlet end of the air pressure regulating valve passes through the top glass sealing plate and connects to the upper space of the inner cavity of the plexiglass cylinder 1 via an air passage pipe. The air compressor 2 is equipped with an air pressure regulating valve, which is used to stably regulate the air supply pressure output by the air compressor 2 to the test set value, thereby controlling the stability of the air pressure above the slurry. During the test, the air compressor 2 is started to output compressed gas. After the compressed gas is regulated to the target pressure value by the air pressure regulating valve, it enters the upper space of the inner cavity of the plexiglass cylinder 1 through the air passage, applying uniform downward air pressure to the surface of the slurry. Figure 1 and Figure 2 As shown, the air pipeline is led out from the air compressor 2 and connected to the top glass sealing plate of the plexiglass cylinder 1 through the air pressure regulating valve. The air pressure regulating valve can stabilize the air pressure above the slurry at 50 kPa. Under the action of this pressure difference, the slurry penetrates the sand layer from top to bottom, and the bentonite particles in the slurry are gradually deposited on the surface of the sand layer to form a mud film.

[0019] In this embodiment, the pressure-sensing end of the pressure gauge 3 passes through the top glass sealing plate and extends into the inner cavity of the plexiglass cylinder 1, for real-time measurement of the gas pressure above the mud in the inner cavity of the plexiglass cylinder 1. The pressure gauge 3 works continuously throughout the experiment, feeding back the real-time value of the gas pressure in the cavity to the operator, so that the operator can accurately control and synchronously record the pressurization conditions in conjunction with the pressure regulating valve.

[0020] In this embodiment, as Figure 1 and Figure 2As shown, the sensing probe of the pore water pressure gauge 4 penetrates the side wall of the plexiglass cylinder 1 and extends into the test medium, directly contacting the pore water inside the cylinder 1. The pore water pressure gauge 4 is used to continuously monitor the changes in pore water pressure within the medium of the plexiglass cylinder 1 during the experiment, simultaneously recording the changes in pore water pressure and drainage volume over time. During pressurized infiltration, the mud permeates the sand layer from top to bottom and gradually forms a mud film. The pore water pressure inside the medium changes dynamically accordingly. The pore water pressure gauge 4 captures this change process in real time through the direct contact between the sensing probe and the pore water, providing mechanical parameter data for analyzing the seepage characteristics during mud film formation.

[0021] In this embodiment, the ground-penetrating radar detection system 8 includes a radar antenna, a radar main unit, and a transceiver control unit. The transceiver control unit is built into the radar main unit and electrically connected to its main control circuit, used to precisely control the timing of the radar antenna's electromagnetic wave transmission pulses and the reception window. The radar main unit is electrically connected to the radar antenna via an RF feed line, driving the antenna to transmit and receive electromagnetic waves according to a set timing sequence. The ground-penetrating radar antenna is replaceable, and there are two types: pulsed radar antennas and stepped radar antennas. Pulsed antennas have two different frequencies to choose from: 1600MHz and 2600MHz. Stepped radar antennas also have two different center frequencies to choose from: 1500MHz and 2500MHz. Depending on the antenna's frequency and detection performance, the radar antenna is either fixed to a dedicated mounting position on top of the test device or installed at a mounting position on the bottom of the test device.

[0022] In this embodiment, the ground-penetrating radar detection system 8 is installed at a preset location. The preset location refers to the outer surface of the top or bottom glass sealing plate of the plexiglass cylinder 1; the specific installation location is determined based on the frequency of the selected radar antenna and its corresponding detection depth and detection effect. Figure 1 As shown, when the detection depth and effect of the ground-penetrating radar antenna are sufficient to cover the entire penetration column test device, the preset position is the top of the test device, that is, the mode of installing the radar antenna only on the top of the test device is adopted. Figure 2As shown, if the radar antenna installed at the top of the test device has a high frequency and cannot achieve sufficient detection depth and effect (i.e., the acquired radar data has very weak signals or even no radar reflection signals in deeper areas), then the preset position is the bottom of the test device. This means that under the same test conditions, an additional supplementary test group needs to be added, with the radar antenna of the supplementary test group installed at the bottom of the test device. Two separate tests are conducted, collecting radar data from the top to the bottom of the device and from the bottom to the top, respectively, to achieve complementarity and improvement of the radar data and solve the problem of insufficient detection depth and effect. Not using two identical radar antennas simultaneously installed at the top and bottom of the test device is to avoid mutual interference between the signals emitted by the two radar antennas and to reduce the overall cost of the device.

[0023] The radar antenna of the ground-penetrating radar detection system 8 is fixed in a preset position using a mounting bracket. The mounting bracket is securely connected to the outer surface of the top or bottom glass sealing plate using threaded fasteners. The central area of ​​the mounting bracket is recessed inward to form a mounting base, reserving positioning space for the radar antenna installation. The radar antenna is embedded in the mounting base, ensuring a seamless fit between the antenna's emitting surface and the outer surface of the top or bottom glass sealing plate, with no gaps between the ground-penetrating radar detection system 8 and the top or bottom of the test apparatus. During installation, the antenna of the ground-penetrating radar detection system 8 requires laser calibration to ensure that the antenna's emitting surface faces the layered interface within the soil column, and that its central normal is as perpendicular as possible to the layered interface, guaranteeing perpendicular electromagnetic wave incidence and achieving optimal signal penetration and reflection. A waterproof and wave-transparent enclosure covers the outer surface and exposed interfaces of the radar antenna, sealingly fitting with the mounting base to provide waterproof protection for the radar antenna under complex and variable laboratory conditions. The wear-resistant protective layer, made of plexiglass, is attached to the outer surface of the waterproof and wave-permeable enclosure, covering all exposed areas to protect against impacts and abrasions in the laboratory environment. Figure 1 and Figure 2 As shown, the thickness S4 of the mounting bracket and protective component 6 is approximately 25mm. The inwardly recessed area of ​​the mounting bracket and protective component 6 provides rigid support and precise positioning for the radar antenna.

[0024] In this embodiment, one end of the transmission cable 7 is electrically connected to the signal output port of the radar host, and the other end of the transmission cable 7 is electrically connected to the signal input port of the industrial control computer. The middle section of the transmission cable 7 is embedded in the cable groove inside the mounting bracket and leads out to the external equipment through the outside of the mounting bracket. The transmission cable 7 is used to connect the ground penetrating radar detection system 8 to the external equipment, and undertakes the functions of transmitting echo signals and supplying power.

[0025] The working process of the permeation column test device for mud film identification in a slurry-water balance shield tunnel is as follows: When the air compressor 2 is started and applies stable air pressure to the inner cavity of the plexiglass cylinder 1 through the air pressure regulating valve, it drives the mud to penetrate downwards into the sand layer and form a mud film on the surface of the sand layer. At the same time, the radar host drives the radar antenna to continuously emit electromagnetic waves and receive reflected echoes through the transceiver control unit. After the electromagnetic waves pass through the top or bottom glass sealing plate and enter the test medium, they are reflected at the interface where the dielectric constant changes abruptly between the mud layer, mud film, permeation zone, and undisturbed sand layer. The radar antenna receives the reflected echo signal, and after the radar host performs preliminary sampling of the echo signal, it transmits the digitized raw echo data to the industrial control computer in real time through the transmission cable 7. The ground-penetrating radar detection system 8 continuously monitors in real time throughout the entire test process, completely recording the radar data during the entire mud film formation process. After the permeation continues for 5 minutes and the mud film no longer changes significantly, the bottom drain valve is closed, and the test ends. The ground-penetrating radar detection system 8 is shut down, and the radar data of the complete test process is saved. At the same time, the pore water pressure data and the device drainage volume are recorded. Based on real-time radar data, key signal changes can also be manually marked and recorded to facilitate subsequent data processing and analysis.

[0026] like Figure 3 As shown, this application also provides a working method for a permeable column test device for identifying mud film in slurry balance shield tunneling. The method includes: S1, sequentially filling the interior of an acrylic cylinder with a gravel filter layer and a layered, compacted sand layer; then closing the drain valve and slowly injecting water in reverse through the drainage channel of the bottom glass sealing plate to fully saturate the sand layer; after saturation, injecting a measured amount of mud slurry onto the top surface of the sand layer; and then sealing the top glass sealing plate onto the top of the acrylic cylinder. It should be understood that a test medium system capable of realistically simulating the geological distribution state ahead of the excavation face during slurry balance shield tunneling is constructed within the inner cavity of the acrylic cylinder 1. This allows the mud slurry to permeate the sand layer from top to bottom under pressure and gradually deposit on the sand layer surface to form a mud film, thus providing a physical simulation basis for the ground-penetrating radar detection system 8 to perform real-time non-destructive detection of the mud film formation process during the test.

[0027] In this embodiment, a gravel filter bed is laid above the bottom glass sealing plate of the plexiglass cylinder. The function of the gravel filter bed is to provide uniform support to the bottom of the sand layer during subsequent pressurized infiltration, while allowing the permeate to flow smoothly through the pores between the gravel layers to the drainage channel of the bottom glass sealing plate, preventing sand particles from clogging the drainage channel with the permeate loss. When laying the gravel filter bed, it is necessary to ensure that the gravel particles are evenly distributed above the bottom glass sealing plate and the surface is flat, so that the sand layer above receives uniform bottom support. The bottom glass sealing plate is fixedly connected to the bottom opening of the plexiglass cylinder 1 in a sealed manner. The drain valve is installed at the outlet of the drainage channel of the bottom glass sealing plate and is threaded to the bottom glass sealing plate. The measuring cylinder is placed below the drain valve, with the outlet end of the drain valve vertically aligned with the mouth of the measuring cylinder, so that the permeate flows into the measuring cylinder after being discharged through the drain valve. Before laying the gravel filter bed, it is necessary to ensure that the drain valve is in the closed state to prevent water from being lost from the drainage channel during subsequent water injection.

[0028] In this embodiment, a layered compaction method is used to prepare a sand layer above the gravel filter layer. The layered compaction method involves dividing the sand into several equal batches, pouring each batch into the inner cavity of the plexiglass cylinder 1, and compacting the surface of each batch after it is poured. This process is repeated until the sand layer reaches a preset height. The purpose of this layered compaction method is to ensure that the density distribution of the sand layer is as uniform as possible throughout its entire height range, avoiding an uneven density distribution caused by pouring a large amount of sand at once, resulting in a loose upper part and a dense lower part. This ensures that the mud can evenly penetrate the sand layer along its depth during subsequent infiltration, allowing a mud film to form uniformly on the sand layer surface. The compaction is achieved by hammering the surface of the sand layer to control its relative density within the target value. Relative density is a physical index characterizing the compaction degree of sand. It is defined as the ratio of the difference between the current void ratio and the void ratio of its loosest state to the difference between the void ratio of its loosest state and the void ratio of its densest state. The calculation formula is as follows: In the formula, For relative density, This represents the current porosity of the sand layer. The void ratio of sand in its loosest state. The void ratio is the sand's most compacted state. The relative density ranges from 0 to 1, with a higher value indicating a denser sand layer. In the experiments of this application, the relative density of the sand layer was controlled within the target range by controlling the amount of sand poured in each batch and the number of hammer blows, in order to simulate the compaction state of the strata in front of the tunnel boring machine in actual engineering.

[0029] In this embodiment, after the sand layer is prepared and compacted, the drain valve is closed, and water is slowly injected in reverse through the drainage channel of the bottom glass sealing plate to fully saturate the sand layer. Reverse slow water injection refers to injecting water upwards from the bottom of the plexiglass cylinder 1. The water first enters the gravel filter layer through the drainage channel, and then slowly seeps from the gravel filter layer upwards into the pores of the sand layer, gradually discharging the air from the pores until the water level rises above the top surface of the sand layer. The purpose of using a bottom-up reverse water injection method, rather than a top-down forward water injection method, is to avoid trapping air inside the sand layer and forming bubbles during the rapid downward seepage of water from the sand layer surface during forward water injection, which would prevent the sand layer from becoming fully saturated. Reverse water injection allows the water to advance layer by layer from the bottom, continuously displacing the air in the pores of the sand layer vertically and discharging it from the top surface, thereby achieving full saturation of the sand layer. The water injection rate needs to be controlled at a slow level to ensure that the water advances smoothly and evenly in the sand layer pores, and to avoid leaving unsaturated areas due to the water rising rapidly along the local dominant channel caused by the water injection rate being too fast.

[0030] In this embodiment, after the sand layer is saturated, a measured amount of mud is injected onto the top surface of the sand layer. The mud injection must be carried out before the water on the top surface of the sand layer is drained to avoid exposing the top surface of the sand layer to air, which would cause a decrease in surface moisture content. The mud is evenly spread along the top surface of the sand layer, forming a mud layer covering the surface of the sand layer. The mud contains bentonite particles, which will infiltrate into the sand layer with the mud during the subsequent pressurized infiltration process and gradually deposit and accumulate at the pore inlets on the surface of the sand layer, forming a mud film.

[0031] In this embodiment, after the mud injection is completed, the top glass sealing plate is sealed and installed on the top of the acrylic cylinder 1. The top glass sealing plate is fixedly connected to the top opening of the acrylic cylinder 1 in a sealed manner, forming a closed test chamber together with the acrylic cylinder and the bottom glass sealing plate. After the top glass sealing plate is installed, the inner cavity of the acrylic cylinder 1 forms a layered structure from bottom to top, consisting of a gravel filter layer, a saturated sand layer, a mud layer, and an air space above the mud layer. The output port of the air compressor 2 is threadedly connected to the inlet end of the air pressure regulating valve through an air passage pipe. The outlet end of the air pressure regulating valve passes through the top glass sealing plate through an air passage pipe and connects to the upper space of the inner cavity of the acrylic cylinder 1, thereby enabling the air compressor 2 to apply controllable air pressure to the air space above the mud layer through the air passage. The pressure sensing end of the pressure gauge 3 passes through the top glass sealing plate and extends into the inner cavity of the acrylic cylinder 1 for real-time measurement of the gas pressure inside the cavity. After the cavity is sealed, the test medium filling and saturation treatment steps are completed, and the device enters the ready state to perform the pressure permeation test.

[0032] In one specific embodiment of this application, a gravel filter bed with a thickness of 3 cm is first laid at the bottom of the acrylic cylinder 1, with a thickness S3 of approximately 30 mm. Then, a sand layer is prepared using a layered compaction method, with a thickness S2 of approximately 120 mm. The relative density of the sand layer is controlled to approximately 80% by hammering the surface. Next, the drain valve is closed, and water is slowly injected from the drain outlet at the bottom of the acrylic cylinder 1 to fully saturate the sand layer. After saturation, approximately 1 L of slurry is injected onto the top of the sand layer. Finally, a top glass sealing plate is installed on the top of the acrylic cylinder 1, and the total height S1 of the inner cavity of the acrylic cylinder 1 is approximately 250 mm. At this point, from bottom to top, the inner cavity of the acrylic cylinder 1 consists of a gravel filter layer with a thickness of approximately 30 mm, a saturated sand layer with a thickness of approximately 120 mm, and a slurry layer above the sand layer. The space between the slurry layer and the top glass sealing plate is the loading space for the subsequent air compressor 2 to apply air pressure via the air pressure regulating valve.

[0033] S2, the air compressor is started to output compressed gas. After the compressed gas is adjusted to the target pressure value by the air pressure regulating valve, it enters the plexiglass cylinder through the air passage to apply uniform downward air pressure to the surface of the mud. Then, the drain valve is opened, and under the action of pressure difference, the mud permeates the sand layer from top to bottom. Bentonite particles in the mud gradually deposit on the surface of the sand layer to form a mud film. Meanwhile, the pore water pressure gauge continuously monitors the changes in pore water pressure in the medium inside the plexiglass cylinder. It can be understood that by applying controllable gas pressure to the upper space of the inner cavity of the plexiglass cylinder 1, a stable pressure boundary condition is established above the mud layer. At the same time, by opening the bottom drain valve to form a vertically connected seepage channel, the mud permeates the saturated sand layer from top to bottom under the drive of pressure difference, simulating the physical process of pressurized mud acting on the excavation face and seeping into the strata during actual slurry-water balance shield tunneling. During this seepage process, bentonite particles in the mud continuously deposit in the pores of the soil, thereby forming a water-resistant or slightly permeable mud film on the surface of the sand layer. With the help of this mud film, the mud-water pressure can be effectively established, achieving a balance between water and soil pressure and ensuring the stability of the excavation face.

[0034] In this embodiment, air compressor 2 is activated to output compressed gas. Air compressor 2 provides an adjustable compressed gas source to pressurize the air above the slurry inside the acrylic cylinder 1. The output port of air compressor 2 is threadedly connected to the inlet end of a pressure regulating valve via an air passage pipe. The pressure regulating valve is located at the output end of air compressor 2 and is used to stably regulate the supply pressure to the test set value. The outlet end of the pressure regulating valve passes through the top glass sealing plate and connects to the upper space of the inner cavity of the acrylic cylinder 1 via an air passage pipe. After the compressed gas is regulated to the target pressure value by the pressure regulating valve, it enters the upper space of the inner cavity of the acrylic cylinder 1 through the air passage, applying uniform downward air pressure to the surface of the slurry. During the air pressure loading process, the pressure sensing end of pressure gauge 3 is inserted through the top glass sealing plate and into the inner cavity of the acrylic cylinder 1 to measure the gas pressure above the slurry in the inner cavity of the acrylic cylinder 1 in real time and provide feedback to the operator. Based on the real-time reading of pressure gauge 3, the operator adjusts the air pressure regulating valve to ensure that the gas pressure inside the chamber is maintained stably at the target pressure value, avoiding pressure fluctuations that could affect the mud infiltration process and the quality of mud film formation. Air pressure loading must be completed and stabilized before the drain valve is opened to ensure that a stable pressure boundary condition is established above the mud layer at the moment the drain valve is opened.

[0035] In this embodiment, after the gas pressure inside the cavity is stabilized at the target pressure value, the drain valve is then opened. The drain valve is installed at the drain channel outlet of the bottom glass sealing plate and is threadedly connected to the bottom glass sealing plate to control the flow of the permeate. After the drain valve is opened, the bottom of the inner cavity of the plexiglass cylinder 1 is connected to the external atmosphere. The pressure at the gravel filter layer at the bottom of the sand layer is close to atmospheric pressure, while the mud layer is subjected to the gas pressure of the target pressure value applied by the air compressor 2. This creates a vertical pressure gradient in the inner cavity of the plexiglass cylinder 1. Under the action of this pressure difference, the mud permeates the sand layer from top to bottom. The liquid phase components (water and fine particle suspension) in the mud flow downward through the pores of the sand layer driven by the pressure gradient, while bentonite particles in the mud with a particle size larger than the throat size of the sand layer pores are trapped on the surface of the sand layer and in the shallow pores. As the infiltration process continues, bentonite particles gradually deposit and accumulate on the surface of the sand layer, forming a dense, thin layer structure with significantly lower permeability than the undisturbed sand layer—the mud film. The mud film refers to the low-permeability, dense thin layer formed by the gradual deposition and accumulation of bentonite particles in the mud on the sand layer surface and in its shallow pores during pressurized infiltration. This thin layer effectively blocks or significantly reduces further infiltration of the mud liquid phase, concentrating the mud pressure on the mud film surface and transmitting it to the sand layer's structural framework. The formation of the mud film is a dynamic evolutionary process. In the early stages of infiltration, an effective mud film barrier has not yet formed on the sand layer surface, resulting in a high mud infiltration rate and significant filtration loss. As bentonite particles continue to deposit, the mud film gradually thickens and becomes denser, its infiltration resistance continuously increases, the mud infiltration rate decreases accordingly, and the filtration loss gradually diminishes. When the mud film thickness and density reach a certain level, the infiltration rate approaches zero, and the mud film formation process becomes stable. After passing through the sand and gravel filter layers, the permeate reaches the drainage channel of the bottom glass sealing plate, flows out through the drain valve, and then enters the measuring cylinder 5. The measuring cylinder 5 is positioned below the drain valve, with the outlet end of the drain valve vertically aligned with the opening of the measuring cylinder 5, so that the permeate flows into the measuring cylinder 5 after being discharged through the drain valve. The operator records the cumulative drainage volume by reading the scale of the measuring cylinder 5. The curve of the cumulative drainage volume over time reflects the filtration rate of the mud and its changing trend, providing a quantitative basis for judging the formation process of the mud film.

[0036] In this embodiment, during the pressurized infiltration process, a pore water pressure gauge 4 continuously monitors the changes in pore water pressure within the medium of the plexiglass cylinder 1. The sensing probe of the pore water pressure gauge 4 penetrates the side wall of the plexiglass cylinder 1 and extends into the test medium, directly contacting the pore water within the plexiglass cylinder 1. This allows for real-time monitoring of pore water pressure changes within the medium of the plexiglass cylinder 1 during the experiment. During infiltration, the mud penetrates the sand layer from top to bottom and gradually forms a mud film, causing dynamic changes in the pore water pressure within the medium. In the initial stage of infiltration, the mud film has not yet formed or is not yet dense, allowing the mud pressure to be transmitted relatively smoothly to the depths of the sand layer, resulting in relatively high pore water pressure at various depths within the sand layer. As the mud film gradually forms and becomes denser, its pressure-blocking effect increases, and the pore water pressure within the sand layer below the mud film gradually decreases and stabilizes. The pore water pressure gauge 4 captures this change process in real time through direct contact between the sensing probe and the pore water, synchronously recording the change in pore water pressure over time, providing mechanical parameter data for analyzing the seepage characteristics during mud film formation. Simultaneously, the pressure gauge 3 continuously measures the gas pressure above the mud in the inner cavity of the plexiglass cylinder 1, ensuring that the pressurization conditions remain stable throughout the infiltration process. Operators simultaneously record the changes in pore water pressure and drainage volume over time. This synchronous monitoring data of multiple parameters provides a complete mechanical and hydraulic parameter basis for subsequent analysis of the mud film formation mechanism, evaluation of mud film quality, and establishment of a quantitative correlation between mud film physical properties and radar signal characteristics.

[0037] In this embodiment, during the continuous infiltration process, the operator judges whether the mud film has formed and stabilized based on the changing trends of the pore water pressure gauge 4 and the cumulative drainage volume of the measuring cylinder 5. When both the reading of the pore water pressure gauge 4 and the drainage volume of the measuring cylinder 5 stabilize and the mud film no longer changes significantly, it indicates that the mud film has fully formed and the infiltration process has reached equilibrium. At this point, the drainage valve is closed, the air compressor 2 is stopped, and the pressurized infiltration and mud film formation steps are completed. After the test, the remaining slurry above the sand layer is carefully removed, and an external mud film can be seen. For subsequent testing, an acrylic cylinder with an inner diameter of 2 cm and a height of 10 cm is vertically inserted into the sand layer to obtain the formation after mud infiltration for actual verification.

[0038] In one specific embodiment of this application, after the sand layer is saturated and approximately 1L of mud is injected, and the top glass sealing plate is installed, the ground-penetrating radar detection system 8 is activated for real-time monitoring. The air compressor 2 is turned on, and the air pressure above the slurry is stabilized at 50kPa via the air pressure regulating valve. Then, the bottom drain valve is opened, and the mud begins to penetrate downwards into the sand layer. During this process, the changes in pore water pressure and drainage volume over time are recorded simultaneously. Throughout the experiment, the ground-penetrating radar detection system 8 continuously monitors the stratification changes of the soil medium within the permeation column test device. Based on real-time radar data, key signal change locations can also be manually marked and recorded, facilitating subsequent data processing and analysis. After 5 minutes of continuous penetration, when the mud film no longer shows significant changes, the bottom drain valve is closed, and the experiment ends. The ground-penetrating radar detection system 8 is shut down, and the radar data for the entire experimental process is saved, along with pore water pressure data and device drainage volume. The industrial control computer saves all raw radar data, dielectric constant inversion images, pore water pressure records, and cumulative drainage volume data for the entire experimental process for subsequent analysis.

[0039] S3. Simultaneously with pressurization, the ground-penetrating radar (GPR) detection system is activated. This system receives and processes the echo signals to determine the actual thickness of the mud film. Essentially, during pressurized infiltration and mud film formation, the GPR detection system 8 performs real-time electromagnetic wave detection on the interface between different dielectric constants within the plexiglass cylinder 1. By receiving reflected echo signals generated at interfaces with different dielectric constants and performing a series of processing and inversion operations on these signals, the actual thickness of the mud film is obtained. This step directly couples the physical process of mud film formation with the radar detection process, enabling real-time, in-situ, non-destructive observation during the dynamic formation of the mud film. It directly acquires the raw radar response data of this specific target, completely recording the electromagnetic response characteristics of the mud film from initial deposition to stable film formation. Throughout the experiment, the GPR detection system 8 continuously monitors the changes in soil stratification within the infiltration column test device. Based on real-time radar data, it can also manually mark and record the locations of key signal changes, facilitating subsequent data processing and analysis.

[0040] In this embodiment, the operator starts the radar host simultaneously with pressurization. The transceiver control unit inside the radar host drives the radar antenna embedded in the mounting base to continuously transmit electromagnetic pulses towards the dielectric layering interface within the plexiglass cylinder 1 according to a preset timing sequence. The preset timing sequence refers to the repetition frequency of the transmitted electromagnetic pulses and the time parameters of the receiving window, pre-set in the transceiver control unit. These parameters are determined based on the type and center frequency of the selected radar antenna. The ground-penetrating radar antenna is replaceable, with two types available: pulsed radar antennas and stepped radar antennas. Pulsed antennas offer two different frequencies: 1600MHz and 2600MHz, while stepped radar antennas also offer two different center frequencies: 1500MHz and 2500MHz. The radar antenna is embedded in the mounting base, and the mounting bracket is fixedly connected to the outer surface of the top glass sealing plate by threaded fasteners. The central area of ​​the mounting bracket is recessed inward to form the mounting base. The transmitting surface of the radar antenna fits seamlessly with the outer surface of the top glass sealing plate, and the central normal direction of the radar antenna is perpendicular to the dielectric layering interface within the plexiglass cylinder 1. During installation, the antenna of the ground-penetrating radar detection system 8 requires laser calibration to ensure that the antenna's emitting surface faces the layered interface within the soil column, and that its central normal is as perpendicular as possible to the layered interface to guarantee perpendicular electromagnetic wave incidence. A waterproof and wave-permeable shell covers the outer periphery and exposed interfaces of the radar antenna, sealing it in place with the mounting base. A wear-resistant protective layer is attached to the outer surface of the waterproof and wave-permeable shell, covering the entire exposed area. After the electromagnetic wave passes through the top glass sealing plate and enters the test medium, it is reflected at the dielectric constant abrupt interface between the mud layer, mud film, permeable zone, and undisturbed sand layer. The radar antenna receives the reflected echo signal. After the radar host performs preliminary sampling of the echo signal, the digitized raw echo data is transmitted in real time to the industrial control computer via transmission cable 7. One end of transmission cable 7 is electrically connected to the signal output port of the radar host, and the other end is electrically connected to the signal input port of the industrial control computer. The middle section of transmission cable 7 is pre-embedded in a cable groove inside the mounting bracket and leads out to external equipment via the outside of the mounting bracket.

[0041] In one embodiment of this application, such as Figure 4 As shown, the ground-penetrating radar detection system receives and processes the echo signal to obtain the actual thickness of the mud film, including: S31, removing the DC component of the echo signal, bandpass filtering, eliminating background interference, and resampling the signal to a standardized size to obtain a two-dimensional radar image reflecting the changes in the stratification of the medium inside the plexiglass cylinder.

[0042] In this embodiment, DC component removal refers to eliminating the zero-frequency component in the radar echo signal caused by the DC bias of electronic devices, bringing the signal baseline to zero and avoiding interference from DC offset in subsequent processing. Bandpass filtering refers to applying a bandpass filter to the echo signal, retaining only the signal components within the effective frequency band matching the center frequency of the selected radar antenna, filtering out high-frequency noise and low-frequency interference. Background interference elimination refers to removing fixed background signals generated by factors such as reflection from the wall of the plexiglass cylinder 1 and antenna coupling, highlighting the effective reflected signals from the internal layering interface of the medium. Signal resampling to normalized size refers to resampling the preprocessed echo signal on the time and amplitude axes to ensure that its data dimensions are consistent with the dataset size used by the deep learning inversion module during training, ensuring that the real-time acquired data can be correctly received and processed by the pre-trained model. The signal preprocessing unit stitches the preprocessed radar signals in the order of acquisition time to synthesize a two-dimensional radar image reflecting the changes in the layering of the medium within the plexiglass cylinder 1, and outputs the two-dimensional radar image to the deep learning inversion module. The horizontal axis of a two-dimensional radar image represents the time series of signal acquisition, and the vertical axis represents the two-way travel time of electromagnetic waves (corresponding to the depth direction of the medium). The grayscale or color values ​​in the image represent the amplitude intensity of the echo signal at each depth location. Strong reflection signals at interfaces with abrupt changes in dielectric constant appear as bright stripes extending along the horizontal axis in a two-dimensional radar image.

[0043] S32, the two-dimensional radar image is output to the deep learning inversion module to obtain a layered structure image containing the spatial distribution of dielectric constant. In this embodiment, the core of the deep learning inversion module is a pre-trained supervised deep learning inversion model. Structurally, this deep learning inversion model mainly comprises an input layer, a hidden layer mapping network, and an output layer connected sequentially. The spatial dimension of the input layer matches the two-dimensional radar image resampled to a standardized size, and is used to receive radar matrix data reflecting electromagnetic wave propagation characteristics. The hidden layer mapping network, as the core feature transformation structure of the model, adopts an encoder-decoder architecture. Its front-end feature extraction layer is responsible for extracting deep abstract features representing electromagnetic wave reflection amplitude, two-way travel time, and waveform dispersion changes from the two-dimensional radar image layer by layer. The back-end feature reconstruction layer is responsible for upsampling and spatially restoring the above-mentioned abstract features based on electromagnetic response, mapping them to dielectric constant features with actual physical meaning. The output layer receives the computational output of the hidden layer mapping network and reassembles it into a pixel matrix corresponding to the actual physical detection space, thereby obtaining a layered structure image containing the predicted dielectric constant values ​​at each depth location.

[0044] After receiving the two-dimensional radar image, the deep learning inversion module calls a pre-trained supervised deep learning model with the aforementioned network structure to perform the inversion operation. The training process of this pre-trained model is as follows: A numerical geological model containing mud film, permeable zone, and undisturbed soil structure is constructed in advance based on the medium distribution within the permeable column device. When constructing the numerical geological model, the thickness variation range of the mud film is pre-defined, and interference consistent with the actual environmental noise characteristics is injected into the simulation signal. A large number of synthetic radar datasets with dielectric constant labels are generated through forward modeling. Supervised training is performed using synthetic radar images as input and dielectric constant images as labels, enabling the model to learn the mapping relationship from radar images to dielectric constant images.

[0045] Specifically, forward modeling refers to calculating the echo signal waveform that a radar antenna should receive in the known dielectric structure and dielectric constant parameters of each layer using electromagnetic wave propagation theory, thereby generating synthetic radar data that corresponds one-to-one with the known dielectric constant distribution. Supervised training involves using synthetic radar images as input samples for the model's input layer and the corresponding dielectric constant images as known labels. By minimizing the error between the predicted output of the model's output layer and the known labels, the backpropagation algorithm iteratively optimizes the weights and bias parameters within the hidden layer mapping network, allowing the model to gradually learn the mapping relationship from radar images to dielectric constant images. When real-time two-dimensional radar images are input into the deep learning inversion module, the module outputs a layered structure image containing the spatial distribution of dielectric constants, and then transmits the inversion results to the result display unit.

[0046] The results display unit presents a real-time image of the dielectric constant layer structure on the industrial control computer screen, allowing operators to interpret the mud film status. The layer structure image of the spatial distribution of the dielectric constant refers to a two-dimensional image with the depth direction as the vertical axis and the acquisition time series as the horizontal axis. The value of each pixel in the image represents the predicted value of the dielectric constant of the medium at the corresponding depth location. Different dielectric constant values ​​are presented in different colors or grayscale levels, making the layered structure of the mud layer, mud film, permeable zone, and undisturbed sand layer visible in the image due to differences in dielectric constant.

[0047] S33, Based on the layered structure image containing the spatial distribution of dielectric constants, the actual thickness of the mud film is determined. In the first embodiment of this application, after the deep learning inversion module outputs the layered structure image containing the spatial distribution of dielectric constants, it delineates the location of the mud film according to the difference in dielectric constant values ​​of each layered region. Then, it counts the pixel height occupied by the region determined to be the mud film in the vertical direction and multiplies it by a fixed pixel-to-actual-scale scaling factor pre-calibrated by the system to linearly calculate the actual thickness of the mud film.

[0048] Specifically, in the dielectric constant prediction map of the analysis medium, the mud film region, due to the dense deposition of bentonite particles displacing pore water and resulting in reduced water content, has a significantly lower dielectric constant than the upper mud layer and the lower saturated sand layer. Therefore, in the dielectric constant image, the mud film region appears as a band with a significantly lower dielectric constant. Operators or automated algorithms, based on the difference in dielectric constant values, identify continuous regions with dielectric constants below a preset threshold as mud film regions. For these identified mud film regions, their pixel height H in the vertical direction (depth direction) is calculated. The pixel-to-actual-scale scaling factor is a fixed conversion factor determined during the system calibration phase. It is calculated by comparing the number of pixels occupied by a standard sample of known physical size in the radar image with its known physical size during radar detection in a permeation column test apparatus. The unit is mm / pixel. Based on the pre-calibrated pixel and actual-scale scaling factor, the actual thickness of the mud film, T = H × scaling factor (mm), can be obtained. For example, if the system's pixel-to-actual-scale scaling factor is 0.1 mm / pixel, and the area identified as mud film occupies a vertical pixel height H of 50 pixels, then the actual thickness of the mud film T = 50 × 0.1 = 5 mm. This implementation method is simple to calculate and fast to process, making it suitable for rapid evaluation scenarios where the accuracy of mud film thickness is not critical.

[0049] However, research has shown that the first embodiment described above has three interrelated structural defects in the real physical scenario of a complex permeation column test.

[0050] First, the permeation column test simulates the physical process of pressurized slurry permeating into the sand layer and forming a mud film during slurry-balanced shield tunneling. In this process, bentonite particles gradually deposit from the slurry side into the depth of the sand layer. The interface between the mud film and the underlying permeation zone is not a sharp, abrupt cut, but rather an objectively existing region where the bentonite particle concentration gradually transitions from high to low. Directly corresponding to the bentonite particle concentration gradient is the water content gradient. The water content in the mud film region is significantly lower than that of the saturated sand layer due to dense blockage, while the water content in the permeation zone region is between the two. These three regions form a continuous but non-linear water content variation curve along the depth direction. The dielectric constant, as a function of water content, also exhibits a gradual rather than abrupt distribution along the depth direction. The first embodiment only uses hard thresholding based on the dielectric constant image to identify the location of the mud film, essentially forcibly discretizing the continuous, gradual physical process into a binary classification with no clear distinction. This approach cannot accurately capture the true location of the dramatic change in water content at the mud film boundary. The transition area is either misclassified as part of the mud film area, leading to an overestimation of its thickness, or excluded from the mud film area, leading to an underestimation of its thickness. The boundary positioning deviation is systematic and cannot be eliminated by simple parameter adjustments.

[0051] Secondly, the echo signals received by ground-penetrating radar essentially record the two-way travel time of electromagnetic waves propagating in the medium, rather than the direct spatial distance. The propagation speed of electromagnetic waves varies significantly in media with different dielectric constants; the higher the dielectric constant, the slower the propagation speed, and the longer the travel time for the same physical thickness, resulting in a larger pixel height on the radar image. In the permeable column test, the mud film region experiences reduced water content due to the dense deposition of bentonite displacing pore water, leading to a significant decrease in dielectric constant compared to the saturated sand layer. Consequently, the propagation speed of electromagnetic waves in the mud film region is faster than in the saturated sand layer. The first embodiment uses a single fixed scaling factor to linearly convert pixel height to physical thickness, implicitly assuming that electromagnetic waves propagate at a uniform speed throughout the entire depth of the permeable column. This assumption clearly does not hold true in the layered medium system of mud film-permeable zone-undisturbed sand layer. Fixed scaling factors are typically calibrated based on the propagation speed of a specific medium (e.g., saturated sand layer). When used to calculate the dielectric constant of a mud film region that significantly deviates from the calibration medium, it inevitably introduces a non-negligible systematic error.

[0052] Finally, from the dielectric constant image output by the deep learning inversion module to the final mud film thickness value, all processing steps in the first embodiment are purely geometric operations—pixel counting and linear scaling—without any step introducing a physical model to verify the reasonableness of the intermediate or final results. When the deep learning model encounters abnormal conditions not fully covered by the training dataset (such as unexpected phenomena like bubble aggregation or mud segregation during the experiment), it may output a dielectric constant prediction value that deviates from physical laws, and the first embodiment has no ability to identify or correct such anomalies.

[0053] To address the aforementioned shortcomings, and focusing on the nonlinear coupling relationship between dielectric constant and water content in a permeation column test scenario, this application proposes a second embodiment, specifically, as follows: Figure 5 As shown, in the second embodiment, determining the actual thickness of the mud film based on a layered structure image containing the spatial distribution of dielectric constants includes: S331, a nonlinear inversion of the water content profile is performed on the spatial distribution of the dielectric constant to obtain the water content depth profile. It should be understood that although the dielectric constant depth profile output by the deep learning inversion module reflects the electromagnetic response characteristics of the medium at each depth in a numerical sense, the dielectric constant itself does not directly correspond to the core physical variable in the mud film formation process: volumetric water content. In the bentonite-sand mixed medium system involved in the permeation column test, the formation of the mud film is essentially a process of bentonite particles depositing and blocking in the pores of the sand layer. The direct physical consequence of this process is a drastic non-uniform change in pore water content along the depth direction. Only by converting the dielectric constant to water content can subsequent steps identify the gradient abrupt change characteristics at the mud film boundary in the space of water content, which has a clear physical meaning. Therefore, it is necessary to first nonlinearly map the dielectric constant depth profile to the water content depth profile. The volumetric water content refers to the ratio of the volume occupied by water per unit volume of soil to the total volume of the soil, ranging from 0 to 1, and is a fundamental physical quantity characterizing the water content state of the soil.

[0054] For the dielectric constant depth profile output by the deep learning inversion module, water content inversion is performed by applying an inverse transformation of the improved Topp empirical equation at each depth location. The classical Topp equation establishes a cubic polynomial relationship between soil dielectric constant and volumetric water content, but its calibration is for general mineral soils and does not consider the additional polarization effect of high specific surface area clay minerals such as bentonite. In the permeation column test, bentonite particles are highly enriched in the mud film region, and the additional dielectric response generated by the bound water between bentonite layers cannot be ignored. Therefore, a bentonite correction term is introduced into the classical Topp equation to construct a modified Topp equation suitable for the experimental medium system: In the formula, For depth position The dielectric constant at that point, Let z be the volumetric water content at depth z. This is a bentonite correction factor. Correction term. The physical meaning is as follows: The high specific surface area of ​​bentonite particles causes the bound water layer adsorbed on their surface to generate additional polarization under the action of an external electromagnetic field. The intensity of this polarization effect first increases and then decreases with the increase of water content. In the low water content region (i.e., the mud film region where bentonite is highly enriched), the proportion of bound water to the total water is large, and the additional polarization effect is significant. In the high water content region (i.e., the saturated sand layer where bentonite is scarce), free water dominates the dielectric response, and the additional polarization effect tends to be weak. The specific values ​​are determined through a calibration test of the dielectric constant of the bentonite-sand mixture used in the test under known moisture content conditions before the permeation column test. For example, in the permeation column test of this application, the bentonite-sand mixture system used is calibrated, and the bentonite correction factor is determined by the calibration test. Typical values ​​range from 4.5 to 8.0, with the specific value depending on the mineral composition and blending ratio of the bentonite. When the bentonite blending ratio is high... Take the larger value when the proportion of bentonite added is low. Take the smaller value.

[0055] For each depth position The above modified Topp equation is considered as about The cubic equation, with input Given the quantities, solve the equations using the numerical root-finding method, taking the values ​​that satisfy the physical constraints. The real root is used as the water content inversion result at that depth. After traversing all depth locations, a complete water content depth profile is obtained. After this stage of processing, the dielectric constant depth profile output by the deep learning inversion module is transformed into a water content depth profile with clear physical semantics, and the transformation process is achieved through a bentonite correction factor. The introduction of this method is specifically adapted to the unique electromagnetic response characteristics of the bentonite-sand mixture in the permeation column test, laying a physical foundation for the subsequent accurate positioning of the mud film boundary in the water content domain. At the same time, the modified Topp equation itself constitutes a layer of physical constraint. If the dielectric constant value at a certain depth position output by the deep learning model causes the equation to have no real roots in the interval [0, 1], it indicates that the predicted value deviates from the physical law, and the system can mark the anomaly and trigger a review accordingly.

[0056] S332, adaptive detection of the mud film boundary based on abrupt changes in water cut gradient is performed on the water cut depth profile to obtain the depth positions of the upper and lower boundaries of the mud film. It should be understood that the water cut depth profile... It contains complete information about the gradual transition of water content between the mud film region and the adjacent medium. The upper boundary (the boundary between the mud layer and the mud film) and the lower boundary (the boundary between the mud film and the permeable zone) correspond to the two spatial locations where the water content changes most drastically along the depth direction. At these two locations, the rate of change of water content is much higher than that in the mud film interior and exterior regions. This characteristic allows the problem of identifying the mud film boundary to be transformed into the problem of detecting water content gradient energy anomalies. Detecting gradient abrupt changes in the water content domain has a direct physical correspondence; the abrupt change in the water content gradient directly reflects the jump in the deposition density of bentonite particles, which is precisely the physical essence of the mud film boundary.

[0057] First, the water content depth profile The first-order gradient along the depth direction is calculated as follows: In the formula, For depth position The moisture content gradient at that location, is the discrete sampling interval in the depth direction. The larger the absolute value, the more drastic the change in water content at that depth.

[0058] Subsequently, to overcome the problem of insufficient adaptability of the fixed threshold to different experimental conditions (different mud mix ratios, different sand layer densities, and different pressurization values), a normalized gradient energy density function was constructed, and the detection threshold was adaptively generated based on the full-profile gradient statistical characteristics. The normalized gradient energy density function is defined as: In the formula, For depth position The normalized gradient energy density at point L is the total depth range covered by the water content depth profile, and the denominator term is... This represents the spatial mean of the gradient energy across the entire profile. This characterizes the ratio of the gradient energy at depth z to the average gradient energy level across the entire profile. At the upper and lower boundaries of the mud film, the water cut undergoes a dramatic jump. The water content will be significantly higher than the average of the entire profile; however, in areas with relatively uniform water content, such as inside the mud film and inside the saturated sand layer, the water content will be significantly higher. It will be close to or below the mean.

[0059] The adaptive detection threshold is defined as: In the formula, for The arithmetic mean across the entire cross-section. for Standard deviation over the entire cross-section This is the threshold sensitivity coefficient, typically ranging from 2.0 to 3.5. The value of is used to balance the sensitivity and noise immunity of mud film boundary detection. Taking a smaller value increases detection sensitivity but may misinterpret gradient fluctuations caused by noise as mud film boundaries. A larger value enhances noise immunity but may miss thin mud films with weak boundary signals. For example, in the permeability column test of this application, when the test medium is medium-coarse sand and the mud mix ratio is conventional, the threshold sensitivity coefficient... A typical value for λ is 2.5, which allows for effective detection of the mud film boundary while maintaining good noise immunity. When the test medium is fine sand and the mud film is expected to be thin, λ can be reduced to 2.0 to improve detection sensitivity. When electromagnetic interference is strong in the test environment, λ can be increased to 3.0 to 3.5 to enhance noise immunity. and All of these are determined by the statistical characteristics of the data from the current test, and can be automatically adjusted according to different test conditions without the need for manual calibration.

[0060] Obtaining an adaptive threshold After that, it will satisfy The shallowest point in the set of depth locations is marked as the upper boundary of the mud film, and the deepest point is marked as the lower boundary of the mud film, as shown below: In the formula, This refers to the depth of the upper boundary of the mud film. This indicates the depth of the lower boundary of the mud film.

[0061] After this step, the upper and lower boundaries of the mud film no longer rely on fixed thresholds set by human experience in the dielectric constant domain, but are adaptively determined by the full-profile statistical characteristics of the water content gradient energy density. The physical meaning of the upper and lower boundaries of the mud film is precisely anchored to the spatial location where the water content undergoes a drastic change, which strictly corresponds to the physical phenomenon of the bentonite particle deposition density changing rapidly at the edge of the mud film. The boundary positioning results obtained in this way have working condition adaptability, eliminating the defect of systematic misjudgment of the transition region caused by the hard threshold segmentation in the first embodiment.

[0062] S333, based on the spatial distribution of dielectric constant, performs adaptive integral estimation of mud film thickness at the depth positions of the upper and lower boundaries of the mud film to obtain the actual thickness of the mud film. It should be understood that the depth position of the upper boundary of the mud film... and the depth of the lower boundary of the mud film These are all pixel-domain depth coordinates, which need to be converted to the actual spatial distance in the physical domain to obtain the actual thickness of the mud film. In the layered medium system of the permeation column test, the propagation speed of electromagnetic waves in each layer varies significantly due to the different dielectric constants. Specifically, the propagation speed of electromagnetic waves at any depth z satisfies the following relationship with the dielectric constant at that location: In the formula, For electromagnetic waves at depth position Propagation speed in the medium The speed of electromagnetic waves in a vacuum (approximately 3 × 10⁻⁶) 8 m / s For depth position The dielectric constant at the location. In the mud film region, the dense deposition of bentonite displaces pore water, reducing the water content and lowering the dielectric constant compared to the adjacent saturated sand layer, resulting in a correspondingly faster electromagnetic wave propagation speed.

[0063] The dielectric constant image output by the deep learning inversion module is based on a certain reference propagation speed (corresponding to the reference dielectric constant). The result of mapping the travel time axis to the depth axis is shown below. When the actual dielectric constant deviates from the reference value, there is a non-linear deviation between the pixel domain depth coordinates and the true depth of the physical domain. The reference dielectric constant is... This refers to the uniform dielectric constant assumption used when converting the two-way travel time axis of the radar signal to the depth axis during the training phase of the deep learning inversion model. This value is typically taken as the dielectric constant of the saturated sand layer occupying the main volume in the test setup. For example, in the permeation column test of this application, the typical dielectric constant of the saturated sand layer is approximately 20 to 25, with a reference dielectric constant... A value of 22 is acceptable. Pixel domain depth With the true depth of the physical domain The pointwise mapping relationship between them is as follows: In the formula, Let z be the physical domain depth corresponding to the pixel domain depth z. This is the reference dielectric constant used during the training of the deep learning inversion model. pixel domain depth The actual dielectric constant at that point. Mapping factor. The physical meaning is: when the actual dielectric constant at a certain depth is lower than the reference value (such as in the mud film area), the actual propagation speed of electromagnetic waves is faster than the reference speed, the physical distance corresponding to the same travel time is greater than the pixel domain distance, the mapping factor is greater than 1, and the physical domain depth is stretched; conversely, it is compressed.

[0064] Therefore, the actual physical thickness of the mud film The difference in actual depth between the upper and lower boundaries of the mud film in the physical domain: In the formula, This represents the actual physical thickness of the mud film.

[0065] In the engineering implementation, the depth direction is sampled discretely, and the mud film interval is set as […]. , The [sample] contains N discrete sampling points, and the dielectric constant of each sampling point is [value]. (i=1,2,…,N), the distance between adjacent sampling points is Then the discretization engineering calculation formula for the above integral is: In the formula, This represents the total number of discrete sampling points within the mud film area. The depth-direction discrete sampling interval. Let be the dielectric constant at the i-th sampling point. The reference dielectric constant is used. Summation term. A point-by-point correction based on the local dielectric constant was applied to the pixel spacing of each sampling point within the mud film region. When the dielectric constant at a sampling point is lower than the reference value (a typical characteristic of dense mud film regions), Greater than The physical thickness increment contributed by this point is greater than its pixel pitch. This corrects the systematic underestimation of such regions by the fixed scaling factor in the first embodiment; conversely, it avoids overestimation when the dielectric constant is higher than the reference value.

[0066] After this step, the calculation of mud film thickness no longer relies on a pre-calibrated fixed scaling factor, but is driven point-by-point by the actual dielectric constant at each depth position within the mud film range. The accelerated propagation effect of electromagnetic waves in the mud film, a dense medium with low water content, is accurately captured and compensated by the variable propagation model. The nonlinear deviation between the pixel-domain geometric distance and the physical domain true distance is eliminated point-by-point during the integration process, and the final mud film thickness has a strict physical meaning.

[0067] The second embodiment described above uses the nonlinear coupling relationship between dielectric constant and water content in the permeation column test scenario as its physical core, systematically eliminating the defects of the first embodiment from three levels and achieving the improvement of the technical objectives.

[0068] At the mud film boundary location level, the nonlinear inversion of the water content profile (step S331) transforms the dielectric constant, an electromagnetic parameter, into water content, a physical variable directly corresponding to the mud film formation process. This allows the dramatic water content jump caused by the abrupt change in bentonite particle deposition density at the mud film boundary to be clearly presented in the water content domain. The gradient energy density adaptive detection (step S332) transforms the mud film boundary identification from the hard threshold segmentation of the dielectric constant in the first embodiment to the detection of water content gradient energy anomalies. The detection threshold is generated driven by the statistical characteristics of each test data. It can be adaptively adjusted under different mud mix ratios, different sand layer densities, and different pressurization conditions, eliminating the systematic boundary deviation and insufficient universality of working conditions caused by setting a fixed threshold based on human experience.

[0069] At the mud film thickness calculation level, the adaptive integral calculation under the variable-speed propagation model (step S333) constructs a velocity correction factor point by point using the actual dielectric constant at each depth position within the mud film interval, adaptively mapping the geometric distance of the pixel domain to the true distance of the physical domain. The nonlinear deviation between the pixel domain thickness and the physical domain thickness caused by the accelerated propagation of electromagnetic waves in the dense region of the mud film is accurately compensated for point by point during the integration process, completely replacing the uniform propagation assumption implicit in the linear conversion with a fixed scaling factor in the first embodiment, and eliminating the systematic thickness calculation error introduced by this assumption.

[0070] At the level of physical constraint verification, the modified Topp equation introduced in step S331 naturally constitutes a physical rationality check while performing water content inversion. If the deep learning model outputs a dielectric constant prediction value that deviates from the physical law under abnormal operating conditions, the modified Topp equation will not be able to obtain a real root in the physically meaningful water content range [0, 1]. The system can automatically mark abnormal data points and trigger a verification mechanism accordingly, thus making up for the weak link of the complete lack of physical constraint verification between the model output and the final result in the first embodiment.

[0071] In summary, the second embodiment of this application replaces the pure geometric mapping link of dielectric constant → hard threshold segmentation → fixed coefficient linear scaling in the first embodiment with a physical driving link of dielectric constant → water content → gradient mutation detection → variable speed correction integration. This simultaneously improves the accuracy and reliability of mud film identification and thickness quantification in permeable column tests from three dimensions: boundary positioning accuracy, thickness calculation accuracy, and robustness under abnormal working conditions. It provides a more accurate data processing foundation for real-time, non-destructive, and intelligent monitoring of mud film status in slurry balance shield tunnels.

[0072] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A permeability column test device for identifying mud film in slurry balance shield tunneling machines, characterized in that, include: Acrylic glass cylinders, air compressors, pressure gauges, pore water pressure gauges, and ground-penetrating radar detection systems; The acrylic cylinder forms a closed test chamber; the output port of the air compressor is connected to the upper space of the inner cavity of the acrylic cylinder through an air pipeline and an air pressure regulating valve; the pressure sensing end of the pressure gauge extends into the inner cavity of the acrylic cylinder; the sensing probe of the pore water pressure gauge passes through the side wall of the acrylic cylinder and extends into the test medium, and is in direct contact with the pore water inside the acrylic cylinder. The ground-penetrating radar detection system is installed at a preset position and includes a radar antenna. The emitting surface of the radar antenna is in close contact with the top surface of the plexiglass cylinder without gaps, and the central normal direction of the radar antenna is perpendicular to the medium layering interface inside the plexiglass cylinder. The permeable column test device for mud film identification in slurry balance shield tunneling also includes a mounting bracket, a waterproof and wave-permeable cover, and a wear-resistant protective layer. The mounting bracket is fixedly connected to the outer surface of the top glass sealing plate by threaded fasteners. The central area of ​​the mounting bracket is recessed inward to form a mounting base. The radar antenna is embedded in the mounting base, and the emitting surface of the radar antenna is in close contact with the outer surface of the top glass sealing plate without gaps. The central normal direction of the radar antenna is perpendicular to the medium layering interface inside the plexiglass cylinder. The waterproof and wave-permeable cover covers the outer peripheral surface and exposed interface of the radar antenna and is sealed with the mounting base. The wear-resistant protective layer is attached to the outer surface of the waterproof and wave-permeable cover, covering the entire exposed area of ​​the waterproof and wave-permeable cover.

2. The permeability column test device for mud film identification in slurry balance shield tunneling according to claim 1, characterized in that, The acrylic cylinder includes an acrylic cylinder body, a top glass sealing plate, and a bottom glass sealing plate. The top glass sealing plate is fixedly connected to the top opening of the acrylic cylinder in a sealed manner, and the bottom glass sealing plate is fixedly connected to the bottom opening of the acrylic cylinder in a sealed manner. The acrylic cylinder body, the top glass sealing plate, and the bottom glass sealing plate form a closed test chamber.

3. The permeability column test device for mud film identification in slurry balance shield tunneling according to claim 2, characterized in that, The output port of the air compressor is threadedly connected to the inlet end of the air pressure regulating valve through an air pipeline; the outlet end of the air pressure regulating valve passes through the top glass sealing plate through an air pipeline and connects to the upper space of the inner cavity of the plexiglass cylinder.

4. The permeability column test device for mud film identification in slurry balance shield tunneling according to claim 2, characterized in that, It also includes a drain valve and a measuring cylinder. The drain valve is installed at the drain channel outlet of the bottom glass sealing plate and is threaded to the bottom glass sealing plate. The measuring cylinder is set below the drain valve, with the outlet end of the drain valve vertically aligned with the mouth of the measuring cylinder, so that the permeate flows into the measuring cylinder after being discharged through the drain valve.

5. The permeability column test device for mud film identification in slurry balance shield tunneling according to claim 4, characterized in that, It also includes transmission cables and an industrial control computer. One end of the transmission cable is electrically connected to the signal output port of the radar host, and the other end of the transmission cable is electrically connected to the signal input port of the industrial control computer.

6. The permeability column test device for mud film identification in slurry balance shield tunneling according to claim 5, characterized in that, The middle section of the transmission cable is embedded in the cable tray inside the mounting bracket and led out to the external equipment through the outside of the mounting bracket.

7. A working method for a permeability column test device for identifying mud film in a slurry-water balance shield tunnel, characterized in that, include: Gravel filter layer and layered compacted sand layer are sequentially filled into the interior of the plexiglass cylinder. Then, the drain valve is closed, and water is slowly injected in reverse through the drainage channel of the bottom glass sealing plate to fully saturate the sand layer. After saturation, a certain amount of mud is injected into the top surface of the sand layer, and then the top glass sealing plate is sealed and installed on the top of the plexiglass cylinder. The air compressor is started to output compressed gas. After the compressed gas is adjusted to the target pressure value by the air pressure regulating valve, it enters the plexiglass cylinder through the air passage to apply uniform downward air pressure to the surface of the mud. Then the drain valve is opened. Under the action of pressure difference, the mud penetrates the sand layer from top to bottom. The bentonite particles in the mud are gradually deposited on the surface of the sand layer to form a mud film. Meanwhile, the pore water pressure gauge continuously monitors the pore water pressure change in the medium inside the plexiglass cylinder. Simultaneously with pressurization and startup, the ground-penetrating radar (GPR) detection system is activated. This system receives and processes the echo signals to determine the actual thickness of the mud film, including: The echo signal is subjected to DC component removal, bandpass filtering, background interference cancellation, and signal resampling to a standardized size to obtain a two-dimensional radar image reflecting the changes in the stratification of the medium inside the plexiglass cylinder; The two-dimensional radar image is output to the deep learning inversion module to obtain a hierarchical structure image containing the spatial distribution of dielectric constant; Nonlinear inversion of the spatial distribution of dielectric constant to obtain the water content depth profile is performed. Adaptive detection of mud film boundaries based on abrupt changes in water content gradient is performed on the water content depth profile to obtain the depth positions of the upper and lower boundaries of the mud film. Based on the spatial distribution of dielectric constant, adaptive integral estimation of mud film thickness is performed at the depth positions of the upper and lower boundaries of the mud film to obtain the actual thickness of the mud film.

Images