Test system and method for simulating water outburst and mud outburst of subsea tunnel under action of vibration

Abstract

The invention relates to the technical field of soil engineering model tests, and discloses a test system and method for simulating water and mud bursting of a subsea tunnel under the action of vibration, and the system comprises a multi-degree-of-freedom servo feedback vibration module which is used for providing a multidirectional vibration loading platform and feeding back a motor current signal in real time; the visual rock-soil water inrush simulation and multi-field loading module is mounted above the vibration table and is used for accommodating a sample and applying confining pressure and water pressure load; the hierarchical multi-source data acquisition module is communicated with the discharge port and the observation window, and is used for acquiring the outburst mixture and monitoring physical field data; and the intelligent control and risk assessment analysis module is connected with each module to regulate and control loading parameters and invert a catastrophe state based on feedback data. According to the method, a self-adaptive optimization loading mechanism based on the improved particle swarm optimization algorithm is constructed, so that collaborative progressive regulation and control of pore water pressure and vibration frequency parameters are realized, and a critical threshold point of water inrush and mud inrush catastrophe can be accurately captured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of soil engineering model test, in particular to a test system and method for simulating water and mud inrush of seabed tunnel under vibration. BACKGROUND

[0002] With the rapid development of cross-sea transportation projects, seabed tunnel construction is facing extremely complex geological environment challenges. Under the dynamic action of train operation vibration or earthquake load, the hidden water channel is prone to be penetrated, causing water and mud inrush disasters, which poses a great threat to engineering safety. In order to explore this disaster mechanism, indoor physical model test has become an indispensable research means. By building a simulation device in the laboratory, the evolution process of deep geotechnical body under fluid-solid-vibration coupling field can be reproduced, which can help engineers identify risk sources in advance and provide empirical evidence for tunnel support design and disaster warning.

[0003] The existing physical simulation test technology mostly uses an electro-hydraulic servo loading system to cooperate with a rigid model box for research. This mature test system can stably output the preset static load and periodic vibration waveform, and better simulate the initial stress state of the tunnel surrounding rock. In terms of monitoring, the conventional method is to pre-bury miniature sensors such as pore water pressure gauges and soil pressure cells in specific deep positions of the model sample. This contact type measurement method can directly obtain the stress and water pressure values of the specified points inside the geotechnical body, and combined with the external high-speed photography equipment to record the final settlement form of the ground surface, so as to obtain more intuitive stress-strain whole process data.

[0004] However, the above conventional technology has obvious logical faults when dealing with sudden geological disaster simulation. The traditional hierarchical fixed step loading mode is too mechanical and cannot perceive the subtle response inside the geotechnical body, which is prone to directly cross the critical threshold between two loading levels to cause disasters, resulting in that the test is difficult to capture the real boundary conditions at the disaster starting moment; the pre-buried sensor can measure internal data, but in the moment of water and mud inrush explosion, the high-pressure fluid wraps and carries the sand at high speed, which is easy to cause the sensor to shift or even the line to break, and the data at the most critical moment is often lost, and the pre-buried itself will disturb the original field of the sample; in addition, relying solely on external observation can only see the results, but cannot understand the process of internal cavity incubation, and the data analysis model lacking physical mechanism constraints is difficult to distinguish which is the real geological mutation signal in the face of mechanical vibration noise in the test table, which is easy to cause the evaluation result to be distorted. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a test system and method for simulating water and mud inrush of a submarine tunnel under vibration action, to solve the technical problems that fixed step loading in the existing physical simulation test is difficult to accurately capture the critical threshold of disaster, pre-embedded sensors are easily impacted by the gushing fluid, resulting in missing of key data, and a pure data-driven model lacks physical mechanism constraints, resulting in poor noise resistance.

[0006] To achieve the above object, the present application is implemented by the following technical solutions: a test system for simulating water and mud inrush of a submarine tunnel under vibration action, comprising: A multi-degree-of-freedom servo feedback vibration module for providing a multi-directional vibration loading platform and feeding back current signals of a driving motor in real time; A visualized rock-soil water inrush simulation and multi-field loading module installed on the multi-directional vibration loading platform for accommodating a rock-soil sample and applying confining pressure load and water pressure load to the rock-soil sample; A hierarchical multi-source data acquisition module connected to an outlet of the visualized rock-soil water inrush simulation and multi-field loading module and aligned with a visual window of the rock-soil sample for acquiring the mixture of the rock-soil sample and monitoring physical field data; An intelligent control and risk assessment analysis module electrically connected with the multi-degree-of-freedom servo feedback vibration module, the visualized rock-soil water inrush simulation and multi-field loading module, and the hierarchical multi-source data acquisition module for regulating and controlling loading parameters and inverting a disaster state based on feedback data.

[0007] Preferably, the multi-degree-of-freedom servo feedback vibration module comprises: A base vibration table fixed to a foundation and having a longitudinal slide rail arranged on a surface thereof; A one-layer platform slidingly installed on the longitudinal slide rail and having a transverse slide rail arranged on a surface thereof; A two-layer platform slidingly installed on the transverse slide rail for carrying the visualized rock-soil water inrush simulation and multi-field loading module; A driving assembly composed of a servo motor, a lead screw, and a connector for independently driving the one-layer platform and the two-layer platform to perform linear reciprocating motion, and the servo motor internally integrated with a monitoring unit for acquiring q-axis torque current data.

[0008] Preferably, the visualized rock-soil water inrush simulation and multi-field loading module comprises: A main box having a rock-soil cavity inside for accommodating the rock-soil sample, and a drainage passage formed in a bottom of the main box; A confining pressure applying device arranged on a side of the main box for applying lateral pressure to the rock-soil sample through a thick tempered glass plate embedded in a side wall of the main box; A water pressure loading subsystem, which is composed of a water tank, a water pressure booster and a water supply pipe, the water supply pipe is connected to the top of the main tank through a dynamic sealing rotary joint, used to apply pore water pressure to the inside of the rock-soil sample.

[0009] Preferably, the hierarchical multi-source data acquisition module comprises: A non-contact field monitoring assembly, which contains an ASC area array camera and an infrared thermal imager aligned with the visual window, the ASC area array camera is used to acquire deformation field data, and the infrared thermal imager is used to acquire temperature field data; A physical substance collection assembly, which contains a funnel-shaped collection port, an exhaust pipe, a flowmeter, a detachable graded sediment collector and a collector at the bottom, which are connected in series; A metering sensing assembly, which contains a micro pressure sensor arranged below different aperture screens in the detachable graded sediment collector, and a weight sensor arranged below the collector.

[0010] Preferably, the intelligent control and risk assessment analysis module comprises: An industrial control computer, which is internally integrated with a data acquisition card, a motion control card and a field bus communication interface card, used to receive monitoring signals from the flowmeter, the micro pressure sensor, the weight sensor, the ASC area array camera and the servo motor, and send control instructions to the water pressure loading subsystem and the driving assembly; A memory, which stores executable instructions, the instructions are configured to execute particle swarm optimization control algorithm, physical neural network analysis model and damage identification logic.

[0011] Preferably, the intelligent control and risk assessment analysis module executes an adaptive search mode, the adaptive search mode comprises: Defining a group composed of multiple particles, mapping the position of each particle in the search space as a vector combination of water pressure loading value and vibration frequency; Monitoring real-time feedback of water inrush flow prediction value, stress overrun risk index and construction disturbance risk index, calculating fitness function value with water inrush risk comprehensive index as the core; According to the deviation of the current position of the particle and the historical best position, and the deviation of the current position of the particle and the best position of the group, combining inertia factor, individual learning factor and overall learning factor to update the velocity vector and position vector of the particle; Using a ramp generator to convert the updated position vector into a smooth control curve, driving the water pressure booster and the servo motor to perform gradual loading until the critical threshold of water inrush and mud inrush occurs is searched.

[0012] Preferably, the intelligent control and risk assessment analysis module performs structural damage identification, which includes: In the initial state without water pressure, the servo motor outputs a fixed frequency excitation, the torque current is collected, and the baseline torque characteristic value is calculated; During the loading process, the q-axis torque current of the servo motor is collected in real time, and the total equivalent moment of inertia and equivalent viscous damping coefficient of the system containing the mass of the material in the rock-soil cavity are calculated using the dynamic balance equation; The deviation of the real-time torque root mean square value from the baseline torque characteristic value is calculated, and when the deviation exceeds the preset range, it is determined that structural damage or hidden cavity formation occurs inside the rock-soil sample.

[0013] Preferably, the intelligent control and risk assessment analysis module runs a physical neural network verification unit, which includes: A deep learning network architecture is constructed, and the Navier-Stokes equation for controlling fluid motion and the Allen-Cahn equation for controlling phase interface evolution are embedded as regularization terms in the loss function; The observation data collected by the flowmeter and the ASC area array camera are input into the deep learning network architecture, and the equation parameters and network weights are trained synchronously by minimizing the mean square error loss function; The flow field and phase field distribution that meet the physical law constraints are output, and the loss function value is monitored, and when the loss function value exceeds the tolerance range, the physical inflection point of the water inrush channel formation is marked.

[0014] Preferably, the intelligent control and risk assessment analysis module runs a time series data prediction unit, which includes: Linear trend prediction of pressure fluctuation data in a short period is performed using an ARIMA model, which includes autoregressive coefficients and moving average coefficients; Nonlinear trend prediction of water inrush flow evolution data in a long period is performed using a long short-term memory network model, which captures long sequence dependencies through a forget gate, an input gate, and a state update mechanism; The linear trend prediction result and the nonlinear trend prediction result are superimposed to output the time series prediction result of water inrush disaster evolution.

[0015] A test method for simulating water and mud inrush in a submarine tunnel under the action of vibration, comprising the following steps: S1, preparing a rock-soil sample to be tested and filling it into a visual rock-soil water inrush simulation and multi-field loading module, and performing saturation treatment on the rock-soil sample to be tested; S2, start the multi-degree-of-freedom servo feedback vibration module to generate vibration load, collect servo motor torque current to establish baseline, and apply confining pressure; S3, a particle swarm optimization algorithm is run by using the intelligent control and risk assessment analysis module, control parameter combinations of water pressure and vibration frequency are generated, and an adaptive loading search is driven by an executing mechanism; S4, in the loading process, the internal structure damage state of the rock-soil sample to be tested is inversed according to the deviation of the real-time torque current and the baseline; S5, the gushing mixture is collected through the hierarchical multi-source data acquisition module, the cumulative weight of the different particle size sediments and the fluid flow rate are measured; S6, the collected data are physically consistent verified by using a physical neural network model, and a disaster risk is evaluated by combining a time sequence prediction model.

[0016] The application provides a test system and method for simulating water and mud inrush of a submarine tunnel under vibration action. 1, the application realizes the coordinated and gradual regulation of the pore water pressure and the vibration frequency parameters by constructing an adaptive optimization loading mechanism based on an improved particle swarm algorithm, can accurately capture the critical threshold point of the water and mud inrush disaster, and effectively avoids the defects that the traditional test generally uses a hierarchical fixed step loading method, directly crosses the critical point due to improper loading span setting, or cannot respond to the real-time state of the sample, leading to disaster simulation distortion, and ensures the high fidelity of the test process to the complex working condition restoration of the submarine tunnel.

[0017] 2, the application establishes a hidden damage inversion model based on the evolution of dynamic parameters by using the q-axis torque current signal inside the multi-degree-of-freedom servo feedback vibration module, can realize non-invasive monitoring of the system equivalent moment of inertia micro-change caused by cavity expansion in the premise of not invading the internal structure of the rock-soil body, overcomes the technical bottlenecks that the existing technology mainly relies on surface displacement meter observation lag, or relies on internal pre-embedded sensors, which is easy to damage the original stress field of the sample and is extremely easy to be destroyed by the gushing fluid, and realizes non-destructive full-time monitoring of the internal hidden damage of the rock-soil body and the incubation process of the water inrush channel.

[0018] 3, the application integrates a physical neural network verification unit at the data processing end, creatively embeds the Navier-Stokes equation and other fluid mechanics laws as regularization constraints into the deep learning architecture, physically consistent cleans and corrects the multi-source heterogeneous monitoring data, compared with the traditional test method which only simply statistically fits the flow rate and gushing amount, lacks physical mechanism constraints, and has poor data noise resistance and is easy to be misjudged by environmental interference, the application obviously improves the explainability of the test data and the robustness of the disaster risk assessment model. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The system principle block diagram of the embodiment of the application is shown in the figure; Figure 2The method flowchart of the embodiment of the present application is shown in the figure; Figure 3 The schematic diagram of the overall mechanical structure of the device of the embodiment of the present application is shown in the figure; Figure 4 The top view of the main body box in the test device of the embodiment of the present application is shown in the figure; Figure 5 The water and mud burst collecting device in the test device of the embodiment of the present application is shown in the figure.

[0020] Among them: 1, base vibration table; 2, a platform; 3, slide rail; 4, servo motor; 5, screw rod; 6, connector; 7, two-layer platform; 8, confining pressure applying device; 9, ASC area array camera; 10, thick tempered glass plate; 11, main body box; 12, water pressure pipe; 13, water supply pipe; 14, water pressure booster; 15, water collecting pipe; 16, water tank; 17, funnel type collecting port; 18, discharge pipe; 19, flowmeter; 20, collector; 21, infrared thermal imager; 22, computer end; 23, geotechnical cavity; 24, micro pressure sensor; 25, detachable graded sediment collector; 26, weight sensor; 100, multi-degree-of-freedom servo feedback vibration module; 200, visual geotechnical water burst simulation and multi-field loading module; 300, graded multi-source data acquisition module; 400, intelligent control and risk assessment analysis module. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0022] Referring to the drawings in the specification of the present application, Figure 1 And Figure 3 , Figure 1 The principle block diagram of the test system suitable for simulating water and mud burst of a submarine tunnel under the action of vibration of the embodiment of the present application is shown in the figure, Figure 3 The schematic diagram of the overall mechanical structure of the device is shown in the figure. The present application provides a test system suitable for simulating water and mud burst of a submarine tunnel under the action of vibration, which mainly comprises a multi-degree-of-freedom servo feedback vibration module 100, a visual geotechnical water burst simulation and multi-field loading module 200, a graded multi-source data acquisition module 300 and an intelligent control and risk assessment analysis module 400.

[0023] The multi-degree-of-freedom servo feedback vibration module 100 is used as a power input end of the whole test system, and is used for simulating transverse wave and longitudinal wave components in different directions of a seismic wave. The multi-degree-of-freedom servo feedback vibration module 100 comprises a base vibration table 1, and the base vibration table 1 is fixed to a foundation as a bottom support structure. A slide rail 3 is arranged on an upper surface of the base vibration table 1, and a first layer platform 2 is installed on the slide rail 3 in a matched mode. An upper surface of the first layer platform 2 is provided with a slide rail 3 which is orthogonally distributed on the slide rail 3 of the base vibration table 1, and a second layer platform 7 is installed on the upper slide rail 3 in a matched mode. The first layer platform 2 and the second layer platform 7 are respectively driven by independent servo motors 4 through a screw rod 5 and a connector 6. A transmission mechanism formed by the servo motor 4 and the screw rod 5 can drive the first layer platform 2 and the second layer platform 7 to reciprocate along the respective slide rails 3, so as to realize unidirectional vibration loading or bidirectional coupling cross vibration loading. A current monitoring unit is integrated in the servo motor 4, which is used for collecting q-axis torque current data in a motor running process in real time, and transmitting the data to an intelligent control and risk assessment analysis module 400.

[0024] The visualized rock water inrush simulation and multi-field loading module 200 is installed above the second layer platform 7 of the multi-degree-of-freedom servo feedback vibration module 100. The visualized rock water inrush simulation and multi-field loading module 200 comprises a main body box 11. A rock soil cavity 23 is arranged in the main body box 11, and the rock soil cavity 23 is used for accommodating a rock soil sample to be tested and a water permeable filling piece. A side wall of the rock soil cavity 23 is made of a transparent material, and a confining pressure applying device 8 is arranged on three side surfaces of the main body box 11. The confining pressure applying device 8 comprises a thick tempered glass plate 10, and the thick tempered glass plate 10 is used as a visual window while applying confining pressure. The visualized rock water inrush simulation and multi-field loading module 200 further comprises a water pressure loading subsystem, which comprises a water tank 16, a water collecting pipe 15, a water pressure booster 14, a water supply pipe 13 and a water pressure pipe opening 12. After water in the water tank 16 is pressurized by the water pressure booster 14, the water is delivered to the water pressure pipe opening 12 at the top of the main body box 11 through the water supply pipe 13, so as to apply a set pore water pressure to the sample in the rock soil cavity 23.

[0025] The hierarchical multi-source data acquisition module 300 is used to collect the physical field data and the outpouring matter generated in the test process. The hierarchical multi-source data acquisition module 300 comprises an ASC area array camera 9 and an infrared thermal imager 21 arranged outside the visual window on the side of the main body box 11. The ASC area array camera 9 is used to record the deformation field and the seepage path of the sample in the rock-soil cavity 23, and the infrared thermal imager 21 is used to monitor the temperature field change. The hierarchical multi-source data acquisition module 300 further comprises a physical matter collection assembly which comprises a funnel-shaped collection port 17, a discharge pipe 18, a flowmeter 19, a detachable hierarchical sediment collector 25 and a collector 20. The funnel-shaped collection port 17 is arranged below the opening at the bottom of the main body box 11 and is connected to the discharge pipe 18 through a pipeline. The flowmeter 19 is arranged on the discharge pipe 18 and is used to measure the instantaneous velocity of the outpouring fluid. The detachable hierarchical sediment collector 25 is arranged at the end of the discharge pipe 18 and is internally provided with a plurality of filter screens with different aperture diameters and a micro pressure sensor 24, which are used to collect and measure the sediment with different particle sizes. The collector 20 is arranged at the bottom end and is provided with a weight sensor 26 at the lower part, which is used to measure the total outpouring amount.

[0026] The intelligent control and risk assessment analysis module 400 takes the computer terminal 22 as the core carrier and is electrically connected with the multi-degree-of-freedom servo feedback vibration module 100, the visual rock-soil water inrush simulation and multi-field loading module 200 and the hierarchical multi-source data acquisition module 300 through a data cable. The intelligent control and risk assessment analysis module 400 is used to receive the real-time monitoring signals from the flowmeter 19, the micro pressure sensor 24, the weight sensor 26, the ASC area array camera 9 and the servo motor 4. The intelligent control and risk assessment analysis module 400 internally runs a particle swarm optimization control algorithm and a physical neural network analysis model, which can inversely calculate the structural damage state inside the rock-soil cavity 23 according to the real-time feedback of the flow rate change rate and the torque current fluctuation of the servo motor 4.

[0027] Referring to the accompanying drawings Figure 2 , Figure 2 The flowchart is a schematic diagram of the test method for simulating the water and mud inrush of a submarine tunnel under the action of vibration according to an embodiment of the present application. The present application provides a test method for simulating the water and mud inrush of a submarine tunnel under the action of vibration. The method is implemented based on the foregoing test system and comprises the following steps: S100, preparing a rock-soil sample to be tested and loading the sample into the rock-soil cavity 23 of the visual rock-soil water inrush simulation and multi-field loading module 200, checking the sealing performance of each connection part and performing saturation treatment on the rock-soil sample; S200, controlling the confining pressure applying device 8 to apply confining pressure to the rock-soil sample to achieve a preset ground stress target value, starting the multi-degree-of-freedom servo feedback vibration module 100 to generate a vibration load at a preset initial frequency, collecting the torque current reference value of the servo motor 4 through the intelligent control and risk assessment analysis module 400 and establishing a vibration feedback baseline; S300, entering the adaptive search mode, the intelligent control and risk assessment analysis module 400 generates a combination of water pressure and vibration frequency control parameters by using a particle swarm optimization algorithm, drives the water pressure booster 14 and the servo motor 4 to perform loading, and monitors the real-time feedback data of the hierarchical multi-source data acquisition module 300 and the servo motor 4 to calculate the fitness value, and automatically searches for the critical threshold condition of water and mud inrush in the solution space; S400, during the loading process, the intelligent control and risk assessment analysis module 400 calculates the equivalent mass change and stiffness change inside the rock-soil cavity 23 according to the deviation of the real-time torque current and the vibration feedback baseline, and identifies the internal structural damage state of the rock-soil sample; S500, after triggering macro water and mud inrush, collecting the outflow mixture through the funnel-shaped collection port 17, measuring the flow rate by using the flowmeter 19, separating the sand of different particle sizes by using the detachable hierarchical sand collector 25, and measuring the cumulative weight of different particle size components by using the micro pressure sensor 24 and the weight sensor 26; S600, all collected data are transmitted to the computer end 22, and the intelligent control and risk assessment analysis module 400 uses a physical neural network model to perform physical consistency verification and time sequence risk assessment on the seepage field process.

[0028] The specific implementation process of the present application will be described in detail below in combination with the above modules and steps.

[0029] Referring to the accompanying drawings Figure 1 and the accompanying drawings Figure 2 , the multi-degree-of-freedom servo feedback vibration module 100 includes a base vibration table 1. The base vibration table 1 serves as a fixed base of the test device and is anchored to the laboratory foundation or a vibration isolation trench through foundation bolts to absorb the reverse impact force generated during vibration loading and ensure the stability of the test reference. A set of slide rails 3 is arranged in parallel on the upper surface of the base vibration table 1, and the slide rails 3 extend horizontally and longitudinally (defined as the X-axis direction). A one-layer platform 2 is installed on the slide rails 3 of the base vibration table 1 through cooperating slide blocks at the bottom of the one-layer platform 2, and can perform linear reciprocating motion in the X-axis direction with low friction coefficient. A set of slide rails 3 is also arranged in parallel on the upper surface of the one-layer platform 2, and the slide rails 3 extend horizontally and transversely (defined as the Y-axis direction), so that the slide rails 3 on the one-layer platform 2 and the slide rails 3 on the base vibration table 1 are orthogonally distributed in space projection, forming a cross-shaped guide rail layout. A two-layer platform 7 is installed on the slide rails 3 of the one-layer platform 2 through cooperating slide blocks at the bottom of the two-layer platform 7, and can perform linear reciprocating motion in the Y-axis direction. Positioning slots for fixing the visual rock-soil water inrush simulation and multi-field loading module 200 are opened on the upper surface of the two-layer platform 7, and a clearance hole is opened at a position corresponding to directly below the rock-soil cavity 23 to prevent the metal platform from blocking the bottom signal transmission or interfering with the physical field distribution.

[0030] In terms of the driving mechanism, the multi-degree-of-freedom servo feedback vibration module 100 is equipped with independent driving components for the first-level platform 2 and the second-level platform 7. Each driving component includes a servo motor 4, a lead screw 5, and a connector 6. The servo motor 4 is fixedly mounted on the lower support structure of the object it drives (i.e., the servo motor 4 driving the first-level platform 2 is fixed to the side bracket of the base vibration table 1, and the servo motor 4 driving the second-level platform 7 is fixed to the extension bracket of the first-level platform 2). The output shaft of the servo motor 4 is connected to one end of the lead screw 5 through a rigid coupling. The connector 6 is constructed as a nut seat for a ball screw nut pair, which is fixedly mounted at the bottom center of the driven platform (first-level platform 2 or second-level platform 7). The lead screw 5 passes through the connector 6 and forms a precision helical transmission engagement with the balls inside, converting the rotational motion of the servo motor 4 into the linear translational motion of the platform. This rigid transmission chain design eliminates the attenuation effect of elastic deformation on the vibration waveform. By controlling the speed, direction, and phase difference of the two sets of servo motors 4, the multi-degree-of-freedom servo feedback vibration module 100 can achieve unidirectional (X-axis only or Y-axis only) vibration loading, or synthesize planar coupled vibration loading with arbitrary trajectories. The specific selection of the servo motors 4 and the lead parameters of the lead screw 5 can be conventionally calculated and configured by those skilled in the art based on the amplitude range and frequency response requirements of the experiment; these are well-known technologies in the field and will not be elaborated upon here.

[0031] The multi-degree-of-freedom servo feedback vibration module 100 not only functions as an actuator but is also configured as a load monitoring device with sensing capabilities, used for benchmark establishment in step S200 and damage identification in step S400. The servo motor 4 integrates a high-precision current sensor and a high-resolution absolute encoder, connected to the intelligent control and risk assessment analysis module 400 via a high-speed industrial fieldbus (such as EtherCAT), enabling the uploading of electrical and motion parameters during motor operation at a high sampling rate (above 1kHz). The servo motor 4 is equipped with a high-performance vector control driver to decouple the stator current, separating the excitation current component (d-axis current) from the torque current component (q-axis current). In permanent magnet synchronous motors, the electromagnetic torque and q-axis current have a strictly linear relationship, and this module is configured in a strong position closed-loop control mode, meaning that regardless of changes in the external load, the controller forces the motor to follow a preset motion trajectory, causing current changes to balance load fluctuations.

[0032] During the experiment, the intelligent control and risk assessment analysis module 400 reads the q-axis torque current of the servo motor 4 in real time. The physical state inside the soil-rock cavity 23 is calculated based on the established dynamic equilibrium equations. The dynamic equilibrium equations for the servo motor 4 driving the load are expressed as follows: ; in, represents the torque constant of the servo motor 4, which is a property inherent to the motor; represents the electromagnetic torque output; represents the torque constant of the servo motor 4, which is a property inherent to the motor; represents the real-time collected q-axis torque current; , , respectively represent the angular displacement, angular velocity and angular acceleration of the motor shaft, which are obtained by real-time feedback of the motor encoder, and under steady-state vibration, the amplitude and frequency thereof are locked as known constants by the control system; represents the total equivalent moment of inertia of the system converted to the motor shaft end, which is determined by the inherent mass of the platform and the real-time mass of the mixture in the rock-soil cavity 23 together; represents the equivalent viscous damping coefficient of the system, which reflects the frictional energy dissipation characteristics between particles in the rock-soil body and the damping of the mechanical transmission chain; represents the equivalent torsional stiffness coefficient of the system; represents the Coulomb friction torque of the mechanical transmission system, which can be regarded as a constant after the running-in period of the system.

[0033] In step S200, the system controls the servo motor 4 to output a sine wave excitation of fixed frequency and amplitude in the initial state without applying the seepage water pressure. At this time, the rock-soil cavity 23 is structurally complete and the mass distribution is uniform, and the intelligent control and risk assessment analysis module 400 records the data in a stable period, identifies the initial equivalent moment of inertia and equivalent damping by using the least squares method, and calculates the baseline torque characteristic value .

[0034] In step S400, when the particle migration occurs due to the erosion of the water flow in the rock-soil cavity 23, resulting in the formation of a hidden cavity, or a large amount of material is lost due to water and mud inrush, the load mass and structural damping of the system will change significantly. At this time, will show a downward trend, and will fluctuate due to the change in flow state. Since the servo system is under closed-loop control, the and its derivative are forcibly maintained unchanged, which inevitably leads to the non-periodic deviation or distortion of the input current . By monitoring the deviation between the real-time torque and the baseline torque , and combining the aforementioned dynamic equation, the mass loss rate in the rock-soil cavity 23 can be quantitatively inverted without opening the main body box 11 or relying on external visual observation, thereby realizing early identification of the incubation stage of the water and mud inrush disaster of the seabed tunnel.

[0035] Referring to the accompanying Figure 1 , the accompanyingFigure 2 and attached Figure 4 The visualization of water inrush simulation and multi-field loading module 200 is fixedly installed on the two-layer platform 7 of the multi-degree-of-freedom servo feedback vibration module 100, and follows the platform to perform multi-directional coupled vibration. The module mainly comprises a rigid pressure-bearing main body box 11, a flexible boundary rock-soil cavity 23, a visualization true triaxial confining pressure loading assembly, and a servo water pressure seepage loading assembly.

[0036] The main body box 11 serves as a pressure-bearing main body structure, is welded from high-strength stainless steel material, and has a wall thickness that is checked by finite elements to meet the rigidity requirement under the maximum internal pressure in the design, so as to prevent plastic deformation under the coupling action of high-frequency vibration and high internal pressure. A rectangular through hole is formed in the bottom of the main body box 11, the size of the through hole matches the cross-sectional size of the rock-soil cavity 23 below, and is coaxially aligned with the clearance hole on the two-layer platform 7, to form a material discharge channel when a water and mud inrush disaster occurs. A quick sliding gate plate (or a quick release bottom cover) driven by an electro-hydraulic servo is installed at the bottom of the main body box 11. The gate plate is in a locked state during the test loading stage, and can bear the vertical load in the rock-soil cavity 23. At the triggering moment of the simulation disaster, the gate plate can be horizontally slid open in milliseconds, instantaneously releasing the bottom constraint, and simulating a sudden geological disaster.

[0037] The rock-soil cavity 23 is located in the center of the internal space of the main body box 11, and is used for accommodating a rock-soil body sample to be tested. The rock-soil cavity 23 adopts a double-layer flexible constraint design, the outer layer is a transparent PVC high-pressure waterproof film, and the inner layer is smeared with vaseline lubricating grease, so as to eliminate the restriction of the rigid boundary on the deformation of the soil body and the interference of the sidewall friction on the stress distribution. The upper end of the PVC high-pressure waterproof film is sealingly connected with the top cover of the main body box 11, and the lower end is sealingly pressed against the edge of the bottom quick sliding gate plate. A rigid porous water-permeable filling piece with a high permeability coefficient is laid on the top of the sample, the filling piece is constructed as a sintered copper sand plate or a porous ceramic plate, and serves as a transmission medium and a flow equalizing plate of the overburden pressure, to ensure that the input pressure water uniformly acts on the top surface of the sample in the form of a plane wave.

[0038] The confining pressure application device 8 and the thick tempered glass plate 10 together constitute a visualized true triaxial confining pressure loading assembly. Observation windows are provided on three sides of the main body box 11, and high-pressure sealing guide sleeves are embedded within these windows. The thick tempered glass plate 10 is embedded within the guide sleeve, and multiple Glyd rings or combined sealing rings are arranged between it and the inner wall of the sleeve to form a dynamic sealing fit structure. The thick tempered glass plate 10 is made of high-purity quartz glass or borosilicate special glass, which acts as a pressure piston to directly press against the PVC waterproof membrane on the outer layer of the soil cavity 23. The confining pressure application device 8 is constructed as a servo hydraulic cylinder or a precision electric screw jack, which is fixed to the reaction frame extending from the outside of the main body box 11. The top of the push rod is rigidly connected to the outer surface of the thick tempered glass plate 10 via a flange. In step S200, the intelligent control and risk assessment analysis module 400 independently controls the confining pressure application devices 8 in three directions, driving the thick tempered glass plate 10 to advance slightly within the guide sleeve, applying anisotropic or isotropic lateral pressure to the soil-rock cavity 23. At this time, the effective stress state inside the sample follows Terzaghi's effective stress principle: ; in, The effective stress tensor of the soil-rock skeleton; The total stress tensor applied by the confining pressure application device 8; The effective stress coefficient; Pore ​​water pressure; For Kroneck's symbol.

[0039] This formula indicates that this module independently controls the confining pressure. With water pressure It can accurately reproduce the soil liquefaction or shear failure phenomenon caused by the sharp decrease in effective stress during water inrush.

[0040] The water pressure seepage loading assembly is responsible for building a high-pressure seepage field, which mainly includes a water tank 16, a water pressure booster 14, a water supply pipe 13, and a water pressure pipe opening 12. The water tank 16 is connected to the low-pressure input end of the water pressure booster 14 through a pipeline. The water pressure booster 14 adopts a servo motor driven plunger pump or a gas-liquid booster pump, which can output constant high-pressure water flow or pulsating water flow according to the instruction. The water pressure pipe opening 12 is arranged on the top sealing cover plate of the main body tank 11 and directly communicates to the upper surface of the water permeable filling member in the rock-soil cavity 23. The water supply pipe 13 adopts an anti-vibration flexible dynamic sealing structure composed of an inner lining of a stainless steel corrugated pipe and an outer covering of a high-strength steel wire braided layer of armored high-pressure hose, and a spherical articulated high-pressure rotary joint is arranged at the end of the hose connected to the main body tank 11. The rotary joint has at least two degrees of freedom of rotation and can compensate for the high-frequency displacement components of the main body tank 11 in the X and Y directions in real time, eliminate pipeline torsional stress, prevent the pipeline from breaking due to repeated bending fatigue, and ensure that the seepage water pressure of up to 10 MPa or more can still be stably delivered to the inside of the rock-soil cavity 23 under severe vibration working conditions.

[0041] Referring to the drawings Figure 1 The hierarchical multi-source data acquisition module 300 is a data acquisition device responsible for obtaining the physical field evolution characteristics and the quantitative information of water and mud inrush substances in the test system. It is installed independently of the multi-degree-of-freedom servo feedback vibration module 100 and is used to perform real-time monitoring in step S300 and substance hierarchical metering in step S500. The module is mainly composed of a non-contact field monitoring assembly and a physical contact hierarchical metering assembly.

[0042] The non-contact field monitoring assembly is arranged on an independent stationary support outside the multi-degree-of-freedom servo feedback vibration module 100 to realize mechanical decoupling from the vibration table and prevent vibration interference from affecting the imaging quality. The assembly includes an ASC area array camera 9 and an infrared thermal imager 21. The ASC area array camera 9 is aimed at the visible window of the thick tempered glass plate 10 on the side of the main body tank 11. In the test preparation stage, the rock-soil body sample is sprayed with a black and white artificial speckle field on the side facing the visible window. The ASC area array camera 9 cooperates with an externally triggered high-frequency flash source to shoot a sequence of speckle images of the sample surface at a sampling rate synchronized or multiplied with the vibration frequency. The intelligent control and risk assessment analysis module 400 uses a digital image correlation algorithm (DIC) to analyze the displacement field and strain field inside the rock-soil body by calculating the gray-scale mutual correlation function of the image sub-regions before and after deformation, and identifies the evolution process of the shear slip band. The infrared thermal imager 21 focuses on the same observation area and uses the difference in specific heat capacity and thermal diffusivity between pore water and rock-soil skeleton to monitor the infrared radiation temperature field of the rock-soil cavity 23 surface in real time. When an advantage seepage channel is formed inside, the water flow with specific temperature tracing characteristics will change the local temperature distribution, and the infrared thermal imager 21 can capture the temperature difference abnormal area in the form of tree branches or strips, thereby mapping the spatial distribution of the seepage path.

[0043] The physical contact-type graded metering component is located below the main tank 11 and is used to collect and analyze the effluent after a water or mud inrush. (Attached) Figure 3 and attached Figure 5 As shown, the component, from top to bottom, includes a funnel-shaped collection port 17, a discharge pipe 18, a flow meter 19, a detachable graded sediment collector 25, and a collector 20. A vibration-resistant flexible connector is provided between the funnel-shaped collection port 17 and the drain hole at the bottom of the main body box 11. This connector is constructed as a corrugated telescopic guide hood made of large-diameter silicone or high-strength canvas. The upper end of the guide hood is fixed to the bottom of the vibrating main body box 11 via a flange, and the lower end is fixed to the stationary funnel-shaped collection port 17. It can accommodate the maximum amplitude displacement of the main body box 11 in the X and Y axes, preventing splashing or leakage of effluent during vibration.

[0044] The discharge pipe 18 is connected to the lower end of the funnel-shaped collection port 17, serving as a fluid transport channel. A flow meter 19 is installed on the vertical straight section of the discharge pipe 18. This flow meter 19 is selected as a full-bore electromagnetic flow meter or a clamp-on Doppler ultrasonic flow meter. Its internal measuring channel has no flow obstruction components, enabling it to adapt to high-concentration, large-particle solid-liquid two-phase flow environments and to measure the instantaneous flow velocity of the mixed fluid in real time. .

[0045] The detachable grading sediment collector 25 is a device for classifying and statistically analyzing sediment particle size, and it is installed in series at the end of the discharge pipe 18. This collector adopts a vertically stacked multi-layered box structure, with each layer containing an independent floating filter frame. Metal filter screens are installed within the filter frames, and the pore size of each layer decreases sequentially from top to bottom (e.g., the first layer has a 5mm pore size, the second layer has a 2mm pore size). A miniature pressure sensor 24 (specifically a cantilever beam or button-type load cell) is installed below the support lug of each floating filter frame, with the sensor base fixed to the inner wall of the box. When the mixed fluid passes through, solid particles of different sizes are trapped on their respective filter screens, while the fluid and fine particles continue to flow downwards through the mesh. The miniature pressure sensor 24 detects the total vertical load on the filter frames and the trapped material in real time and converts the signal into mass data, which is then transmitted to the computer 22. The collector 20 is located directly below the detachable grading sediment collector 25 and is used to collect the remaining mud-water mixture (mainly water and extremely fine particles) that has passed through all the filter screens. Collector 20 is placed on weight sensor 26 (constructed as a high-precision electronic weighbridge) to measure the final cumulative total mass.

[0046] With the above structure, the system can establish a mathematical model of material transport in the process of sudden water and mud inrush. At any given time... Total outflow mass Obtained by summing up the parts: wherein, represents the number of filter layers; represents the real-time net increment measured by the micro pressure sensor 24 under the layer filter, which is the current reading minus the baseline value of the initial self-weight of the filter frame and the impact force of water flow, reflecting the mass of particles with a particle size greater than the pore size of the layer; represents the mass of fluid in the collector 20 measured by the bottom weight sensor 26.

[0047] At the same time, combined with the flow rate measured by the flowmeter 19 and the known internal cross-sectional area of the discharge pipe 18, the average density evolution curve of the fluid can be calculated: The formula characterizes the dynamic change characteristics of the sediment content in the water and mud inrush process, and can distinguish different disaster stages such as clear water gushing, sediment-laden water gushing and debris flow gushing. The hierarchical multi-source data acquisition module 300 provides complete material conservation boundary conditions for subsequent physical consistency verification through the synchronous acquisition of the above physical quantities.

[0048] Referring to the attached Figure 1 and the attached Figure 2 , the intelligent control and risk assessment analysis module 400 is the control and data processing unit of the test system, which is connected to the multi-degree-of-freedom servo feedback vibration module 100, the visualized rock water inrush simulation and multi-field loading module 200 and the hierarchical multi-source data acquisition module 300 through electrical lines. The module includes an industrial control computer (IPC), which is internally integrated with a central processor, a memory and various interface cards. The interface cards include multi-channel high-speed analog quantity acquisition cards (DAQ), motion control cards and industrial field bus communication interface cards (such as EtherCAT or CANopen interface cards). The memory stores computer program instructions executable by the central processor, and the instructions are configured to implement data acquisition, closed-loop control, parameter optimization and physical model verification logic.

[0049] ​​At the level of hardware signal interaction, the intelligent control and risk assessment analysis module 400 receives the voltage or current signals of the flowmeter 19, the micro pressure sensor 24 and the weight sensor 26 through the analog input interface; receives the image data stream of the ASC area array camera 9 and the infrared thermal imager 21 through the high-speed digital I / O interface or the special image acquisition interface; establishes real-time communication with the driver of the servo motor 4 through the industrial field bus to read the position and speed information fed back by the motor encoder and the q-axis torque current data fed back by the internal current loop. At the same time, the module sends pressure control instructions to the water pressure booster 14 and motion control instructions to the servo motor 4 through the analog output interface or bus instructions.

[0050] In the adaptive search mode of step S300, the intelligent control and risk assessment analysis module 400 runs the embedded improved particle swarm optimization (PSO) algorithm unit to automatically find the critical threshold of water and mud inrush in the preset feasible region. The group is defined to consist of particles, the th particle is represented by a vector , where , respectively, correspond to the water pressure loading value and the vibration frequency . The position of each particle represents a potential loading condition solution. The velocity of the particle is represented by a vector .

[0051] The intelligent control and risk assessment analysis module 400 updates the velocity and position of the particle according to the following formula: ; ; wherein is the iteration number; is the inertia factor, the value range is set to 0.8 to 1.2, which is used to balance the global search and local search ability; is the individual learning factor, which represents the weight of the particle adjusting the state according to the historical best position ; is the global learning factor, which represents the weight of the particle adjusting the state according to the group best position ; and are random numbers in the range of [0, 1]. After calculating the new position vector , the intelligent control and risk assessment analysis module 400 adjusts the new target water pressure and frequency​ The smooth control curve is generated, and the water pressure booster 14 and the servo motor 4 are driven through the interface card to perform gradual loading, preventing non-physical step impact on the test system due to parameter mutation.

[0052] In order to quantitatively evaluate the disaster risk under the current working condition, the intelligent control and risk assessment analysis module 400 constructs an adaptability function with the water inrush risk comprehensive index as the core . This function optimizes the simple path length evaluation index to a multi-physical field coupling evaluation index, and its calculation formula is as follows: ; Among them, , , represent the weight coefficients of each item; represent the water inrush flow predicted by the physical neural network, which is based on the real-time reading of the flowmeter 19 and the trend extrapolation; represent the stress overrun risk index, which is represented by the normalized servo motor 4 torque current deviation characterization, reflecting the stress damage degree of the internal skeleton of the rock-soil body; represent the construction disturbance risk index, which is derived from the product of the vibration amplitude and frequency of the multi-degree-of-freedom servo feedback vibration module 100, representing the external disturbance energy level.

[0053] Based on the calculated , the system introduces a threshold warning mode. The intelligent control and risk assessment analysis module 400 extracts the risk index extreme value under the safety-danger critical state and sets it to (for example, set the critical value corresponding to the water inrush probability of 30%). The system compares the monitored risk index with in real time: when , it is determined to enter the warning area. Among them, when , a yellow warning signal is output, and the system automatically reduces the loading step; when , a red warning signal is output, and the system triggers the stop or safety protection logic.

[0054] When performing step S600 data verification and evaluation, the intelligent control and risk assessment analysis module 400 calls the physical neural network (PINNs) verification unit and the time series data prediction unit.

[0055] The physical neural network verification unit is a kind of deep learning architecture that integrates physical law constraints. Its core is to embed the partial differential equation (PDE) that controls the evolution of the water inrush of the rock-soil mass as a regularization term into the loss function. This unit integrates the Navier-Stokes equation and the Allen-Cahn equation to handle non-Darcy fluid motion and dynamic evolution of the cement two-phase interface under high Reynolds number.

[0056] Specifically, for the fluid motion in the water inrush channel, the network is constrained by the two-dimensional Navier-Stokes equation: ; ; where, denotes the component of the velocity; denotes the component of the velocity; denotes the fluid pressure; , is a fluid dynamics parameter. For fluid mass conservation, the incompressible fluid continuity equation constraint is introduced, and the implicit function is assumed to satisfy , so that the continuity equation is automatically satisfied.

[0057] At the same time, in order to track the nonlinear interface between the water phase and the mud phase during the water inrush process, the network also introduces the Allen-Cahn equation constraint: ; where, is a phase field parameter. The intelligent control and risk assessment analysis module 400 synchronously trains the equation parameters and neural network weights by minimizing the mean square error loss function, ensuring that the output flow field and phase field not only conform to the observation data of the flowmeter 19 and the ASC area camera 9, but also strictly follow the above physical laws. When the calculated loss function value exceeds the preset tolerance range, it indicates that the flow state has changed in nature, and the system automatically marks this time point as the physical inflection point of the water inrush channel formation.

[0058] After completing the physical consistency verification, the time series data prediction unit performs fusion prediction on the future disaster trend. This unit adopts the strategy of superimposing linear components and nonlinear components for output.

[0059] For short-term linear trend prediction (such as pressure fluctuations within 1-7 seconds), the ARIMA model is used: where, is the time series data, is an autoregressive coefficient, is a moving average coefficient, is an error term.

[0060] For the nonlinear complex trend of medium and long term (such as the evolution of water inrush flow within 7-30 seconds), a long short-term memory network (LSTM) model is adopted. The LSTM controls the information flow through the gating mechanism, and the forgetting gate The calculation formula is: ; wherein, is an activation function, is a weight matrix, is a hidden state at the previous moment, is a current input. The input gate and the candidate state jointly determine the update of information: ; ; The final state is updated to . The fusion strategy uses ARIMA to capture linear trends, uses LSTM to filter noise and remember long sequence dependencies, thereby outputting high-precision water inrush disaster time series prediction results.

[0061] In order to further illustrate the synergistic working process of the technical solutions described in the application, a specific working scene example will be described below.

[0062] In this embodiment, the test system is used to simulate the water and mud inrush disaster evolution process of a certain submarine tunnel fault fracture zone under the action of seismic waves. First, in the test preparation stage, the technical personnel prepares the reconstructed rock-soil sample composed of gravel soil and clay according to the on-site survey data, and fills it into the rock-soil cavity 23 of the visual rock-soil water inrush simulation and multi-field loading module 200. The inner layer of the rock-soil cavity 23 is smeared with vaseline and attached with a PVC waterproof film, and the outer layer is constrained by a thick tempered glass plate 10. Subsequently, the intelligent control and risk assessment analysis module 400 instructs the confining pressure applying device 8 to start, and applies a constant confining pressure of 5MPa to the rock-soil cavity 23 through the thick tempered glass plate 10, to simulate the initial ground stress environment of the submarine stratum. At the same time, the inside of the sample is pre-primed with water through the water supply pipe 13, so that the sample reaches a saturated state, but a high-pressure seepage water head is not applied. In this state, the intelligent control and risk assessment analysis module 400 drives the servo motor 4 in the multi-degree-of-freedom servo feedback vibration module 100 to drive the first layer platform 2 and the second layer platform 7 to perform low-frequency sinusoidal scanning vibration with a frequency of 5Hz and an amplitude of 2mm. In this process, the intelligent control and risk assessment analysis module 400 continuously collects the q-axis torque current data fed back by the servo motor 4, and solves the initial equivalent moment of inertia of the current system load using the dynamic equilibrium equation and reference torque characteristic value This completes the establishment of the system dynamics baseline.

[0063] Upon entering the experimental loading phase, the intelligent control and risk assessment analysis module 400 activates the particle swarm optimization algorithm unit. Based on the set search boundaries (e.g., water pressure range 0-10 MPa, frequency range 0-50 Hz), the algorithm initializes and generates a set of loading parameter particle swarms. Module 400 generates smooth control commands via a ramp generator, driving the water pressure booster 14 to gradually increase the water pressure to the initial set value (e.g., 2 MPa), and drives the servo motor 4 to output random vibration signals simulating seismic waves. At this time, the multi-degree-of-freedom servo feedback vibration module 100 applies shear vibration to the soil-rock cavity 23, causing a reduction in the interlocking force between the soil-rock skeleton particles; simultaneously, the visualization of soil-rock water inrush simulation and the seepage pressure applied by the multi-field loading module 200 generate seepage force in the pores, attempting to break through the skeleton constraints.

[0064] During loading, the graded multi-source data acquisition module 300 is in a continuous monitoring state. The ASC area array camera 9 captures speckle images of the sample side at a frame rate of 100fps, while the infrared thermal imager 21 simultaneously monitors the temperature field. When the water pressure rises to the critical region (e.g., 4.5MPa) and the vibration frequency reaches near the resonant frequency of the soil and rock mass, the fine particles inside the soil and rock mass begin to migrate with the water flow, forming micro-seepage channels invisible to the naked eye. At this time, although no macroscopic water inrush phenomenon is observed externally, the load characteristics of the servo motor 4 have changed. Due to the migration of internal particles, the damping coefficient... Equivalent stiffness caused by descent and local liquefaction To maintain the preset vibration waveform, the q-axis torque current of servo motor 4 is reduced. High-frequency jitter causes a deviation between the real-time RMS torque value and the reference value. The voltage increased rapidly. The intelligent control and risk assessment analysis module 400 detected this electrical signal anomaly and calculated the fitness function value. If the water pressure rises sharply, the system will automatically determine that the sample has entered the critical period of water incubation and record the combination of water pressure and frequency at this time as the critical threshold for catastrophic events.

[0065] With the loading continues to maintain or slightly increase, the internal structure of the rock and soil eventually destabilizes, and macro water and mud inrush occurs. The rapid sliding gate at the bottom of the main box 11 opens, and the mixture of water and mud in the rock and soil cavity 23 is spouted out under the dual action of gravity and water pressure. The outflow passes through the funnel-shaped collecting port 17 into the discharge pipe 18, and the flow meter 19 measures the flow rate in real time and suddenly increases to a peak value. The mixed fluid then enters the detachable graded sand collector 25, large-diameter gravel is intercepted by the upper coarse filter screen, medium-diameter sand is intercepted by the lower fine filter screen, and mud flows into the bottom collector 20. The micro pressure sensors 24 in each layer and the bottom weight sensor 26 synchronously upload the mass data to the computer terminal 22. At the same time, the infrared thermal imager 21 captures the tree branch-shaped low temperature anomaly zone on the surface of the sample (assuming that the temperature of the injected water is lower than the ambient temperature), and the strain field calculated by the ASC area array camera 9 shows a through shear slip zone.

[0066] After the test is completed, the intelligent control and risk assessment analysis module 400 calls the physical neural network verification unit, and the time series data of the collected pressure , flow and total mass are substituted into the physical consistency verification loss function for calculation. If the value remains low before the water inrush occurs, it indicates that the seepage conforms to Darcy's law; if the value suddenly jumps at the moment of water inrush, it indicates that the flow state changes to non-Darcy turbulent flow. The system finally integrates the above-mentioned multi-source information to generate a complete test report containing critical water pressure-frequency threshold, water inrush path evolution cloud diagram and outflow grading curve, thereby revealing the disaster mechanism of water and mud inrush in submarine tunnel under the coupling action of vibration field and seepage field.

[0067] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A test system for simulating water and mud inrush of a submarine tunnel under the action of vibration, characterized in that, include: A multi-degree-of-freedom servo feedback vibration module is used to provide a multi-directional vibration loading platform and provide real-time feedback of the current signal of the drive motor. A visualization module for simulating water inrush in soil and rock and a multi-field loading module are installed on the multi-directional vibration loading platform to accommodate soil and rock samples and apply confining pressure loads and water pressure loads to the soil and rock samples. A graded multi-source data acquisition module is connected to the outlet of the visualization rock and soil water inrush simulation and multi-field loading module and aligned with the visual window of the rock and soil sample. It is used to collect the mixture protruding from the rock and soil sample and monitor the physical field data. The intelligent control and risk assessment analysis module is electrically connected to the multi-degree-of-freedom servo feedback vibration module, the visualized soil and rock water inrush simulation and multi-field loading module, and the hierarchical multi-source data acquisition module, respectively, and is used to adjust the loading parameters and invert the disaster state based on the feedback data.

2. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 1, characterized in that, The multi-degree-of-freedom servo feedback vibration module includes: A foundation vibration table is fixed on the foundation, and the surface of the foundation vibration table is provided with longitudinal slide rails; A single-layer platform is slidably mounted on the longitudinal slide rail, and the surface of the single-layer platform is provided with a transverse slide rail; The second-layer platform is slidably installed on the transverse slide rail and is used to support the visualization of rock and soil water inrush simulation and multi-field loading module; The drive assembly consists of a servo motor, a lead screw, and a connector, which independently drive the first-layer platform and the second-layer platform to perform linear reciprocating motion. The servo motor has an integrated monitoring unit for collecting q-axis torque and current data.

3. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 1, characterized in that, The visualization-based soil and water inrush simulation and multi-field loading module includes: The main body box has an internal soil cavity for accommodating the soil sample, and a drainage channel is provided at the bottom of the main body box. A confining pressure application device is installed on the side of the main body box, and applies lateral pressure to the soil and rock sample through a thick tempered glass plate embedded in the side wall of the main body box; The water pressure loading subsystem consists of a water tank, a water pressure booster, and a water supply pipe. The water supply pipe is connected to the top of the main tank through a dynamic sealing rotary joint and is used to apply pore water pressure to the interior of the soil and rock sample.

4. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 1, characterized in that, The hierarchical multi-source data acquisition module includes: A non-contact field monitoring component includes an ASC area array camera and an infrared thermal imager aligned with the viewing window. The ASC area array camera is used to acquire deformation field data, and the infrared thermal imager is used to acquire temperature field data. The physical material collection assembly includes a funnel-shaped collection port, a discharge pipe, a flow meter, a detachable graded sediment collector, and a bottom collector connected in series. The metering sensing component includes miniature pressure sensors disposed below filters of different pore sizes within the removable graded sediment collector, and a weight sensor disposed below the collector.

5. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 1, characterized in that, The intelligent control and risk assessment analysis module includes: The industrial control computer integrates a data acquisition card, a motion control card, and a fieldbus communication interface card. It is used to receive monitoring signals from the flow meter, the miniature pressure sensor, the weight sensor, the ASC area array camera, and the servo motor, and to send control commands to the water pressure loading subsystem and the drive assembly. a memory storing executable instructions configured to execute a particle swarm optimization control algorithm, a physical neural network analysis model, and damage identification logic.

6. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 5, characterized in that, The intelligent control and risk assessment analysis module executes an adaptive search mode, which includes: defining a swarm consisting of a plurality of particles, mapping a position of each particle in a search space as a vector combination of water pressure loading values and vibration frequencies; monitoring real-time feedback of water inrush flow prediction values, stress overrun risk indexes, and construction disturbance risk indexes, and calculating a fitness function value with a water inrush risk comprehensive index as a core; updating a velocity vector and a position vector of the particle according to deviations of the current position of the particle and a historical best position and deviations of the current position of the particle and a swarm best position, and combining an inertia factor, an individual learning factor, and a global learning factor; using a ramp generator to convert the updated position vector into a smooth control curve to drive the water pressure booster and the servo motor to perform gradual loading until a critical threshold of water inrush or mud inrush occurrence is searched.

7. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 5, characterized in that, The intelligent control and risk assessment analysis module executes structural damage identification, which includes: in an initial state without water pressure, controlling the servo motor to output a fixed frequency excitation, collecting torque current and calculating a reference torque characteristic value; in a loading process, collecting a q-axis torque current of the servo motor in real time, and calculating a total equivalent moment of inertia of a system including a mass of material in the rock-soil cavity and an equivalent viscous damping coefficient using a dynamic balance equation; calculating a deviation of a real-time torque root mean square value from the reference torque characteristic value, and determining that structural damage or a hidden cavity is formed inside the rock-soil sample when the deviation exceeds a preset range.

8. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 5, characterized in that, The intelligent control and risk assessment analysis module runs a physical neural network verification unit, which includes: building a deep learning network architecture, embedding Navier-Stokes equations for controlling fluid motion and Allen-Cahn equations for controlling phase interface evolution into a loss function as regularization terms; inputting observation data collected by the flowmeter and the ASC area array camera into the deep learning network architecture, and synchronously training equation parameters and network weights by minimizing a mean square error loss function; outputting a flow field and a phase field distribution satisfying physical law constraints, and monitoring the loss function value, and marking a physical inflection point of water inrush passage formation when the loss function value exceeds a tolerance range.

9. The test system for simulating water and mud inrush of a submarine tunnel under vibration action according to claim 5, characterized in that, The intelligent control and risk assessment analysis module runs a time series data prediction unit, which includes: using an ARIMA model to perform linear trend prediction on pressure fluctuation data in a short period, the ARIMA model including autoregressive coefficients and moving average coefficients; using a long short-term memory network model to perform nonlinear trend prediction on water inrush flow evolution data in a long period, and capturing long sequence dependencies through a forget gate, an input gate, and a state update mechanism; superimposing the linear trend prediction result and the nonlinear trend prediction result to output a time series prediction result of water inrush disaster evolution.

10. A test method for simulating water and mud inrush of a submarine tunnel under vibration, applied to the test system for simulating water and mud inrush of a submarine tunnel under vibration according to any one of claims 1-9, characterized in that, including the following steps: S1, prepare the rock-soil sample to be tested and load it into the visual rock-soil water inrush simulation and multi-field loading module, and saturate the rock-soil sample to be tested; S2, start the multi-degree-of-freedom servo feedback vibration module to generate vibration load, collect servo motor torque current to establish a baseline, and apply confining pressure; S3, use the intelligent control and risk assessment analysis module to run the particle swarm optimization algorithm to generate a control parameter combination of water pressure and vibration frequency, and drive the actuator to perform adaptive loading search; S4, during the loading process, the internal structure damage state of the rock-soil sample to be tested is inversed according to the deviation of the real-time torque current and the baseline; S5, collect the outburst mixture through the hierarchical multi-source data acquisition module, and measure the cumulative weight of different particle size sediments and the fluid flow rate; S6, use the physical neural network model to perform physical consistency verification on the collected data, and combine the time series prediction model to evaluate the disaster risk.

Images