Deep underground continuous wall mechanical method obstacle clearance intelligent equipment and construction method

Abstract

The application discloses a kind of deep underground continuous wall mechanical method barrier removal intelligent equipment and construction method, comprising: vertical positioning system, including full-rotation drilling machine, high-precision verticality sensor and multi-section adjustable steel casing, for establishing vertical channel and real-time monitoring adjustment verticality;Horizontal rock breaking execution system is set in the bottom of the vertical positioning system;Intelligent sensing and navigation system, including three-dimensional detection module, inertial navigation positioning module and multi-source sensing monitoring module, for detecting underground obstacle form, accurately positioning equipment posture and real-time monitoring rock breaking parameter;Intelligent control and decision system, signal connection intelligent sensing and navigation system and horizontal rock breaking execution system.The application realizes "accurate, efficient, low disturbance, digitized" intelligent barrier removal operation, improves deep barrier removal positioning accuracy, avoids adaptive rock breaking ability, low intelligent degree and barrier removal backfill disconnection.

Description

Technical Field

[0001] This invention relates to the field of underground construction engineering technology, and in particular to an intelligent equipment and construction method for mechanical obstacle removal of deep underground continuous walls. Background Technology

[0002] As urban underground space development deepens, newly constructed subway shield tunnels face increasingly severe technical challenges when traversing existing subway stations, diaphragm walls, interlocking piles, and other obstacles. When shield tunnels need to pass through diaphragm walls exceeding 50 meters in depth, achieving high-precision, low-disturbance obstacle removal within a limited space has become a bottleneck restricting further development of urban underground space.

[0003] Currently, traditional obstacle removal methods mainly include manual chiseling, core drilling, and vertical grinding with a full-rotation drilling rig. However, these methods have the following technical drawbacks in practical applications: 1. Limited Clearance Depth and Accuracy: Traditional full-rotation drilling rigs or vertical casing methods face significantly increased challenges in verticality control at depths exceeding 35 meters, making precise horizontal positioning and removal difficult. When obstacles are located beneath existing structures, traditional methods can easily damage them and cannot achieve "zero-space" operation. The verticality control accuracy of current full-rotation drilling rigs is generally 1 / 400 (approximately 0.25%), and the cumulative horizontal deviation can exceed 200mm at deep depths, failing to meet the clearance accuracy requirements of shield tunnels.

[0004] 2. Poor adaptability: Deep diaphragm walls often use high-strength concrete of C35 or higher, with high reinforcement ratios, large rebar diameters, and steel joints. The cutting parameters (speed, feed force, cutting mode) of existing equipment cannot be dynamically adjusted according to the actual obstacle strength, which easily leads to drill jamming, uneven wear, or equipment overload, resulting in low construction efficiency or even equipment damage.

[0005] 3. Low level of intelligence: Existing technologies lack a real-time perception and feedback mechanism for the shape of underground obstacles, equipment posture and ground disturbance. They mainly rely on the experience of operators to make judgments. During the construction process, it is impossible to form a visual digital archive, the obstacle clearing effect is difficult to quantify and verify, and there are safety risks for personnel to go down into the well.

[0006] 4. Disconnection between obstacle removal and backfilling: Existing methods mostly adopt a step-by-step operation mode of "clearing obstacles first and then backfilling". If the cavities formed by obstacle removal are not backfilled in time, they are prone to cause ground subsidence and secondary disturbance to the surrounding environment. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings mentioned above by providing an intelligent equipment and construction method for mechanical obstacle removal in deep underground continuous walls, achieving "precise, efficient, low-disturbance, and digital" intelligent obstacle removal operations, improving the positioning accuracy of deep obstacle removal, avoiding poor adaptive rock breaking ability, low level of intelligence, and disconnection between obstacle removal and backfilling.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: an intelligent equipment and construction method for mechanical obstacle removal of deep underground diaphragm walls, comprising: The vertical positioning system, including a full-rotation drilling rig, a high-precision verticality sensor, and a multi-section adjustable steel casing, is used to establish a vertical channel and ensure that the vertical channel deviation is less than 0.1% through closed-loop feedback control of the high-precision verticality sensor. A horizontal rock-breaking execution system is installed at the bottom of the vertical positioning system and performs horizontal rock-breaking operations inside the steel casing. It includes a telescopic arm, a rotating base, and a multi-functional integrated rock-breaking head. The multi-functional integrated rock-breaking head is installed at the bottom of the steel casing and connected to the rotating base through the telescopic arm. The intelligent sensing and navigation system includes a 3D detection module, an inertial navigation and positioning module, and a multi-source sensing and monitoring module. It is used to perform 3D morphological detection of underground obstacles within the steel casing, accurately locate the horizontal rock-breaking execution system, and monitor cutting force, vibration, and temperature parameters in real time during the rock-breaking process. The 3D detection module is installed at the front end of the multi-functional integrated rock-breaking head. The inertial navigation and positioning module uses a fiber optic inertial measurement unit combined with a ground total station for real-time calculation. The multi-source sensing and monitoring module includes a cutting force sensor, a vibration sensor, and a temperature sensor, which are respectively installed on the spindle, housing, and hydraulic lines of the milling module. The signals collected by the 3D detection module and the multi-source sensing and monitoring module are packaged by the downhole data preprocessing unit and transmitted to the ground control station via industrial Ethernet. The ground control station performs multi-source data fusion calculation and issues attitude correction commands to the inertial navigation and positioning module based on the calculation results. The inertial navigation and positioning module receives commands from the ground control station and performs real-time calibration and positioning calculation of the equipment attitude. The adaptive rock-breaking algorithm module adopts a two-layer control architecture, including: The upper layer is a deep neural network prediction model, which takes the obstacle thickness, reinforcement density, historical cutting force spectrum and current depth stratum information of the three-dimensional detection module as input variables, and outputs the optimal rock breaking mode combination, initial rotation speed ω0, initial feed speed v0 and initial water jet pressure p0. The three-dimensional detection module 31 is a three-dimensional sonar detection module or a three-dimensional radar detection module. The lower layer is a fuzzy PID controller, which takes the deviation e between the actual cutting force and the target cutting force, the rate of change of the deviation ec, and the vibration amplitude A as input variables, and outputs the speed regulation Δω, the feed rate regulation Δv, and the water jet pressure regulation Δp. The prediction output of the deep neural network prediction model is used as the initial set value of the fuzzy PID controller. The real-time adjustment of the fuzzy PID controller is fed back to the horizontal rock breaking execution system, forming a two-level hierarchical control of "macro mode selection - micro parameter fine-tuning". The initial rotational speed ω0, initial feed rate v0, and initial water jet pressure p0 are superimposed with the corresponding speed adjustment Δω(t), feed rate adjustment Δv(t), and water jet pressure adjustment Δp(t) to form real-time rotational speed ω(t), real-time feed rate v(t), and real-time water jet pressure p(t), which are used as the final input parameters of the horizontal rock-breaking execution system (2). Their time-domain expressions are as follows: ; The intelligent control and decision-making system also includes a digital twin model, which maps real-time rock breaking parameters, equipment posture, and obstacle damage status to a three-dimensional visualization interface, archives data from each obstacle clearing operation, and continuously optimizes the deep neural network model through incremental learning. The intelligent control and decision-making system comprises a vertical positioning system, a horizontal rock-breaking execution system, an intelligent sensing and navigation system, and an intelligent control and decision-making system connected sequentially to form a closed-loop control circuit. The intelligent sensing and navigation system transmits three-dimensional detection signals and multi-source sensor monitoring signals to the intelligent control and decision-making system. The intelligent control and decision-making system generates control commands based on these signals and sends them to the horizontal rock-breaking execution system to dynamically adjust rock-breaking parameters according to the real-time morphology of underground obstacles. The intelligent control and decision-making system includes a ground control station and an adaptive rock-breaking algorithm module deployed thereon, used to construct a digital twin model and dynamically adjust rock-breaking parameters based on the adaptive algorithm. The real-time waste residue treatment system includes a negative pressure slag suction port, a spiral conveying pipeline, and a ground screening device, which is used to realize simultaneous rock breaking and slag removal operations.

[0009] Furthermore, the multi-functional integrated rock-breaking head includes a turret-type mounting platform and a high-speed rotary milling module, a static splitting module, and a high-pressure water jet assisted cutting module mounted on it. The turret-type mounting platform achieves rapid switching between the modules through hydraulic drive. The intelligent control and decision-making system automatically switches the rock-breaking module through the turret-type mounting platform based on the cutting force mutation signal or vibration abnormal signal fed back by the multi-source sensor monitoring module, and starts a three-level collaborative operation mode of milling, splitting, and water jet cutting.

[0010] Furthermore, the high-speed rotary milling module includes a planetary gear reducer, a high-speed hydraulic motor, and a cylindrical milling cutter head; The diameter of the cylindrical milling cutter head is adjustable. The cylindrical milling cutter head is in the shape of a hollow cylinder, and the cutting teeth are arranged only around the outer circumference of the cylinder end face. The inside of the cylinder is a hollow slag discharge channel. The outer diameter of the cylindrical milling cutter head is 500mm-1500mm and the wall thickness is 20mm-50mm. The cutting teeth arranged around the outer circumference reduce the cutting torque when breaking rock in the horizontal direction compared to a full-disc tooth structure.

[0011] Furthermore, the static splitting module includes multiple splitting gun groups arranged in a ring around the circumference of the multi-functional integrated rock-breaking head, a high-pressure hydraulic pump station, and a radial expansion sleeve. Each splitting gun group contains 3-5 splitting guns, which are evenly distributed along the circumference. The splitting gun has a tapered guide head at the front end and adopts a multi-stage expansion structure inside, with a single gun splitting force ≥2000kN; The radial expansion sleeve generates a uniform radial expansion force under high-pressure hydraulic drive, which is used for vibration-free crushing of high-strength concrete.

[0012] Furthermore, the high-pressure water jet assisted cutting module includes an ultra-high pressure plunger pump, a three-axis adjustable rotating water jet nozzle, and an abrasive mixing chamber; The three-axis adjustable rotating water jet nozzle can adjust the spray angle within the range of ±45° horizontally and ±30° vertically. The front end of the three-axis adjustable rotating water jet nozzle is equipped with an automatic anti-clogging device. The abrasive mixing chamber has a built-in abrasive supply pump. When the three-dimensional detection module identifies the main reinforcement signal, the intelligent control and decision-making system automatically adds abrasive to the high-pressure water flow to form an abrasive water jet to cut the high-strength steel bars.

[0013] The intelligent construction method for mechanical obstacle removal of deep underground diaphragm walls is characterized by the following steps: Step S1: Vertical channel establishment and precise orientation. The full-rotation drilling rig, in conjunction with the steel casing, is vertically lowered to the predetermined depth of the diaphragm wall. Closed-loop control using a high-precision verticality sensor ensures that the vertical channel deviation is less than 0.1%. Step S2: 3D obstacle modeling and path planning, lower the horizontal rock breaking execution system to the working depth, start the 3D detection module to detect the shape and reinforcement information of the underground continuous wall, build a 3D visualization model and plan the optimal rock breaking path; Step S3: Adaptive rock breaking and dynamic control. Rock breaking is initiated along the planned path. The intelligent sensing and navigation system provides real-time feedback on cutting force, vibration, and temperature data. The adaptive rock breaking algorithm module dynamically adjusts the rock breaking parameters based on the digital twin model. Step S4: Verification of obstacle removal effect and synchronous backfilling. The obstacle removal area is re-measured using a three-dimensional detection module. After confirming that there are no residual protrusions, pressure-maintaining grouting and backfilling are carried out simultaneously. Step S5: Equipment recovery and data archiving. The horizontal rock breaking execution system and steel casing are recovered, and the entire process data is automatically archived and used for incremental model learning.

[0014] The intelligent construction method for mechanical obstacle removal of deep underground diaphragm walls for equipment includes step S3 as follows: The deep neural network prediction model receives detection data from the three-dimensional detection module and predicts and outputs rock breaking mode combinations and initial setting parameters. The rock-breaking operation is carried out along the planned path. The fuzzy PID controller monitors the cutting force deviation and the rate of change of deviation in real time, and automatically adjusts the rotation speed, feed rate and water jet pressure. When a sudden change in cutting force or abnormal vibration is detected, the rock-breaking module is switched via the turret-type installation platform to initiate the collaborative operation mode.

[0015] Furthermore, in step S4, the pressure-maintaining grouting backfill uses cement-water glass dual-liquid grout, and the grouting pressure is controlled at 0.3-0.5MPa.

[0016] The beneficial effects of this invention are reflected in: 1. This invention provides high-precision and high-depth obstacle removal capabilities: By combining fiber optic gyroscope verticality closed-loop control with inertial navigation positioning, it achieves millimeter-level precision removal of underground continuous walls at depths exceeding 50 meters, avoiding accidental damage to the existing main structure.

[0017] 2. This invention, adaptive rock breaking and intelligent decision-making: adopts a two-layer control architecture of "DNN prediction + fuzzy PID real-time control", realizing full-coverage control from macro mode selection to micro parameter fine-tuning. To verify the effectiveness of the above two-layer control architecture, a simulation model of the rock breaking process control system was established in the Matlab / Simulink environment. In the simulation model, the controlled object is the hydraulic motor-milling head system (the transfer function is obtained based on the actual engineering parameters). The interference signal simulates typical working conditions such as sudden change in concrete strength (C30-C80) and steel bar obstruction (step change in cutting force) during the clearing of underground continuous wall obstacles. The simulation compares three control strategies: (1) traditional PID control; (2) single fuzzy PID control; (3) the two-layer control of DNN prediction + fuzzy PID of this invention. Simulation results show that when encountering a sudden change in concrete strength (from C50 to C80), the cutting force overshoot of traditional PID control reaches 35%-45%, with a settling time of 3-5 seconds. The overshoot of single fuzzy PID control is reduced to 20%-25%, with a settling time of about 1.5-2.5 seconds. However, the dual-layer control architecture of this invention, due to the early identification of the strength change trend and the preset initial parameters by the DNN prediction model, can control the overshoot within 10%-15%, and shorten the settling time to 0.8-1.2 seconds. In the simulated steel bar obstruction condition, traditional PID control is prone to drill jamming due to integral saturation (the probability of drill jamming in the simulation is about 65%). However, the dual-layer control architecture of this invention, through timely tool retraction and parameter adjustment of fuzzy PID, can reduce the probability of drill jamming to below 25%. Based on the above simulation results, the dual-layer control architecture of this invention reduces the cutting force fluctuation range by more than 40% and the risk of stuck drill by more than 60% compared with the traditional single PID control. The above simulation data is obtained based on the control system model established on the Matlab / Simulink platform. The control parameters will be further verified and optimized through field industrial tests in the future.

[0018] 3. This invention features multi-functional integration and collaborative operation: The turret-type integrated three-module design enables a single device to handle various obstacles such as concrete, steel bars, and high-strength interlayers. The three-level collaborative operation mode avoids the problems of single-method bottlenecks or low efficiency. Among them, the cylindrical milling cutter head adopts an outer circumferential tooth arrangement design, with the cutter teeth arranged only around the outer circumference of the cylindrical end face. The hollow interior forms a slag discharge channel. Compared with the full-disc tooth arrangement structure, the cutting torque requirement is reduced by more than 30%, and it is conducive to the circulation of mud to carry waste slag, making it less likely to cause tooth jamming or drill bit jamming. Theoretical and experimental basis for reduced cutting torque: The difference in cutting torque between peripheral toothed structures and full-disc toothed structures can be explained by cutting mechanics analysis. In horizontal rock breaking scenarios, the cutting resistance of the milling cutter head mainly comes from the interaction between the cutter teeth and the concrete. In a full-disc toothed structure, all the cutter teeth (usually 80-120) participate in cutting simultaneously, and the total cutting torque is the sum of the torques of the cutting forces of each cutter tooth. In contrast, in a peripheral toothed structure, only the cutter teeth around the outer circumference of the end face (usually 24-36) participate in cutting, and there are no cutter teeth in the central area. During horizontal rock breaking, after the concrete in the central area of ​​the cross-section is cut around the periphery, the stress is released due to the release of lateral constraints, and it will naturally break and fall off from the cross-section without the need for direct cutting by cutter teeth.

[0019] To verify the above theoretical analysis, indoor scaled-down cutting tests were conducted. The test conditions were: concrete specimen grade C50, cutting depth 50mm, feed rate 20mm / min. Comparative cutting tests were conducted using scaled-down milling cutters with peripheral toothed structure and full-disc toothed structure. The test results showed that the measured cutting torque of the peripheral toothed structure was reduced by 32%-38% compared with that of the full-disc toothed structure, verifying the torque reduction effect of the peripheral toothed structure in horizontal rock breaking scenarios. The above analysis is based on rock cutting mechanics theory (Evans cutting theory), and the test data were obtained by the applicant through testing.

[0020] 4. In this invention, the radial expansion sleeve in the static splitting module generates a uniform and controllable radial expansion force through hydraulic drive, which can realize vibration-free and low-noise crushing of high-strength concrete. It is particularly suitable for environments such as existing stations and tunnels that are sensitive to vibration. At the same time, the multi-stage expansion structure of the radial expansion sleeve can output a splitting force of more than 2000kN in narrow boreholes, effectively crushing large pieces of residual concrete and avoiding impact disturbance.

[0021] 5. This invention features low disturbance and high safety: By combining "vertical casing wall protection + horizontal precision rock breaking + synchronous backfilling" with remote intelligent control, soil disturbance is greatly reduced, and fully mechanized safe operation without personnel entering the well is achieved.

[0022] 6. This invention features full-process digitalization and self-evolution capabilities: It establishes a full-process digital twin model from detection and construction to acceptance. The self-updating mechanism based on incremental learning enables the equipment to automatically optimize parameter strategies after multiple operations, greatly reducing the dependence on operator experience and providing reliable data support for the refined construction and record management of urban underground spaces. Attached Figure Description

[0023] Figure 1 This is a three-dimensional structural view of the present invention; Figure 2 This is a three-dimensional structural view of the horizontal rock-breaking execution system and intelligent sensing and navigation system of the present invention; Figure 3 This is a three-dimensional structural view of the turret-type installation platform of the present invention; Figure 4 This is a three-dimensional structural view of the multifunctional integrated rock-breaking head and multi-source sensing monitoring module of the present invention. Figure 5 This is a front view of the multifunctional integrated rock-breaking head of the present invention; Figure 6 This is a control architecture diagram of the adaptive rock-breaking algorithm module of the present invention; Figure 7 This is a flowchart of the construction method of the present invention.

[0024] In the picture: 1. Vertical positioning system; 11. Full-rotation drilling rig; 111. Hydraulic correction mechanism; 12. High-precision verticality sensor; 13. Multi-section adjustable steel casing; 2. Horizontal rock-breaking actuator system; 21. Telescopic boom; 22. Rotating base; 23. Multifunctional integrated rock breaking head; 231. Turret-type installation platform; 232. High-speed rotary milling module; 2321. Star gear reducer; 2322. High-speed hydraulic motor; 2323. Cylindrical milling cutter head; 233. Static splitting module; 234. High-pressure water jet assisted cutting module; 2341. Ultra-high pressure plunger pump; 2342. Three-axis adjustable rotary water jet nozzle; 2343. Abrasive mixing chamber; 3. Intelligent sensing and navigation system; 31. Three-dimensional detection module; 32. Inertial navigation and positioning module; 33. Multi-source sensing and monitoring module; 4. Intelligent control and decision-making system; 5. Real-time waste residue treatment system. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1-7 This invention discloses an intelligent equipment and construction method for mechanical obstacle removal of deep underground continuous walls, including: a vertical positioning system 1, a horizontal rock breaking execution system 2, an intelligent sensing and navigation system 3, an intelligent control and decision-making system 4, and a waste residue real-time treatment system 5. The vertical positioning system 1 includes a full-rotation drilling rig 11, a high-precision verticality sensor 12, and a multi-section adjustable steel casing 13. It is used to establish a vertical channel and control the vertical channel deviation to be less than 0.1% through closed-loop feedback control of the high-precision verticality sensor 12. The high-precision verticality sensor 12 adopts a fiber optic gyroscope, which is installed at the bottom of the steel casing 13. The sampling frequency is not less than 100Hz and the measurement accuracy is better than 0.01°. The full-rotation drilling rig 11 is equipped with a hydraulic correction mechanism 111, which can automatically adjust the verticality of the casing according to the sensor feedback. The steel casing 13 is in the form of a multi-section connection and extends vertically into the formation. The high-precision verticality sensor 12 is located near the bottom of the steel casing 13. The horizontal rock-breaking execution system 2 is set at the bottom of the vertical positioning system 1 and performs horizontal rock-breaking operations inside the steel sleeve 13. It includes a telescopic arm 21, a rotating base 22 and a multi-functional integrated rock-breaking head 23. The multi-functional integrated rock-breaking head 23 is set at the bottom of the steel sleeve 13 and connected to the rotating base 22 through the telescopic arm 21. The intelligent sensing and navigation system 3 includes a 3D detection module 31, an inertial navigation and positioning module 32, and a multi-source sensing and monitoring module 33. It is used to detect the shape of underground obstacles, accurately locate the equipment's attitude, and monitor rock-breaking parameters in real time. The 3D detection module 31 is installed at the front end of the multi-functional integrated rock-breaking head 23 and uses a 3D sonar / radar detection module with a detection frequency range of 100kHz-500kHz. It can penetrate concrete to a depth of up to 500mm and has an imaging resolution better than 10mm. The multi-source sensing and monitoring module 33 includes a cutting force sensor, a vibration sensor, and a temperature sensor, which are respectively installed on the spindle, housing, and hydraulic lines of the milling module. The inertial navigation and positioning module 32 uses fiber optics. The inertial measurement unit (IMU), with built-in attitude calculation and Kalman filtering algorithms, is used to receive attitude correction commands from the ground control station and perform real-time calibration and positioning calculation of the equipment attitude. Its positioning error is periodically corrected by combining with the ground total station. The signals collected by the three-dimensional detection module 31 and the multi-source sensing monitoring module 33 are converted by A / D and conditioned by the downhole data preprocessing unit, and then transmitted to the ground control station via industrial Ethernet. The ground control station performs fusion calculation on the multi-source sensing data and sends the calculated equipment attitude deviation to the inertial navigation and positioning module 32 in the form of attitude correction commands, forming a closed-loop signal interaction link of "acquisition-upload-calculation-send-execution". The adaptive rock-breaking algorithm module adopts a two-layer control architecture, including: The upper layer is a deep neural network prediction model. The obstacle thickness, reinforcement density, historical cutting force spectrum and current depth stratum information of the three-dimensional detection module 31 are used as input variables. The output is the optimal rock breaking mode combination, initial rotation speed ω0, initial feed speed v0 and initial water jet pressure p0. The three-dimensional detection module 31 is a three-dimensional sonar detection module or a three-dimensional radar detection module. The lower layer is a fuzzy PID controller, which takes the deviation e between the actual cutting force and the target cutting force, the rate of change of the deviation ec, and the vibration amplitude A as input variables, and outputs the speed regulation Δω, the feed rate regulation Δv, and the water jet pressure regulation Δp. The predicted output of the deep neural network prediction model is used as the initial set value of the fuzzy PID controller. The real-time adjustment of the fuzzy PID controller is fed back to the horizontal rock-breaking execution system 2, forming a two-level hierarchical control of "macro mode selection - micro parameter fine-tuning". In the aforementioned two-layer control architecture, the initial rotational speed output by the deep neural network prediction model... Initial feed rate and initial water jet pressure The speed regulation amount output in real time by the fuzzy PID controller Feed rate adjustment and water jet pressure regulation After being superimposed, the corresponding rotational speed is formed, which acts on the horizontal rock-breaking actuator (2). Real-time feed rate and real-time water jet pressure Its time-domain expression is: ; in, , , The deep neural network prediction model determines the rock breaking mode and initial parameter selection at the macro level based on the obstacle thickness, reinforcement density, historical cutting force spectrum and current depth stratum information obtained by the three-dimensional detection module (31). , , The parameters are fine-tuned at the micro level by the fuzzy PID controller based on the deviation e between the actual cutting force and the target cutting force, the rate of change of the deviation ec, and the vibration amplitude A. The final parameters after the superposition of the two parameters drive the actuators of the horizontal rock breaking system 2 in real time, thereby realizing the two-level hierarchical control of "macro mode selection and micro parameter fine-tuning". The intelligent control and decision system 4, vertical positioning system 1, horizontal rock-breaking execution system 2, intelligent sensing and navigation system 3 and intelligent control and decision system 4 are sequentially connected to form a closed-loop control loop. The intelligent sensing and navigation system 3 transmits three-dimensional detection signals and multi-source sensor monitoring signals to the intelligent control and decision system 4. The intelligent control and decision system 4 generates control commands based on the signals and sends them to the horizontal rock-breaking execution system 2 to dynamically adjust the rock-breaking parameters according to the real-time shape of the underground obstacle. The intelligent control and decision system 4 includes a ground control station and an adaptive rock-breaking algorithm module deployed on it, which is used to build a digital twin model and dynamically adjust the rock-breaking parameters based on the adaptive algorithm. The intelligent control and decision-making system 4 also includes a digital twin model, which maps real-time rock breaking parameters, equipment posture, and obstacle damage status to a three-dimensional visualization interface, archives data from each obstacle clearing operation, and continuously optimizes the deep neural network model through incremental learning. The waste residue real-time treatment system 5 includes a negative pressure slag suction port, a spiral conveying pipeline and a ground screening device, which is used to realize the simultaneous operation of rock breaking and slag cleaning.

[0027] In a specific embodiment, the multi-functional integrated rock breaking head 23 includes a turret-type mounting platform 231 and a high-speed rotary milling module 232, a static splitting module 233, and a high-pressure water jet assisted cutting module 234 mounted on it. The turret-type mounting platform 231 achieves rapid switching of each module through hydraulic drive. The intelligent control and decision-making system 4 automatically switches the rock breaking module through the turret-type mounting platform 231 based on the cutting force mutation signal or vibration abnormal signal fed back by the multi-source sensor monitoring module 33, and starts the three-level collaborative operation mode of milling, splitting and water jet cutting to achieve rapid switching of each operation module. The multi-functional integrated rock breaking head 23 has a turret-type mounting platform 231 inside, and the three functional modules are mounted on the turret at a 120° angle and achieve rapid cutting through hydraulic motor drive.

[0028] In one embodiment, the high-speed rotary milling module 232 includes a planetary gear reducer 2321, a high-speed hydraulic motor 2322, and a cylindrical milling cutter head 2323. The high-speed rotary milling module 232 is installed at the first station of the turret mounting platform 231. The cylindrical milling cutter head 2323 has an adjustable diameter and is in the shape of a hollow cylinder. The cutting teeth are arranged only around the outer circumference of the cylinder end face. The inside of the cylinder is a hollow slag discharge channel. The outer diameter of the cylindrical milling cutter head 2323 is 500mm-1500mm, and the wall thickness is 20mm-50mm. The cutting teeth are made of cemented carbide or diamond composite material. The cutting teeth arranged around the outer circumference reduce the cutting torque when breaking rock in the horizontal direction compared to the full-disc tooth structure. Selection criteria for the outer diameter range (500mm-1500mm): The outer diameter range of the 2323 cylindrical end mill head is determined based on a comprehensive consideration of the following factors: 1. The common thickness of diaphragm walls is 600mm-1200mm. The outer diameter of the milling cutter head should not be less than the thickness of the diaphragm wall to ensure that the target section can be completely removed in a single stroke. 2. The inner diameter of the steel sleeve 13 limits the maximum outer diameter of the milling cutter head. In this embodiment, the inner diameter of the steel sleeve is 1600mm-1800mm, and the outer diameter of the milling cutter head needs to be at least 100mm smaller than the inner diameter of the steel sleeve to ensure free rotation and mud circulation space. 3. When the outer diameter is less than 500mm, the single rock-breaking coverage area is too small (less than 0.2m²), which seriously affects the construction efficiency; 4. When the outer diameter is greater than 1500mm, normal rotation operation is not possible due to the limitation of the inner diameter of the steel sleeve. Therefore, an outer diameter range of 500mm-1500mm is a reasonable choice that balances the obstacle clearing efficiency and equipment space constraints. The selection criteria for the wall thickness range (20mm-50mm): The wall thickness of the 2323 cylindrical milling cutter head is determined by strength calculations based on the cutting torque and bending moment borne by the cutter head. According to the formula for the torsional strength of a thin-walled cylinder in mechanics of materials, τ=T / (2πr²t) (where T is the cutting torque, r is the average radius of the cylinder, and t is the wall thickness), under the same cutting torque, the wall thickness is inversely proportional to the stress in the cylinder. Calculations show that under the condition of a maximum design cutting torque Tmax=120kN·m: 1. When the wall thickness is less than 20mm, the maximum stress of the cylinder exceeds the allowable stress of the material (in this embodiment, the cutter head material is 42CrMo, and the allowable stress [σ] = 850MPa), which poses a risk of structural failure. 2. For wall thicknesses between 20mm and 50mm, the cylinder stress meets strength requirements, with a safety factor of 1.5-3.0; 3. When the wall thickness is greater than 50mm, although the strength redundancy increases, the weight of the cutter head increases significantly (the weight increases by about 15%-20% for every 10mm increase in wall thickness), which affects the horizontal control accuracy and dynamic response performance of the telescopic arm 21. Considering structural strength, control accuracy and economy, a wall thickness of 20mm-50mm is the optimal selection range.

[0029] In the embodiments, the effects of the outer diameter range and wall thickness range of the cylindrical milling cutter head 2323 on the obstacle clearing efficiency, slag removal performance, cutting torque, and wear life are illustrated below:

[0030] In a specific embodiment, the static splitting module 233 includes multiple splitting gun groups arranged in a ring around the circumference of the multi-functional integrated rock breaking head 23, a high-pressure hydraulic pump station, and a radial expansion sleeve. The static splitting module 233 is installed at the second station of the turret-type installation platform 231, including four splitting gun groups arranged in a ring around the circumference of the rock breaking head, a high-pressure hydraulic pump station, and a radial expansion sleeve. The front end of the splitting gun body is provided with a conical guide head, and the internal structure adopts a three-stage expansion structure. The splitting force of a single gun is ≥2000kN. The radial expansion sleeve is fitted outside the expansion section of the splitting gun and is made of high-strength alloy steel. Its inner wall slides in fit with the wedge block of the multi-stage expansion structure. When the high-pressure hydraulic pump station injects ultra-high pressure hydraulic oil into the splitting gun, the wedge block advances axially, forcing the radial expansion sleeve to expand uniformly outward in the radial direction. The maximum expansion amount is 3mm-5mm, generating a radial splitting force of more than 100MPa, causing concrete or rock to crack and break from the inside. This process is vibration-free and low-noise, avoiding impact disturbance to the surrounding strata. The calculation basis of radial splitting force: The radial splitting force F is calculated based on the rated working pressure P of the hydraulic system and the effective area A of the splitting gun plunger, that is, F=P×A.

[0031] In this embodiment, the rated working pressure of the hydraulic system is P=70MPa, the diameter of the splitting gun plunger is D=190mm, and the effective area of ​​the plunger is A=π×D² / 4=28353mm². After calculation, the theoretical splitting force of a single gun is F=70MPa×28353mm²≈1985kN. Considering the mechanical efficiency coefficient η=0.95-0.98, the actual output splitting force of a single gun is F_actual=F×η≈1900-1950kN. When 4 sets of splitting guns (each set contains 3-5 splitting guns) act on the concrete section at the same time, the total splitting force can reach more than 7600kN, which is enough to cause through cracks in C50-C80 high-strength concrete. The above splitting force values ​​are derived from theoretical calculations based on the rated parameters of the hydraulic system and the piston area. Those skilled in the art can obtain different splitting force outputs by adjusting the working pressure of the hydraulic system (50-100MPa) and the piston diameter (150-220mm) according to actual engineering needs. All of these are direct variations of the technical solution of this invention. Each splitting gun group contains 3-5 splitting guns, which are evenly distributed along the circumference. The splitting gun has a tapered guide head at the front end and adopts a multi-stage expansion structure inside.

[0032] In one embodiment, the high-pressure water jet assisted cutting module 234 includes an ultra-high pressure plunger pump 2341, a three-axis adjustable rotating water jet nozzle 2342, and an abrasive mixing chamber 2343; The three-axis adjustable rotating water jet nozzle 2342 can adjust the spray angle within the range of ±45° horizontally and ±30° vertically. The front end of the three-axis adjustable rotating water jet nozzle 2342 is equipped with an automatic anti-clogging device, and the maximum spray pressure reaches 400MPa. The abrasive mixing chamber 2343 can automatically add abrasives such as garnet sand for cutting high-strength steel bars. The abrasive mixing chamber 2343 has a built-in abrasive supply pump. When the three-dimensional detection module 31 identifies the main reinforcement signal, the intelligent control and decision system 4 controls the automatic addition of abrasive to the high-pressure water flow to form an abrasive water jet to cut the high-strength steel bar. The high-pressure water jet auxiliary cutting module 234 is installed at the third station of the turret-type installation platform 231.

[0033] In one embodiment, the abrasive mixing chamber 2343 automatically adds abrasive for cutting high-strength reinforcing bars. The technical principle is as follows: pure high-pressure water jets (with a maximum jet pressure of 400 MPa) have a good crushing effect on brittle materials such as concrete. However, for tough metal materials such as high-strength reinforcing bars (e.g., HRB400 grade, tensile strength ≥ 400 MPa), due to the low density and light mass of water, the impact momentum is limited, making it difficult to effectively destroy the metal lattice structure. To solve this problem, the abrasive mixing chamber 2343 has a built-in abrasive supply pump and flow regulating valve. When the three-dimensional detection module 31 identifies the presence of main reinforcing bar signals within the underground continuous wall during rock breaking, the intelligent control and decision-making system 4 determines that the current working condition needs to be switched to the reinforcing bar cutting mode, and then sends a start command to the abrasive mixing chamber (2343). The abrasive supply pump operates at a set flow rate (2.5-3... in this embodiment). High-hardness abrasive particles such as garnet sand (80-120 mesh, Mohs hardness 7.5-8.0) are injected into a high-pressure water stream at a flow rate of 0 kg / min. The high-pressure water generated by the ultra-high-pressure plunger pump 2341 carries the abrasive particles and is ejected through the triaxial adjustable rotating water jet nozzle 2342, forming a solid-liquid two-phase high-speed jet. The abrasive particles are accelerated to supersonic speed (Mach number 2-3) and the reinforcing steel is removed by high-frequency micro-cutting and erosion. The single cutting depth can reach 35-45 mm, thereby achieving precise and efficient cutting of high-strength reinforcing steel with a diameter of 32 mm. After the cutting is completed, the intelligent control and decision system 4 judges that the reinforcing steel has been cut according to the retest signal of the three-dimensional detection module 31, and then controls the abrasive supply pump to stop feeding to avoid abrasive waste and excessive wear of the nozzle. Through the above-mentioned automatic abrasive addition and intelligent control, the high-pressure water jet assisted cutting module 234 can work in conjunction with the high-speed rotary milling module 232 and the static splitting module 233, enabling a single device to have the ability to handle all working conditions, including large-area concrete demolition, fixed-point cutting of steel bars, and vibration-free splitting of steel joints. The universes of discourse for cutting force deviation e and deviation change rate ec are divided into 7 levels: negative large (NB), negative medium (NM), negative small (NS), zero (ZO), positive small (PS), positive medium (PM), and positive large (PB). A triangular membership function is adopted, and the fuzzy rule base is formulated based on expert experience, with a total of 49 rules. The core rules include: If e is NB and ec is NB, then the speed adjustment Δω is PB (significantly increasing the speed), and the feed rate adjustment Δv is NB (significantly decreasing the feed rate). If e is ZO and ec is ZO, then Δω is ZO and Δv is ZO (keeping the current parameters); If the vibration amplitude exceeds the threshold and e is PS, then Δω is NS and Δv is NS (slightly reduce the rotational speed and feed rate to avoid the resonance frequency). The centroid method is used to resolve fuzzy logic, and the precise adjustment amount is output to the hydraulic servo valve and frequency converter. The digital twin model is developed based on the Unity3D engine. It receives data such as rock-breaking head coordinates, cutting force, vibration, and temperature in real time and presents the rock-breaking process in the form of 3D animation. After each operation, the system automatically uploads "working condition characteristics - control parameters - obstacle clearing effect" as a set of training samples to the cloud database. The DNN model performs incremental learning once a week to continuously optimize the prediction accuracy.

[0034] In one embodiment, the intelligent control and decision-making system 3 also includes a digital twin model, which maps real-time rock breaking parameters, equipment posture, and obstacle damage status to a three-dimensional visualization interface, archives data from each obstacle clearing operation, and continuously optimizes the deep neural network model through incremental learning.

[0035] In the test case To verify the practical engineering effect of the intelligent mechanical obstacle removal equipment and construction method for deep underground diaphragm walls of this invention in deep obstacle removal operations, a field construction test was conducted on an underground diaphragm wall project of an existing operating subway station auxiliary structure in a certain city's rail transit shield tunnel section. The obstacle removal design depth was 45m, the wall thickness was 800mm, the concrete grade was C50, and the reinforcement ratio was 4.2% (the main reinforcement was Φ32mm). HRB400 grade steel bars, arranged in a double layer and bidirectional pattern with a spacing of 150mm, and horizontal distribution bars of Φ20mm with a spacing of 200mm. The strata from top to bottom are: miscellaneous fill layer (thickness 3.0m), silty clay layer (thickness 8.5m), silty clay layer (thickness 13.0m), silty sand layer (thickness 11.0m), silty clay layer (thickness 6.5m), and moderately weathered mudstone layer (thickness 3.0m). The groundwater level is about 2.0m deep. Obstacle removal requirements: A continuous wall section with a width of 3.6m and a height of 6.0m needs to be removed within the shield tunneling area. After obstacle removal, the allowable deviation of the wall bottom elevation is ±50mm, the allowable deviation of the horizontal positioning is ±30mm, and the ground settlement control value is less than 10mm.

[0036] 1. Vertical channel establishment and precise orientation After the full-rotation drilling rig 11 is in place, the first section of the multi-section adjustable steel casing 13 has an outer diameter of 1200mm and a wall thickness of 30mm. It is connected in 9 sections (5m each) with a total length of 48m. It is vertically pressed into the center point of the obstacle removal. A high-precision verticality sensor 12 (fiber optic gyroscope, sampling frequency 100Hz, measurement accuracy better than 0.01°) is used to monitor the verticality of the casing throughout the process. The verticality is checked every 5m when the casing sinks. When the deviation exceeds 0.05%, the hydraulic correction mechanism 111 of the full-rotation drilling rig 11 is activated for adjustment. After the vertical channel was established, the test showed that the verticality deviation of the casing was 0.03%-0.06%, which is better than the design allowable value of 0.1%. The bottom elevation of the casing was -48.03m (the design value is -48.0m). The deviation between the casing center and the design center was +6mm (the allowable deviation is ±20mm). The borehole diameter was 1202-1207mm (the design value is ≥1200mm). The vertical channel took 4.2 hours to establish. The digital twin model synchronously recorded and displayed the verticality change curve of the casing sinking process.

[0037] 2. 3D obstacle modeling and path planning The horizontal rock-breaking system 2 was lowered to a working depth of -45.5m (with a 0.5m protective layer reserved). The 3D detection module 31 (3D sonar / radar, detection frequency 300kHz, concrete penetration depth ≥500mm, imaging resolution ≤10mm) was activated to scan and model the target continuous wall. The 3D detection results are as follows: wall thickness 796-804mm (design 800mm), 4 layers of Φ32mm main reinforcement (2 layers on the soil-facing side and 2 layers on the soil-repellent side), layer spacing 180mm, measured main reinforcement protective layer thickness 45-52mm, Φ20mm horizontal distribution reinforcement spacing 195-210mm (design 200mm), estimated concrete strength 52.8-55.3MPa (design C50), wall bottom elevation -47.95m (design -47.5m), and one obstacle was detected on the left edge of the clearing area. The steel joint (280mm wide, 450mm high) was constructed using a digital twin model based on 3D detection data within 30 seconds, clearly displaying the distribution of reinforcing bars, the location of the steel joint, and the cloud map of concrete strength distribution. The intelligent control and decision-making system 4 planned a "honeycomb" rock-breaking path based on the 3D model, with a grid size of 300mm×300mm, avoiding areas with dense main reinforcement and the location of the steel joint (leaving a 50mm safety distance). The total path length was 19.2m. The DNN prediction model (with obstacle thickness of 800mm, reinforcement density of 4.2%, historical cutting force spectrum, and current depth stratum information as input) output the initial rock-breaking parameters: the rock-breaking mode was "high-pressure water jet assisted milling as the main method", the initial rotation speed was 110rpm, the feed rate was 18mm / min, and the water jet pressure was 350MPa.

[0038] 3. Adaptive rock breaking and dynamic control The rock-breaking operation is carried out along the planned path. The digital twin model synchronously maps the rock-breaking head coordinates, cutting force, vibration, temperature and other data throughout the process, and presents the rock-breaking process in real time in the form of three-dimensional animation. High-speed rotary milling module 232 operating section (path 0-13.0m): The cylindrical milling cutter head 2323 uses a specification of 800mm outer diameter and 30mm wall thickness. In the initial stage (path 0-2.5m), which is a uniform C50 concrete section, the actual cutting force is stable at 78~88kN, the vibration amplitude is 0.7~1.1mm / s, and the cutter tooth temperature is 62~76℃. In the 2.5-5.0m section, a dense area of ​​horizontal ribs (spacing 195mm) is encountered, and the cutting force increases from 85kN to 118kN. The fuzzy PID controller automatically performs the adjustment: the rotation speed is increased to 148rpm, the feed rate is reduced to 10mm / min, and the water jet pressure is adjusted. The cutting force was adjusted to 375 MPa and then dropped to 98-108 kN. The vibration amplitude was controlled within 1.4 mm / s. In the 5.0-9.0 m section, the cutting force entered the uniform concrete zone in the middle of the wall. The cutting force stabilized at 82-92 kN. The digital twin model showed that the rock breaking head trajectory deviated from the planned path by less than 5 mm. In the 9.0-13.0 m section, the cutting force approached the bottom of the wall (-46.0--48.0 m). The concrete strength increased slightly (estimated value 54-55 MPa), and the cutting force rose to 105-120 kN. The controller automatically reduced the feed speed to 8 mm / min and increased the rotation speed to 155 rpm, and the process was completed smoothly. High-pressure water jet assisted cutting module 234 operating section (path 13.0-15.5m): The three-dimensional detection module 31 provides an early warning of a dense area of ​​Φ32mm main ribs ahead (double layer with a spacing of 150mm). The DNN model determines that it is necessary to switch to the high-pressure water jet assisted cutting mode. The system automatically switches to the high-pressure water jet assisted cutting module 234 through the turret-type installation platform 231, starts the abrasive mixing chamber 2343, and the pressure of the ultra-high pressure plunger pump 2341 rises to 380MPa. The three-axis adjustable rotating water jet nozzle 2342 adjusts the spray angle within the horizontal ±30° range to perform fixed-point cutting of the main ribs. The cutting time for each main rib is 45-60s. After the cutting is completed, it switches back to the milling module to continue the operation. The measured water jet cutting parameters are: spray pressure 375-385MPa, abrasive (garnet sand) addition flow rate 2.5-3.0kg / min, and single cutting depth 35-45mm. Static splitting module 233 operating section (path 15.5-17.0m): Encountering a steel section joint (280mm wide, 32mm flange thickness), the DNN model determined that neither milling nor water jetting could handle it efficiently. The system switched to static splitting module 233 via turret mounting platform 231. Four sets of splitting guns were arranged in a ring around the steel section (4 splitting guns per set, 16 guns in total). The maximum expansion of the radial expansion sleeve reached 4.2mm, generating a radial splitting force of over 120MPa. The steel section joint was split into 3 pieces within 15 minutes. The splitting process was vibration-free and low-noise, and the ground monitoring vibration velocity was less than 0.5mm / s. The following are the statistics of cutting force and control data throughout the process: Spindle speed initial setting: 110 rpm, actual fluctuation range: 80-165 rpm, stable value after adjustment: 100-155 rpm, control accuracy: ±5 rpm; Feed rate initial setting: 18 mm / min, actual fluctuation range: 6-22 mm / min, stable value after adjustment: 8-15 mm / min, control accuracy: ±1 mm / min; Water jet pressure initial setting: 350 MPa, actual fluctuation range: 300-390 MPa, stable value after adjustment: 350-380 MPa, control accuracy: ±5 MPa. MPa, cutting force target value 85kN, actual fluctuation range 62-135kN, after adjustment stable value 82-118kN, fluctuation range reduced by 42%, vibration amplitude actual fluctuation range 0.5-2.8mm / s, after adjustment stable value ≤1.6mm / s, over-threshold time less than 3%, cutting tooth temperature fluctuation range 55-115℃, after adjustment stable value 62-95℃, peak temperature reduced by 18%. Compared with traditional manual or single PID control (historical data under the same working conditions), fuzzy PID control reduces the cutting force fluctuation range by 42% and reduces the risk of stuck drill by more than 60%.

[0039] 4. Verification of obstacle removal effectiveness and simultaneous backfilling After rock breaking was completed, the 3D detection module 31 re-measured the cleared area, and the digital twin model generated a verification report on the clearing effect: the maximum residual protrusion in the cleared area was 3-7mm (allowable value <10mm), the elevation deviation of the cleared bottom surface was +12-+22mm (allowable value ±50mm), the horizontal deviation of the cleared side was +5-+15mm (allowable value ±30mm), the thickness of the residual concrete at the bottom of the wall was 8-25mm (allowable value <50mm), the residual length of the main reinforcement was 5-18mm (allowable value <30mm), and the steel joint removal rate was [not specified]. After confirming 100% completion and the absence of residual protrusions, pressure-maintaining grouting and backfilling were initiated simultaneously. Cement-water glass dual-liquid grout was used, with grouting pressure controlled at 0.32-0.48 MPa (control value 0.3-0.5 MPa), grouting volume of 2.8 m³ (theoretical cavity volume 2.5 m³, filling coefficient 1.12), maximum surface settlement of 3.2-4.8 mm (control value <10 mm), and convergence of adjacent tunnel segments of 0.6-1.2 mm (control value <5 mm). Simultaneous backfilling effectively prevented ground collapse, and ground settlement was controlled within 5 mm.

[0040] 5. Equipment recycling and data archiving The horizontal rock-breaking execution system 2 and the multi-section adjustable steel casing 13 were recovered. All equipment operating parameters, geological sensing data, cutting force curves, control records, and construction images were automatically archived to a digital twin model database. The digital twin model generated a three-dimensional digital as-built archive, including a three-dimensional model of the cleared area (before / after comparison), time history curves of cutting force / rotation speed / feed rate / water jet pressure throughout the process, fuzzy PID control event time axis markers, statistics on the operation time and efficiency of each module, and ground settlement monitoring data curves. The operation time and efficiency statistics for each module are as follows: high-speed rotary milling module 232 operation path length 13.0m, operation time 185min, average clearing efficiency 1.69m² / h, high-pressure water jet... The flow-assisted cutting module 234 has a working path length of 2.5m, a working time of 52min, and an average obstacle clearing efficiency of 1.15m² / h (including rebar cutting). The static splitting module 233 has a working path length of 1.5m, a working time of 28min, and an average obstacle clearing efficiency of 1.29m² / h (for steel splitting). The milling finishing working path length is 2.2m, a working time of 35min, and an average obstacle clearing efficiency of 1.51m² / h. The total working path length is 19.2m, the total time is 300min (5.0h), the overall obstacle clearing efficiency is 1.54m² / h, and the total time for the entire obstacle clearing process is 5.0h. This is more than 60% more efficient than the traditional method (manual chiseling + core drilling) which takes 12-16h.

[0041] The experimental verification results of the digital twin model throughout the construction process are as follows: the real-time 3D visualization mapping refresh rate is 12-15Hz (index ≥10Hz), the screen latency is <80ms, the rock breaking head coordinate positioning accuracy is +4-+9mm (index ±10mm) after total station calibration, the average error between the DNN cutting force prediction value and the measured value is 11.3% (index <15%), the vibration spectrum real-time analysis frequency resolution is 0.5Hz (index ≤1Hz), the maximum path deviation is 7mm (alarm threshold 10mm), all 68 sensor parameters are completely archived, and the incremental learning effect verification shows that after the completion of this operation, the system automatically uploads "working condition characteristics-control parameters-obstacle clearing effect" as a set of training samples to the cloud database. Compared with the initial DNN model without incremental learning, the accuracy of the DNN model in selecting the rock breaking mode in this experiment reached 92.3% (the initial model was 87.5%), and the rationality score of the initial parameter setting was improved by 15.6%.

[0042] The above test results show that: First, the full-rotation drilling rig, combined with fiber optic gyroscope closed-loop control, controls the verticality deviation of the casing at a depth of 45m within 0.06%, far superior to the 0.1%-0.3% of traditional methods. Second, the DNN prediction + fuzzy PID dual-layer control architecture reduces the cutting force fluctuation range by 42% and the risk of stuck drill by more than 60%. The digital twin model enables real-time visualization monitoring of all elements of the rock breaking process. Third, the turret-type three-module design allows a single device to handle various obstacles such as concrete, densely reinforced areas, and steel joints, achieving a comprehensive obstacle clearing efficiency of 1.54m² / h, which is more than 60% higher than traditional methods. Fourth, the static splitting module breaks steel joints without vibration, and synchronous grouting backfilling controls surface settlement within 4.8mm, with minimal impact on the surrounding environment and existing structures. Fifth, the digital twin model fully records the data of the entire process from detection, construction to acceptance. The generated digital as-built archive provides reliable data support for the refined construction and management of urban underground space, and the incremental learning mechanism continuously optimizes the prediction accuracy of the DNN model.

[0043] The intelligent construction method for mechanical obstacle removal of deep underground diaphragm walls is characterized by the following steps: Step S1: Vertical channel establishment and precise orientation. The full-rotation drilling rig 11, together with the steel casing 13, is vertically lowered to the predetermined depth of the diaphragm wall. The high-precision verticality sensor 12 is used for closed-loop control to ensure that the vertical channel deviation is less than 0.1%. Step S2: 3D obstacle modeling and path planning, lower the horizontal rock breaking execution system 2 to the working depth, start the 3D detection module 31 to detect the shape and reinforcement information of the underground continuous wall, build a 3D visualization model and plan the optimal rock breaking path; Step S3: Adaptive rock breaking and dynamic control. Rock breaking is initiated along the planned path. The intelligent sensing and navigation system 3 provides real-time feedback of cutting force, vibration and temperature data. The adaptive rock breaking algorithm module dynamically adjusts the rock breaking parameters based on the digital twin model. Step S4: Verification of obstacle removal effect and synchronous backfilling. The obstacle removal area is re-measured by the three-dimensional detection module 31. After confirming that there are no residual protrusions, pressure-maintaining grouting and backfilling are carried out simultaneously. Step S5: Equipment recovery and data archiving. The horizontal rock breaking execution system 2 and steel casing 13 are recovered, and the entire process data is automatically archived and used for incremental model learning.

[0044] In a specific embodiment: taking an existing underground continuous wall with a depth of 45 meters, a thickness of 800 mm, a concrete grade of C50, and a high reinforcement ratio as an example, a certain subway shield tunnel section needs to pass through it. Step S1: Vertical Channel Establishment and Precise Orientation The full-rotation drilling rig 11 lowered the steel casing 13 to -44.5 meters, "leaving a 0.5-meter protective layer". During the lowering process, the fiber optic gyroscope monitored the verticality in real time, and the deviation was controlled within 0.05%. Step S2: 3D obstacle modeling and path planning The rock-breaking head is lowered to a depth of 23 to -44.5 meters. The three-dimensional sonar module 31 is activated to scan the underground continuous wall and construct a three-dimensional model. After receiving the model data, the DNN model predicts the output: the initial mode is "high-pressure water jet assisted milling as the main mode", the initial rotation speed is set to 120 rpm, the feed rate is 20 mm / min, the water jet pressure is 350 MPa, and the system plans a "honeycomb" rock-breaking path based on the three-dimensional model to avoid the main reinforcement dense area. The path grid size is 300 mm × 300 mm. Step S3: Adaptive Rock Breaking and Dynamic Control The rock-breaking head 23 cuts along the planned path. The fuzzy PID controller monitors the cutting force in real time. When the cutting force rises from the preset 80kN to 120kN, it is determined that the main reinforcement is densely packed. The controller automatically executes the following: the water jet pressure is adjusted to 380MPa, the rotation speed is increased to 150rpm, and the feed speed is reduced to 10mm / min. At the same time, the vibration sensor detects that the vibration amplitude exceeds the threshold. The system actively executes the vibration avoidance strategy and reverses the rotation speed to 145rpm to destroy the resonance frequency. When encountering a 32mm diameter main reinforcement bar, the system automatically switches to the high-pressure water jet module 234 via the turret-type installation platform 231, starts the abrasive mixing chamber 2343, and performs point cutting on the reinforcement bar. After the cutting is completed, it switches back to the milling module to continue the operation. After passing through the dense reinforcement bar area, the controller automatically restores to the initial parameters and enters the high-efficiency milling mode. Step S4: Verification of obstacle removal effect and simultaneous backfilling After the rock breaking was completed, the three-dimensional sonar module 31 re-measured the cleared area to confirm that there were no residual protrusions greater than 10mm. At the same time, the grouting pipeline was started, and cement-water glass double liquid grout was used to backfill the cleared cavity with pressure. The grouting pressure was controlled at 0.3-0.5MPa to ensure that the ground settlement was controlled within 5mm. Step S5: Equipment Recycling and Data Archiving The horizontal rock-breaking execution system 2 and steel casing 13 were recovered. The equipment operation parameters, geological perception data and construction records of the whole process were automatically archived to generate digital as-built archives. The data of this operation was uploaded to the cloud as new training samples. The DNN model performed incremental learning to optimize the ability to predict the initial parameters of similar diaphragm walls in the future.

[0045] In one embodiment, step S3 includes: The deep neural network prediction model receives detection data from the three-dimensional detection module 31 and predicts and outputs rock breaking mode combinations and initial setting parameters. The rock-breaking operation is carried out along the planned path. The fuzzy PID controller monitors the cutting force deviation and the rate of change of deviation in real time, and automatically adjusts the rotation speed, feed rate and water jet pressure. When a sudden change in cutting force or abnormal vibration is detected, the rock breaking module is switched via the turret-type installation platform 231 to start the collaborative operation mode.

[0046] In a specific embodiment, the pressure-maintaining grouting backfill in step S4 uses cement-water glass double-liquid grout, and the grouting pressure is controlled at 0.3-0.5MPa.

[0047] In one embodiment, reference Figure 6 The flowchart is arranged from top to bottom.

[0048] First layer, upper layer: The input box on the left is labeled "Input: obstacle thickness, reinforcement density, historical cutting force spectrum, and formation information"; Draw a rectangle in the middle and label it "Deep Neural Network Prediction Model (DNN)"; The output box on the right is labeled "Output: Optimal rock breaking mode combination, initial rotation speed, initial feed rate, initial water jet pressure"; Second layer, lower layer: Draw an input box on the left: labeled "Input: Cutting force deviation e, deviation change rate ec, vibration amplitude"; Draw a rectangle in the middle and label it "Fuzzy PID Controller"; The output box is drawn on the right: labeled "Output: Speed ​​adjustment Δω, feed speed adjustment Δv, water jet pressure adjustment Δp"; Feedback loop: Draw an arrow downwards from the lower output, pointing to "Actuator (Hydraulic Servo Valve / Inverter)"; A dashed arrow is drawn upwards from the actuator, pointing to the "Multi-source Sensing Monitoring Module (Feedback)"; Arrows are drawn from the multi-source sensing and monitoring module, pointing to the lower-level input, forming a closed loop.

[0049] Data flow: The upper layer output is connected to the lower layer input by an arrow, indicating the initial parameter settings; Bottom label: Draw a rectangle labeled "Digital Twin Model (Real-time Mapping and Data Archiving)" and connect it to the upper and lower layers with bidirectional arrows.

[0050] In one embodiment, reference Figure 7 Overall layout: Five main step boxes are arranged from top to bottom and connected by arrows: Step S1: A rectangle containing the text "Vertical Channel Establishment and Precise Orientation," with the sub-steps labeled in smaller print below: "Full Rotation Drill Rig Sinking → Fiber Optic Gyroscope Closed-Loop Correction → Verticality < 0.1%"; Step S2: A rectangle containing the text "3D obstacle modeling and path planning", with the sub-steps labeled in smaller print below: "3D sonar detection → 3D model construction → honeycomb path planning"; Step S3: A rectangle containing the text "Adaptive Rock Breaking and Dynamic Control", with the sub-steps labeled in smaller print below: "DNN Predicts Initial Parameters → Fuzzy PID Real-Time Control → Switches to Water Jet Module When Encountering Reinforcing Steel → Collaborative Operation"; Step S4: A rectangular box containing the text "Verification of Obstacle Clearing Effect and Synchronous Backfilling", with the sub-steps labeled in smaller print below: "3D Sonar Re-measurement → Pressure-maintaining Grouting Backfilling → Ground Settlement Control"; Step S5: A rectangle containing the text "Equipment Recycling and Data Archiving," with the sub-steps labeled in smaller print below: "Recycle Equipment → Generate Digital As-Built Archives → Cloud Incremental Learning"; Branching and Looping: A decision branch can be added within step S4: "Retest passed? → Yes: Proceed to S5; No: Return to S3 for local repair". Side label: On the right side of the flowchart, use a dashed box to label “Digital Twin Synchronous Monitoring”, and connect it with steps S1 to S5 with a double-headed arrow.

[0051] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0052] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0053] Additionally, "multiple" refers to two or more.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent mechanical obstacle removal device for deep underground continuous walls, characterized in that, include: The vertical positioning system (1) includes a full-rotation drilling rig (11), a high-precision verticality sensor (12), and a multi-section adjustable steel casing (13), which is used to establish a vertical channel and make the vertical channel deviation less than 0.1% through closed-loop feedback control of the high-precision verticality sensor (12); A horizontal rock-breaking execution system (2) is set at the bottom of the vertical positioning system (1) and performs horizontal rock-breaking operations inside the steel casing (13). It includes a telescopic arm (21), a rotating base (22) and a multi-functional integrated rock-breaking head (23). The multi-functional integrated rock-breaking head (23) is set at the bottom of the steel casing (13) and connected to the rotating base (22) through the telescopic arm (21). The intelligent sensing and navigation system (3) includes a three-dimensional detection module (31), an inertial navigation and positioning module (32), and a multi-source sensing and monitoring module (33), used to perform three-dimensional morphological detection of underground obstacles within the steel casing (13), to accurately locate the horizontal rock-breaking execution system (2), and to monitor the cutting force, vibration, and temperature parameters during the rock-breaking process in real time; the three-dimensional detection module (31) is installed at the front end of the multi-functional integrated rock-breaking head (23), the inertial navigation and positioning module (32) uses a fiber optic inertial measurement unit and combines it with a ground total station for real-time calculation, and the multi-source sensing and monitoring module (33) Module (33) includes a cutting force sensor, a vibration sensor and a temperature sensor, which are respectively installed on the spindle, housing and hydraulic pipeline of the milling module. The signals collected by the three-dimensional detection module (31) and the multi-source sensing monitoring module (33) are packaged by the downhole data preprocessing unit and transmitted to the ground control station via industrial Ethernet. The ground control station completes the multi-source data fusion calculation and issues attitude correction instructions to the inertial navigation and positioning module (32) based on the calculation results. The inertial navigation and positioning module (32) is used to receive the instructions from the ground control station and perform real-time calibration and positioning calculation of the equipment attitude. The adaptive rock-breaking algorithm module adopts a two-layer control architecture, including: The upper layer is a deep neural network prediction model. The obstacle thickness, reinforcement density, historical cutting force spectrum and current depth stratum information of the three-dimensional detection module (31) are used as input variables. The output is the optimal rock breaking mode combination, initial rotation speed ω0, initial feed speed v0 and initial water jet pressure p0. The three-dimensional detection module (31) is a three-dimensional sonar detection module or a three-dimensional radar detection module. The lower layer is a fuzzy PID controller, which takes the deviation e between the actual cutting force and the target cutting force, the rate of change of the deviation ec, and the vibration amplitude A as input variables, and outputs the speed regulation Δω, the feed rate regulation Δv, and the water jet pressure regulation Δp. The prediction output of the deep neural network prediction model is used as the initial setting value of the fuzzy PID controller. The real-time adjustment of the fuzzy PID controller is fed back to the horizontal rock breaking execution system (2), forming a two-level hierarchical control of "macro mode selection - micro parameter fine-tuning". The initial rotational speed ω0, initial feed rate v0, and initial water jet pressure p0 are superimposed with the corresponding speed adjustment Δω(t), feed rate adjustment Δv(t), and water jet pressure adjustment Δp(t) to form real-time rotational speed ω(t), real-time feed rate v(t), and real-time water jet pressure p(t), which are used as the final input parameters of the horizontal rock-breaking execution system (2). Their time-domain expressions are as follows: ； The intelligent control and decision system (4) is formed by the vertical positioning system (1), the horizontal rock breaking execution system (2), the intelligent sensing and navigation system (3) and the intelligent control and decision system (4) being connected in sequence to form a closed-loop control loop. The intelligent sensing and navigation system (3) transmits three-dimensional detection signals and multi-source sensing monitoring signals to the intelligent control and decision system (4). The intelligent control and decision system (4) generates control commands based on the signals and sends them to the horizontal rock breaking execution system (2) to dynamically adjust the rock breaking parameters according to the real-time shape of the underground obstacle. The intelligent control and decision system (4) includes a ground control station and an adaptive rock breaking algorithm module deployed thereon, which is used to construct a digital twin model and dynamically adjust the rock breaking parameters based on the adaptive algorithm. The intelligent control and decision-making system (4) also includes a digital twin model, which is used to map real-time rock breaking parameters, equipment posture, and obstacle damage status to a three-dimensional visualization interface, and archive the data of each obstacle clearing operation, and continuously optimize the deep neural network model through incremental learning. The waste residue real-time treatment system (5) includes a negative pressure slag suction port, a spiral conveying pipeline and a ground screening device, which is used to realize the simultaneous operation of rock breaking and slag cleaning.

2. The intelligent equipment for mechanical obstacle removal of deep underground continuous walls according to claim 1, characterized in that: The multi-functional integrated rock breaking head (23) includes a turret-type mounting platform (231) and a high-speed rotary milling module (232), a static splitting module (233), and a high-pressure water jet auxiliary cutting module (234) mounted on it. The turret-type mounting platform (231) realizes rapid switching of each module through hydraulic drive. The intelligent control and decision-making system (4) automatically switches the rock breaking module through the turret-type mounting platform (231) based on the cutting force mutation signal or vibration abnormal signal fed back by the multi-source sensor monitoring module (33), and starts the three-level collaborative operation mode of milling, splitting and water jet cutting.

3. The intelligent mechanical obstacle removal equipment for deep underground continuous walls according to claim 2, characterized in that: The high-speed rotary milling module (232) includes a planetary gear reducer (2321), a high-speed hydraulic motor (2322), and a cylindrical milling cutter head (2323). The diameter of the cylindrical milling cutter head (2323) is adjustable. The cylindrical milling cutter head (2323) is in the shape of a hollow cylinder. The cutting teeth are arranged only around the outer circumference of the cylinder end face. The inside of the cylinder is a hollow slag discharge channel. The outer diameter of the cylindrical milling cutter head (2323) is 500mm-1500mm and the wall thickness is 20mm-50mm. The cutting teeth arranged around the outer circumference reduce the cutting torque when breaking rock in the horizontal direction compared to a full-disc tooth structure.

4. The intelligent equipment for mechanical obstacle removal of deep underground continuous walls according to claim 2, characterized in that: The static splitting module (233) includes multiple splitting gun groups, a high-pressure hydraulic pump station, and a radial expansion sleeve arranged in a ring around the circumference of the multi-functional integrated rock breaking head (23). Each splitting gun group contains 3-5 splitting guns, which are evenly distributed along the circumference. The splitting gun has a tapered guide head at the front end and adopts a multi-stage expansion structure inside, with a single gun splitting force ≥2000kN; The radial expansion sleeve generates a uniform radial expansion force under high-pressure hydraulic drive, which is used for vibration-free crushing of high-strength concrete.

5. The intelligent mechanical obstacle removal equipment for deep underground continuous walls according to claim 2, characterized in that: The high-pressure water jet assisted cutting module (234) includes an ultra-high pressure plunger pump (2341), a three-axis adjustable rotating water jet nozzle (2342), and an abrasive mixing chamber (2343). The three-axis adjustable rotating water jet nozzle (2342) can adjust the spray angle within the range of ±45° horizontally and ±30° vertically. The front end of the three-axis adjustable rotating water jet nozzle (2342) is equipped with an automatic anti-clogging device. The abrasive mixing chamber (2343) has a built-in abrasive supply pump. When the three-dimensional detection module (31) identifies the main reinforcement signal, the intelligent control and decision system (4) controls the automatic addition of abrasive to the high-pressure water flow to form an abrasive water jet to cut the high-strength steel bars.

6. A construction method for intelligent mechanical obstacle removal of deep underground diaphragm walls according to any one of claims 1-5, characterized in that, Includes the following steps: Step S1: Vertical channel establishment and precise orientation. The full-rotation drilling rig (11) and steel casing (13) are vertically lowered to the predetermined depth of the diaphragm wall. The vertical channel deviation is ensured to be less than 0.1% through closed-loop control using a high-precision verticality sensor (12). Step S2: Three-dimensional obstacle modeling and path planning, lower the horizontal rock breaking execution system (2) to the working depth, start the three-dimensional detection module (31) to detect the shape and reinforcement information of the underground continuous wall, construct a three-dimensional visualization model and plan the rock breaking path; Step S3: Adaptive rock breaking and dynamic control, start rock breaking along the planned path, intelligent sensing and navigation system (3) feeds back cutting force, vibration and temperature data to intelligent control and decision system (4) in real time, and the adaptive rock breaking algorithm module dynamically adjusts rock breaking parameters based on digital twin model; Step S4: Verification of obstacle removal effect and synchronous backfilling. The obstacle removal area is re-measured by the three-dimensional detection module (31). After confirming that there are no residual protrusions, pressure-maintaining grouting and backfilling are carried out simultaneously. Step S5: Equipment recovery and data archiving. The horizontal rock breaking execution system (2) and steel casing (13) are recovered, and the data of the whole process are automatically archived and used for incremental model learning.

7. The intelligent construction method for mechanical obstacle removal of deep underground diaphragm walls according to claim 6, characterized in that, Step S3 includes: The deep neural network prediction model receives the detection data from the three-dimensional detection module (31) and predicts and outputs the rock breaking mode combination and initial setting parameters. The rock-breaking operation is carried out along the planned path. The fuzzy PID controller monitors the cutting force deviation e and the deviation change rate ec in real time, and automatically adjusts the rotation speed, feed rate and water jet pressure. When a sudden change in cutting force or abnormal vibration is detected, the rock breaking module is switched through the turret-type installation platform (231) to start the three-level collaborative operation mode of milling, splitting and water jet cutting.

8. The intelligent construction method for mechanical obstacle removal of deep underground diaphragm walls according to claim 6, characterized in that: In step S4, the pressure-maintaining grouting backfill uses cement-water glass dual-liquid grout, and the grouting pressure is controlled at 0.3-0.5MPa.

Images