CN-121993209-A - Intelligent dynamic regulation tunneling method for shield of multiple karst cave sections

CN121993209ACN 121993209 ACN121993209 ACN 121993209ACN-121993209-A

Abstract

The invention relates to the technical field of underground engineering construction, in particular to an intelligent dynamic regulation tunneling method for a shield with multiple karst cave sections, which comprises S1, an advanced detection step; the method comprises the steps of S2, data fusion, S3, risk assessment, S4, karst cave treatment, S5, dynamic tunneling, S6, attitude control and monitoring, wherein a multi-source detection means is adopted, and existing investigation data are combined, so that a high-precision geological model is built in real time, and stratum structures and hidden danger information before shield tunneling are mastered in an omnibearing manner. The simulation pre-judgment of the tunneling process is performed by a digital twin technology, the perspective capability is provided for subsequent operation, the predictability of complex geological risks such as karst cave, faults and aquifers is fundamentally improved, the risk management is realized by ' steering ' pre-precaution ' after treatment, multi-source heterogeneous data are input into an AI decision model, the dynamic quantitative evaluation of hidden troubles such as stratum mutation, mud burst and water burst is realized, and a measurable risk level is formed.

Inventors

JIA JIANWEI
NIU BEN
CUI HUICHAO
HUA DONGSHENG
YE YOULIN
ZHANG YANNIAN

Assignees

中国建筑第六工程局有限公司
中建六局交通建设有限公司

Dates

Publication Date: 20260508
Application Date: 20251028

Claims (7)

1. The intelligent dynamic regulation tunneling method for the multi-karst cave section shield is characterized by comprising the following steps of: The geological radar, the acoustic wave sensor and the drilling probe are deployed and used for detecting stratum structures and karst cave distribution in the range of 50-80 meters in front of the shield tunneling machine, so that stratum space distribution and hidden danger information are obtained; Carrying out multi-source fusion and three-dimensional modeling processing on the stratum space distribution and hidden danger information to obtain a real-time geological model; the AI decision model is adopted to analyze the real-time geological model and the sensor data of the shield tunneling machine and is used for evaluating the risk levels of karst cave and aquifer to obtain dynamic risk indexes; performing high-pressure grouting or freezing reinforcement treatment on the dynamic risk indexes, and filling karst cave and forming a supporting structure to obtain a reinforced stratum body; automatically adjusting the propelling force, the opening pressure and the slurry injection amount according to the reinforced stratum body to obtain the pressure in the balance surface; And carrying out jack differential adjustment and fiber bragg grating monitoring treatment on the pressure in the balance surface to obtain a tunneling posture and a deformation state.
2. The intelligent dynamic regulation tunneling method of the multi-karst cave section shield according to claim 1, wherein the AI decision model is an LSTM neural network to perform time sequence prediction on the real-time geological model to obtain the hidden danger level; The formula of the LSTM neural network is as follows: C t ＝σ(W o h t +b o ) h t ＝tanh(W h x t +U h h t-1 +b h ) Wherein xt is an input vector, and dimension nx is a multisource fused geological feature vector at the t moment, and the multisource fused geological feature vector comprises multisource sensing data of stratum permeability coefficient, porosity, front detection radar echo intensity, on-site shield tunneling machine propulsion, torque and water level pressure; ht is a hidden state vector, and the dimension nh is the memory and output of the LSTM unit at the t moment and reflects the stratum evolution trend; the weight matrix and the bias for calculating the hidden state are obtained through offline training of historical engineering data; Weights and biases for mapping hidden states to risk indices; sigma (& gt) Sigmoid activates a function, and maps the output into a hidden danger level probability value between (0 and 1); and t, the step number of the current shield tunneling period corresponds to time or the driving mileage.
3. The intelligent dynamic regulation tunneling method of the multi-karst cave section shield according to claim 1, wherein the automatically adjusting the propulsion force, the opening pressure and the slurry injection amount according to the bearing capacity of the reinforced stratum to obtain the pressure in the balance plane comprises the following steps: Automatically adjusting the propulsion by adopting a closed-loop self-adaptive propulsion control formula; the closed-loop self-adaptive propulsion control formula is as follows: ΔP(t)=P model (t)-P Actual measurement (t) The F pushing (t) is actual pushing force, and the unit kN is the output tunneling pushing force of the shield machine at the moment t; F, taking the reference as the reference propulsion, wherein the unit kN is a recommended thrust value under normal and stable stratum conditions, and the recommended thrust value is preset by a geological model; ΔP (t) is the face pressure deviation in kPa; Pmodel (t) is a recommended shield plane hydraulic pressure obtained by digital twin simulation; P is actually measured (t), namely the shield surface hydraulic pressure is measured in real time by a site mud or soil pressure sensor; Kp is proportional gain, dimensionless, response strength to current front pressure deviation; Ki is the integral gain, in kN/(kPa s) the response to the accumulated deviation; The face pressure deviation from the start of tunneling to time t is integrated.
4. The intelligent dynamic regulation tunneling method for the multi-karst cave section shield according to claim 1, wherein data of the geological radar and the acoustic sensor are used for inverting formation porosity and permeability coefficients and obtaining enhanced geological characteristic parameters.
5. The intelligent dynamic regulation tunneling method of the multi-karst cave section shield according to claim 1, wherein the high-pressure grouting adopts a layered injection technology, and quick setting cement slurry is injected layer by layer to control the slurry diffusion range and form a continuous supporting layer, so that the reinforced stratum body is obtained.
6. The intelligent dynamic regulation tunneling method of the multi-karst cave section shield according to claim 1, wherein the jack differential regulation and fiber bragg grating monitoring treatment are performed on the pressure in the balance plane to obtain tunneling postures and deformation states, and the method comprises the following steps: And carrying out jack differential adjustment and grouting linkage correction on the tunneling attitude and the deformation state by adopting inclination sensor data and a convergence instrument measurement result to obtain the tunneling attitude and the deformation state.
7. The intelligent dynamic regulation tunneling method of the multi-karst cave section shield according to claim 1, further comprising: when the dynamic risk index is higher than a preset threshold value, starting an emergency depressurization and directional drainage measure; when the tunneling gesture and the deformation state exceed the early warning range, the advancing speed is automatically reduced, and the optimization parameters are simulated again.

Description

Intelligent dynamic regulation tunneling method for shield of multiple karst cave sections Technical Field The invention belongs to the technical field of underground engineering construction, and particularly relates to an intelligent dynamic regulation tunneling method for a shield with multiple karst cave sections. Background In the field of tunnel and underground engineering construction, the shield method is widely applied to projects such as urban subways, railway tunnels, municipal pipe galleries and the like by virtue of the advantages of high efficiency, safety, environmental protection and the like. However, when shield construction encounters multiple karst cave sections, complex geological conditions present significant challenges to engineering construction. Karst cave is a special geological structure under the ground, the distribution of which has randomness and uncertainty, and the size, shape, filling condition, connectivity with aquifers and the like of the karst cave are greatly different. In the traditional shield tunneling process, the geological detection means for multiple karst cave sections are relatively single, and mainly depend on earlier geological investigation data. However, these data often have difficulty in comprehensively and accurately reflecting the actual stratum structure and karst cave distribution in front of the shield machine due to factors such as limited investigation range, insufficient precision, dynamic change of geological conditions and the like. In the tunneling process, the shield tunneling machine may suddenly encounter an undetermined karst cave, causing a series of serious problems, such as out-of-control attitude of the shield tunneling machine, cutter head damage, water burst sand burst and the like, not only seriously affecting the construction progress, but also causing huge economic loss and potential safety hazard. In this regard, the application provides an intelligent dynamic regulation tunneling method for a multi-karst cave section shield, which is used for solving the problems. Disclosure of Invention The invention aims to provide an intelligent dynamic regulation tunneling method for a multi-karst cave section shield so as to solve the problems in the background technology. In order to achieve the above purpose, the present invention provides the following technical solutions: the intelligent dynamic regulation and control tunneling method for the shield of the multiple karst cave sections comprises the following steps: An advanced detection step, namely deploying a geological radar, an acoustic wave sensor and a drilling probe, and detecting stratum structures and karst cave distribution in the range of 50-80 meters in front of a shield tunneling machine to obtain stratum space distribution and hidden danger information; A data fusion step, namely carrying out multi-source fusion and three-dimensional modeling processing on the stratum space distribution and hidden danger information to obtain a real-time geological model; The risk assessment step is to analyze the real-time geological model and the sensor data of the shield tunneling machine by adopting an AI decision model and is used for assessing the risk levels of karst cave and aquifer to obtain dynamic risk indexes; A karst cave treatment step, namely carrying out targeted high-pressure grouting or freezing reinforcement treatment on the dynamic risk indexes, and filling the karst cave and forming a supporting structure to obtain a reinforced stratum body; A dynamic tunneling step, namely automatically adjusting the propelling force, the opening pressure and the slurry injection amount according to the reinforced stratum body to obtain the pressure in the balance surface; And a posture control and monitoring step, namely performing jack differential adjustment and fiber bragg grating monitoring treatment on the pressure in the balance surface to obtain a tunneling posture and a deformation state. Preferably, the AI decision model is an LSTM neural network, so as to perform time sequence prediction on the real-time geological model, thereby obtaining the hidden danger level; The formula of the LSTM neural network is as follows: Ct＝σ(Woht+bo) ht＝tanh(Whxt+Uhht-1+bh) The geological feature vector after multisource fusion at the t moment comprises multisource sensing data of stratum permeability coefficient, porosity, front detection radar echo intensity, on-site shield tunneling machine propulsion, torque and water level pressure; ht (hidden state vector, dimension nh) is the memory and output of the LSTM unit at the t moment, reflecting the stratum evolution trend; Wh epsilon Rnh x nx is used for calculating a weight matrix and bias of a hidden state, and the weight matrix and bias are obtained through offline training of historical engineering data in a scene; Wo e R1 x nh, bo e R weight and bias for mapping hidden state to risk index; sigma (& gt) Sigmoid activates a function, and ma