CN-121981519-A - Digital twinning-based TBM card machine risk assessment index dynamic prediction method

CN121981519ACN 121981519 ACN121981519 ACN 121981519ACN-121981519-A

Abstract

The invention relates to the technical field of tunnel construction and construction, and particularly discloses a dynamic prediction method of a TBM card risk assessment index based on digital twinning, which comprises the following steps of 1, constructing a TBM card risk factor association network; step 2, establishing a TBM card risk assessment index system and a static TBM card Bayesian network structure, step 3, constructing an index weight analysis model based on a dynamic Bayesian network, and step 4, dynamically calculating and predicting the index weight driven by TBM monitoring data. According to the invention, real-time monitoring data of equipment such as cutter head rotating speed, torque and the like which are critical to the risk of machine blocking in the TBM construction process are fully utilized, accurate prediction of the risk factor index of the TBM machine blocking is realized, the index weights at different moments are dynamically adjusted by utilizing a dynamic Bayesian network model, the problems that the importance degree change of factors cannot be considered in the existing risk assessment, the time-space transfer effect and the like are solved, and the accuracy, timeliness and construction adaptability of the risk assessment of the machine blocking are remarkably improved.

Inventors

ZHOU JUNWEI
ZHU QING
CHEN JIANFENG
JIANG QIAN
DING YULIN
LUO XUN

Assignees

川藏铁路有限公司
西南交通大学
中国国家铁路集团有限公司

Dates

Publication Date: 20260505
Application Date: 20251217

Claims (8)

1. A dynamic prediction method of a TBM card machine risk assessment index based on digital twinning is characterized by comprising the following steps: Step 1, constructing a TBM card risk factor association network; step 2, establishing a TBM card risk assessment index system and a static TBM card Bayesian network structure; Step 3, constructing an index weight analysis model based on a dynamic Bayesian network; And 4, dynamically calculating and predicting the index weight driven by the TBM monitoring data, namely combining the field construction monitoring data and expert experience, adopting a hierarchical analysis method and an entropy weight method to determine the comprehensive weight of the index at the moment T=t as a priori probability value, determining the comprehensive weight of the dynamic node at the next moment as a transition probability, calculating the conditional probability of the node by using a Noisy-or gate model, carrying out forward dynamic probability reasoning by combining a dynamic Bayesian network, and outputting the dynamic probability value of the TBM card risk assessment index at the latest moment by continuously and iteratively updating.
2. The dynamic predicting method for the TBM card risk assessment index based on the digital twinning of claim 1 is characterized in that in step 1, specifically, a TBM card risk accident report and a risk event are collected based on a complex network theory, important factors and causal relations among factors affecting the TBM card risk are extracted according to a large language model, the risk factors are extracted as nodes in a network, the nodes with the causal relations are connected as edges in the network, and a TBM card risk factor association network is constructed by adopting a node-edge structure.
3. The dynamic prediction method for the TBM card machine risk assessment index based on the digital twinning is characterized in that TBM card machine conditions are divided into three types of a card cutter, a card shield and a card support shoe, the most important factors affecting the card machine risk degree comprise geological risk factors and equipment risk factors, the geological risk factors are tunnel face mud burst sand, surrounding rock collapse, rock explosion and soft rock large deformation, and among the equipment risk factors, cutter rotation speed, cutter torque and cutter penetration degree affect the card cutter, support shoe pressure affects the card support shoe, and tunneling thrust and pushing speed affect the card cutter, the card shield and the card support shoe simultaneously.
4. The dynamic predicting method for the digital twin-based TBM card risk assessment index of claim 1 is characterized by comprising the following steps: step 2.1, selecting risk factors related to geological risks and equipment risks as TBM machine risk assessment indexes based on a TBM machine risk factor related network, grading the assessment indexes according to a hierarchical analysis method, and constructing a TBM machine risk assessment index system; And 2.2, simplifying the constructed TBM card risk factor association network according to a TBM card risk assessment index system, extracting key network nodes, and constructing to obtain a static Bayesian network.
5. The dynamic prediction method of the TBM card risk assessment index based on the digital twin system of claim 4 is characterized in that assessment indexes are classified according to a hierarchical analysis method, wherein the index classification hierarchy comprises a target layer, a criterion layer and an index layer, the target layer is a TBM card risk state R, the criterion layer consists of a risk category affecting the risk state R and comprises geological risk and equipment risk, the index layer comprises a geological risk factor index and an equipment risk factor index, the geological risk factor index comprises a surrounding rock grade B 1 , a large amount of groundwater B 2 and a large amount of sand B 3 , and the equipment risk factor index comprises a cutter disc rotating speed C 1 , a cutter disc torque C 2 and a cutter disc penetration C 3 .
6. The dynamic predicting method for the digital twin-based TBM card risk assessment index of claim 1 is characterized by comprising the following steps: establishing a risk state transition network according to the relation among risk factors, namely determining a time slice, finding a key path causing a card accident according to a TBM card Bayesian network, extracting dynamic nodes with the state changing along with time from the key path, and establishing the risk state transition network according to the relation among the risk factors; Step 3.2, constructing a dynamic Bayesian network model, namely constructing a Bayesian network model of a dynamic TBM card machine on the basis of a static Bayesian network, namely, constructing a Bayesian network before construction, namely, a time node t, a dynamic Bayesian network during construction, namely, an intermediate time node and a Bayesian network after construction, namely, a final time node according to a state transfer network according to time slices; And 3.3, constructing an index weight analysis model, namely inputting the prior probability of each node at the time T=t, the node transition probability and the conditional probability of the father node from the time T=t to the time T=t+1 in a dynamic Bayesian network model, and dynamically outputting the index weight predicted value at the time T=t+1 according to the Bayesian network principle.
7. The dynamic prediction method of the TBM card risk assessment index based on the digital twin system of claim 6 is characterized in that the prior probability of the node is a comprehensive weight result of each node obtained by calculating construction monitoring data at the time of T=t, the node transition probability is a comprehensive weight result of a dynamic node at the time of T=t+1, and the node conditional probability is a conditional probability value obtained by converting the node grade evaluation result into a condition probability value by using a triangle fuzzy number principle.
8. The dynamic predicting method for the digital twin-based TBM card risk assessment index of claim 1 is characterized by comprising the following steps: Step 4.1, preprocessing TBM monitoring data and expert experience data, and carrying out standardized processing on index data in a predicted time period by using a range variation method for the TBM monitoring data; Step 4.2, calculating parameters of a dynamic Bayesian network model, calculating objective weights of all indexes by using the preprocessed monitoring data according to an entropy weight method, calculating expert experience data according to a hierarchical analysis method to obtain subjective weights of all indexes, and obtaining comprehensive weights of all nodes at the time of T=t and comprehensive weight results of dynamic nodes at the time of T=t+1 as prior probabilities of the dynamic Bayesian network model at the time of T=t; Calculating the comprehensive weight of each dynamic node at the next moment, namely the transition probability of the node; Calculating the conditional probability of each child node by using the risk probability value according to Noisy-or gate model to obtain a corresponding conditional probability table; And 4.3, forward reasoning of a dynamic Bayesian network model, namely inputting the prior probability, the transition probability and the conditional probability of the nodes into the dynamic Bayesian network, and calculating to obtain posterior probability values of the nodes, namely, the weight predicted values of the indexes in the time period.

Description

Digital twinning-based TBM card machine risk assessment index dynamic prediction method Technical Field The invention relates to the technical field of tunnel construction, in particular to a digital twin-based TBM card risk assessment index dynamic prediction method. Background At present, tunneling technology mainly comprises a drilling and blasting method and a full-face tunneling machine (Tunnel Boring Machine, hereinafter referred to as TBM). TBM is the most advanced underground excavation and tunneling equipment in the world at present, and is the factory pipeline tunnel construction equipment integrated by systems such as machinery, electricity, liquid, light, gas and the like. As large-scale underground engineering equipment integrating tunneling, supporting and slag discharging, the method has gradually replaced the drilling and blasting method because of the advantages of green, safety, high mechanical degree, high tunneling speed, short construction period, good surrounding rock stability, small labor force demand and the like in construction, becomes a construction method which is preferentially adopted in the construction of deep and long tunnels of underground engineering, and is widely applied to the underground engineering construction of departments such as traffic, energy, water conservancy and municipal administration. In the TBM construction method, the interaction relation between a machine-surrounding rock-supporting system is extremely strong, a deep tunnel project is positioned in a stratum, and the complexity of the stratum makes the risk of machine blockage in the underground construction process difficult to express by using very accurate and quantized data, so that the method has larger uncertainty and ambiguity. Thus, the tunneling safety and efficiency problems of tunnel construction are a significant technical challenge when facing complex geological conditions. In addition, due to the huge body of the TBM and the lag of surrounding rock support, the tunneling parameters in construction are often controlled by means of artificial experience, and subjectivity is high. When the TBM is subjected to bad geology, the tunneling scheme and the control parameters of the TBM are difficult to timely and effectively adjust, machine blocking accidents are easy to occur, safety of mechanical equipment and constructors is threatened, and great economic loss and construction period delay are caused. Therefore, how to utilize TBM construction monitoring data and construction risk evaluation methods, pay attention to factor importance degree changes and space-time transfer effects, and dynamically adjust risk factor indexes and weights thereof so as to achieve the aim of accurately evaluating the risk of machine blocking is an urgent problem to be solved. In the TBM construction risk evaluation process, the weight occupied by each risk evaluation index in the evaluation process directly influences the accuracy of the risk evaluation result. In order to more accurately and scientifically determine the risk evaluation index weight, the selection of an evaluation index weighting method is important. At present, the method for determining the risk evaluation index weight is mainly divided into three types, namely, 1) a subjective weighting method, wherein the subjective weighting method is a method for determining the risk evaluation index weight according to subjective experience of a decision maker or expert scholars. The subjective weighting methods commonly used at present are expert scoring, delphi, expert investigation, analytic hierarchy process, two-term coefficient method, loop ratio scoring and the like. The method utilizes the abundant experience of decision makers and expert scholars to the greatest extent, but the result of the evaluation index weighting is often subjectively influenced by a plurality of factors such as the knowledge structure, working experience, preference degree and the like of the decision makers or the expert scholars, and the importance degree of the risk evaluation index cannot be reflected in all aspects. 2) Objective weighting method, namely, the objective weighting method is mainly a method for determining the weight of risk evaluation indexes according to the size of resolution information provided by the evaluation indexes and the correlation among the indexes. The currently commonly used objective weighting methods include a factor analysis method, a principal component analysis method, a CRITIC method, an entropy weighting method, a variation coefficient method, a gray correlation analysis method and the like. The method greatly improves the objectivity of the risk evaluation index weight, but ignores the rich experience of expert scholars, and the result of the evaluation index weighting sometimes has the condition of being inconsistent with the actual result and has stronger dependence on samples. 3) The combined weighting method is a comprehensive met