CN-121976811-A - Multi-objective collaborative decision optimization method for slurry shield machine cutterhead and scraper configuration

CN121976811ACN 121976811 ACN121976811 ACN 121976811ACN-121976811-A

Abstract

The invention discloses a multi-objective collaborative decision optimization method for slurry shield machine cutterhead and scraper configuration. The method comprises the steps of firstly constructing a dynamic geologic map through multi-source sensor data fusion, predicting the abrasion trend of the scraper by utilizing a depth time sequence convolution network, then adopting a meta-evolution hierarchical optimization architecture, carrying out self-adaptive population initialization on a macroscopic layer according to geologic partitions, carrying out co-evolution on a microscopic layer based on the mechanical interaction relation of the scraper, and finally optimizing algorithm parameters in real time through a dynamic multi-target weight adjustment mechanism and a reinforcement learning strategy self-adaptive module. The invention breaks through the limitation of the traditional empirical layout, solves the difficult problems of multi-objective, strong constraint and nonlinear optimization, can automatically generate the optimal scraper layout scheme taking the tunneling efficiency, the economic cost and the service life into consideration on the premise of ensuring the stress balance of the cutterhead, and improves the intelligent level and the comprehensive benefit of shield construction.

Inventors

WANG HUAWEI
LIU SIJIN
LIU PENG
ZHANG JIYING
ZHAO JIAHONG
SHEN LAN
TIAN YE

Assignees

中铁十四局集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260107

Claims (10)

1. The multi-objective collaborative decision optimization method for the configuration of the cutterhead and the scraper of the slurry shield machine is characterized by comprising the following steps of: Based on geological-wear real-time perception, carrying out fusion and feature extraction on the multi-source sensor data to obtain a dynamic geological map; generating a coding sequence for guiding the optimization of the scraper layout according to the dynamic geologic map optimization evolution algorithm parameters; When the method is used, the coding sequence and the real-time construction data are combined, meta-evolution layering optimization is carried out, the meta-evolution layering optimization comprises macro-layer geological partition self-adaption and micro-layer scraper co-evolution, and a preliminary optimization scheme is generated by combining dynamic multi-objective weight adjustment and reinforcement learning strategy self-adaption; And performing Pareto front screening and engineering constraint verification on the preliminary optimization scheme, outputting a final scraper layout scheme, and evaluating and feeding back optimization parameters.
2. The method of claim 1, wherein the fusing and feature extraction of the multi-source sensor data to obtain the dynamic geologic map comprises: acquiring torque vibration data, slurry pressure data, acoustic geological detection data and electromagnetic wave geological radar data in the tunneling process of the shield tunneling machine to form a multi-source sensor data stream; Performing time domain alignment and frequency domain transformation on the multi-source sensor data stream to obtain a time alignment sampling sequence and an initial frequency domain representation; Based on the time alignment sampling sequence, extracting signal arrival time distribution, a signal intensity decay curve and a time pulse width expansion coefficient to form a time domain feature vector; based on the initial frequency domain representation, calculating an energy spectrum and extracting spectrum energy distribution, frequency shift characteristics and frequency domain coherence to form a frequency domain characteristic vector; And carrying out fusion processing on the time domain feature vector and the frequency domain feature vector by adopting a Bayesian geological inference model to generate dynamic geological type distribution data, geological strength profile data and geological heterogeneity index data, and integrating to obtain a dynamic geological map.
3. The method of claim 2, further comprising, after obtaining the dynamic geologic map, doctor blade wear timing predictions: Constructing a multi-mode time sequence training data set containing geological features, operation parameters and abrasion labels based on the historical and real-time construction data; model training is carried out on the multi-mode time sequence training data set by adopting a deep time sequence convolution network, so as to obtain a pre-trained abrasion prediction model; And inputting real-time geological data and tunneling parameters into a pre-trained abrasion prediction model, and outputting the abrasion prediction data of the scraper in a future time period through forward propagation calculation.
4. The method of claim 2, further comprising macro-layer geologic partition adaptive population initialization after the dynamic geologic map is obtained: based on the dynamic geologic map, dividing the cutter head working face into a plurality of geologic characteristic partitions by adopting a clustering algorithm; Aiming at each geological characteristic partition, calculating a multi-target weight coefficient of partition specificity through fuzzy reasoning according to geological intensity and abrasive data of the geological characteristic partition; And respectively carrying out population initialization on each partition based on the partition specific multi-target weight coefficient to generate a plurality of partition initial population data.
5. The method of claim 1, wherein the microscopic layer doctor co-evolves, comprising: Carrying out mechanical interaction relation analysis on individuals in the global population, and calculating an interaction intensity matrix among the scrapers; Based on the interaction intensity matrix, adopting a community discovery algorithm to carry out dynamic group division to generate a plurality of co-evolution scraper groups; in each co-evolution scraper group, calculating a shared fitness value in the group by adopting a fitness sharing mechanism; Based on the interaction intensity matrix, the crossover and mutation operation is guided, the gene recombination driven by the interaction intensity is carried out, and the new generation of scraper layout scheme data is generated.
6. The method of claim 1, wherein the dynamic multi-objective weight adjustment comprises: collecting current construction state data, including real-time geological conditions, scraper abrasion state and engineering progress data; based on the current construction state data, calculating the preliminary importance scores of all optimization targets through fuzzy logic reasoning; Normalizing the preliminary importance scores to obtain dynamic multi-target weight vectors at the current moment; Carrying out comprehensive fitness evaluation on a new generation of scraper layout scheme by utilizing the dynamic multi-target weight vector to obtain a population fitness evaluation result; Carrying out rapid non-dominant sorting and crowding degree calculation on the population fitness evaluation result to generate Pareto optimal solution set data of the current generation; And comparing the Pareto optimal solution set of the current generation with the historical front data, and analyzing to obtain the Pareto front evolution trend data.
7. The method of claim 6, wherein reinforcement learning strategy adaptation comprises: constructing a historical decision database containing construction states, optimized actions and performance rewards; based on a historical decision database, carrying out strategy training on the deep Q network to obtain a trained strategy decision model; Inputting the current construction state and Pareto front evolution trend data into a strategy decision model, and generating a meta-learning adjustment instruction through Q value calculation; and carrying out online self-adaptive updating on the crossover probability, the variation probability and the selection pressure of the genetic algorithm according to the meta-learning adjustment instruction.
8. The method of claim 1, wherein outputting the final doctor blade layout scheme comprises: judging whether the iterative times of the algorithm reach the maximum value or whether the Pareto front converges, and returning to continue to optimize if the iterative times of the algorithm do not reach the maximum value; Selecting a group of optimal scraper configuration parameters from the Pareto optimal solution set of the final generation according to engineering preference or dynamic weight; And carrying out scheme formatting treatment on the optimal scraper configuration parameters, and outputting a final scraper layout scheme comprising the polar diameter, polar angle and type of each scraper.
9. The method of claim 8, wherein the schema formatting process comprises: converting the scraper configuration parameters into a spatial position distribution diagram under polar coordinates; Trimming and conflict detection are carried out on the position of the scraper according to the constraint of the cutter head structure and the requirements of the installation process; And generating a technical file, a drawing or a numerical control machining code which can be executed by a workshop.
10. A multi-objective collaborative decision-making system for a slurry shield machine cutterhead and scraper configuration for implementing the method of any one of claims 1-9, comprising: the geological-wear real-time sensing module is used for multi-source data acquisition, fusion and wear prediction; the meta-evolution layering optimization module is used for macro-partition self-adaption and micro co-evolution; the dynamic decision and strategy self-adaption module is used for weight adjustment and reinforcement learning control; And the scheme generation and output module is used for Pareto solution set screening and engineering scheme formatting.

Description

Multi-objective collaborative decision optimization method for slurry shield machine cutterhead and scraper configuration Technical Field The invention belongs to the technical field of compound shields, and particularly relates to a multi-objective collaborative decision optimization method for cutter head and scraper configuration of a slurry shield machine. Background Along with the continuous acceleration of urban progress in China and the continuous increase of development demands of underground spaces, the large-diameter shield tunnel engineering is increasingly widely applied to the construction of infrastructures such as railways, highways, subways, water conservancy and municipal comprehensive pipe galleries. The shield machine is used as core equipment for tunneling, and the performance of a cutting system of the shield machine is directly related to the efficiency, cost and safety of engineering construction. The scraper is used as a key cutting part of a cutter head of the slurry shield machine, and plays an irreplaceable role in the process of rock and soil stripping, slag soil improvement and flow. Scientific and reasonable scraper layout design can not only remarkably improve the tunneling efficiency of the shield machine, prolong the service life of the cutter and reduce the construction cost, but also effectively improve the stress balance of the cutterhead and enhance the adaptability of the shield machine under complex geological conditions. At present, a plurality of technical routes are formed by the optimization research of the cutter layout of the shield machine. In the early development stage, engineering experience summary and geometric constraint modeling are mainly relied on. The method aims at improving the wear uniformity of the cutter, and based on a wear coefficient method, a tunneling coefficient model and Archimedes spiral track planning, basic principles such as plane symmetry layout are provided, so that a foundation is laid for subsequent research. Along with the development of numerical simulation and mechanical analysis technology, an optimization method based on mechanical balance gradually becomes a research key point, and the method focuses on reducing the radial unbalance force and overturning moment of the cutterhead, adopts a multi-objective Pareto non-inferior layering genetic algorithm to carry out collaborative layout optimization on cutters such as a scraper and a hob, and therefore stability and reliability of operation of the cutterhead are remarkably improved. In recent years, a multi-objective optimization and intelligent decision technology is introduced into the field, a part of schemes combine a multi-objective evolutionary algorithm with a fuzzy decision method to construct a parameter optimization model taking propulsion speed, tunneling specific energy and cutter abrasion loss as core targets, and scientific basis is provided for actual construction management and control. Non-dominant sequencing genetic algorithms such as NSGA-II and the like are also used for coordinating contradiction between tunneling efficiency and cutter abrasion, and the optimization of cutter changing cost is realized while the single tunneling distance is improved. In addition, the emerging intelligent algorithms such as the gray wolf optimization algorithm, the particle swarm optimization algorithm and the like show good global optimizing capability in the refined search of a single target by simulating a natural optimizing mechanism, and are beneficial to improving the local excessive wear phenomenon. Based on ash correlation analysis theory, a polar angle fine adjustment mechanism is introduced, and a multi-index correlation evaluation is combined, so that a new thought is provided for multi-parameter collaborative optimization of the integral cutterhead. However, through extensive analysis of the prior art, it has been found that there are significant shortcomings in doctor blade layout optimization studies compared to the systematicness and depth of the doctor blade layout optimization study. Most of the existing researches are in the experience arrangement or geometric principle level, mainly focus on stress analysis and wear rule simulation of the scraper, but are obviously insufficient in research on algorithm optimization modeling, and lack in depth fusion of mathematical modeling and intelligent optimization algorithm. Particularly in the field of multi-objective collaborative optimization solution, the prior art is still in a preliminary exploration stage, and accurate quantization and global optimization of a layout scheme are difficult to realize. Most of current academic researches and engineering practices tend to simplify the optimization targets into a single index, for example, only pursuing the maximization of tunneling distance or only paying attention to the minimization of tool changing cost, and the method ignores the complex relationship of mutual res