粒子群优化
工程类
过程(计算)
计算机科学
控制工程
机器学习
操作系统
作者
Xiaojun Li,Sicheng Zhao,Yi Shen,Gang Li,Hehua Zhu
标识
DOI:10.1016/j.tust.2023.105040
摘要
The timely and appropriate adjustment of operational control strategy is necessary for efficient tunnelling and hazard prevention in tunnel boring machine (TBM) projects. Because TBM tunnelling parameters have specific features, such as spatiality, real-time, and constraint pluralism, the just-in-time (JIT) optimization of the tunnelling parameters is still challenging. In this study, we propose an integrated parameter optimization approach to provide a JIT operational control strategy for TBM tunnelling, consisting of two parts: (i) a machine-learning based rock–machine mapping model with TBM operational parameters and real-time geological information as input and tunnelling loads as output, updated by an out-of-core retraining method. (ii) a JIT optimization model, considering excavation efficiency as the optimization objective and construction safety and tool wear as multivariate constraints. Light gradient boosting machine (LightGBM) and particle swarm optimization (PSO) algorithms are merged in the modelling process. To validate the method and implement the in situ service, a database was established, consisting of more than 685,000 time-series data collected from the Pearl River Delta Water Resources Allocation Project, which is divided into 14 sections to simulate the data flow update. Combined with an intelligent platform, the proposed integrated parameter optimization approach is conducive for meeting the requirements of the efficiency, the accuracy and the stability of providing the JIT operational control strategy. This approach was deployed as a module in the PRDWRA TBM Tunnel intelligent construction safety control platform. This module provides JIT operational control strategy suggestions for project participants who are quite conscious of the efficiency and safety of projects.
科研通智能强力驱动
Strongly Powered by AbleSci AI