Due to the challenges posed
by varying factors such as environmental conditions, equipment performance, and
operational procedures, achieving an optimal balance between safety,
efficiency, and cost in tunnel excavation is complex. To address this problem,
a case study of a section of the Wuhan rail transit is conducted, proposing a
hybrid intelligent framework that integrates categorical boosting(CatBoost) and
nondominated sorting genetic algorithm(NSGA)-Ⅲ.
This framework aims to develop an intelligent control decision-making system for construction parameters, considering factors such
as advance speed, tunneling specific energy, and tool wear. The approach begins
with the use of the CatBoost model to predict advance speed, specific energy,
and tool wear, which then informs the fitness function for control targets.
Following this, the CatBoost-NSGA-Ⅲ algorithm is employed for multi-objective
optimization of construction parameters based on the fitness function derived
from the CatBoost model. Optimal parameter combinations are subsequently
selected from the multiple Pareto optimal solutions using a fuzzy decision
method. This approach supports intelligent prediction and optimization of
construction parameters for shield tunneling. Key findings include: (1) The
CatBoost model provides accurate predictions with a goodness of fit exceeding
0.9. (2) Multi-objective optimization using the
CatBoost-NSGA-Ⅲ algorithm,
combined with fuzzy decision-making, determines the
optimal scheme. Compared to the average measured data, this scheme reduces
specific driving energy and tool wear by 5.3% and 13.5%, respectively, while
increasing advance speed by 6.3%, thereby enhancing intelligent management and
decision-making for shield tunneling.
CHEN Libo
,
ZHANG Mingshu
,
CHEN Haiyong
,
WU Xianguo
,
CAO Yuan
.
Analysis and Optimal Control of Shield Tunnel Construction Parameters Using Categorical Boosting-Nondominated Sorting Genetic Algorithm-Ⅲ
[J]. Tunnel Construction, 2024
, 44(8)
: 1587
-1598
.
DOI: 10.3973/j.issn.2096-4498.2024.08.006
