Induction of tunnel reinforcement selection rules by using decision tree technique

摘要

Tunnel reinforcement method such as fore-poling and steel pipe roof is quite often employed to prevent the local failure of tunnel face caused by the uncertainty of ground condition. Due to the uncertainty of ground and environment, design of tunnel reinforcement is empirical as the factors affecting local stability of tunnel are hardly evaluated and their influences have rarely been quantified against the stability of tunnel. Recently, fuzzy approximate reasoning, neural network, and multiobjective optimization approaches have been proposed to determine the stability of tunnel and select the tunnel supports or reinforcements. Adoption of these complicate approaches can reduce the effects of uncertainty induced from ground conditions and environment on the tunnel design in some extent, however, has limitation on developing a rule-based decision support system. With advance of information technology effective use of the converged wireless network and personal digital assistant (PDA) has been developed (Lee et al., 2006) in construction site. The expert system in combination with PDA and wireless network requires a rule-based system to facilitate the rapid revision and update of the tunnel design. The rule-based system can be efficiently programmed on a small computing tool such as PDA. Accordingly, we focus on the generation of tunnel reinforcement selection rules which will be coded on a PDA. In this study the generation of tunnel reinforcement selection rules follows decision tree techniques.