Abstract: Based on the statistics from the field study of construction programs in Tianjin city and Zhejiang Province, the ABM was used to explore the approach to optimization of safety input scheme. In the Netlogo software environment, it ran with the Monte Carlo method for the sensibility ranking and finding the variation tendency of safety performance. The results show that when safety inputs and especially inputs of safety precaution increase, the majority of safety performance has a significant rise with a distinct marginal diminishing effect, that inputs of safety precaution obviously have synergy effects with safety equipment inputs and safety supervisors' salaries, that accident fatalities impose upper limits on the effect of safety equipment inputs, which indicates the existence of Pelzman effect, and that in most cases, there is an optimal value in total expense can reach an optimal value as the safety inputs increase.

Key words: safety input, safety performance, Netlogo, agent based model(ABM), Monte Carlo method