Comparison of Construction Alternatives Using Matched Simulation Experiments

Abstract

The comparison of alternative construction methods is one of the principal reasons for using simulation to model construction processes. The efficiency and effectiveness of such comparisons can be greatly improved by the prudent use of “matched pairs,” a variance reduction technique based on dedicated and fully synchronized random number streams. The basic methodology is illustrated by using the STROBOSCOPE simulation system to compare two alternative construction methods for rock tunneling [Conventional versus the New Austrian Tunneling Method (NATM)]. For this example the effects are dramatic. The probability of identifying and choosing the cheaper construction method based on a single run increases from 55% to 96%, the variance of the cost difference decreases by two orders of magnitude, and the 95% confidence interval for the true cost difference given by 4,000 independent runs can be obtained by performing only seven replications using matched pairs. Besides this improvement in statistical efficiency, the use of matched pairs is a necessity for this example in order to compare the alternatives on a logical and equitable basis.