Optimization using simulation and response surface methodology with an application on subway train scheduling

This paper presents a technique based on discrete‐event simulation and response surface methodology to model and then optimize the schedule of subway train travels. The aim of this study is to find appropriate headways–time intervals between the travels of two consecutive trains–at different hours in order to optimize average passenger travel time and rate of carriage fullness. For physical reasons and the observance of safety standards, an increase in the train speed in order to decrease average passenger travel time may not exceed some specified limits. One of the ways to decrease this average is to appropriately adjust headways of trains in the schedule. For this purpose, a metamodel of multinomial type is fitted to the data obtained from simulation tests to describe the relation between input variables (headways) and output variables (average passenger travel time in system and carriage fullness rate), and then optimal combinations of input variables are obtained using a weighed metric method and sequential quadratic programming.