A surrogate-based optimization method for the issuance of passenger evacuation orders under ship fires

Abstract

In order to reduce assembly time and improve the computational efficiency under ship fires, a surrogate-based optimization technique is proposed to estimate the optimal time of the issuance of evacuation orders, which determines the passenger response times. The Legendre polynomial chaos expansion technique is employed to construct the surrogate model of passenger assembly time with response time parameters, and validated with the original computer evacuation model. The genetic algorithm (GA) is subsequently applied to find the optimal response time combination. A 3-storey passenger ship is presented and four different fire scenarios that involve typical fire locations are considered to address the validity of this proposed method. The results suggest that this proposed method can be used to determine the optimal and the worst response time parameters with a significantly reduced number of evaluation samples. The relative error of assembly time from the surrogate model follows the normal distribution. For evacuation zone with the longest average travel distance, passengers should be informed of the immediate evacuation. For evacuation zones with short average travel distances, the optimal response time may be larger than the worst solution. This work provides a new method to make a reasonable evacuation planning in ship fires.