Please use this identifier to cite or link to this item:
|Title:||Improved optimisation method using genetic algorithms for mass transit signalling block-layout design|
|Authors:||Chang, C.S. |
|Citation:||Chang, C.S.,Du, D. (1998). Improved optimisation method using genetic algorithms for mass transit signalling block-layout design. IEE Proceedings: Electric Power Applications 145 (3) : 266-271. ScholarBank@NUS Repository.|
|Abstract:||The application of an improved optimisation technique to an existing blocklayout design process is proposed for equi-block 72-aspect (H ≥ 2) mass transit signalling systems. The approach is built upon previous published work by applying genetic algorithms (GA) in place of conventional gradient search methods. The genetic algorithms have been proven theoretically and empirically for providing multiple-point search, as well as robust and global convergence in complex search spaces. Such characteristics have enabled GAs to simplify the original approach and to broaden its scope for dealing with either changes of the objective function or changes of the signalling scheme. In the proposed formulation, a railway line between two stations is divided into three sections, f namely: a constraint section, a stretchable section and a critical section. Since each of these sections is prescribed with a different headway design criterion, GA is applied separately to each section to optimise the layout of block joints and their positions within each section. Train performance simulation is then used to verify each optimised signalling design against the overall requirements of mass transit signalling systems operation. It is also used to calculate the objective function for each candidate solution as generated by GA. Comparative studies are presented to show improvements made by GAs over conventional optimisation techniques. © IEE, 1998.|
|Source Title:||IEE Proceedings: Electric Power Applications|
|Appears in Collections:||Staff Publications|
Show full item record
Files in This Item:
There are no files associated with this item.
checked on Nov 24, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.