Failure of assets reduce the output of production systems but are unaccounted when schedules are made. This reduction in output can be significant when assets are operating in remote areas as it takes time for maintenance crews to arrive. Increasing the frequency of scheduled maintenance can reduce the risk of failure but also decreases availability and production. Hence there is a need to optimise maintenance to maximise asset availability. The aim of this project is to develop a method to maximise the availability of assets in a real-world system by changing the scheduled maintenance program. This is achieved using discrete event simulation and a genetic algorithm developed using Java. This approach is applied to a case study of a locomotive fleet using data from 118 locomotives over a 24-month period. The reliability of the six subsystems of the locomotives are estimated by analysing their maintenance records.
The six locomotive subsystems analysed all exhibit infant mortality. This results in the validated simulation and genetic algorithm suggesting a run-to-failure strategy. This increases the rate of failure but only increases the downtime due to maintenance activities by 20%. Experimental results suggests this change leads to an increase of availability and production of 70%. Thus, adjustments to preventative maintenance scheduling could result in significant increases in production. The method developed can be used as a guide to perform maintenance optimisation on any system.
Alastair Chin completed his for his Mechanical Engineering and Computer Science Masters project. This was cosupervised by Lyndon While in Computer Science and Melinda Hodkiewicz in Engineering and data was provided by BHP.