Roster Modelling for FIFO in the mining, oil and gas sector using genetic algorithms and neural networks

Chief Investigator

Australia currently has approximately 130,000 Fly-In/Fly-Out (FIFO) mining workers and 70,000 construction workers nationally as at 2011 (Australian Government report, 2011). The vast majority of FIFOs are on two week rosters and 45% time at work, which equates to 10.4 million plane trips (i.e., 5.2 million return trips) per year, approximately 84,000 room nights, 250,000 meals per day and other associated supply logistics such as cleaning and administration. Currently, the systems that manage the rostering of FIFO worker's travel, accommodation, meals, cleaning and administration, have limited ability to assist with determining the most efficient roster model. A 3% waste across transport and accommodation equates to millions of dollars (e.g., approximately 300,000 flights, 2,500 rooms and 7,500 meals). Any reduction in waste would represent substantial saving. The ability to efficiently schedule resources and maintain sufficient staff coverage, rest periods etc. has the capacity to improve the productivity of FIFO enterprises. The aim of this research is to explore the use of artificial intelligence to build an optimised travel, accommodation and roster model based on varying sets of seed data and supervised and unsupervised neural network training. Neural networks can be used to optimise rosters, transport and accommodation using a subset of all required parameters. Two approaches will be investigated to determine the most effective and efficient AI model. These Please provide a brief outline of the proposed project and deliverables Researchers in Business Application Form v4 Page 11 of 15 include neural networks with 1) supervised training using cloud based historical data or 2) unsupervised training using Genetic algorithms. The outcome of this project will be an artificially intelligent solution to optimisation modelling of workforce logistics in the mining, oil and gas sector. The product will be an intuitive, automated roster modelling tool that will benefit Osmotion's current and future clients by offering waste reduction and economic value through optimising the FIFO roster model.

01 Sep 2014 - 30 Apr 2015

N/A