The paper offers an intelligent approach to analyze and determine the design parameters minimizing the total cost and achieving the desired performance measures in the maintenance float systems. The expected total cost in a maintenance float system includes the cost of lost production, the cost of repair persons and the cost of standby machines. The developed design procedure integrates simulation, metamodel and genetic algorithms. Neural networks are able to approximate functions based on a set of sample data, i.e. construct metamodels from simulation results in this study. The objective of metamodels is to predict simulation responses in order to significantly reduce the amount of simulation runs. The predictive performance of neural metamodels comparably outperforms the traditional regression metamodels. The neural metamodels are further extended to formulate a decision model for optimizing the maintenance float systems by using genetic algorithms.

No abstract