ForeSim-BI: A predictive analytics decision support tool for capacity planning

This paper proposes a decision support tool for maintenance capacity planning of complex product systems. The tool - ForeSim-BI - addresses the problem faced by maintenance organizations in forecasting the workload of future maintenance interventions and in planning an adequate capacity to face that expected workload. Developed and implemented from a predictive analytics perspective in the particular context of a Portuguese aircraft maintenance organization, the tool integrates four main modules: (1) a forecasting module used to predict future and unprecedented maintenance workloads from historical data; (2) a Bayesian inference module used to transform prior workload forecasts, resulting from the forecasting module, into predictive forecasts after observations on the maintenance interventions being predicted become available; (3) a simulation module used to characterize the forecasted total workloads through sets of random variables, including maintenance work types, maintenance work phases, and maintenance work skills; and (4) a Bayesian network module used to combine the simulated workloads with historical data through probabilistic inference. A linear programming model is also developed to improve the efficiency of the decision-making process supported by Bayesian networks. The tool uses real industrial data, comprising 171 aircraft maintenance projects collected at the host organization, and is validated by comparing its results with real observations of a given maintenance intervention to which predictions were made and with a model simulating current forecasting practices employed in industry. Significantly more accurate forecasts have been obtained with the proposed tool, resulting in an important cost saving potential for maintenance organizations.