Valuing data in aircraft maintenance through big data analytics: A probabilistic approach for capacity planning using Bayesian networks

Abstract

Capacity planning is an important problem faced by aircraft Maintenance, Repair and Overhaul (MRO) organizations given the uncertainty of maintenance workloads. Despite the considerable amount of data generated and stored during the planning process, these have yet to provide a decisive competitive advantage to aircraft MROs. This paper addresses this problem by exploring Bayesian networks (BNs) as a big data and predictive analytics (BDPA) tool to cope with the uncertainty on both scheduled and unscheduled maintenance workloads and to improve the MROs capacity planning decision-making process based on incomplete information. The BNs were developed from a real industrial dataset referring to 372 aircraft maintenance projects of a Portuguese MRO and comprise information variables representing typical information collected during the planning process and hypothesis variables representing the workloads required to be estimated. The benefits of applying BNs as a BDPA tool in aircraft maintenance are demonstrated through examples referring to capacity planning, but also sales planning, using real maintenance data. The BDPA tool based on BNs is generic and can be applied to the maintenance capacity planning process of any MRO, allowing accurate estimations and more informed decisions to be made when compared to current practices, which are based on descriptive statistics of past maintenance workloads.