Intermittent demand forecasting of aerospace rotable parts. A framework for unpredictable flight patterns

Demand forecasting of aerospace spare parts has a high impact on aircraft maintenance operations, aircraft serviceability and companies’ profitability. Traditional forecasting methods used in the industry utilise past consumption for forecasting future demand, often overlooking operational data. Incorporating fleet usage-data for capturing service variability in demand forecasting methods is crucial in operations with unpredictable flight patterns, like military aircrafts, business jets, and different air services, like air ambulances, search and rescue or policing operations. In this paper, we present the development of a machine learning (ML) framework for the forecasting of aerospace rotable components, generally life limited or inspected regularly. Different traditional and ML-based forecasting methods are reviewed, the impact of different service-related features is analysed, and a framework for addressing the potential service variability of an asset during its lifetime is proposed. The framework is validated with historical spare parts consumption of a European maintenance, repair and overhaul (MRO) service centre, achieving the most accurate demand forecast in 99.2% of the stock keeping units (SKUs) analysed compared to traditional baselines.