In aviation Maintenance, Repair, and Overhaul (MRO) systems, spare parts inventory management faces highly stochastic demand, significant tail risks, supply delays, and complex contractual constraints, making it difficult for traditional deterministic or risk-neutral models to balance high service levels with low costs. This paper proposes a multi-stage stochastic programming inventory optimization model incorporating Conditional Value-at-Risk (CVaR) to enhance economic efficiency and robustness. The model employs a scenario generation framework based on moment matching and entropically regularized optimal transport that preserves inter-part correlations through Pearson coefficients and efficiently reduces scenario trees via Sinkhorn distance, accurately capturing demand volatility and extreme tail events. It further incorporates non-anticipativity constraints in a multi-stage mixed-integer stochastic program, along with rolling-horizon mechanisms, tiered pricing, and contract restrictions to enable dynamic ordering decisions under sequentially revealed information. A Mean-CVaR optimization criterion with a tunable risk aversion parameter is introduced to balance expected cost against extreme losses. Numerical experiments on typical spare parts from a commercial airline’s B737 fleet show that the risk-averse model modestly increases expected cost relative to the risk-neutral version but substantially reduces tail risk. Stochastic strategies outperform deterministic approaches, with value-of-stochastic-solution metrics confirming clear advantages under uncertainty, while model stability and robustness are validated through in-sample and out-of-sample testing. This study offers airlines a practical framework for developing cost-effective and safe spare parts inventory policies in uncertain environments.
Yang et al. (2026) studied this question.