Uncertainty and Business-Aware Remaining Useful Life Estimation for Semiconductor Manufacturing

Semiconductor manufacturing relies on tightly interconnected components, soearly identification of the assets most likely to fail is essential to preventa single breakdown from disrupting the entire production pipeline. Maintenanceplanning must therefore balance unexpected failures against prematurelyinterrupted operating life. We present a Predictive Maintenance (PdM) frameworkcombining Deep Learning (DL) sequence models and Simoultaneous Quantile Regression (SQR) for uncertainty-aware Remaining Useful Life (RUL) estimation and risk-aware maintenance decisions. Several architectures arecompared on ion-milling data from the 2018 PHM Data Challenge (PHM18), including architectures basedon State Space Models (SSM), usingprediction and business metrics: Unexpected Breaks (UB), Unexploited Lifetime (UL), and a cost-weightedobjective. Diagonal State Spaces (S4D) delivers the best RUL estimates across quantiles and,relative to Preventive Maintenance (PvM) baselines, substantially lowers business cost by avoidingsystematically early interventions. The results support uncertainty-aware,cost-sensitive maintenance planning in semiconductor production.