Towards predictive maintenance of lithography systems: Robust fault diagnosis via LPV-to-LTI reformulation

Fault diagnosis plays a critical role in predictive maintenance of high-precision mechatronic systems, such as lithography machines, where unplanned downtime can cause substantial economic losses. Traditional model-based methods rely on fixed-parameter linear time-invariant (LTI) models and require accurate mathematical representations, while purely data-driven techniques often suffer from sensitivity to noise, large data requirements, and poor robustness under distribution shifts. These limitations become more pronounced in linear parameter-varying (LPV) systems, where system dynamics inherently change with operating conditions, making it difficult to distinguish between normal parameter variations and genuine faults.To address this challenge, this paper proposes a novel framework that reformulates the fault diagnosis problem of LPV systems with small parameter variations as an uncertain LTI problem. Within this framework, a fault detection filter with robustness enhancement is formulated to balance the H_i / H _∞ performance index, maximizing fault sensitivity while maintaining resilience to disturbances, modeling errors, and parameter variations.The feasibility and effectiveness of the proposed method are validated through simulations on the long-stroke module of wafer scanners in lithography systems.