Adaptive, Dynamic Life Models of Drive Train Clutch Systems

ALPHATECH proposes, in conjunction with our partners, Rolls-Royce North America, a novel approach to the prognostic modeling of drivetrain clutch systems, combining dynamic modeling of transients with physics-of-failure based hazard modeling. Inparticular, our effort is focused towards the F35B LiftFan clutch, to which our models and techniques will be applied. The goal is to predict clutch problems before they impact vehicle performance, enabling the vision of autonomic logistics throughprecision sensing and prognosis. Our approach entails two components: a modeling part and an estimation part. The model extends existing LiftFan clutch dynamic models to address the transient engagement dynamics, and incorporates severalphysics-of-failure models for different types of failures. The estimation component uses a Markov approach to compute the current and future fault probabilities, and incorporates a Bayesian data fusion scheme to combine the model-based estimates withanomaly detection estimates. The Markov approach dynamically adapts the hazard used in the life prediction using actual measurements. Further, our physical modeling approach addresses the unique features of this clutch, which has extremely high powerdensity and transient loads. The combined modeling and estimation solution has the potential to provide high-accuracy predictions of remaining useful life. Advanced prognostic models are urgently needed for the LiftFan clutch, which provides an immediatecommercial avenue for this research. More generally, drivetrain clutches of the disk type are found on a vast diversity of power transmission systems, from helicopters to cars. An improved prognostic capability in the area of clutch life would bemarketable in such areas.