Bayesian inference for fault-tolerant control

Abstract

In this contribution, we present initial developments in view of model-based fault-tolerant control (FTC). In this context, we use an original method based on the Kalman-filter by which fault detection, diagnosis and accommodation is possible provided that an accurate model is available. Since this is not generally true, we attempt to alleviate this necessity by means of accounting for uncertainty, in both model as well as in the measurements used for fault diagnosis. Our preliminary results are focused on the diagnosis step in the FTC scheme.