Highly fault-tolerant thrust estimation for gas turbine engines via feature-level dissimilarity design

Thrust is a critical yet unmeasurable performance parameter of gas turbine engines during flight. Existing thrust estimation methods lack fault tolerance to single-sensor drift failures. This study proposes a highly fault-tolerant thrust estimation (HFTE) method, which employs a feature-level dissimilarity design (FLDD) to develop three highly dissimilar modules, acting as virtual thrust sensors. Aside from three reliable parameters—Mach number, low-speed shaft speed, and high-speed shaft speed—any single-sensor drift failure only affects one module, while the others function normally. The desired high fault tolerance is achieved by computing the median of their outputs. The innovations include: (1) Introducing FLDD to design three mutually backup virtual thrust sensors, ensuring fault tolerance to single-sensor drift failures. (2) Developing a novel feature selection strategy for FLDD, selecting three highly dissimilar, low-failure-risk feature sets for thrust estimation. Hardware-in-the-loop and ground testing experiments verify the effectiveness and engineering applicability of the HFTE method.