Robust fault detection for a jet engine using robust l1 Estimation

Abstract

This paper considers the application of robust l1 estimation to robust fault detection for a jet engine. It reviews the design of robust l1 estimators based on multiplier theory and the resulting fixed threshold approach to fault detection (FD). Although this methodology has been established, this paper investigates its application to a realistic system. As the jet engine is a highly nonlinear system, a linear representation of the system incorporates uncertainty to account for modeling errors. Due to the modeling errors and some unmeasurable disturbances, it is difficult to distinguish between the effects of an actual fault and those caused by uncertainty and disturbances. Of the various types of faults that may occur, it is typical for a jet engine to experience a drift in the sensor reading, where the error between the actual and sensed values increases over time. It is the aim of a robust FD system to be sensitive to faults, such as this, while remaining insensitive to uncertainty and disturbances. In addition to uncertainty in the system and output matrices A and C, respectively, this paper considers uncertainty in the input matrix B, whereas current theory does not apply to this uncertainty. Using a discrete-time, linear uncertain model of a jet engine and implementing an LQG controller, the closed-loop system is formulated. A fault is represented by simulating sensor drift in the system. The results of FD using robust l1 estimation with a fixed threshold are demonstrated for the closed-loop system with very positive results.