Failure Detection and Compensation for Aircraft Inertial System

Abstract

This chapter is dedicated to techniques for ensuring fault tolerance in redundant aircraft sensors involved in computation of flight control laws. The objective is to switch off the faulty sensor and to compute a reliable (a.k.a. as “consolidated”) parameter using data from valid sensors, in order to eliminate any anomaly before propagation in the control loop. The benefit of the presented method is to improve the consolidation process with a fault detection and isolation approach when only few sources (less than three) are valid. Different techniques are compared to accurately detect any behavioral change of the sensor outputs. The approach is tested on a recorded flight dataset. This chapter is dedicated to fault detection and isolation of redundant aircraft sensors involved in the computation of flight control laws. The objective is to switch off the erroneous sensor and to compute a so-called consolidated parameter using data from valid sensors, in order to eliminate any anomaly before propagation in the control loop. We will focus on oscillatory failures and present a method for integrity control based on the processing of any flight parameter measurement in the flight control computer (FCC) like, e.g., anemometric and inertial data. One of the main tasks dedicated to the FCC is the flight control laws (FCL) computation which generates a command (position order) to servo-control each moving surface (see Fig. 5.1). The comparison between the pilot commands (or the piloting objectives) and the aircraft state is used for FCL computation. The aircraft state is measured by a set of sensors delivering, e.g., anemometric and inertial measurements that characterize the aircraft attitude, speed, and altitude. The data is acquired using an acquisition system composed by several dedicated redundant units (usually three). The FCC receives three redundant values of each flight parameter data from the sensors and must compute unique and valid flight parameters required for the FCL computation. This specific data fusion processing, called “consolidation,” classically consists of two simultaneous steps (Fig. 5.2): selection or computation of one unique parameter from the three available sources, and, in parallel, monitoring of each of the three independent sources to discard any faulty one. As a consequence, the consolidation allows reliable flight parameters computation with the required accuracy by discarding any involved failed source.