NEURAL NETWORK-BASED VIRTUAL SENSORS IN FLIGHT CONTROL SYSTEMS

摘要

This paper presents a study aiming at reducing the cost of hardware replication in flight-critical systems, in particular for what concerns the Flight Control System (FCS) sensors. Compliance with the safety requirements is a significant driver in the definition of the total costs of a FCS and the current approach is generally based on the definition of a proper level of hardware redundancy, impacting on both the affordability and complexity of the system. An alternative approach, based on the adoption of soft computing techniques to implement an analytical redundancy, is presented and some results are shown. The objective of the study concerns in particular the possibility of generating additional and synthetic signal sources to be used to validate the physical ones.