Fault Detection and Flight Data Measurement. Demonstrated on Unmanned Air Vehicles Using Neural Networks

You are here: Homepage > Search

Fault Detection and Flight Data Measurement. Demonstrated on Unmanned Air Vehicles Using Neural Networks - Conclusions and Future Work

Samy, Ihab / Gu, Da-Wei

This book aims to exploit existing aircraft technologies to reduce costs in UAVs. The technologies include a NN-based SFDIA scheme tested on a nonlinear UAV model, and a FADS system tested on a MAV. In industry, sensor faults are generally detected based on physical redundancy and/or limit value checking techniques. However such methods can suffer from high instrumentation costs, slow fault detection times and a high sensitivity to sensor noise. Over the years model-based SFDIA schemes have been proposed to overcome the drawbacks of traditional SFDIA methods. However the theory has generally targeted linear, fixed model based methods. Unfortunately such methods can be limited to linear, time-invariant (LTI) systems. Novel methods consider the use of NNs due to their nonlinear and adaptive structures. Fault detection techniques have been applied to large manned aircrafts, underwater vehicles, and autonomous helicopters, while few have been extended to fixed wing UAVs. Work carried out using NN-based methods includes. The work presented in this book is distinct from previous research in that a NN-based SFDIA scheme is tested on a UAV application. Model-based methods are an invaluable alternative to traditional approaches (such as physical redundancy) especially for UAVs due to weight and cost restrictions.