I. Introduction ircraft are extremely complex dynamic systems whose safe operation depends upon a large number of very different sub-systems. Upset conditions in any of these sub-systems may result in very dangerous situations, often with catastrophic endings1. Increasing the safety of aircraft operation has in recent years become a major objective for the aerospace engineering community2. As part of the efforts to achieve this goal, control systems capable of accommodating abnormal conditions have been developed using a variety of technologies3, 4. Most of the proposed approaches for fault tolerant control laws rely on some form of failure detection and identification5, 6 (FDI) with twofold action. One is to increase pilot situational awareness and the other is to trigger specific compensatory actions from the automatic control system. It should be noted that most of the research efforts in the FDI area have focused on individual classes of failure and did not address the need for integrated FDI schemes. State estimation or observer-based schemes have been widely proposed5-8 for actuator FDI relying on Kalman or other classes of filters. Artificial Neural Networks (ANN) have also been extensively used9-12 to solve the FDI problem for aerospace systems. Alternative approaches for FDI and pilot awareness enhancement were also proposed based on inductive learning13.

The attempt to integrate FDI for a large diversity of aircraft sub-systems and over-extended areas of the flight envelope poses significant challenges due to the complexity and extremely high dimensionality of the problem. Adequate tools are necessary to develop a comprehensive solution. In recent years, a new concept–the artificial immune system (AIS)-based FDI-has emerged as a very promising tool and has been used in an extremely wide variety of dynamic systems 14-18. This rather new biologically-inspired technique has specific characteristics that allow it to deal with complex multidimensional problems and large amounts of information. In this paper, the development of an Immunity-Based Failure Detection and Identification (IBFDI) scheme for aircraft actuator failures is described. Within this research effort, real flight data from a small jet UAV have been used for the first time to design and validate an artificial immune system (AIS)-based FDI scheme. It is expected that the AIS approach provides adequate tools to handle the multi-dimensionality of the integrated aircraft FDI problem. The next section will briefly outline the basic background of the AIS applied to failure detection pertaining to aircraft sub-system failure. In Section III, the UAV used and the test flight scenario are described along with the procedures for data acquisition and processing. The design of the FDI scheme is discussed in Section IV. The FDI results and the analysis of the FDI performance are presented in Section V. Some conclusions are summarized in Section VI. Finally, acknowledgements and the list of the cited bibliography complete this paper.