Progressive Certification Process Based on Living Systems Architecture Using Learning Capable Controllers

To expedite and facilitate the certification process of a new aircraft against different climatic conditions, a new methodology has been proposed that makes use of combined advantages of both neural networks and fuzzy logic. The idea is to devise a learning capable control system that helps to decrease the number of flight tests to a minimum and, therefore, decrease the cost of the flight-test program for initial certification of an aircraft. In this approach, the certifying authorities test the learning capability of the aircraft in some preselected climatic conditions. During service life the aircraft extends its knowledge base with some suitable data collected during each flight; until it reaches a certain level of maturity, as it is the case with a living system. Moreover, with some management techniques, aircraft of the same type can share data with one another. To show the different capabilities of this approach, three autolanding controllers in the presence of different strong wind patterns have been studied, namely, a classical proportional– integral–derivative (PID), a hybrid neuro-PID, and finally an adaptive network-based fuzzy inference system-PID. It is shown that intelligent architectures are suitable tools to fulfill the proposed idea because of their learning capabilities, robustness and generalization properties.