Air-to-air refueling (AAR) has been commonly used in military jet applications. Recently, civilian applications of AAR have been garnering increased attention due to the high cost of air travel, which is largely dictated by the cost of jet fuel. There are two types of AAR approaches:​ probe-drogue and flying boom systems. This work explores the probe-drogue AAR system in commercial applications. Typical AAR applications deploy a drogue connected to a long flexible hose behind a moving aircraft tanker. The drogue is connected to a probe in a receiver aircraft before initiating fuel transfer and is retracted back into the tanker when the fuel transfer is completed. In order to ensure a safe and efficient refueling operation sophisticated systems need to be developed to accommodate the turbulences encountered, particularly in respect to vibration reduction of the flexible hose and drogue. The objective of this work is to develop a probe-drogue system for helicopter AAR applications. The first project is to make a preliminary design of a new AAR system for helicopter refuelling from a modified AT-802 tanker aircraft.

The second project is to model and control the drogue system under different turbulence conditions. As the real AAR system is not available, a flexible structure will be used instead to explore strategies for adaptive drogue system control. In order to achieve adaptive control of the flexible structure, an adaptive neuro-fuzzy (NF) controller is developed to suppress the vibration of perturbed flexible structures under variable dynamics conditions. A new hybrid training technique based on the bisection particle swarm optimization (BPSO) algorithm is proposed to optimize the NF controller parameters recursively to accommodate changes to system dynamics. To solve the issue of controller response to nonlinearities inducing additional vibrations in the steady state solution space, a fuzzy boundary function is proposed to shape the control signal output. The parameters related to output suppression are optimized simultaneously during recursive NF system training. The effectiveness of the proposed NF controller and hybrid training method is verified by experimental tests under different system dynamics conditions by placing mass blocks at different locations on the flexible beam. Experimental tests have shown that the proposed adaptive NF controller and hybrid training method outperform other related control schemes in terms of overshoot, undershoot, and settling time.