Multidisciplinary Trim Analysis Using Improved Optimization, Image Analysis, and Machine Learning Algorithms

Abstract

A multiobjective design optimization methodology is used to determine the trim controls that minimize power required, noise, and blade loads of a coaxial-pusher rotorcraft, and to quantify the trade-offs among those three objectives in the form of 3-dimensional Pareto frontiers. A moderate-fidelity simulation model is used, which includes blade flexibility and a free vortex rotor wake model. A hybrid optimizer is developed, which starts with a genetic algorithm and radial basis function-based response surfaces, and ends with a gradient-based refinement. A new gradient-based method for constrained multiobjective optimization is developed, based on an extension of the method of feasible directions. A new technique for the automatic interpretation of rotor maps, based on image analysis and k-means clustering is presented. A new technique based on a k-nearest neighbor algorithm predicts trimmability. These two techniques reduce the need for analyst intervention during the optimization and improve accuracy. Results are presented for a 6- and an 8-control effector coaxial configuration in high speed flight.