Neural Network Models for the On-Board and Individual Blade Control of Helicopter Rotors

Abstract

Listen

Translate article

Neural Networks are evaluated to model OnBoard Bla de Control concepts for helicopter rotors. Computational Fluid Dynamic simulations of active flap and active twist concepts have been run using the compressible Navier Stokes solver OVERFLOW. Neural Network models were then made as function of Mach number and mean angle of attack for the change in lift, drag, and pitching moment and their associated time constants. These models show the ability to capture the nonlinear effects of stall and shocks. These models are now suitable for incorporation into rotorcraft flight simulation software. I. Introduction elicopters are versatile vehicles that can vertically take off and land, hover, and perform maneuvers at very low forward speeds. These characteristics make them unique for a number of civilian and military applications. However, the radial and azimuthal variation of dynamic pressure causes rotors to experience adverse phenomena such as transonic shocks and 3-D dynamic stall. Adverse interactions such as blade vortex interaction and rotor- airframe interaction may also occur. These phenomena contribute to noise and vibrations. A variety of techniques have been proposed for reducing the noise and vibrations and for improving handling qualities. These include on-board control (OBC) devices, individual blade control (IBC), and higher harmonic control (HHC). Addition of these devices adds to the weight, cost, and complexity of the rotor system and reduces reliability of operations. Simpler OBC concepts will greatly alleviate these drawbacks and enhance the operating envelope of vehicles. IBC and OBC concepts may be modeled using physics-based computational fluid dynamics (CFD) and computational structural dynamics (CSD) tools that are coupled to each other. However, these approaches are expensive and not suitable for the design of controllers. Reduced order models are highly desirable for designing these devices and for assessing the handling qualities of helicopters that employ these devices. Development of reduced order models is hampered by the fact that the flow field is highly non-linear and spans a wide range of Mach numbers and mean angles of attack. II. Approach A three-step approach is under development for modeling IBC and OBC devices, in particular active twist and trailing edge flap devices mounted on rotor blades. The first step involves conducting 2-D and 3-D non-linear simulations to generate a large database of steady state lift, drag, and pitching moments as a function of angle of attack, Reynolds number, and Mach number. The second step involves modeling the evolution of lift, drag, and pitching moment as a function of time for the active twist and trailing edge control concepts. The third step involves the use of neural network curve fits of the computed static and unsteady flow data. The fourth and final step is the incorporation of these neural network based models in rotorcraft flight simulation software and comparisons with high fidelity simulations. In this work, results from the first three steps are discussed. Incorporation of the neural network models in flight simulation software are also being done. For details, the reader is referred to Ref. 1. The computational fluid dynamics analyses used in this study were all done using the compressible Navier-Stokes computational fluid dynamics code OVERFLOW 2 Versions 2.0y and 2.1y. OVERFLOW, developed by NASA,