• Projects
  • Active Flow Control
  • Adaptive Reconfigurable Control
  • Diagnostics and Prognostics
  • Intelligent Guidance and Trajectory Reshaping
  • Real-Time Modeling and Prediction
    • Next-Generation Fuel Controller
    • Helicopter Advanced Control Technologies
    • Neural Network Limit Avoidance System for Rotorcraft
  • Tools for Healthcare Assessment

Helicopter Advanced Control Technologies

Abstract

Barron Associates is involved in the Helicopter Advanced Control Technology (HACT) program to enhance handling qualities and improve mission effectiveness by designing and developing neural network systems to improve the degree of accuracy in helicopter flight control parameters that will ensure proper flying performance.

Problem

The Helicopter Advanced Control Technology (HACT) program is developing, demonstrating(via flight test) and quantifying the benefits of the HACT Flight Control System (HFCS). The HFCS will employ active, digital flight control technologies to enhance handling qualities and improve mission effectiveness. Central to the control law design is a requirement for certain helicopter flight parameters that cannot be sensed in flight (e.g., main rotor inflow ratio) to be estimated or predicted from other, measurable parameters (e.g. air density, collective pitch). Accurate prediction of these parameters is necessary to ensure proper flying qualities.

Solution

Barron Associates is using set-point condition (i.e., quasi-static limit value) training data in tandem with the expected forms of functional relationships (as derived from helicopter physics) to develop robust neural network structures that are capable of on-line prediction of several parameters relevant to helicopter flight control such as thrust coefficient, coning and flapping in both the main and tail rotors. These polynomial models not only have to provide accurate estimates of critical parameters, but also must match known partial derivatives. Thus far, Barron Associates has constructed models, utilizing structure-learning techniques, with a high degree (i.e., R-squared greater than .95) of accuracy.