Abstract: Constrained nonlinear model predictive control (NMPC) is applied to a turbo-shaft engine based on a nonlinear integrated helicopter/turbo-shaft engine simulation system with a reliable confidence level. First, a new method called multiple-output recursive reduced least square support vector regression (RRLSSVR) is proposed to build an engine dynamic predictive model. In the flight envelope of altitude 0-5 km and forward speed 0-75 m/s, the predictive model has a satisfying accuracy within 5‰. Subsequently, considering the comprehensive influences of rotor torque, engine fuel flow, gas turbine rotor speed, power turbine rotor speed and their constraint conditions, a rolling optimizer is designed with the sequential quadratic programming (SQP) algorithm. Then, in order to track the constant reference signal with no error, a correcting step is utilized by adding a deviation integral to the output of the predictive controller. Finally, in comparison with the cascade PID controller, the proposed predictive controller can decrease the droop or overshoot of power turbine rotor speed remarkably during the maneuver flight simulations of the helicopter.

Key words: helicopter, turbo-shaft engine, nonlinear model predictive control, support vector regression, sequential quadratic programming, cascade PID control