Robust multivariable control of an engine-dynamometer system

Presented in this paper is the robust multivariable feedback controller design of a highly coupled, nonlinear diesel engine-dynamometer system. The performance goal is to maximize the closed loop reference tracking of prespecifled engine speed and torque curves by reducing the output variations due to system nonlinearities/uncertainty, and loop interactions through feedback control. The difficulties in controlling engine-dynamometer systems include the large degree of loop interactions, induction-to-power delays, combustion uncertainties, and engine nonlinearities. The performance goal is realized by balancing the bandwidths of the loop transfer functions to avoid excessive loop interactions in the closed-loop system. Standard high-gain and high-bandwidth solutions cannot be used for this design since the system contains pure delays and the controller implementation has sample rate limitations. The engine-dynamometer models used for controller design are developed from spectral estimation techniques and step responses where the effects of the system nonlinearities are captured in a parametric uncertain format. The controllers are designed using a sequential frequency domain approach. The controllers are implemented on an 8.3L turbocharged diesel engine and eddy current dynamometer maintained at Cummins Engine Company, Columbus, IN. The controllers are evaluated based upon the ability of the closed-loop system to track step inputs and a transient test cycle.