A Nonlinear State-Space Model of Diesel Propulsion Plant Operation Using Neural Nets

Abstract Cycle-mean-value, quasi-steady, thennodynamic models of the slow-speed, two-stroke turbocharged marine Diesel engine is a valuable simulation tool. Engine operation models of this kind comprise of two differential equations and a nonlinear, perplexed algebraic system for the engine/turbine/compressor torques. Neural nets are proposed in order to establish the functional dependence of the torque variables upon the state variables (engine/turbo rpm) and the control input (fuel index). In effect a state-space model is obtained which is decomposed to a nonlinear nominal model and a linear uncertain perturbation model, convenient for the application of robust control methods.