A cycle-mean-value, quasi-steady, thermodynamic model of slow-speed, two-stroke marine Diesel engines, used for performance prediction and engine-propeller/turbocharger matching, is converted into a power-plant analytic model. The dynamic part of the cycle-mean model consists of the two first-order differential equations for the cycle-mean crankshaft and turbocharger shaft rotational accelerations. This form implies a state-space formulation of the power-plant modeling approach. However, engine, turbine and compressor torques have to be calculated through the solution of the algebraic part of the model, which consists of a nonlinear, perplexed algebraic system of equations not analytically solvable. This inhibits the formulation of the power-plant state-space description. By approximating the torque maps, generated by the thermodynamic model, with neural nets, explicit functional relationships are obtained. Identification of the power-plant operating regimes through linearization and decomposition is performed. In effect, a supervisory power-plant controller structure, applicable to real-time control and diagnostics, is proposed, incorporating the nonlinear state-space description of the plant.

This content is only available via PDF.
You do not currently have access to this content.