Modellbasierte optimale Mehrgrößenregelung eines aufgeladenen Dieselmotors mittels Methoden der nichtlinearen Optimierung

Abstract

Considering the increasingly restrictive emission standards for diesel engines, the automotive industry is in need of innovative ideas and solutions for engine control problems. This work addresses the challenge of determining optimal controllers for multiple, mutually dependent actuators in the engine gas system. The problem is posed as a parameter identification problem for a system of ordinary differential equations that model the gas system against measurements of an actual engine. The standard approach is to apply an NLP solver to a shooting technique, however this gives rise to several local minima when applied to large data sets. An alternative approach is studied, which performs the integration of the discretised ode system within the optimisation. The resulting high-dimensional optimisation problem is solved by a sparse NLP solver. Based on the nonlinear dynamic model a linear-quadratic regulator is designed and validated with a rapid prototyping system on a test vehicle.