Modelling and development of an engine airpath coordinated control structure

Aim of this dissertation is the improvement of a benchmark airpath control logic, adopted on a 3-litre, four cylinders diesel engine, with a new control architecture, based on the gain scheduling control technique. Thesis work demonstrates the effectiveness of the control logic, highlighting the performance improvement that the exploitation of such schemes offers. Motivating purpose of the work lays behind the enforcement of stricter emissions requirements, along with the technological need for constantly developing more efficient solutions for the improvement of engine performance. The work is divided in three main steps. • Identification of the mathematical models, representing the physical behaviour of the airpath. • Development of a tailored gain scheduling control architecture, based on engine models obtained at previous step. • Validation of the control results, refinement, and comparison with benchmark experimental control logic. These general procedures are applied to the Exhaust Gas Recirculation valve, also known as EGR, and to the Variable Geometry Turbocharger, referred to as VGT. Engine operating range, defined by engine rotation speed and load, is split into twenty subregions. System identification is performed basing on simulated engine airpath plant, developed on GT-Power environment; inputs to the system are EGR and VGT actuators positions and outputs are air mass flow rate at intake manifold and boost pressure. SISO identifications are performed, one per each actuator and each subregion. Adopted model families for identification purposes are ARX, ARMAX, OE and state-space (SS). Gain scheduling control technique involves the development of several compensators of the same family, with coefficients that vary depending on engine map region. Twelve out of twenty identified models are chosen so as to be the basis of the gain scheduling logic, with a view to avoiding repetitions. These models are referred to as Control Engine Operating Points, or CEOPs, and are screened basing on similarities between identified SS models. The design of the control logic is performed through PI compensators: one per each of the twelve CEOPs is designed, via an iterative gains tuning, both for the EGR and the VGT actuation logics. Discerning the most fitting controller in each engine operating situation is done via engine mapping: each engine current state is assigned to the most adequate controller model. Three assignment procedures are proposed and illustrated: vicinity mapping, affinity mapping and k-means clustering. A range of anti-windup architectures is studied and compared, with a view to preventing saturation phenomena. Among these, integral anti-windup scheme achieving faster desaturation response is equipped on the controllers. Filters are applied to the VGT action and to sudden scheduled gain switches. Testing is performed adopting different engine cycles, progressively spanning from calm to aggressive, and a fine tuning of the control parameters is performed. A thorough evaluation of the obtained control performances and a comparison with the benchmark control directly provided by the engine ECU manufacturer is done, highlighting the tracking improvement that is achieved via the designed control architecture. Controller design and validation is performed progressively enriching the operativity range of the control action, assessing its strength and limitations. Several techniques are compared throughout the design, so as to choose the best one to address each control issues.