PhD Thesis
Model-Based Optimization of Combustion-Engine Control
Gabriel Turesson
Abstract
The work presented in this thesis is motivated by the need to reliably operate a compression-ignition engine in a partially premixed combustion (PPC) mode. Partially premixed combustion is a low temperature combustion concept, where the ignition delay is prolonged to enhance fuel-air mixing in the combustion chamber before the start of combustion. A premixed combustion process, in combination with high levels of exhaust-gas recirculation (EGR), gives low combustion temperatures, which decrease NOx and soot formation. Lowered combustion temperatures also reduce heat-transfer losses which increase the thermodynamic engine efficiency. The ignition delay is, however, determined by chemical reactions rates, which are sensitive to temperature, gas-mixture composition, fuel properties and fuel-injection timing. This sensitivity makes PPC more challenging to operate compared to conventional diesel combustion. Challenges related to PPC include an increased sensitivity to operating conditions, decreased combustion-timing controllability, high pressure-rise rates, and low combustion efficiency at low engine loads. These challenges put high demands on the engine control system that needs to be able to adjust fuel-injection timings and durations to compensate for the combustion sensitivity.
Therefore, this thesis investigates closed-loop combustion control for reliable PPC operation. The feedback loop from pressure-sensor measurement to fuel-injection actuation is studied in particular. A common theme for the controllers presented is the use of models in the controller design. Either to evaluate controller performance in simulation, or to optimize engine performance online. The principle of model predictive control is used for its ability to incorporate modeled system behavior in the controller design, and to control multi-variable systems with input and output constraints.
The problem of tuning robust and noise insensitive combustion-timing controllers, and its dependence on fuel reactivity is studied in simulation. Simulation results reveal a nonlinear relation between start of injection and combustion timing that depends on both load and fuel reactivity. Optimization is used to find robust and noise-insensitive controller gains. Guidelines for combustion-timing controller tuning are also presented.
Low-order autoignition models are evaluated and compared for the purpose of model-based controller design. The comparison shows that a simple autoignition model is sufficient for control of the ignition delay when the cylinder-charge properties are varied. This model is then used by a model predictive controller to simultaneously control ignition delay and combustion timing in transient engine operation, using both gas-exchange and fuel-injection actuation.
The effects of pilot injection on the combustion processes are characterized experimentally. Experimental results show that a pilot injection can decrease the main-injection ignition delay and maintain the pressure-rise rate at an acceptable level. This is utilized by a presented fuel-injection controller that manages to control both combustion timing and pressure-rise rate.
Strategies for improving the low-load performance of PPC are studied experimentally, where results show that the selection of injection timings and the use of a pilot injection are important when maximizing the combustion efficiency. The suggested low-load controller demonstrated a 9 % efficiency increase during transient engine operation.
This thesis also investigates the design of a controller that utilizes the degrees of freedom enabled by multiple injections to efficiently fulfill constraints on cylinder pressure, NOx emissions and exhaust temperature. A simulation study shows a potential 2 - 4 % indicated efficiency increase when two injections are used instead of one. These findings motivated the design of a hybrid multiple-injection controller that changes the number of injections depending on operating conditions. The controller designed was capable of reproducing the found efficiency increase experimentally with respect to constraints on pressure and NOx emissions.
A model-predictive pressure controller is also introduced. The controller predicts how the cylinder pressure varies with fuel injection by taking advantage of the estimated heat-release rate and a cylinder-pressure model. This feature was used to adjust fuel-injection timings, durations, and number of injections, for efficient constraint fulfillment in transient engine operation. Experimental results demonstrate that the pressure controller can also be used for tracking of cycle-resolved in-cylinder pressure trajectories, as well as finding the most efficient combustion timing.
Heat-release analysis is an essential component in the pressure-sensor feedback loop. Methods for calibrating heat-release model parameters with the use of engine data, and methods for detecting combustion timings are therefore discussed in the thesis.
The experimental results presented were conducted on a heavy-duty Scania D13 engine with a modified gas-exchange system. The fuel used was a mixture (by volume) of 80 % gasoline and 20 % n-heptane, to elevate the fuel octane number and allow for longer ignition delays.
Keywords
Model Predictive Control (MPC), Partially Premixed Combustion, Pressure Sensor Feedback, Model Based Control, Particle Filter, Multiple Fuel Injections, Gasoline Compression Ignition, Model Predictive Control (MPC), Partially Premixed Combustion (PPC), Low Temperature Combustion, Model Based Control, Multiple Fuel Injections, Pressure Sensor Feedback, Particle Filter, Gasoline Compression Ignition
PhD Thesis Department of Automatic Control, Lund University, Sweden, May 2018.
Download Model-Based Optimization of Combustion-Engine Control.pdf