Multi-objective and multi-model shape optimization of turbocharger turbines over real-world drive cycles for low carbon vehicles

Abstract

Turbocharging is the established method for downsizing internal combustion (IC) engines to lower CO2 emissions and fuel consumption while meeting the desired performance. Turbochargers for automotive engines commonly utilize radial turbines for exhaust energy extraction. However, the design of a turbocharger turbine is subject to conflicting requirements. A crucial consideration when matching a turbocharger to an engine is the ability to meet the specified low-end torque target while minimizing the turbine inlet pressure (particularly at high engine speed) to reduce the engine pumping work. Conventionally, the matching procedure used in the industry relies on experimentally measured compressor and turbine performance maps to model turbocharger operation within engine cycle simulation software. In this way, the compressor and turbine configuration that best meets the specified customer requirements is down-selected. Thus, only existing turbine geometries can be evaluated during the conventional matching process. This makes it a passive process as the turbine aerodynamic performance and inertia cannot be modified during the matching evaluations. Ideally, what is needed is a framework that physically models both the turbine and engine with sufficient accuracy and allows turbine geometric changes to be accounted for.

To this end, the objective of this work is to establish a novel and fast-running framework that allows turbine shape optimization based on engine-level objectives and constraints, and understand from a fluid dynamic perspective why a given turbine design is better for the engine. An in-house reduced-order model (meanline code) to estimate aerodynamic performance and a neural network-based inertia prediction tool for radial turbines are developed. These are integrated in a validated engine model to provide a framework for modelling the engine-turbine interaction using a numerically inexpensive technique. It allows the effect of turbine geometric changes on inertia and aerodynamic performance to be reflected in the exhaust boundary conditions and thereby in the overall performance of the engine. A genetic algorithm is employed within the framework, providing an opportunity for single-objective (for example, weighted cycle-average BSFC) or multi-objective (for example, weighted cycle-averaged BSFC and engine transient response) shape optimization of turbine meridional geometry. The framework has been applied to a Renault 1.2L turbocharged gasoline engine to minimize the fuel consumption and therefore CO2 emissions, while meeting a sensible transient response constraint. Turbine shape optimization was carried out over a cluster of weighted part-load operating points that represent the World harmonized Light vehicles Test Cycle (WLTC). The design candidates lying on the Pareto front present improvements of up to 0.4% in the weighted cycle-averaged fuel consumption, and up to 8% in transient response. Dynamic vehicle simulations over the WLTC are used to confirm the improvement observed in fuel consumption. Based on the meridional parameters obtained from the 1D optimization, 3D designs are created for both the turbine housing and the rotor. Finally, CFD evaluation and experimental testing are performed to verify the performance of optimized designs. 3D CFD predictions showed good agreement with experimental results, lying within the range of experimental uncertainty. The CFD analysis also showed a significant reduction in secondary flow features in the optimized design compared with the baseline turbine.

While the developed framework can be used to improve existing turbine designs, it also facilitates the development and optimization of `tailor-made' turbines for new low carbon engine projects. Even though, for the particular case described, the optimization process indicates a moderate 0.2--0.4% reduction in the weighted cycle-averaged BSFC, this would translate to a reduction of at least 270,000 tonnes of CO2 considering the lifetime of all GDI engines manufactured each year in the EU. Thus, the developed turbine optimization framework has a massive potential, especially because it requires no new or additional technology.