Experimental characterisation of turbocharger turbine exit flow deviation under steady, pulsating and transient flows, for improved 1D modelling of engine transient performance

Abstract

Modern IC engine-driven machines must adhere to ever stricter regulation of their pollutant footprint, which now includes emissions certification over transient duty cycles, as well as steady-state. At the same time, OEMs strive to make their products ever more attractive in terms of critical customer requirements, e.g., fuel economy and transient response, to offer machines that are more productive and use less fuel. However, the need to improve performance while meeting new emissions standards often pulls in opposite directions, massively increasing the required design, test and calibration effort. A key strategy for meeting this challenge is increased use of digital simulation, in place of costly and time consuming engine testing. Consequentially, there is a great need for modelling tools capable of accurately simulating both steady-state and transient engine operation, while reducing reliance on test data.

The engine's air and exhaust system plays a major role in its transient performance, and the corresponding reduced order model (ROM) is an important part of any engine cycle simulation used for performance and emissions prediction. This typically comprises sub-models for ducting, intercoolers, after-treatment, and the turbocharger compressor and turbine components. The main complexity in the turbine sub-model comes from the need to accommodate the physics of unsteady flow. This causes turbine instantaneous operation to occur away from the manufacturer's steady performance map, while introducing out-of-phase boundary conditions. It also introduces time-dependent flow features in the rotor, which affect turbine torque.

To date, turbine sub-models are almost exclusively dependent on steady-state supplier maps. This is due both to unavailability of geometrical data from turbocharger suppliers, and gaps in the reduced-order modelling of turbine flow physics. Even if the turbine sub-model accounts for blade geometrical parameters, via a meanline model, inability to correctly model rotor unsteady flow leaves it far from being fully predictive, and still very much reliant on empirical data.

The turbine torque equation in a meanline model requires the exit flow angle. This depends on the exit flow deviation, which is any departure of the flow from the blade angle at the trailing edge. It occurs due to imperfect flow guidance and the presence of secondary flow structures in the blade passage, giving rise to sub-optimal blade pressure differences, hence lower torque. Accurate turbine performance prediction must therefore account for exit flow deviation. Ideally it would be predicted by the model, rather than being an input, but this requires an understanding of how it is influenced by the time-varying rotor boundary conditions.

In this work, extensive measurements of the exit flow deviation, and its spatial and temporal variation during severely off-design operation typical of engine transients, have been performed on the low pressure radial turbine (Ø = 68 mm) of a commercial two-stage turbocharging system for a 7-litre industrial diesel engine. The outcomes are threefold. Firstly, a novel exit flow field measurement system, employing a combination of constant temperature and constant current hot wire anemometry, was designed and commissioned. This enabled a survey of exit flow deviation between 18-72 % of rotor exit span and covering 210° of azimuth, all in close proximity to the rotor trailing edge. Secondly, this survey revealed how exit flow deviation varies across a range of steady, and engine-like pulsating and transient-pulsating flows. Steady-state results showed that despite widespread use of a constant (and often zero) value in meanline codes, the exit flow deviation angle varies widely under off-design conditions. This observed variation resulted in a 7.9 % increase in the predicted time-to-torque of a validated 1D engine model. Over a pulse cycle the mass flow-averaged deviation angles vary between -5° and -8° depending on the pulse amplitude. Finally, a novel framework has been developed for analysis of exit flow field data, deriving a physically grounded exit flow angle for application in turbine meanline models.

In conclusion, it is recommended that all turbine models in transient engine simulations account for exit flow deviation. To not do so results in transient response predictions that vary by up to ± 16 %, severely compromising confidence in such simulations. Fortunately, this thesis provides an extensive database of engine-realistic experimental data and a supporting analytical approach on which to base improved turbine models, ultimately leading to significantly more accurate and reliable virtual product development tools.