Cavitation Phenomena in Diesel Injection Systems Gauss Centre for Supercomputing e.V.

COMPUTATIONAL AND SCIENTIFIC ENGINEERING

Cavitation Phenomena in Diesel Injection Systems

Principal Investigator:
Christian Egerer

Affiliation:
AER/TU München (Germany)

Local Project ID:
LRZ Project ID: pr86ta / HLRS Project ID: LESCAV

HPC Platform used:
SuperMUC (LRZ) / Hornet/Hermit (HLRS)

Date published:

Modern Diesel injection systems exceed injection pressures of 2000 bar in order to meet current and future emission regulations. By accelerating the flow through an injection nozzle or throttle valve pressure in the liquid can drop below vapor pressure, initiating local evaporation (hydrodynamic cavitation). The advection of vapor cavities into regions where the static pressure of the surrounding liquid exceeds vapor pressure leads to a sudden re-condensation or collapse of vapor cavities. The surrounding liquid is accelerated towards the center of the cavities and strong shock waves are emitted. The resulting pressure loads can lead to material erosion. For optimization of future fuel injectors the ability to predict cavitation and cavitation erosion during early stages of the design is desirable.

Motivation

Modern Diesel injection systems exceed injection pressures of 2000 bar in order to meet current and future emission regulations. By accelerating the flow through an injection nozzle or throttle valve pressure in the liquid can drop below vapor pressure, initiating local evaporation (hydrodynamic cavitation).

The advection of vapor cavities into regions where the static pressure of the surrounding liquid exceeds vapor pressure leads to a sudden re-condensation or collapse of vapor cavities. The surrounding liquid is accelerated towards the center of the cavities and strong shock waves are emitted. The resulting pressure loads can lead to material erosion. For optimization of future fuel injectors the ability to predict cavitation and cavitation erosion during early stages of the design is desirable.

High Performance Computing

Time-resolved computations are required since cavitation dynamics are highly unsteady. The numerical time step is limited by the ratio of grid size to the speed of sound of the fluid. Since the speed of sound in liquids is large, e.g. 1500 m/s for water, and cell sizes required to resolve dominant features of the flow field are small, the numerical time step is on the order of nanoseconds. On the other hand, convective time scales in injection systems are on the order of milliseconds. Thus, covering all relevant time scales and gathering statistical data requires millions of time steps.

Results

The investigated throttle resembles the control valve of a fuel injector. We consider two operating points, which can be discriminated by the pressure difference across the throttle. For small pressure differences we observe vortex cavitation in the detached shear layer at the inlet of the throttle. By decreasing the outlet pressure we observe a transition to sheet cavitation at the inlet of the throttle, compare bottom row of Fig. 1. Additionally, cavitation occurs in the corner vortices and the center of the throttle. Interaction between cavitation and turbulence can be studied by comparing vortical structures, see top row of Fig. 1. Cavitation damps the development of wall turbulence. Comparison with experimental observations showed that our simulation correctly predicts the transition of cavitation and turbulence structures and that the mean cavitation length agrees well [1]. Our results are a milestone in the simulation of turbulent cavitating flows and for the first time reproduce accurately experimental reference data.

Future Work

Future work will focus on transient effects due to the dynamic movement of the injector needle. The framework for enabling such simulations has been extended by developing an immersed boundary method that is able to handle moving geometries in compressible liquid flows [2].

Acknowledgements

• German Research Foundation (DFG) for financial support under contract no. AD 186/20-1
• Leibniz Supercomputing Centre (LRZ) for computing time on SuperMUC
• High Performance Computing Center Stuttgart (HLRS) for computing time on Hermit and Hornet

References

1. Egerer, C.P., et al., Large-eddy simulation of turbulent cavitating flow in a micro channel. Physics of Fluids, 2014. 26(8): p. 085102.
2. Örley, F., et al., Cut-element based immersed boundary method for moving geometries in compressible liquid flows with cavitation. Journal of Computational Physics, 2015. 283: p. 1-22.

Research Team & Contact Information

Christian Egerer, Felix Örley, Steffen Schmidt
Dr. Stefan Hickel and Prof. Nikolaus Adams
Institute of Aerodynamics and Fluid Mechanics
Technische Universität München
http://www.aer.mw.tum.de
email: christian.egerer@aer.mw.tum.dek

Tags: TUM CSE LRZ HLRS