Petascale simulations of cloud cavitation collapse

Abstract

The generation of a gaseous bubble, its violent collapse and rebound in a liquid that we believe is almost incompressible is another beauty of nature known as cavitation. The phenomenon appears in a wide range of applications---desired or undesired---where it is observed in arrangements involving not just one but many bubbles, broadly termed cloud cavitation. Experimental and theoretical insights to cloud cavitation are limited due to the spatial and temporal scales as well as the high non-linearity inherent to the physics of cavitation. A remedy is found in recent technological advances of HPC architectures that enable the numerical investigation of this phenomenon. Past as well as many recent modeling approaches employ complexity reductions to cope with the high cost of fully resolved simulations. The expense of such simplifications are reduced accuracy or even inaccurate predictions of certain dynamics in the multi-phase flow. The first part of this thesis, therefore, develops a computational framework that is designed for large scale stencil computations on recent HPC systems. The framework carefully implements optimizations that enable massively parallel simulations. The second part of the thesis is devoted to the development of a high throughput, compressible multi-phase flow solver for petascale simulations of cloud cavitation. We simulate the collapse of an unprecedented cloud with 12500 bubbles and perform an extensive analysis of the involved microjet formation and shock wave formation in bubbly liquids. Finally, the technology developed in this thesis is used to perform a parametric study of cloud cavitation collapse with particular emphasis on the kinetic energy allocation within the bubble cloud that is affected by local bubble deformations. The results of the study are further used to asses the limitations of reduced order models commonly used in practice.