Conquer the terabyte scale: post-processing of high resolution unsteady CFD data for turbomachinery analysis

Abstract

The post-processing tool POST has been developed at the DLR Institute of Propulsion Technology specifically for turbomachinery applications. In this paper, we present strategies employed to meet the challenges posed by modern unsteady CFD simulations. The focus is laid on parallelization with dynamic load balancing capabilities and the premise of data reduction at the earliest stage to ensure high performance but low resource consumption. Freely definable analysis surfaces and lines are created with methods provided by the Visualization Toolkit (VTK) by Kitware. User-specific analysis chains are handled by a flexible scheduling system minimizing hard disk access and memory usage. In order to achieve compatibility with a wide range of tools, both CGNS and Tecplot file formats are supported. To demonstrate the capabilities of POST, the tool is applied to analyze the data obtained from an unsteady whole annulus simulation of the first two transonic stages of the 4.5-stage high pressure research compressor rig 250. In the CFD simulation approximately 120 million cells distributed over 1679 blocks were used, generating a total of 6 terabytes of data to be processed within industrial design time frames.