FlowSimulator

Kurzfassung

Over the last years, multi-disciplinary simulations got more and more important. Especially, in the field of aerospace science and engineering, the aim is to 'fly the virtual aircraft', i.e. to gather as much information about future aircraft as possible before their first flight, including aerodynamics and aero elastics as well as economic data (e.g. the range of new aircraft). The underlying simulations incorporate tools from different vendors and research facilities which have to be coupled. Right now, this is often done by simple shell scripts, user interaction and a lot of disk I/O and copying across the network which slows down the complete process. To overcome this, the FlowSimulator framework has been developed by Airbus, DLR; EADS, and others as a new backbone for scientific computations in the field for aerodynamics and associated topics. It consists of the FlowSimulator DataManager (FSDM) which is a library providing fully-parallelized data structures for numeric simulations. This library can be extended using a (pseudo-)plug-in mechanism and is largely interfaced to Python. On top of this, there is the FlowSimulator Control layer. This is a pure Python library helping the user to define work flows in an easy way. With this library, the whole scientific and industrial simulation process can then be controlled and monitored by a Python script. The talk starts with some motivational background about large multi-disciplinary aerospace simulations in science (DLR) and the aerospace industry (Airbus and EADS). It describes the basic concepts of the FlowSimulator framework with focus on the Data Manager and Control Layer. The talk with show, how Python is being used and describes the advantages and drawbacks for using Python in massive parallel applications. Also, Python example scripts are being presented, which demonstrates their ease of use for solving complex simulation tasks.