The proposed workflow starts from the initial and released CAD representation of a 3D geometry. The mesh generation service will be used to generate the mesh, which will then be input to the CFD solver running on the Cloud, possibly on a different infrastructure. The mesh generation and solver service on the Cloud run in a transparent way for the user. The process will in many cases need to be run iteratively to reach the optimal design. For this experiment, the meshing setup and CFD setup are carried out on the user’s local machine. The monitoring of the solver convergence will be through a web interface, either through a dedicated thin client or through a web browser. In this experiment, small and medium sized problems are tackled. CAD input files will typically range for a few megabytes to tens of megabytes. Meshes and CFD results are expected to range from tens to hundreds of megabytes.
Description of the infrastructure
- Input data and models by Stellba
- HPC Cluster and Cloud Infrastructure by Arctur
- Test and Usage scenario by University og Nottingham.
Detailed description of the demonstrator
CFD simulation on the cloud addresses the needs of manufacturing industries at different levels:
- For SMEs, CFD cloud simulation enables the access to computational resources that can otherwise not be made available in the company, due to infrastructure and maintenance issues. It also allows casual CFD users, who cannot afford to buy a complete CFD package, to have timely access to CFD capabilities as well as externalized computer resources.
- For larger companies, CFD cloud simulations enable large simulations to be performed on complex geometries, by accessing appropriate large resources on-demand, as well as absorbing CFD usage peak loads, when a new product design requires many CFD analyses and validations.
Innovation and novelty (business perspective): Cloud enabled CFD solutions open up many opportunities to extend the use of CFD in both small and large companies, creating new opportunities to develop better products in a shorter time frame. On the cloud CFD with PLM support speeds up both reuse of earlier work and tracking of performed analyses when proof of liability is needed. Competitiveness in a global market and with strict requirements on transparency, interoperability and quality will be strengthened.
Expected results (technical perspective): The main achievement of the experiment is to demonstrate the effectiveness of CFD computation on demand, requiring dynamic allocation of parallel computer resources in the cloud, in a way fully transparent to the user. The experiment will also contribute to show how PLM support improves result quality and process performance.
For the specific turbine applications of the end user, it would be of additional value to compute characteristic curves or even hill charts, which include series of CFD simulation runs, re-meshing or – when possible – a change of the node coordinates in the mesh at runtime while keeping the same mesh topology (which is usually the case). The position of the guide vanes depend on the operating point, and for Kaplan turbines, the runner blades can be rotated as well, further increasing the complexity of the problem. There is a certain relation between the position of guide vanes and runner blades depending on flow rate and current water levels, which ensures that the machine is operating in the respective best efficiency point. It would be of great benefit if this relation could be computed automatically by using the power and comfortable services of cloud-based techniques, seamlessly integrated in the engineer’s workflow.
Jotne EPM Technology, University of Nottingham, Arctur, Stellba
Modelling & Simulation