Pierre Sullivan

University of Toronto 
Project Title: Using high performance computing for improved energy production and pollutant mitigation
Industry Partners: Custom Flow Sim, others
Platform:  Blue Gene/Q

Advanced Manufacturing

Large-Eddy Simulation, or LES, has become the method of choice for computationally-intensive simulations resolved to the necessary scales. Traditional Reynolds-averaged methods (RANS), although useful, require significant assumptions that compromise the fidelity of the flow physics obtained. Direct Numerical Simulation (DNS) which resolves all the scales remains a prohibitive method due to its computational requirements. LES bridges between RANS and DNS, where the energetic large scales are resolved and computed directly whereas the smaller more universal scales are modelled. By modeling the subgrid scales within the inertial subrange, it is possible to extract high-fidelity flow information that can be used to improve local conditions. The separation between the large scales of motion from the small ones is possible with a filtering operation at a width of , a characteristic length scale within the turbulence inertial subrange; below this size, the contribution of small scale eddies are captured with a subgrid scale model (SGS). However, LES is computationally intensive and micro-climate modeling is beyond the capability of most desktop computers. As well, until recently LES models were either lab-developed codes or commercial codes. Lab-developed codes are difficult to transfer to industry partners as they are not necessarily client-friendly. While there exists excellent commercial codes, using these on multi-processor machines is prohibitively expensive. To model this flow, OpenFoam, an open-source turbulence code, will be used.