Tacoma is primarily a turbomachinery CFD solver. Tacoma is configured to read in a grid representing a single airfoil passage in a turbomachine annulus, and reproduce that domain for as many passages as is desired to calculate – up to the number of airfoils in the full annulus. The calculations intended for the Director’s Discretion grant begin with two airfoil passage grids – representing a high-pressure turbine nozzle vane and rotor blade. The nozzle blade grid will be approximately 30 million computational points, while the rotor blade grid will be approximately 12 million. For the turbine of the test case, there are 46 nozzle vanes per annulus & 92 rotor blades per annulus, resulting in a total of 2.5 billion computational nodes, if the entire annulus is simulated. For unsteady analyses, it is desired to have approximately 25,000 grid points per processing core, or 100,000 cores for the full annulus calculation. Tacoma is configured to convenien
tly allow for automatic partitioning of the computational domain into sub-domain sizes of this size. For periodicity, solutions require the same tangential extents of the mix of nozzles & rotors – that of a 1:2 ratio. Changing the size of the computational model therefore requires only mofifying a handfull of inputs to Tacoma, in order to change from a 1:2 count (54 million grid points) to a 2:4 (108 million), to a 3:6 (162 million), etc., all the way to the 46:92 (2.5 billion). As such, the case will be a convenient vehicle for assessing the scalability of Tacoma up to ~100,000 processing cores, which exceeds 20% of the current Jaguar machine.
|Source||Hours||Start Date||End Date|