Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Sibendu Som Project Name: CAT-engine-modeling Division: ES Project title: Development of sub-grid turbulent combustion models for LES under engine conditions Associated funding: DOE Office of vehicle technologies under the small business voucher program Other Systems: Science: This project will use the computational fluid dynamics (CFD) software CONVERGE to simulate cycle-to-cycle variation (CCV) in spark-ignited gasoline engines. The primary objective of the proposed research is to understand the impact of model settings, and simulation strategies on capturing CCV in spark-ignited (SI) gasoline engines through large-eddy simulation (LES)-based CFD. The ultimate objective is to develop best practices for industry-users of CONVERGE to run CCV simulations that accurately characterize the variation in the SI engines they intend to optimize, while also being computationally efficient. Project description: In the previous year, we worked with Convergent Science, Inc. (CSI), the makers of CONVERGE to identify the engine dataset to simulate and the version of CONVERGE to benchmark for this project. We are using a dataset from spark-ignition engine experiments on a Ford single-cylinder, direct-injected engine at Argonne. We performed multiple cycles of LES for this engine running on the LCRC cluster, Blues to simulate CCV, first using a base mesh size of 4 mm. We optimized various settings for the turbulence model, adaptive meshing and embedding, and solver settings, based on computational bottlenecks identified through preliminary simulation. We are currently running simulations with finer mesh resolutions (i.e. higher fidelity) of 2 mm and 1 mm base mesh sizes as well to understand the impact of mesh resolution on capturing CCV. In FY18, we intend to explore the impact of introducing numerical perturbations in the simulations through multiple restarts, to evaluate whether this artificial numerical perturbation adversely affects the overall CCV prediction from multiple consecutive cycles. This is of importance for industry users of CONVERGE who might have restrictions on maximum walltime a simulation can run, before it needs to be restarted. Further, we also intend to explore the potential of capturing CCV through multiple concurrent simulations, each with an initial perturbation introduced through a synthetic turbulent field. We would like to evaluate whether this concurrent simulation approach yields similar CCV values as running multiple cycles sequentially, as the concurrent approach has the potential of reducing overall turnaround time for CCV simulations by orders of magnitude on leadership-scale systems like Theta and Aurora. Industry partnership: Convergent Science, Inc. Project URL: http://verifi.anl.gov/ Current FY Hours Used: undetermined amount New FY Requested allocation: 800000 Q1: 200000 Q2: 200000 Q3: 200000 Q4: 200000 Justification: Our recent publication (J. Kodavasal, K. Harms, P. Srivastava, S. Som, S. Quan, K.J. Richards, M. Garcia, “Development of stiffness-based chemistry load balancing scheme, and optimization of I/O and communication, to enable massively parallel high-fidelity internal combustion engine simulations,” Journal of Energy Resource Technology; JERT-16-1022, 2016) together with MCS and Convergent Science discusses the improvements to the Converge tool that has resulted in significant improvement in scaling. Currently, we are able to scale the code up to 4096 processors on Mira with about 70% scaling efficiency for a fixed mesh size. This was achieved due to the implementation of: (1) MPI I/O, (2) Improved Communication, (3) METIS load balancing scheme, (4) Development of a new chemistry load balancing scheme. The above changes are not only expected to benefit calculations on Mira but also help on computing clusters like Bebop and Blues. Storage requirements: 5 TB Thank You, The LCRC Accounts System