Application of High Performance Computing for Simulating Cycle to Cycle Variation in Dual Fuel Combustion Engines

Project Description

In collaboration with Convergent Science Inc. (CSI), GE developed a dual-fuel combustion CFD model to develop and evaluate concepts for improved performance and reduced emissions. In addition to improving performance and emissions, the capability to predict dilute combustion stability limits (knock and misfire) is important to design for operational stability and reliability. Key to developing this capability is being able to run several (~100) sequential engine cycles. With this capability, the serial nature of cyclic dispersion can be captured and thus the longer-timescale combustion oscillations and dynamics can be simulated. CFD provided 3D space distribution of combustion related quantities can greatly aid in understanding combustion stability limits. This research is important as these computations are important in accelerating the adoption of natural gas for locomotive propulsion. This is an urgent need driven by the huge natural gas boom in Un ited States, reduction of carbon emissions, reduction in transportation fuel cost. This also aids the national security by reducing the dependence on foreign oil. But, the computational overhead and clock time for simulating 100+ engine cycles will be overwhelming. Running 100s of sequential cycles on a supercomputer would take 3-6 months to simulate. ORNL reported a novel approach for utilizing supercomputers to capture long-timescale features of dilute combustion, without the need to simulate several engine cycles in series. This methodology includes carefully guided, concurrent, single cycle simulations to capture the cycle-to-cycle variation that is typical of a dual fuel combustion scenario. GE wishes to learn and evaluate this methodology as part of the current collaborative project. That is one of the key objectives of the current project. The other key objective is to understand the scalability of this algorithm so we collect enough data to apply for large allocations through INCITE and ALCC grants.