Many modern engine systems are designed using one-dimensional computational fluid dynamics (1D CFD). This same technique can be used to model these systems in real time. This real-time model can be used to create virtual sensors in places where due to environmental or cost reasons physical sensors would not be practical. Achieving real-time performance of the CFD model requires more throughput than is available on single processor systems, so an Field Programmable Gate Array (FPGA) is employed. By employing an FPGA, we can synthesize and reconfigure our system to ensure determinism and lower resource usage. We instantiate several dedicated processing cores for parallel processing of sub-volumes. The number of cores can be configured to support up to 500 fluid volumes, more than enough for common 1D CFD models used in engines. This paper evaluates the feasibility of such a system and evaluates the complexity of such models against the GT-SUITE simulation software.

This content is only available via PDF.
You do not currently have access to this content.