Asynchronous Communication Pattern for Reducing Simulation Time in Engine Performance Models

Abstract

Abstract Demand is constantly increasing for high-fidelity models that execute quickly. To increase the fidelity of aircraft engine simulations, low-fidelity thermodynamic models are often interfaced with high-fidelity models used for capturing 2D and 3D effects. Unfortunately, interfaces and high-fidelity models are typically slow compared to rapid lower-fidelity thermodynamic models. This paper presents an approach for managing the interface between a low-fidelity model and high-fidelity model that, under certain circumstances, significantly reduces simulation time. The typical solver in a thermodynamic engine simulation sequentially iterates through each component in the system until a converged solution is reached. This type of communication pattern is also known as synchronous communication. Alternatively, an asynchronous communication pattern partitions the model into independent groups of components. Each partition contains its own solver, and each partition converges independently of the others. By partitioning a high-fidelity subsystem model, which was developed in an external tool, along with its associated interface, the solver variables associated with the subsystem model can be separated from the main solver. This implementation can reduce the number of calls to the high-fidelity subsystem and reduce overall simulation time. To demonstrate the asynchronous communication pattern, and to explore the approach’s impact on simulation time, a turbine-based combined cycle (TBCC) engine performance model was developed using the Numerical Propulsion System Simulation (NPSS) software. The NPSS model was interfaced with a high-fidelity afterburner model that was developed in an external tool. Using an asynchronous communication pattern, the engine performance model converged roughly 2.2x faster than using the conventional synchronous communication pattern for the same model. However, it was also shown that the asynchronous communication pattern is not beneficial in all scenarios, because it requires more total solver iterations. In this paper, the authors discuss the implementation and application of the asynchronous communication pattern, results, and conclusions.