Designing propulsion system architectures to meet next generation requirements requires many tradeoffs be made. These trades are often between performance, risk, and cost. For example, the core of an engine is the most expensive and highest risk area of a propulsion system design. However, a new core design provides the greatest flexibility in meeting future performance requirements. The decision to upgrade or redesign the core must be justified by comparison with other lower risk options. Furthermore, for turboshaft applications, the choice of compressor, whether axial or centrifugal, is a major decision and trade with the choice being heavily driven by both current and projected weight and performance requirements. This problem is confounded by uncertainty in potential benefits of technologies or future performance of components. To address these issues this research proposes the use of a Bayesian belief network (BBN) to extend the more traditional robust engine design process. This is done by leveraging forward and backward inference to identify engine upgrade paths that are robust to uncertainty in requirements performance. Prior beliefs on the different scenarios and technology uncertainty can be used to quantify risk. Forward inference can be used to compare different scenarios.
The problem will be demonstrated using a two-spool turboshaft architecture modeled using the Numerical Propulsion System Simulation (NPSS) program. Upgrade options will include off the shelf, derivative engine (fixed core) with no technologies, derivative engine with new technologies, a new engine with no technologies, and a new engine with new technologies. The robust design process with a BBN will be used to identify which engine cycle and upgrade scenario is needed to meet performance requirements while minimizing cost and risk. To demonstrate how the choice of upgrade and cycle change with changes in requirements, studies are performed at different horsepower, ESFC, and power density requirements.