Sayan Ghosh, Govinda Anantha Padmanabha, Cheng Peng, Valeria Andreoli, Steven Atkinson, Piyush Pandita, Thomas Vandeputte, Nicholas Zabaras, and Liping Wang

J. Mech. Des. Feb 2022, 144(2)

While designing a blade for an industrial gas turbine  (IGT), one needs to consider multiple performance characteristics like aerodynamic efficiency, durability, and safety. The sequential and iterative nature of the process leads to extremely long cycle times and precludes a deeper understanding of the design space. Thus, leading to potentially unrealized efficiency and performance. To overcome these challenges, we demonstrate a framework that explicitly models design as a function of desirable performance characteristics. This framework leverages probabilistic machine learning for explicit inverse design (PMI), resulting in optimal use of resources while inferring optimal designs. In this work, PMI’s capabilities are highlighted on an inverse aerodynamic design of three-dimensional turbine blades.