Session: 37-07 Radial - Performance 2

Submission Number: 177728

Lead Time Reduction in Centrifugal Compressor Performance Mapping With Geometry-Informed Machine Learning 

Centrifugal compressors are finding a growing role in the energy transition thanks to high efficiency, robustness and flexibility. Nevertheless, to readily meet market challenges, a fast and accurate performance prediction remains pivotal during new stage development. Against this backdrop, although CFD analyses offer detailed insights, their computational time and effort hinder scalability across large design spaces. Building upon the two-step approach introduced in 2023 - based on a reduced-order model and an evolutionary algorithm - the research proposes a machine learning approach that efficiently finds implicit geometric representations aimed at further reducing time and cost of performance mapping. Indeed, by integrating geometric descriptors directly into machine learning pipeline, the model learns performance-relevant features with minimal reliance on additional CFD simulations, without compromising prediction accuracy. From a theoretical standpoint, the present research shows how geometry-informed machine learning can be a game changer to speed-up predictions. Practically, the study provides practitioners with an industrial-grade, easy-to-use tool for performance mapping. The approach is corroborated on a wide centrifugal compressor stage family for high-efficiency applications. Results reveal strong agreement with CFD outcomes. The present study lays the foundation for future advancements in data-driven turbomachinery design, with the potential to reduce product development cycles in energy-critical applications.

Presenting Author: Marco Bicchi Baker Hughes

Presenting Author Biography: Marco Bicchi is a lead design engineer in the centrifugal compressor new product development team at Baker Hughes. In 2023, he earned a Ph.D. in Industrial Engineering from the University of Florence, with a dissertation centered on the integration of artificial intelligence into the aerodynamic design of compressor stages. From 2023 to 2025, Marco assumed the role of R&D reference engineer for Inquiry-to-Order (ITO) support, contributing to customer bids for new units and upgrade solutions. Currently, he leads the technical coordination for compressed air energy storage applications. His main interests include the aerodynamic development of compressors and the industrial implementation of AI-driven design methodologies.

Authors:

Marco Bicchi Baker Hughes
Laure Barriere Baker Hughes
Andrea Panizza Baker Hughes
Leonardo Pulga Baker Hughes