Design of Experiment on a fast coupled cycle code with meanline design and map prediction capability enables rapid exploration of design space for single stage or multistage (back-back or inline) heat pumps with different refrigerants (using RGP files)  to optimize the key design choices such as rpm, tip diameter, etc. This places the design in the optimum region in the trade-off between heating capacity and operating range. 

The full 3D blade shape, which produces optimum efficiency, is created using Physics-Enhanced Machine Learning, based on 3D Inverse Design. This method uses a small (less than 50 points) high-fidelity training dataset to accurately predict the overall performance of candidate blade shapes, then produces paradigm-shifting Pareto optimal designs which solve the conflicting multi-point performance requirements.

This webinar will demonstrate the workflow to set up and run this study - showing how easy it is to create a Physics-Enhanced Machine Learning system for even the most complex real-gas applications.