A high-performance code for the use of spectral pod on the analysis of turbomachinery high fidelity simulations

This paper presents a High-Performance Data Analytics procedure for computing Spectral Proper Orthogonal Decomposition (SPOD) from high-fidelity simulations. The code is implemented in Python, leveraging the Dask library for ease of programming and efficient parallel computing. The method is applied to Large Eddy Simulation data of a low-pressure turbine blade, showing the ability to extract the fundamental dynamics and their link to loss generation mechanisms. The mathematical framework behind the SPOD implementation in Python and the decomposition of the total pressure transport equations are discussed. Scalability tests using an exemplary dataset demonstrate a very high efficiency of the implementation. Results from SPOD are compared with a classical decomposition framework such as proper orthogonal decomposition and the paper discusses the limitations and advantages of SPOD. The procedure is open-access and available in an online repository, promoting accessibility and collaboration within the scientific community.