Abstract
The subject of the present paper is the development of a general procedure to calculate the terms of the total pressure transport equations using a High-Performance Data Analytics (HPDA), based on Proper Orthogonal Decomposition (POD) and leveraging on High Performance Computing. This method is applied to data obtained from high fidelity simulations of low pressure turbine (LPT) blades in order to separate the different loss contributions and to visualize the associated structures to the fluid dynamics phenomena that occur inside the passage. This procedure is developed in Python environment because it easily allows parallel computing.
This paper discusses the mathematical framework behind the decomposition of total pressure transport equation and its implementation in Python. The scalability tests are performed for two exemplary datasets and compared to previous implementation showing a considerable speed-up. The procedure applied to Large Eddy Simulation (LES) data shows significant improvement over traditional approaches. It provides more detailed information about the phenomena associated with the generation of losses in the turbine blades, allowing for quick identification of where these losses occur. The HPDA code can be applied to all high fidelity simulations (LES and DNS) in order to get more information from simulations that are extremely expensive, allowing the full exploitation of such large datasets. In addition, to demonstrate the ease of code’s implementation, the data obtained from the POD are compared with data obtained from Fourier decomposition to validate the procedure. The procedure is open access and available in an online repository.