Data-Driven Prediction of Unsteady Vortex Phenomena in a Conical Diffuser

The application of machine learning to solve engineering problems is in extremely high demand. This article proposes a tool that employs algorithms for predicting the frequency response an unsteady vortex phenomenon, precessing core (PVC), occurring conical diffuser behind radial swirler. model input parameters are two components time-averaged velocity profile at cone inlet. An empirical database was obtained using fully automated experiment. associates multiple inlet profiles with pressure pulsations measured diffuser, which caused by PVC swirling flow. In total, over 103 different flow regimes were varying swirl number and angle diffuser. Pressure induced detected fluctuations sensors residing on opposite sides A classifier constructed Linear Support Vector Classification (Linear SVC) experimental data. based average allows one predict emergence accuracy (99%). By training regression artificial neural network, predicted error no more than 1.01 5.4% power pulsations, respectively.