An artificial neural network approach to investigate cavitating flow regime at different temperatures

Identification of cavitating regime is an important issue in a wide range of fluid dynamic systems. The cavitation behavior is affected by several parameters, as the operating pressure and the fluid temperature. In the present study the cavitating behavior of water inside an orifice was analyzed by images analysis and by pressure signals. Four cavitation regimes were characterized: no-cavitation, developing cavitation, super cavitation and jet cavitation. A three-layer Elman neural network was designed to predict the cavitation regime, from the frequency content of the pressure fluctuations, recorded upstream and downstream the internal orifice. Cavitation regimes were successfully predicted. The designed neural networks were useful also to underline the influence of each operating parameter on the phenomena under investigation; in particular it was possible to identify the frequency ranges that characterize the different cavitation regimes and the influence of the fluid temperature.