Nonintrusive measurement of interfacial area and volumetric fraction in dispersed two‐phase flows using a neural network to process acoustic signals—A numerical investigation

A new methodology for measuring the volumetric fraction and interfacial area in two-phase flows is proposed in this study, based on neural networks processing the responses obtained from an acoustic interrogation signal. The geometrical distribution of the phases within the flow is mapped by the local acoustic propagation velocity, which is considered in the governing differential equation. This equation is solved numerically by the finite difference method with boundary conditions reproducing the excitation/measurement strategy. A significant number of propagation velocity distributions were considered in the solution of the differential equation to construct a database from which the neural model parameters could be adjusted. Specifically, the neural model is constructed to map the features extracted from the signals delivered by four acoustic sensors placed on the external boundary of the sensing domain, into the corresponding void fraction and interfacial area. These features correspond to the amplitudes and the times of arrival of the first three peaks of the acoustic wave. Numerical results showed that the neural model can be trained in a reasonable computational time and is capable of estimating the values of the volumetric fraction and the interfacial area of the examples of the test set. Copyright © 2008 John Wiley & Sons, Ltd.