Abstract
A new approach has been introduced for separation of oil-in-water emulsion by using ultrasound standing wavefield. A neural networks model was used to simulate changes in the size of droplet during treatment. Model outputs were then validated and generalization capability was evaluated. For each network, the optimum values of isotropic spread were obtained by minimizing the root mean square error and maximizing the corresponding coefficient. It was found that the predicted values were in good agreements with experimental results. Also, increasing voice speed was demonstrated to predict size of emulsion particles more efficiently and accurately.