Artifcial neural network approach for the prediction of terminal falling velocity of non-spherical particles through Newtonian and non-Newtonian fluids

Abstract

The investigation of the terminal falling velocity of non-spherical particles is currently one of the most promising topics in sedimentation technology due to its great signifcance in many separation processes. In this study, the potential of Artifcial Neural Networks (ANNs) for the prediction of nonspherical particles terminal falling velocity through Newtonian and nonNewtonian (power law) liquids was investigated using 361 experimental data. ANNs emerged as the most popular non-linear mathematical models due to their good prediction, simplicity, flexibility and the large capacity which moderate engineering endeavor, and the availability of a large number of training algorithms. The developed ANN model demonstrated the acceptable values for the prediction of terminal falling velocities such as the determination coefcient ( R2), MSE, and MRE which were equal to 0.9729, 0.0023, and 21%, respectively. In an investigation on terminal falling velocity and drag coefcient of spherical and non-spherical particles, it was found that the terminal falling velocity of non spherical particles to spherical particles was 0.1.