Artificial Neural Network (ANN) Model for Prediction of Mixing Behavior of Granular Flows

• Published in: International journal of computational methods in engineering science and mechanics 2007-04, Vol.8 (3), p.149-158
• Main Authors: Mujumdar, Ajit, Robi, P. S., Malik, Moinuddin, Horio, Masayuki
• Format: Article
• Language: English
• Subjects: Discrete Element Method (DEM); Granular Flows Mixing; Neural Network; Sectorial Container
• ISSN: 1550-2287; 1550-2295
• DOI: 10.1080/15502280701252495

Mixing and segregation behavior of granular flows inside a particulate system comprising an oscillating sectorial container is predicted by an artificial neural network (ANN) model. By employing discrete element method (DEM), numerically simulated characteristics of a sectorial container, which is subjected to harmonic angular oscillations, are trained for the development of a neural network model. Binary system of particles is simulated and degree of mixing is estimated by varying different parameters, such as particle size ratio (1:1 to 1:3), frequency of oscillations (1 to 4 Hz), amplitude of oscillations (30 to 60°), volume filling fraction (0.04 to 0.24), and number of cycles (1 to 20). The learning of ANN is accomplished by feed forward back propagation algorithm. It is found that mean mixing concentration predicted by the neural network model developed in this work is in a good agreement with the simulated values. Percentage error predicted by ANN model is less than ± 8% for 82 out of the 90 data values. Development of the neural network model and its use for the prediction of the outcome of the system (especially in cases where several operating parameters, which determine the outcome of the system, have a non-linear relationship with each other) is believed to be an accurate and computationally inexpensive way of understanding the behavior of the system.