Prediction of fluid pattern in a shear flow on intelligent neural nodes using ANFIS and LBM

Abstract

Prediction of fluid pattern inside chemical mixing tanks and reactors is very challenging, mainly when scale-up and optimization of devices are essential due to enormous computational time and experimental efforts. This work recommends a new prediction tool to understand fluid behavior on the smart neural nodes. The adaptive neuro-fuzzy inference system (ANFIS) is used to learn the lattice Boltzmann method (LBM) data and predict fluid patterns based on its understanding. The anticipated results with the integration of LBM and ANFIS method indicated a good agreement with existing computational fluid dynamics results. The results show that almost similar fluid pattern occurs on neural nodes in the ANFIS method compared by LBM on lattice unit when shear flow applies on the top and bottom of fluid structure. This finding is very promising to avoid substantial computational time or experimental efforts in the optimization of different chemical devices. Prediction of the shear flow and optimization of boundary conditions to get proper droplet size distribution or bubble size distribution requires heavy computational time. Therefore in this work, ANFIS approach besides the LB method was used to replicate the flow between two parallel plates (vortex structure) in a short computational time. The current overview also shows the ability of the ANFIS method as a machine learning tool to learn how the fluid is disturbed by shear flow. The input data are used as big data during the learning process, and the intelligence of the algorithm is examined based on the total percentage of training data.