Polynomial neural network-based group method of data handling algorithm coupled with modified particle swarm optimization to predict permeate flux (%) of rectangular sheet-shaped membraneAnirban Banik, Mrinmoy Majumder, Sushant Kumar Biswal, and Tarun Kanti Bandyopadhyay Department of Civil Engineering, National Institute of Technology Agartala, Jirania, India
E-mail: tarunkantibanerjee0@gmail.com Received: 4 January 2021 Accepted: 22 August 2021 Abstract: The paper discussed a novel approach of polynomial neural network-based group method of data handling coupled with modified particle swarm optimization (PSO) algorithm to predict the permeate flux of rectangular sheet-shaped membrane. In this regard, permeate flux is considered to be a model output, whereas operating pressure, pore size, and feed velocity were considered as model input. It was reported that the polynomial neural network (PNN)-based group method of data handling (GMDH) model has a performance index (PI) value of 0.723 which is higher than in other methods like artificial neural network (ANN), multi-linear regression analysis (MLR), and adaptive neuro-fuzzy inference system (ANFIS). In this study, three typical mathematical test functions were used to evaluate performance of modified PSO. The findings showed that the modified PSO algorithm exhibits better global search ability, computational speed, and stability. Process optimization resulted in optimum operating conditions for operating pressure, pore size, and feed velocity (were 543.69 Pa, 1.40 µm, and 0.179 m/s), respectively. Sensitivity analysis established pore size to be the most significant parameter. Keywords: Artificial neural network; Adaptive neuro-fuzzy inference system; Group method of data handling; Modified particle swarm optimization; Membrane; Multi-linear regression analysis Full paper is available at www.springerlink.com. DOI: 10.1007/s11696-021-01838-x
Chemical Papers 76 (1) 79–97 (2022) |
Sunday, November 24, 2024 
|
|||
© 2024 Chemical Papers |
|
|||
