Artificial neural networks: a new tool for prediction of pressure drop of non-Newtonian fluid foods through tubes Original Research Article

Pressure gradients and the corresponding mass flow rates of five different non-Newtonian fluid foods: 1% solutions of sodium alginate and CMC, 1.5% CMC solution, two different tomato ketchups, oyster sauce, in four different diameter stainless steel tubes ranging from 7.51 to 16.34 mm i.d. were recorded using a continuous recording type tube flow viscometer capable of operating in both transient and continuous flow modes. The fluids were pseudoplastic in nature and followed the power law model. The flow was confined to the laminar flow regime and appreciable slippage occurred in all cases. Commercially available artificial neural networks based on back-propagation and generalized regression algorithm were applied to predict the pressure gradients in tube flow providing mass flow rate, consistency coefficients and flow behavior indices obtained from a low shear rate rotational viscometer, mass density and tube diameters as inputs. The net predicted values closely followed the experimental ones with an average absolute error below 5.44%.