A neural network approach for non-iterative calculation of heat transfer coefficient in fluid–particle systems

A non-iterative procedure was developed using an artificial neural network (ANN) for calculating the fluid-to-particle heat transfer coefficient, hfp, in fluid–particle systems. The problem considered has relevance in agitation processing of cans containing liquid/particle mixtures where fluid temperature is time-dependent. In developing the ANN model, two configurations were evaluated: (i) the input parameters (fluid and particle temperatures) and output parameters (Biot number, Bi) were taken initially on a linear scale, and (ii) input/output parameters were transformed using logarithmic and arctangent scales. The second configuration yielded an optimal ANN model with eight neurons in each of the three hidden layers. This configuration was capable of predicting the value of Bi in the range of 0.1 to 10 with an error of less than 2%. The ANN model used information about experimental transient temperatures of the fluid and particle center and predicted Bi. The Bi/heat transfer coefficients evaluated using the developed ANN model were in close agreement with those evaluated using the numerical method under a wide range of experimental conditions.