The use of artificial neural network to predict exergetic performance of spray drying process: A preliminary study

A feedforward artificial neural network (ANN) was applied to predict the exergetic performance of a microencapsulation process via spray drying. The exergetic data was obtained from drying experiments conducted at different inlet drying air temperatures, aspirator rates (drying air flow rates), peristaltic pump rates (mass flow rates), and spraying air flow rates as inputs for ANN. A multilayer perceptron (MLP) ANN was utilized to correlate the output parameters (inlet exergy, outlet exergy, lost exergy, destructed exergy, entropy generation, exergy efficiency, and improvement potential rate) to the four inputs parameters. Various error minimization algorithms, transfer functions, number of hidden neurons, and training epochs were investigated to find the optimum ANN model. The MLP ANN with Levenberg-Marquardt error minimization algorithm, logarithmic sigmoid transfer function, 20 hidden neurons, and 100 training iterations was selected as the best topology to map the exergetic performance of microencapsulation process according to statistical parameters and model simplicity. The model predicted exergetic parameters of spray drying process with R2 values greater than 0.98 indicating the fidelity of the selected network. Accordingly, the selected ANN model can be applied to determine the exergy efficient drying conditions to achieve a sustainable spray drying process.

► Exergetic prediction of spray drying process by multilayer perceptron neural network.
► Profound effects of network parameters on estimation power of applied topologies.
► Acceptable prediction power of selected topology compared with mathematically constitutive models.