The present study was designed for the experimental and modeling investigations onthermodynamic analysis of drying onion slices using four microwave power levels(100, 350, 500, and 750 W) and the four samples thicknesses (2.5, 5, 7.5, and10 mm). A multilayer feedforward artificial neural network was employed in order topredict energy and exergy performance of the dryer and the prediction success werecompared using three evaluation criteria. The average values for energy efficiencyand specific energy loss ranged from 13.52% to 37.94% and from 1.35 to 7.43 MJ/kgwater, respectively. Findings showed that the exergy efficiency changed from11.79% to 30.84%. In addition, the statistical analysis revealed that higher microwavepower levels and the thinner samples significantly (p< .05) enhanced both the energyand exergy efficiencies. The obtained results for exergy improvement potential(accounted for 37.83%–69.41% from the total exergy inlet) indicated that the dryingprocess has good potential for exergy performance improvement. Based on themodeling outcomes, the energy efficiency was well predicted by an artificial neuralnetwork with a topology of 3–18–18–1 and LM training algorithm and thresholdfunction of Tan–Tan–Lin. However, the best topology for the exergy efficiency pre-diction had 3–20–16–1 structure, LM training algorithm and Log–Log–Lin transferfunction (R2of .94).