Data-driven optimization of spray dryer design

Spray dryers are widely used in the food, pharmaceutical, and chemical industries to reduce product weight and extend shelf life; however, their high energy demand and wall deposition issues make accurate performance prediction essential for improving efficiency. Previous studies mainly examined small laboratory spray dryers with few input parameters and limited ranges, often developing surrogate models without optimization. In contrast, this study considers a larger pilot-scale system with a wider range of operational and geometric parameters, develops a surrogate model, and performs comprehensive multi-objective optimization. This study uses a validated numerical framework to develop Random Forest and Feedforward Neural Network surrogate models. The neural network showed superior accuracy and was applied for multi-objective optimization using Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and a hybrid PSO-GA approach, with the hybrid method performing best. Two optimization scenarios were investigated. In the operational optimization, improvements were achieved solely by adjusting process variables. When both operational and geometric variables were optimized simultaneously, the optimal design reduced energy consumption by 33.47%, increased outlet droplet mass flow rate by 83.23%, and decreased wall deposition by 55.65% relative to the baseline configuration.