Computational Intelligence-Based Optimisation of Drying Schedules for High-Quality Plantain Flour Production in Southern Nigeria

Plantain flour production is a high-value agro-processing activity in southern Nigeria, with quality critically dependent on the drying schedule employed. However, the multi-objective nature of drying optimisation, which simultaneously requires minimisation of energy consumption, drying time, and quality loss, makes it intractable for conventional single-objective optimisation approaches. This study applies computational intelligence methods, specifically multi-objective genetic algorithms and artificial neural network surrogate modelling, to develop optimised drying schedules for high-quality plantain flour production in a pilot-scale forced-air dryer operated at the Federal Institute of Industrial Research Oshodi, Lagos State. The theoretical contribution is a hybrid surrogate-optimisation architecture in which artificial neural network models trained on experimental drying data serve as the objective function evaluators within the genetic algorithm optimisation loop, dramatically reducing the computational burden compared to physics-based model evaluation. Experimental data covering drying temperature (40 to 80 degrees Celsius), air velocity (0.5 to 2.5 m/s), slice thickness (3 to 7 mm), and initial moisture content (65 to 80 percent wb) were collected from a full factorial experiment with 162 runs. The artificial neural network models predicted moisture content trajectory, colour, total phenolic content, vitamin B6 retention, and energy consumption, with prediction R-squared values above 0.97. The Pareto-optimal drying schedules identified by the multi-objective genetic algorithm achieved 18 percent energy savings compared to conventional fixed-temperature protocols while maintaining premium flour quality specifications. The study contributes an original computational intelligence framework for multi-objective food process optimisation. Keywords: computational intelligence, drying optimisation, plantain flour, multi-objective genetic algorithm, artificial neural network