Application of Machine Learning Models for Hot Air Dryer Design in Food Drying Process

Food drying is an essential process for preserving food. Two critical parameters for determining the performance of a hot air dryer are air volume flow rate (V_o in cubic meter per hour) and heater power (Q in kilowatt). Calculating these parameters can be time-consuming due to the complexities involved in moist air thermodynamics. This study investigates the application of machine learning models for designing a hot air dryer system in the food industry. Input parameters include outdoor temperature (t₁), outdoor relative humidity (RH₁), heater outlet temperature (t₂), exhaust relative humidity (RH₃), and the amount of evaporated moisture (W, kilogram per hour). The predicted outputs are V_o and Q. Six machine learning models were employed using the auto mode of RapidMiner: generalized linear model, deep learning, decision tree, random forest, gradient boosted trees (GBT), and support vector machine (SVM). The results showed that GBT was the most suitable model for predicting V_o with an R² of 0.994 and a relative error (RE) of 3.2%. In predicting Q, all six models achieved R² values greater than 0.99 and RE values below 5.1%, with the SVM being the most accurate, achieving an R² of 1 and an RE of 0.4%. These findings highlight the potential of machine learning to effectively handle complex and skewed data, particularly in the food drying industry, where W was identified as the most influential factor in determining both V_o and Q.