Modeling and optimization of spray drying process parameters using artificial neural network and genetic algorithm for the production of probiotic (Lactiplantibacillus plantarum) finger millet milk powder

Abstract Consumers have been more inclined towards functional food due to various health issues. Non‐dairy‐based probiotic milk could be the potential alternative to fulfill consumer demand. Spray drying of functional health drinks could enhance the shelf life and reduce the transportation cost of developed food powder. The current study was focused on the modeling and optimization of spray‐drying process parameters using an artificial neural network (ANN) coupled with a genetic algorithm (GA). The data so obtained using ANN‐GA was also compared with the data obtained using response surface methodology (RSM) coupled with desirability function (DF). The results indicated that the ANN model was better at predicting the response parameters compared to the RSM model with a higher correlation coefficient ( R ) of .9997, .9994, .9964, and .9992 for training, testing, validation, and all datasets, respectively. The optimum conditions obtained using RSM‐DF were 160.41°C of inlet air temperature, 33.77% of maltodextrin content, and 138.79 mL/h of feed rate while that for ANN‐GA were 160.87°C, 20%, and 200 mL/h. The RSM‐DF method proved to be better for the optimization of response parameters. Therefore method selection for modeling and optimization of process and response parameters must be based on fulfilling the specific criteria. Practical Applications Non‐dairy milk production has gained popularity in the area of research and product development. Various process protocols have been reported to produce high‐quality non‐dairy milk to fulfill the demand for a vegan diet. However, nutrient bioavailability, longer shelf life, product stability, and consumer acceptability are hard to obtain from cereal‐based non‐dairy milk. The current study contributes to produce spray‐dried milk powder followed by fermentation at optimum drying conditions with good quality and stability. The process has been modeled and optimized in the study using advanced statistical tools such as ANN‐GA and RSM‐DF. They can effectively predict the quality parameters and determine the optimal conditions for new experimental data.