Estimation of Biosurfactant Production Parameters and Yields Without Conducting Additional Experiments on a Larger Production Scale

Abstract

In this study, a Plackett-Burman design was applied to investigate critical factors for surface tension. After adding a new factor called production scale, a central composite design (CCD) was constructed to examine nonlinear relations among factors and surface tension. An artificial neural network (ANN) was trained using data from CCD experiments. The ANN with the best structure of 5-6-1 was then tested with different unseen data sets. The predictions from ANN were within the 95% confidence interval (CI), even for a larger production scale, deter-mined by using the replicates. A genetic algorithm (GA) was developed to check how the values of the factors vary if the production scale was set to a user-defined value. Based on the validation experiments, it was observed that the 95% confidence interval of surface tension was 36.83 +/- 1.00 mN m-1 while pH 8, temperature 35 degrees C, incubation time 12 h, and amount of inoculum 2.30%, respectively, for the production scale of 600 mL. The proposed methodological approach with the integration of ANN and GA is considered to make a serious eco-nomic contribution as it allows predicting a proper setup for larger production scales without conducting additional experiments.