Multiobjective simultaneous optimization of biosurfactant process medium by integrating differential evolution with artificial neural networks

Abstract

A method of differential evolution (DE) integrated with artificial neural networks (ANN) is derived for modelling and optimization of a biosurfactant process producing rhamnolipid by Pseudomonas aeruginosa. A central composite rotatable design (CCRD) data is used to develop multiple regression and ANN response surface models in order to integrate them with DE for optimizing the medium compositions. The DE with global search operators explores the search space of the response surface models and finds the optimum medium compositions that maximize the rhamnolipid productivity. A multiobjective simultaneous optimization strategy that integrates ANN model with DE search is found to compromise for biomass concentration and maximize the rhamnolipid activity as 55.9 mg/L (R² = 0.914) with an optimized medium compositions of glucose=24.079; NH₄NO₃=3.28; KH₂PO₄=0.24; yeast extract=7.95 and MgSO₄.7H₂O=2.69. The experimental rhamnolipid activity of 56 mg/L obtained using the optimized medium compositions are close to the predicted rhamnolipid activity. These findings demonstrate that the ANN-DE integrated multi objective optimization strategy is quite effective for simultaneous optimization of biochemical and biotechnological processes.