Optimization and modeling of lactic acid production from pineapple waste

Abstract

Despite a great deal of research work on lactic acid fermentation in the past, the production of lactic acid from pineapple waste fermentation using immobilized cells has yet to be investigated. In this study lactic acid was produced from liquid pineapple waste fermentation by Lactobacillus delbrueckii entrapped in calcium alginate gel using batch fermentation systems. Lactic acid production by Lactobacillus delbrueckii was evaluated under immobilized cell fermentation conditions. The factors considered in the experimental design include pH, temperature, concentration of sodium alginate, cultivate size and bead diameter. The substrate concentration used throughout the experiment is 31.3 g/L. The glucose concentration and product formation were analyzed using high performance liquid chromatography (HPLC) and the cell numbers were determined by plate counting method. The experiment results revealed that the bead diameter the most important factor influencing production of lactic acid followed by Na-alginate concentration, pH and temperature. Maximum production, 30.27 g/L of lactic acid is obtained when using 2.0 %w/v sodium alginate concentration of bead diameter 1.0 mm at an initial pH of 6.5 at 37oC and 5 g of cultivate, thus reflecting the optimum conditions. Kinetics of the immobilized fermentation was analyzed based on batch growth model in terms of specific growth rate, yield constant or substrate utilization and rate of product formation. Results indicate an average µmax in the region of 0.09033 h-1 obtained at optimum conditions. For 2 liter fermentation, the Na-alginate immobilized cells produced 0.606g/L lactic acid/g/L glucose. The µnet calculated was 0.033 hour-1. Multilayer Perceptron (MLP) network was used in this study to predict the relationship between cell number and glucose concentration, between cell number and lactic acid concentration and between glucose concentration and lactic acid concentration at various temperatures using. It is found that the performance of MLP model is greatly influenced by the data sets used. The optimum structures of the MLP models are 1-8-1, 1-6-1 and 1-10-1 and the optimum transfer functions for hidden and output layer are Logsig and Tansig.