Optimization of Lipase Catalysed Synthesis of Sugar Alcohol Esters Using Taguchi Method and Neural Network Analysis

Xylitol esters were successfully prepared from reactions of xylitol with free fatty acids (stearic acid, palmitic acid and capric acid), using enzymes as biocatalyst in an organic solvent system. Such a selective enzymatic route obviates the fastidious regime of substrate activation required in chemical synthesis. Preliminary detection and identification of reaction products were facilitated by thin layer chromatography (TLC). Subsequent quantitative studies on main effects of parameters governing the reactions based on conversion of fatty acid were conducted by titration analysis. Enzyme screening showed Novozym 435 to be the most proficient biocatalyst for the reactions. By changing one parameter at a time, the optimum condition was found to be at 60°C and 1:1 molar ratio with 4 g molecular sieve and 0.12 g enzyme in 30 ml solvent in 18 h reaction time. The synthetic reaction was optimized by Taguchi method based on orthogonal array to evaluate the effect of each parameters and interactive effects of reaction parameters including temperature, time, amount of enzyme, amount of molecular sieve, amount of solvent, and molar ratio of substrates (xylitol: fatty acid). Optimal condition derived from Taguchi method showed that, the method couldreduce reaction time, molecular sieve amount and enzyme amount with high conversion (96%, 92%, 88% for xylitol stearate, xylitol palmitate, and xylitol caprate, respectively) compared to conventional one variable at a time approach. Artificial Neural Network (ANN) method was also employed for the estimation of esterification yield in enzymatic synthesis of xylitol esters. Various feedforward neural networks were performed using different learning algorithms. The best algorithm was found to be Levenberg–Marquardt (LM) for a network composed of two hidden layers with six and seven neurons in the first and second layers, respectively for xylitol stearate and xylitol palmitate and also seven and five neurons in the first and second layers for xylitol caprate, with hyperbolic tangent sigmoid transfer function. Both models suggested good quality predictions for all independent variables (reaction time, temperature, amount of enzyme, substrate molar ratio, amount of molecular sieve and amount of solvent) in terms of the percentage conversion of xylitol esters. A high coefficient of determination (R2) (>0.9) and a low mean squared error (MSE) for training and testing data implied the good generalization of the developed models for predicting the reaction conversion. In order to make the synthesis process more environmentally friendly, the reactions between xylitol and capric acid, caprylic acid, and caproic acid, were also performed in solvent-free system. In this system,similar insolvent system, three methods including one variable at a time, Taguchi method and ANN were used for optimization and prediction of percentage of conversion. By varying one parameter at a time, the suitable temperature was 60°C and reaction time was 29 h (for xylitol caprate and xylitolcaprylate), 18 h (xylitol caproate) with 0.2 g enzyme, 0.14 g molecular sieve (for xylitol caprate and xylitol caprylate) and 0.2 g (for xylitol caproate), molar ratio 0.5 (xylitol: capric acid and caproic acid) and 0.33 (xylitol: caprylic acid). By using Taguchi method and ANN, the actual conversions of esters achieved were 74% (xylitol caprate), 69% (xylitol caprylate), and 60.5% (xylitol caproate) with a small amount of enzyme and molecular sieve, which matched well with the predicted values and indicated good conversion and effectiveness of the reactions without adding organic solvent for synthesis of xylitol esters. A comparison between one variable at a time, Taguchi method and artificial neural network shows that both Taguchi and ANN can reduce the amount of enzyme, amount of molecular sieve, reaction time and molar ratio in solvent based and solvent free system. So,this optimization method can help to save time and cost of the process. In addition comparison of statistical measures and performances between Taguchi method and ANN shows that ANN was slightly better than Taguchi for data fitting and estimation capabilities. Analysis of product was carried out by Fourier transform-infrared spectroscopy (FT-IR), Gas chromatography-Mass spectrometry (GC-MS), and Nuclear Magnetic Resonance (NMR). Some physicochemical properties were also analyzed. The overall results on the physicochemical studies showed that xylitol ester exhibited good physicochemical characteristics. The analyses revealed the suitability of the ester produced in industrial application.

Preview PDF FS_2011_13_F.pdf Download (589kB) | Preview

Actions (login required)

View Item