Integrating Taguchi method and artificial neural network (ANN) for improving non-edible biodiesel yield

Yeoh, Kean Weng (2024) Integrating Taguchi method and artificial neural network (ANN) for improving non-edible biodiesel yield. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Integrating Taguchi method and artificial neural network (ANN) for improving non-edible biodiesel yield.pdf - Submitted Version Download (1MB)

Abstract

Hevea brasiliensis also known as rubber seed oil (RSO) and Jatropha curcas oil (JCO) are non-edible feedstock used in biodiesel production. In this study, the integration of the Taguchi Method and Artificial Neural Network (ANN) are used to maximize and predict biodiesel yield of RSO and JCO feedstock using different waste shells as catalyst via two-step transesterification microwave irradiation. The Taguchi Method is utilized to design the optimum experiments with 5 factors (catalyst type, catalyst loading, methanol to oil molar ratio, reaction time and microwave power) at 3 levels of experiment parameters using orthogonal array (OA). The highest biodiesel yield is 96.2% achieved by combination of Perna Viridis (PV) catalyst, 12 wt.% catalyst loading, 1:9 methanol to oil molar ratio, 7 minutes of reaction time and 350 W microwave power for RSO while highest biodiesel yield is 95.26% achieved by combination of Corbicula Fluminea (CF) catalyst, 12 wt.% catalyst loading, 1:15 methanol to oil molar ratio, 9 minutes of reaction time and 400 W microwave power for JCO. Optimizations of experiment were done through SNR and ANOVA analysis to achieve the optimum combinations. ANN with single hidden layer using Levenberg-Marquardt back-propagation algorithm achieved R2 of 0.99953 for RSO and 0.99736 for JCO indicate excellent linear regression predictions for the biodiesel yield. The high value of linear regression shows that ANN with a quick propagation algorithm is an appropriate approach for biodiesel conversion prediction.

Actions (login required)

View Item