Use of Central Composite Design and Artificial Neural Network for Predicting the Yield of Biodiesel

Abstract

Abstract The correlation and prediction of optimum process parameters in biodiesel production is useful in obtaining high rate of conversion of vegetable oil to biodiesel as well as in process control. In this study, the correlation of the operating parameters such as reaction time, reaction temperature, stir speed, catalyst concentration and methanol-oil ratio at a pressure of 400 kPa for the production and prediction of the optimum biodiesel yield during biodiesel production was carried out using the Central Composite Design (CCD) and Artificial Neural Network (ANN). By adjusting networks and initializing weights, the neural network was iteratively trained with the aid of the Levenberg Marquardt algorithm in MATLAB R2018b environment using experimental results from the central composite design. From the analysis of results obtained, the correlation coefficients, adjusted and predicted R as well as R squared were close to1. In addition, predicted values from the central composite design and neural network show good correlation results when compared to the experimental data. The validation of the neural network was done with arbitrary values and random selection of process parameters. The observed output in terms of percentage yield of biodiesel from the trained network fell within the range of experimental results, thus, indicating that the network is an efficient tool for correlating and predicting process parameters for biodiesel production.