Modelling of Sabahan Coffee Bean Roasting Process using Optimized FOPDT Function

Abstract

This paper presents the FOPDT model development of coffee roasting process in a hot air coffee roaster. In this study, Coffee beans from Ranau, Sabah, Malaysia were selected as the raw material. The empirical data were collected using open loop test to capture dynamic behaviour of the bean temperature under roasting condition. First order plus dead time (FOPDT) structure and its vector parameters such as gain, time constant and dead time were estimated using linear regression technique. The optimum model parameters were achieved by minimizing the sum of squared error (SSE) between the measured temperature data and the generated data predicted from the model at every data point using Nelder-Mead (N-M) optimization algorithm. Model validation results indicated that the developed model can successfully provide good output prediction in comparison to actual process data with R-squared error equal to 0.991 and Root Mean Squared Error (RMSE) equal to 6.364 °C.