| Summary: | Palm oil refining is a process of removing the unwanted compounds from crude palm oil to produce a refined, bleached and deodorized (RBD) palm oil. It consists of three main processes which are degumming, bleaching and deodorization. In this study, an adaptive neuro-fuzzy inference system was developed to predict the FFA content in the RBD palm oil. The FFA content is non-linearly correlated to the input quality and the operating condition at the refining processes. Feed flow rate, FFA content (input), moisture, IV, phosphoric acid, bleaching earth, bleaching temperature, bleaching vacuum pressure, deodorizer temperature, deodorizer vacuum pressure, heating temperature, sparging steam pressure, and pre-stripper sparging steam were used to build the ANFIS model. Subtractive clustering method was selected for the ANFIS model. The performance of the model was evaluated using RMSE and R2 value. This study demonstrates that ANFIS are useful for estimating the FFA quality at the palm oil refining process.
|
|---|