Prediction of fatty acid composition of vegetable oils based on rheological measurements using nonlinear models

The aim of this study was to model the fatty acid composition of vegetable oils by using the adaptive neuro fuzzy inference system (ANFIS) and artificial neural network (ANN) based on rheological measurements. For this aim, seven different refined vegetable oils (hazelnut, soybean, sunflower, olive, canola, corn, and cotton seed) were used. Olive oil had the highest viscosity value of 0.067?Pa?s. Principal component analysis (PCA) was performed to determine the correlations among the fatty acids of the oils to reduce parameters which were modeled. PC1 composed of C18:0, C18:1, C18:2, and C20:0 as PC2 composed of C16:0, C18:3, and C22:0. After determination of the fatty acid composition and the rheological properties of the oils, ANN and ANFIS models were established. The inputs were selected as oil type, shear rate and shear stress and the outputs were C16:0 and C18:2. The root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R2) were used for testing the accuracy of the models. This comparison showed that ANFIS and ANN achieved a satisfactory prediction for fatty acid composition of the vegetable oils studied and ANFIS is better than ANN. Practical applications: Determination of the fatty acid composition of vegetable oils is important for the assessment of the purity of the oils as well as the nutritional quality. The fatty acid composition of the oils affects their rheological characteristics. Rheological measurements are easier than chromatographic analysis. Therefore, we aimed to establish a model for prediction of the fatty acid composition of the oils based on their rheological behaviors.