Prediction of HVO content in HVO/diesel blends using FTIR and chemometric methods

• Quantification of hydrotreated vegetable oil (HVO) content in HVO/diesel blends.
• Development of predictive models using FTIR and multivariate calibration.
• Preparation of 226 samples using four diesel fuels and two HVOs.
• Different approaches for selection of calibration and validation sets.
• Strong dependence on kind of preprocessing techniques for some calibration models.

Predictive models based on chemometric processing of FTIR spectra were created in order to develop a method for the determination of hydrotreated vegetable oil (HVO) in petroleum diesel/HVO blends. The blends were prepared in the entire concentration range of 0–100 wt.%. The set of samples was divided into calibration and validation sets. Four various approaches to the sample distribution were employed. There were predicted unknown content of known HVO in known diesel (approach 1), unknown content of known HVO in unknown diesel (approach 2), unknown content of unknown HVO in known diesel (approach 3) and finally unknown content of unknown HVO in unknown diesel (approach 4). The spectra obtained from FTIR measurement were processed using a software. Partial least squares regression (PLS) and principal component regression (PCR) were used to develop the predictive models. Optimal combination of preprocessing techniques was sought. In order to find suitable parameters of the preprocessing methods, hundreds of various models were tested. Although the optimal combination of preprocessing techniques was found for the most, some of them were strongly affected by preprocessing. The models with suitable preprocessing were able to predict the HVO content in given validation samples, however, the prediction of the HVO content in unknown “real” samples may be less accurate even incorrect because of the dependence on the kind of preprocessing.