Quality assessment of the results obtained by multivariate curve resolution analysis of multiple runs of gasoline blending processes

Abstract

This work presents a quality assessment study of the results obtained by multivariate curve resolution of multiple runs of gasoline blending processes. This evaluation is performed by noise propagation studies using resampling methods and by calculation of feasible bands due to rotational ambiguities. Results showed that combination of matrix augmentation and use of constraints strategies during alternating least squares optimization minimize both rotational ambiguities and error propagation effects in the resolved solutions. The proposed approach is tested for simulated and experimental data sets obtained in the study of blending processes of gasolines monitored by NIR spectroscopy.