Refined oil mixtures can be quantified using terahertz-absorption-coefficient spectra and dualistic linear regression fitting. However, when this method was used to quantify mixtures of 90# and 97# gasolines, the absolute error between the real and fitted value was large (25%), and this was due to the component similarity between 90# and 97# gasolines. To solve this problem, the present research addresses the possibility of developing a method that would allow direct, simple, and accurate determination of the 97# gasoline content in gasoline mixtures using a terahertz time-domain pulse coupled to a multipara-meter-combined analysis. The multiparameter represents the time delay and amplitude of the first transmission dip and peak in the time-domain pulse. The relationship between these four parameters and the 97# gasoline content in gasoline mixtures was thoroughly investigated, and four distinct calibration models for quantifying gasoline mixtures were built using least square fitting. To enable the development of an informative and accurate calibration model, the four individual models were given proper weights and combined. The weight was determined by the cosine-optimal method, which aimed to determine the most proper weight under the condition of the cosine of the angle between the fitted content vector and the real content vector that reaches the maximum. This method allows the determination of 97# gasoline content in gasoline mixtures with a low absolute error (6%), resulting in predictions that are more accurate and precise than those obtained by the terahertz-absorption-coefficient spectra and dualistic linear regression fitting.
展开▼
