| Edible oil adulteration represents one of the prominent contradictions of our current food safety with recent "waste oil" incidents emerging in an endless stream, which has raised great concern of the whole society. Adulterated edible oil not only loses its original nutritive value, but also may damage consumers’ health seriously under some circumstances. However, traditional detection methods for edible oil mainly depend on chemical testing that requires tedious pre-treatment process, which is time-consuming, laborious and expensive. So it is imminent to develop novel and efficient detecting methods to deal with this tough situation.Among various kinds of detective techniques, Raman spectroscopy stand out conspicuously for its remarkable advantages, such as simplicity, high throughout. In the regards of the serious shortcomings such as low resolution, overlapping-peaks in traditional Raman spectroscope, we proposed a novel strategy with the systematic fusion of N-way partial least square methods and multi-scale resolution, extracting information from two dimensional correlation spectra accurately. Our work has tested and refined the multi-scale strategy by drawing upon the quantitative analysis of dynamic multiple spectral data sets to yield challenges representative of those encountered in common Raman spectral analysis of oil adulteration. This proposed strategy has improved the accuracy of olive adulteration identification significantly with refined division of two dimensional correlation’s time and frequency domain by weighting way. Satisfactory calibration results suggest the multi-scale strategy is promising to obtain reliable identification of olive adulteration with higher spectral resolution. The main contents of this paper includes the following sections.1.Based on the discussion on the strengths and weaknesses of the Raman spectroscope and Two Dimensional Correlation Spectroscopy in the olive oil adulteration, we consider pre-treatment process and multivariate calibration as a complementary integration, which combines two dimensional correlation spectroscopy and multi-scale strategy.2.We make further discussion on the rules of the two dimensional correlation spectroscopy. With refined division of two dimensional correlation’s time and frequency domain, we also make a distinction among signals, background and noise by making multi-scale model. The accuracy of olive adulteration model was improved by making sub-models with signals under different scales and adjusting the models’ weightings. This strategy makes full use of important information and restrains noise as well in two dimensional correlation spectroscopy.3.We try to classify olives which adulterate soybean oil and sunflower oil with various kinds of qualitative identification methods. This strategy makes full advantage of the features in the two dimensional correlation spectroscopy and synthesizes the modeling results under different scales. Comparing with other qualitative methods, it obtains more accurate result without leakage of information. We get a higher accuracy from 96.25 percentage to 100 percentage and lower root mean square error from 0.3756 to 0.0607.4.We try to make a reliable and stable quantitative model with two dimensional correlation spectroscopy in olive oil adulteration analysis, and then make a comparison with the other calibration methods, such as Partial Least Squares, Multi-way Partial Least Squares for their performance. For soybean oil adulteration, we get a lower root mean square error from 0.003640 to 0.001278 by 10 percentage. Meanwhile, for sunflower oil adulteration, it is from 0.012257 to 0.004569 by 15 percentage.The results suggest that MM-2DCOS-NPLS not only makes full advantage of the features in two dimensional correlation spectroscopy with higher spectral resolution, but also restrains the noise and baseline’s interference. With its great adaptability and reliability, it provides a promising tool for distinguishing edible oil adulteration. |
PDF Full Text Request