Product identification and unit space construction by MT method using waveform data

Abstract

Near-infrared spectroscopy（NIR）is a simple, quick, non-contact, non-destructive inspection method that can recognize the quality of Kampo Medicines herbal medicines. Because of its high consistency, it can be expected to find applications in the quality inspection of many such medicines. In this study, 13539 lots of 37 commercially manufactured medicines were analyzed by NIR, and the Mahalanobis-Taguchi（MT）method was used to determine the how well NIR discriminated among the lots. Unit spaces for all 37 medicines were created by the same method, thereby avoiding such time-consuming tasks as selecting separate factors for each medicine. The resulting overall discrimination rate for these 37 medicines was 99.17 ％, and we were able to create highly discriminative unit spaces for product identification and detection of abnormalities. However, it also became clear that there were misclassification problems for certain specific medicines, and remaining issues concerning the creation of the unit spaces.