| Summary: | 碩士 === 輔仁大學 === 化學學系 === 83 === The two-way multicomponent analysis for the data generated from
HPLC with Diode Array Detection (LC-DAD) is advantageous of
peak resolution without the need for training sets. The new
Heuristic Evolving Latent Projection (HELP) exempt from the
plagued obs- curity of component analysis in the process of
data partition. This achievement has no doubt marked a new
frontier for chemo- metrics, but remained largely a simulation
so far. In this study, implementation of HELP by LC-DAD and
LCEC is demonstrated here. With LC-DAD, qualitative results for
recon- structed peaks from overlapping peaks with match factor
≧ 999.0 and quantitative results of R=0.99 for calibration
curves and a relative standard deviation of 1.3% are
attainable. Another attempt of multidetection to generate the
data for HELP is the implementation of a pseudo multichannel
electrochemical detector (PMD). Several improvements are made
to provide sen- sitive data with smaller noise level.
Nevertheless, the present PMD though in every respect superior
to the its predecessors hardly rivals LC-DAD as far as data
quality is concerned. Qualitatively it gives match factor ≧
994.0, but quantitatively the reconstructed results are far
from adequate, with a relative standard deviation of 13%.
|
|---|
