Identification of mineral components from near-infrared spectra by a neural network.
-
- TANABE Kazutoshi
- National Institute of Materials and Chemical Research
-
- UESAKA Hiroyuki
- Fujitsu Ltd.
-
- INOUE Tsuneshi
- Dowa Mining Company Ltd.
-
- TAKAHASHI Hiroyuki
- Dowa System Engineering Ltd.
-
- TANAKA Soichiro
- Dowa Engineering Ltd.
Bibliographic Information
- Other Title
-
- ニューラルネットワークによる近赤外スペクトルからの鉱物成分の同定
- ニューラル ネットワーク ニ ヨル キンセキガイ スペクトル カラ ノ コウブ
Search this article
Abstract
A system to identify mineral components from near-infrared spectra by applying a neural network technique was examined. Reflective spectral data at 240 wavelength points for the wavelength range between 1300 and 2400 nm were entered into the input layer of a three-layered neural network trained by the error-back-propagation method. Spectra of various kinds of pure and mixed samples were used for the training, and the mineral components contained in the test samples were examined. As a result, a neural network to identify six kinds of mineral components with a probability of nearly 100% was constructed, and the possibility to develop a system to identify mineral components rapidly is demonstrated.
Journal
-
- BUNSEKI KAGAKU
-
BUNSEKI KAGAKU 43 (10), 765-769, 1994
The Japan Society for Analytical Chemistry