(Theme session) Academic exchange session in English

[1301-08-03] Rare Earth Element-related heavy mineral sand identification using AI

○Okhala Muacanhia^{1[Student presentation: Doctoral course]}, Natsuo Okada², Yoko Ohtomo², Youhei Kawamura² (1. Division of Sustainable Resources Engineering, Graduate School of Engineering, Hokkaido University, 2. Division of Sustainable Resources Engineering, Faculty of Engineering, Hokkaido University)

司会：パク　イルファン（北海道大学）

Keywords:Machine learning, heavy mineral, monazite, hyperspectral images, rare earth elements

Rare Earth Element (REE) are critical in varies fields ranging from communication, manufacturing to green technology. Mineral sand deposits sometimes contain a concentrated amount of REE-bearing minerals such as monazite, xenotime and zircon and it is crucial to identify and separate them more efficiently and accurately. Here, we investigated the effect of grain size on the spectrum of heavy minerals sand and how this affects the accuracy of classification using machine learning and hyperspectral imaging in the visible and near-infrared range. It was proven that the spectral reflectance depends significantly on the particle size with finer particles displaying higher reflectance compared with coarser fractions. Epidote, rutile, staurolite, zircon display a monotonically increase in the reflectance spectrum from 600 nm. Tourmaline, pleonaste and ilmenite, on the other hand, exhibit u-shaped profiles dipping between 500 and 900 nm. Monazite, particularly, reveals a reflectance profile prominent absorption feature associated with the presence of trivalent ions of Nd in the 580, 740, 800 and 870 nm bands and in the bands between 700 and 900 nm because of the presence of Sm. Machine learning algorithms of SVM and CNN can efficiently identify monazite based on its spectral behaviour. The identification of monazite was accurate in all grain sizes because of the characteristic spectrum, which did not alter in the spectrum of non-pulverized sample. The particular and consistent absorption features of monazite can be used as indicator of REE-bearing in heavy mineral assemblages.