Abstract
A gradation method based on quartz lascas (lumps) transparency level is proposed. The samples were irradiated by transmitting light, and the images histograms were processed by artificial neural networks. Additionally, the results were compared to conventional classification methods, including density and visual analysis. The network designed with backpropagation architecture using 4 hidden layers of 10 neurons yielded to a relative error <24% in relation to manual classification, indicating a good agreement to the miners criteria. Furthermore, the implementation of competitive learning with 5 neurons resulted in correct discrimination of samples regarding their optical characteristics with a completely non-subjective approach.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call