Please use this identifier to cite or link to this item:
|Title:||Functional materials simulation||Authors:||Koh, Kenn Kyon Xyn||Keywords:||DRNTU::Engineering||Issue Date:||2016||Abstract:||Zeolites are well defined structures containing elements such as aluminum, silicon and oxygen in their framework. Their unique three dimensional structures contain micro sized pores which allow cations to be trapped inside. Zeolites have good adsorption, ion exchange and catalysis properties which enable zeolites to be used in a wide range of applications such as waste water treatment, chemical catalysis and cleaning agents due to their cage-like structure. New zeolites are still being discovered and there are still millions of hypothetical zeolites to be proven. Characterization of such zeolites involves instrumental analysis involving X-ray diffraction and scanning electron microscopy to establish the lattice parameters of the unit cell and involves cross referencing to zeolite structures of known framework types. However with the large amount of framework types discovered, comparison with each individual framework type will take a long time thus it is required that the time taken involved be reduced such that new zeolites takes lesser time to be listed in the International Zeolite Association database. By using machine learning algorithms in the WEKA software, a framework type predictor is built to perform prediction on zeolites with unknown framework type to narrow down the number of zeolites to cross reference when determining whether a synthesized zeolite is a new or reoccurring zeolite. The prediction models built using various classifiers produced highly accurate predictions on a training set with accuracy ranging from 98.53% to 100%. This high accuracy allowed us to identify the zeolite ZSM-5 Calcined as MFI framework type and Beta Polymorph A as BEA framework type correctly. Zeolites such as ETS-10 and PKU-16 had framework types not listed and had no similar framework types present in the database were also identified due to the varying predictions across individual models.||URI:||http://hdl.handle.net/10356/66386||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||MSE Student Reports (FYP/IA/PA/PI)|
Page view(s) 50116
checked on Oct 21, 2020
checked on Oct 21, 2020
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.