Date of Award

2012

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Environmental Engineering

First Advisor

Chang, Shoou-Yuh Dr.

Abstract

The objective of this thesis research is to apply two artificial neural network (ANN) methods, back-propagation neural network (BPN) and radial basis function generalized regression neural network (RBFGRNN) in two environmental engineering case studies to explore their ability to modeling the complex environmental engineering systems. The traditional environmental engineering systems modeling are frequently using the physical-based modeling methods. . Their performance is decided by the quantity of samples and quality of sampling methods, and it is also based on the physical laws they obeyed and the system knowledge they explored. But ANN offers a unique and alternative solution to bridge the cause and effect without knowing the detailed relationship between each other. Two case studies are used to verify the performance of ANNs, landfill leachate flow rate modeling in Greensboro and total phosphorus concentration modeling in Te-Chi reservoir. The testing coefficient of determination R 2 of BPN applied in landfill leachate flow rate modeling is 0.728 and that in total phosphorous concentration modeling is 0.992. The testing coefficient of determination R 2 of RBFGRNN applied in landfill leachate flow rate modeling is 0.823 and in total phosphorous concentration modeling is 1. These results proved the ANNs are qualified to model complex environmental engineering systems modeling problems.

Download

Share

COinS