Masters Theses

Date of Award

8-1996

Degree Type

Thesis

Degree Name

Master of Science

Major

Civil Engineering

Major Professor

Karen C. Chou

Committee Members

Richard M. Bennett, J. Wesley Hines

Abstract

The nation's infrastructure system is aging rapidly, but resources for the restoration of the infrastructure system are usually very limited. Therefore, decisions for allocating these resources must be made logically, consistently and effectively. Current evaluation methods rely heavily on the use of professional judgement and experience. This can lead to inconsistencies due to the varying opinions between engineers. Neural networks have the ability to make predictions based on a rational and consistent analysis of existing information. In this thesis, bridge rating and evaluation data were used to demonstrate the use of neural networks for infrastructure system evaluation. Data collected by the Tennessee Department of Transportation were used for developing a neural network that is capable of simulating bridge evaluation. Multi-layer feed forward networks were developed using the Levenberg-Marquardt training algorithm. All the neural networks considered consistency of at least one hidden layer of neurons. Hyperbolic tangent transfer functions were used in the first hidden layer and log-sigmoid transfer functions were used in the remaining hidden layers and output layer. The recommended neural network consisted of three hidden layers. This network contained three neurons in the first hidden layer, two neurons in the second hidden layer and one neuron in the third hidden layer. Neural network performance was compared with multi-variate linear regression analysis. The network performed well based on a target error of 10%. The results of this study indicate that the potential for using neural networks for evaluation of all types of infrastructure systems is very good.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS