Uncertainty propagation analysis of artificial neural network (ANN) approximated function using numerical and analytical method

Kamal Ariffin, Mohamad (2009) Uncertainty propagation analysis of artificial neural network (ANN) approximated function using numerical and analytical method. Faculty of Mechanical Engineering, Universiti Malaysia Pahang .

Preview

PDF Kamal_Ariffin_Mohamad.pdf Download (5MB)

Abstract

This thesis is to investigate the uncertainty analysis using numerical sequential perturbation method and analytical Newton approximation method. The objective of this project to propose the a new technique using numerical sequential perturbation in calculating uncertainty propagation compare to the use of analytical Newton approximation method in application where the unknown function is approximated using artificial neural network ANN. The process to determine uncertainty have five step including begin from selected function, randomize the data, function approximation and applied the numerical method in ANN and lastly determine percent of error between numerical with ANN and compare with the analytical method. The ANN was applied in MATLAB software. From the uncertainty analysis, was define that three major figure the end of this project. First figure shown the average error between numerical and analytical method without ANN are 0.03%. Second figure average error of function approximate the mass flow rate compare the actual value is 0.03%. The application with numerical method with ANN gives small uncertainty propagation error compare with analytical method where the error is 1.2% is the last graph of this project. The new technique will be approving to determine the uncertainty analysis using artificial neural network (ANN). This technique also can be applied for application in laboratory or industrial field.

Item Type:	Undergraduates Project Papers
Additional Information:	For FULL TEXT, email umplibrary.ir@gmail.com. Enclosed with full name, institution, country, position (student, lecturer, researcher, or others) and the purpose of obtaining full text thesis.
Uncontrolled Keywords:	Neural networks (Computer science)
Subjects:	Q Science > QA Mathematics
Faculty/Division:	Faculty of Mechanical Engineering
Depositing User:	Rosfadilla Mohamad Zainun
Date Deposited:	10 Feb 2011 04:28
Last Modified:	03 Mar 2015 07:48
URI:	http://umpir.ump.edu.my/id/eprint/1093
Download Statistic:	View Download Statistics

Actions (login required)

View Item