Uncertainty analysis for the unknown function using artificial neural network (ANN) approximated function

Siti Hajar, Mohd Noh (2010) Uncertainty analysis for the unknown function using artificial neural network (ANN) approximated function. Faculty of Mechanical Engineering, Universiti Malaysia Pahang.


Download (947kB)


This thesis deals with the finding of uncertainty analysis for the unknown function from experimental data by using Neural Network Approximation. The objective of this thesis is to estimates the uncertainty value for the unknown function where Artificial Neural Network (ANN) approximated function join together with sequential perturbation method will be applied. The thesis describes the uncertainty analysis techniques which are analytical (Newton Approximation) method and numerical (Sequential Perturbation) method to predict the uncertainty value and build up the new function from the experimental data via Fortran program using non-linear regression. The approach in analyzing uncertainty of Nusselt number is approximate the function via ANN using feed-forward and backpropagation network with four inputs and output were randomly generated. Finally, uncertainty outcome through sequential perturbation with ANN will be compare with the outcome using analytical method. Percentage error between both methods shall be compute to prove that uncertainty analysis for unknown function using sequential perturbation with ANN can also be use. From the results, average percentage error between Newton approximation (analytical method) and sequential perturbation (numerical method) retrieved is 5.52395×10-4 %. Meanwhile, the average percentage error between actual Nusselt number produced and approximated Nusselt number is 0.955373 %. However the main focus of this study is to determine whether sequential perturbation with ANN approximated function can be apply or not to estimate the uncertainty for the unknown function. The average percentage error between sequential perturbation with ANN and Newton approximation (analytical method) is 3.563%. Therefore, the objective is achieved.

Item Type: Undergraduates Project Papers
Uncontrolled Keywords: Neural networks (Computer science) System analysis
Subjects: Q Science > QA Mathematics
Faculty/Division: Faculty of Mechanical Engineering
Depositing User: Syed Mohd Faiz
Date Deposited: 27 Sep 2011 02:34
Last Modified: 22 Jun 2021 07:06
URI: http://umpir.ump.edu.my/id/eprint/1909
Download Statistic: View Download Statistics

Actions (login required)

View Item View Item