White Rose University Consortium logo
University of Leeds logo University of Sheffield logo York University logo

Investigating the Effect of Variable Mass Loading in Structural Health Monitoring from a Machine Learning Perspective

Rahim, Sharafiz (2018) Investigating the Effect of Variable Mass Loading in Structural Health Monitoring from a Machine Learning Perspective. PhD thesis, University of Sheffield.

Available under License Creative Commons Attribution-Noncommercial-No Derivative Works 2.0 UK: England & Wales.

Download (32Mb) | Preview


This study is centered around investigating the effects of operational loading variations caused by various fuel tank loading on an aircraft wing from the perspective of Structural Health Monitoring (SHM) and Machine Learning (ML) paradigm. The main goal is to detect and identify various damage severities under the influence of various loading conditions. To perform the task, a vibration response from an aircraft wing structure is first acquired and measured through accelerometers placed around the structure, which the Frequency Response Function is then computed. The structure’s most important characteristics that is its natural frequencies are extracted using Principal Component Analysis (PCA), which the principal components data becomes the input of an Artificial Neural Network model in aim to predict various damage severities under the effects of various loading variables. The work comprised of two main parts; the first part is a Vibration Based Damaged Detection (VBDD) experiment performed on a replicated wing box structure, which is attached with two pseudo-fuel-tanks located on top of the structure. The findings are then concluded and supported with a final vibration test performed on a real and full-scale aircraft wing which its fuel tank loading is varied extensively throughout the test. Kernel PCA (KPCA) techniques are introduced into the current work with aim to improve data separation from different data groups mainly from smaller damage groups. Besides that, the goal is also trying to minimize false negative damage detection due to the effects of loading variables. The finding from this study has highlighted that nonlinear PCA by kernel Gaussian PCA can improve the chance of detecting damage as well reducing the false negative damage detection. The study also provides a data insight by exploring the data structure obtained from the wing box through Gaussian Mixture Model (GMM), which the first two principal components are considered in building the GMM model. This study intends to serve as a data exploratory framework from a statistics and ML perspective in the interest of SHM with the concern of the structure when exposed to various loading conditions. On other note, it is beneficial to recognize that there is significant numbers in research related to the discrimination of the effects of operational loading and environmental conditions in the field of SHM [1]–[6]. This study, however, aims to provide a finding from the effects of fuel tank loading changes on damage detection. It provides statistical models (based on PCA and KPCA algorithms) and machine learning through ANN architecture as its primary solution to the damage detection of the aircraft wing structure. Nevertheless, this study aims to fill-in the gap in the research area of damage detection under the influence of operational loading changes produced inside the fuel tank of an aircraft wing.

Item Type: Thesis (PhD)
Academic Units: The University of Sheffield > Faculty of Engineering (Sheffield)
The University of Sheffield > Faculty of Engineering (Sheffield) > Mechanical Engineering (Sheffield)
Identification Number/EthosID: uk.bl.ethos.772900
Depositing User: Mr Sharafiz Rahim
Date Deposited: 07 May 2019 09:10
Last Modified: 25 Sep 2019 20:07
URI: http://etheses.whiterose.ac.uk/id/eprint/23793

You do not need to contact us to get a copy of this thesis. Please use the 'Download' link(s) above to get a copy.
You can contact us about this thesis. If you need to make a general enquiry, please see the Contact us page.

Actions (repository staff only: login required)