Artificial neural networks and adaptive neuro-fuzzy inference systems for structural damage identification using vibration data / Seyed Jamalaldin Seyed Hakim

Seyed Hakim, Seyed Jamalaldin (2015) Artificial neural networks and adaptive neuro-fuzzy inference systems for structural damage identification using vibration data / Seyed Jamalaldin Seyed Hakim. PhD thesis, University of Malaya.

[img] PDF (Full Text)
Restricted to Repository staff only until 01 January 2018.

Download (13Mb)


    The main objective of this study is to develop and demonstrate the potential of artificial neural networks (ANNs) and adaptive neuro-fuzzy inference system (ANFIS) based damage identification techniques for damage localization and severity prediction in I-beam and steel girder bridge structures using modal properties. In this research, experimental modal analysis and numerical simulations of these structures were carried out to generate dynamic parameters of structures and also to investigate the applicability of ANNs and ANFIS for improved structural damage identification. Vibration data from a scaled down steel girder bridge deck and nine I-beams structures with regard to different damage scenarios for each structure were measured to obtain the first five natural frequencies and mode shapes of the structures. Single and multiple damage cases which include double, triple and quad damages were induced in I-beam structures, while only single damage was inflicted in the girder bridge at different locations. Also, numerical modeling of these structures and computation of the responses were carried out using commercial software. In this research, a combination of natural frequencies and mode shapes for the first five modes of these structures were selected as the input parameters for damage identification purpose. In damage identification using ANNs, five individual networks corresponding to mode 1 to mode 5 were trained, and then a method based on neural network ensemble was proposed to combine the outcomes of the individual neural networks to a single network. Based on this study, ANNs were able to detect the severity and location in single and multiple damages accurately. Some insignificant errors for numerical datasets due to modeling errors of the structure and some less accurate results due to the existence of node points in the structure were demonstrated. The ensemble network produces better damage identification outcomes than the individual networks and shows high accuracy of damage identification predictions. Besides that, results show the ANFIS model could identify the severity and location of damage in I-beam and girder bridge structures with high level of accuracy and demonstrated that the outcomes of ANFIS were very close to targets and the developed ANFIS model can be applied as a very strong tool for identification of damage. By incorporating ANNs and ANFIS techniques, the potential and accuracy of damage identification can be improved and some significant major problems of conventional methods can be overcome. According to the results of the comparative study, although both ANNs and ANFIS presented good predictions, ANNs were very sensitive to insufficient and noisy datasets as compared to ANFIS. However, ANFIS provided a structure for the combination of fuzzy logic and ANNs and was less sensitive to insufficient and noisy data and showed more flexible technique than ANN. The comparative study showed that, although in some cases both techniques demonstrated high level of predictions, the ANFIS showed a superior capability to damage predictions using vibration datasets of structures. In conclusion, the ANFIS technique outperformed the ANN and demonstrated the best performance with lowest AE and highest correlation coefficient.

    Item Type: Thesis (PhD)
    Additional Information: Thesis (Ph.D.) -- Faculty of Engineering, University of Malaya, 2015
    Uncontrolled Keywords: Artificial neural networks; Adaptive neuro-fuzzy; Inference systems; Structural damage; Identification; Vibration data
    Subjects: T Technology > TA Engineering (General). Civil engineering (General)
    Divisions: Faculty of Engineering
    Depositing User: Mrs Nur Aqilah Paing
    Date Deposited: 19 Oct 2015 16:21
    Last Modified: 19 Oct 2015 16:21

    Actions (For repository staff only : Login required)

    View Item