Structural damage detection using artificial neural networks and least square support vector machine with particle swarm harmony search algorithm. (1st January 2014)

• Record Type: Journal Article
• Authors: Ghiasi, Ramin; Torkzadeh, Peyman; Noori, Mohammad
• Is Part Of: International journal of sustainable materials and structural systems. Volume 1: Number 4(2014)
• Page Start: 303
• Page End: 320
• Publication Date: 2014-01-01
• Subjects: structural damage -- steel structures -- damage detection -- artificial neural networks -- ANNs -- least squares -- support vector machines -- LS–SMV -- particle swarm harmony search -- PSHS -- distance criterion -- wavelet packet decomposition -- feature extraction -- particle swarm optimisation -- PSO -- intelligent classification; Materials -- Environmental aspects -- Periodicals; Sustainable engineering -- Periodicals; Sustainable buildings -- Periodicals; 620.1
• DOI: 10.1504/IJSMSS.2014.068798
• Languages: English
• ISSNs: 2043-8621

Abstract

The study presented herein compares the performance of structural damage detection using artificial neural networks (ANNs) and least square support vector machines (LS–SMVs). Structural response signals under ambient vibration are processed according to wavelet energy spectrum for feature extraction. The feature vectors are used as inputs to both classifiers based on ANNs and LS–SVMs. LS–SVM parameters along with the selection of input features are optimised using particle swarm harmony search (PSHS) algorithm with a distance evaluation fitness function. The PSHS that has been introduced in this paper is a new hybrid meta–heuristic algorithm for improving the accuracy and the convergence rate of harmony search (HS) algorithm. The effectiveness of different feature extraction methods and different optimisation algorithms are investigated. This investigation shows that although performance of both classifiers is improved by employing PSHS–based selection, for most cases considered, the classification accuracy of LS–SVM is better than ANN. Furthermore, the results demonstrate the efficiency and the robustness of PSHS.