ABSTRACT

Civil structures are usually prone to damage during their service life and it leads them to loss their serviceability and safety. Thus, damage assessment can guarantee the integrity of structures. As a result, a structural damage detection approach including two main components, a set of accelerometers to record the response data and a data mining (DM) procedure, is widely used to extract the information on the structural health condition. In the last decades, DM has provided numerous solutions to structural health monitoring (SHM) problems as an all-inclusive technique due to its powerful computational ability. This paper presents the first attempt to illustrate the data mining techniques (DMTs) applications in SHM through an intensive review of those articles dealing with the use of DMTs aimed for classification-, prediction- and optimization-based data mining methods. According to this categorization, applications of DMTs with respect to SHM research area are classified and it is concluded that, applications of DMTs in the SHM domain have increasingly been implemented, in the last decade and the most popular techniques in the area were artificial neural network (ANN), principal component analysis (PCA) and genetic algorithm (GA), respectively.

INTRODUCTION

Structural systems in civil engineering such as tall buildings, long hydraulic structures, and long span bridges are damage-prone under different loadings such as fatigue, aging, overloading, earthquakes and other natural disasters during their service life (Duan and Zhang 2006; Hakim and Razak 2014a; Khanzaei et al. 2015; Ghaedi et al. 2016, 2017a, b). Existence of damage can disturb functionality and safety of the structure. Therefore, damage detection is one of the most important factors in order to guarantee the integrity and safety of civil structures (Kanwar et al. 2007; Xu et al. 2012; Hakim and Razak 2014b; Hanif et al. 2016).

Many damage identification techniques have been applied to civil structures. For instance, visual inspections are the most common damage detection techniques. However, they have been used in detecting damage of many structures; they are time consuming and costly. These techniques cannot also be used for continuous monitoring of structures (Razak and Choi 2001; He 2008).