Nonlinearity detection using new signal analysis methods for global health monitoring.

Statistical pattern recognition has emerged as a promising and practical technique for data-based health monitoring of civil structures. This paper intends to detect nonlinearity changes resulting from damage by some simple but e ective signal analysis methods. The primary idea behind these methods is to use measured time-domain vibration signals based on exploratory data analysis without applying any feature extraction. First, statistical moments and central tendency measurements on the basis of the theory of exploratory data analysis are considered as damage indicators to monitor their changes and detect any substantial variations in measured vibration signals. Subsequently, cross correlation and convolution methods are proposed to measure the similarity and overlap between the measured signals of the undamaged and damaged conditions. The main innovation of this study is the capability of the proposed signal analysis methods for implementing nonlinear damage detection without any feature extraction. Numerical and experimental models of civil structures are employed to demonstrate the e ectiveness and performance of the proposed methods. Results show that nonlinearity changes caused by damage lead to reductions in the values of cross correlation and convolution methods. Moreover, some statistical criteria are applicable tools for the global structural health monitoring. [ABSTRACT FROM AUTHOR]