For the condition assessment issues based on SHM, it is an effective and convenient way to learn and compare the response pattern of the structure in various conditions. The basic idea of this method known as a statistic or probabilistic view, is that the response of the damaged structure would be different from the normal one. This enables us using statistic or machine learning tools to assess the conditions. Cable tension is one of the most important measurements because stay cables are critical and vulnerable structural components for cable-supported bridges. Many kinds of sensors can obtain cable tension. Additionally, the identification algorithms for obtaining cable tension have been proposed by using extended Kalman Filter (EKF) or Blind Source Separation (BSS) method. Moreover, it is obviously that each vehicle load could be regarded as a test of the bridge and cable tension as the received signal, therefore, it will be useful to learn the pattern of cable tension under normal operation condition. The monitored cable tension data from Nanjing 3rd Yangtze River Bridge’s SHM system is employed in this study. A simple method to separate vehicular spike based on the tension signal characteristic is proposed. It is observed that the cable tension of up/down-river stay cables shows six different distribution patterns which correspond to vehicle loading on six different lanes. Based on two natural assumptions, GMM (Gaussian Mixture Model) method is employed to cluster the six patterns. The condition of cables can be assessed based on the variation of six patterns.
doi: 10.12783/SHM2015/355