Classification with cooperative semi-supervised learning using bridge structural health data

In the process of bridge structural health monitoring, the parameters monitored by sensors includes strain, vibration, distortion, cable tension etc.. Classification of each parameter can reflect the change of bridge structural health to some extent. According to feature of parameter data, solving methods, namely, improved single-view cooperative-training semi-supervised learning method and multi-view cooperative-training semi-supervised learning method with disagreement are proposed based on problems of single parameter classification and multiple parameters classification separately. The former exploits single-view cooperative-training semi-supervised learning method to classify for single parameter data. The latter takes multiple parameters classification by using different single-view to constitute multi-view, the key of which is to exploit Co-training with disagreement to deal with the problem of two views classification. Bridge structural health information can be obtained effectively through bridge structural data classification with these methods, based on which bridge would be maintained effectively.