Engineer ZHAO Xiukuan proposes a new active semi-supervised procedure to perform fault classification for machine condition monitoring. The effectiveness of the procedure is verified by its application to bearing diagnosis and gear fault detection.

Semi-supervised learning methods use large amounts of unlabeled data together with labeled data to build better classifiers. Since semi-supervised learning requires less human effort and achieves higher accuracy, it is of great interest both in theory and in practice. In the paper, he mainly focuses on the manifold regularization semi-supervised method, which is one graph based method. Graph-based semi-supervised methods define a graph where the nodes are labeled and unlabeled examples in the dataset and the edges (may be weighted) reflect the similarity of examples.

Fig. Procedure of A-LapSVM with two moons type data set. (Image by ZHAO)

Zhao et al. An effective procedure exploiting unlabeled data to build monitoring system. Expert Systems with Applications, 2011, 38（8）：10199–10204（Download Here）