Condition monitoring system design with one-class and imbalanced-data classifier

As machine´s condition plays a central role in today´s manufacturing industry for assuring process stability and productivity, it is crucial to efficiently monitor the machine status to acquire desired performance. However, in many practical applications, due to economic and technical difficulties, there is only normal operating data or small amount of faulty data available, which will render classic classifiers unusable for this case. To address this problem, this paper presents a design of condition monitoring system with one-class classifier and classifier for imbalanced data. Based on the general description of condition monitoring system, support vector data description (SVDD) and Mahalanobis-distance threshold (MDT) are introduced as one-class classifier and classifier for imbalanced data, respectively. Then they are embedded into a design of condition monitoring system in LabVIEW software and hardware configuration, and a prototype is then developed to show the rationality and feasibility of the design idea.