SDBOSR: Separable Decision Boundary based Open Set Recognition for Manufacturing Equipment Fault Classification

Equipment fault may cause serious losses to environmental, economic, and safety. To reduce losses, the requirements for industrial equipment diagnostic systems are rapidly increasing. However, because, industrial equipment is usually complex and large in number, it is difficult to define all types of fault in advance. To reduces the loss, it is necessary to accurately classify the higher the severity level of the fault. In this paper, we propose an algorithm that can classify a fault type of equipment and classify the unknown type based on the open ser recognition method. The proposed method was applied to the AHU dataset and showed improved results compared to another algorithm.