Real-time bearing fault probability estimation using onsite learning for industrial robots

Inhalt:
This paper presents a real-time fault detection method of a rolling bearing which is one of the major components of industrial robots. The feedback current signal of industrial robots generally consists of a control model and a noise, and a defected bearing adds a specific frequency noise to the feedback current signal. We classify the feedback current signal into various frequency bands using filters and analyse the classified signal to extract features which represent the characteristic of the signal to detect the fault of a bearing. Otherwise most of industrial robots repeat specific motion patterns in various circumstances. We present onsite learning and real-time detection method which exploits the characteristic of industrial robots to improve the versatility of the proposed method. We operated a normal reduction gear until to be defected to gather the normal and the defected data under identical condition to verify the proposed method.