Condition Based Monitoring on Robotic Gearbox

In today’s world, robots are becoming more and more influential. Health care, agriculture, food preparation, manufacturing and military are industries with utilization and demand of robots to boost efficiency and convenience for customer and provider. Although the demand and presence are high, the maintenance of these robots are still preventive rather than predictive. It is challenging to achieve the latter especially in the manufacturing world where the robots’ movement include high speed variations. In the market, the maintenance of the robots is done with manners utilizing statistical parameters or specific preprogrammed movements to detect if a damage is present. This needs to be handled in order to move to the predictive maintenance where the down time is only done when it is needed making it less costly and more efficient. When rotating equipment is involved, Condition Based Monitoring is an established way of getting these data, to analyze and build decisions on them. In this master thesis, the application of different sensors and analysis technologies for automatization and robotics shall be investigated. This research work aims to have an initial step to infer on the state of the robots and specifically the gearboxes inside. The methods of monitoring are vast, but this work uses vibration signals where different sensor implementations shall be tested. Vibration measurements were made on high precision gearboxes with and without defects. Analysis of the data with various techniques is done in addition to an SKF developed algorithm. After which Machine Learning, specifically anomaly detection, is attempted to prove its viability in this task. The gearboxes used in robots are complicated but intriguing, different implementations have and are used to achieve the requirement of the industry. In this work, the gearbox being tested is the Nabtesco RV-42N, a 2- stage hypocycloidal gearbox, mostly used for robots with 20Kg payload.