Detecting feeble position oscillations from rotary encoder signal in an industrial robot via singular spectrum analysis

Abstract

Position signal faces several weak oscillations due to mechanical flaw and faults occurred in the systems. These oscillations can be identified by the encoders that determine the performance and health condition of the machine. Nevertheless, also the concerned oscillation, rotary encoder signal also includes some measurement noise and a significant trend. These trends are typically of several orders, greater in activities than the involved amplitude oscillations, making it tough to detect the small oscillations except deformation of the signal. In addition, the oscillations can be problematic, and magnitude adjusted in unstable conditions. Singular spectrum analysis (SSA) is proposed to overcome this issue. A numerical emulation is demonstrated to show the efficiency of the approach. It indicates that SSA outperforms ensemble empirical mode decomposition (EEMD), empirical mode decomposition, and complete EEMD with adaptive noise in ability and accuracy. Moreover, during the movement of the robotic arm, encoder signals from the robot are analysed to determine the sources of oscillations in joints. The suggested technique is proven to be reliable and feasible for an industrial robot.