Robust Fault Detection and Isolation using Bond Graph for an Active-Passive Variable Serial Elastic Actuator

摘要

A robot is a complex machine, comprising mechanism, actuators, sensors, and electrical system. It is, therefore, hard to guarantee that all the components can always function normally. If one component fails, the robot might harm humans. In order to develop the active-passive variable serial elastic actuator (APVSEA) [1] that can detect the occurrence of any component fault, this paper uses bond graph to design a robust fault detection and isolation (RFDI) system. When the robot components malfunction, the RFDI system will execute suitable isolation strategies to guarantee human safety and use zero-gravity control (ZGC) to simultaneously compensate for the torque caused by gravity. Thus, the user can consistently interact with the robot easily and safely. From the experimental results, the RFDI system can filter out uncertain parameters and identify the failed component. In addition, the zero-gravity control can lessen potentially physical damage to humans.