Fault tolerance for industrial actuators in absence of accurate models and hardware redundancy

This paper investigates Fault-Tolerant Control for closed-loop systems where only coarse models are available and there is lack of actuator and sensor redundancies. The problem is approached in the form of a typical servomotor in closed-loop. A linear model is extracted from input/output data to describe the system over a frequency range. Two methods based on the Kalman Filter and Statistical Change Detection techniques are proposed for detecting degradation faults and component failures, respectively. Finally, a reference correction setup is used to compensate for degradation faults.