This paper is concerned with the vibration suppression control
of a one-link flexible arm robot. In our control system,
vibration suppression is realized by an additional feedback of
a strain gauge sensor attached to the arm besides motor angle.
However, a sensor fault may degrade a control performance and
make the control system unstable at its worst. In our control
system, a strain gauge sensor is prone to fail in comparison with
an encoder because it is attached directly to the arm. Then, we
propose a fault-tolerant control system for a disconnection fault
of a strain gauge sensor. We focus on a reaction force observer
and show that a strain gauge sensor signal can be estimated
by multiplying an appropriate gain to reaction force. Moreover,
adaptation algorithm is applied to update of this gain to improve
estimation accuracy and estimation robustness. A strain
gauge sensor fault is detected by monitoring an estimation error
between a sensor signal and an estimated value, and the control
system works with the estimated one after fault detection.
However, the control system may be unstable after exchanging
the sensor signal for the estimated one since the estimated
value includes estimation error and estimation delay, and these
effects are not considered at the controller design. Therefore, a
fault-mode controller taking into account the estimation error
and the estimation delay should be designed and the control
systems should switch from the normal-mode controller to the
fault-mode controller after fault detection to maintain the stability
and the control performance. Loop shaping design procedure
(LSDP) is adapted as a controller synthesis to design both the
normal-mode and the fault-mode controller. Both controllers are
designed according to the same controlled plant with different
weighting functions in consideration of the estimation error and
the estimation delay. As a result, it is possible to design the faultmode
controller easily in comparison with constructing another
sensor-less control system.