Page 1088 - The Mechatronics Handbook

Failure Identification
A fault tolerant manipulator should be capable of identifying a failure, as well as tolerating the failure.
The failed component (mechanical system) of a parallel manipulator, e.g., a failed joint sensor, could be
identified via the manipulator controller using the information provided by the sensors of the device. A
joint sensor fault detection scheme for a class of fault tolerant parallel manipulators, based on redundant
sensing of joint displacements and the comparison of forward displacement solutions, was presented in
(Notash, 2000).
While the failure of active joints could be identified based on the information provided by the sensor(s)
on the corresponding joint, failure of passive joints could be identified by monitoring the overall perfor-
mance of the manipulator in the software. For a given parallel manipulator, the criteria for failure should
be incorporated in the simulation software. For example, the loss of DOF due to workspace boundary
could be monitored (similar to the joint limits and branch interference) and the manipulator could be
stopped before it reaches its envelope to prevent potential failure and damage to the device. As well, all of
the potential special (uncertainty) configurations of the manipulator should be identified, and the closeness
to these singularities should be monitored as the device moves around within its workspace.
Fault Tolerance Through Redundancy
The fault tolerant capabilities of parallel manipulators could be improved by employing appropriate
redundancies. Redundant sensing has been investigated for improving the fault tolerance capabilities of
parallel manipulators, for simplifying the forward displacement analysis of these manipulators, and for
facilitating fixtureless calibration of these devices. Redundancy in actuation has been considered for
eliminating the uncertainty configurations of parallel manipulators. More work is required to develop
methodologies for identifying the failed components of parallel manipulators with elements of redun-
dancy, and compensating for their failures. For parallel manipulators, redundancy could be incorporated
as redundant DOF (mobility), redundant sensing, and redundant actuation.
Redundant DOF could be achieved by incorporating additional joints into the parallel manipulator.
A redundant DOF requires one more actuator on the parallel manipulator. This additional actuator is
not considered as a redundant one because its failure will result in the failure of the parallel manipulator
due to the loss of a required actuation. Redundancy in sensing could be obtained by sensing the existing
passive unsensed joints of the manipulator, by adding a redundant passive sensed branch, or by using
an external sensor such as a vision system. It should be noted that the information redundancy is achieved
by redundant sensing, as well as by providing the task description of the manipulator, such as the Cartesian
trajectory of the end effector (for robot path planning and machining operation). Redundancy in actu-
ation could be accomplished by actuating the passive joints of the manipulator, or by adding an active
branch (in addition to employing dual actuators).

Tool Condition Monitoring
An important element of the automated process control function is the real-time detection of cutting
tool failure, including both wear and fracture mechanisms in machining operations. The ability to detect
such failures online would allow remedial action to be undertaken in a timely fashion, thus ensuring
consistently high product quality (quality of surface finish and dimensional precision) and preventing
potential damage to the process machinery. The basic premise of any automated, real-time tool condition
monitoring system is that there exists either a directly measurable, or a derived parameter, which can be
related to advancing tool wear and/or breakage. Information about tool wear, if obtained online, can be
used to establish tool change policy, adaptive control, economic optimization of machining processes,
and full automation of machining processes.
In the ideal case, the system should be able to detect levels of wear well below those at which the tool
would have to be replaced and should also be sensitive to relatively small changes in the level of wear.
The latter characteristic would provide the system with the potential to “trend” the wear pattern and
predict the amount of useful life left in the tool (allowable wear limit reached).

©2002 CRC Press LLC