Page 359 - Sensors and Control Systems in Manufacturing

314
Cha p te r
S i x

its operational frequency. The component of that signal represents
that change in voltage needed to keep the system at quadrature to
follow the system change in stress. That signal provides the informa-
tion on changing stress.
The UDVSS can be moved around on the specimen to map out
the stress field, and by rotating the probe, one can determine the
direction of a stress. In addition, the probe is easily calibrated. The
UDVSS should find wide acceptance among manufacturers of aero-
space and automotive structures for stress testing and evaluation of
designs.

6.26 Predictive Monitoring Sensors
Serving the CIM Strategy
Computer-integrated manufacturing technology can be well-served
by a predictive monitoring system that would prevent a large num-
ber of sensors from overwhelming the electronic data monitoring
system or a human operator. The essence of the method is to select
only a few of the many sensors in the system for monitoring at a
given time and to set alarm levels of the selected sensor outputs to
reflect the limit of expected normal operation at the given time. The
method is intended for use in a highly instrumented system that
includes many interfacing components and subsystems—for exam-
ple, an advanced aircraft, an environmental chamber, a chemical pro-
cessing plant, or a machining work cell.
Several considerations motivate the expanding effort in imple-
menting the concept of predictive monitoring. Typically, the timely
detection of anomalous behavior of a system and the ability of the
operator or electronic monitor to react quickly are necessary for the
continuous safe operation of the system.
In the absence of a sensor-planning method, an operator may be
overwhelmed with alarm data resulting from interactions among
sensors rather than data directly resulting from anomalous behavior
of the system. In addition, much raw sensor data presented to the
operator may by irrelevant to an anomalous condition. The operator
is thus presented with a great deal of unfocused sensor information,
from which it may be impossible to form a global picture of events
and conditions in the system. The predictive monitoring method
would be implemented in a computer system running artificial intel-
ligence software, tentatively named PREMON. The predictive moni-
toring system would include three modules: (1) a causal simulator,
(2) a sensor planner, and (3) a sensor interpreter (Fig. 6.28).
The word event in Fig. 6.28 denotes a discontinuous change in the
value of a given quantity (sensor output) at a given time. The inputs
to the causal simulator would include a causal mathematical model
of the system to be monitored, a set of events that describe the initial