Prognostic Algorithms for Condition
Monitoring and Remaining Useful Life
PhD thesis, National University of Ireland Maynooth.
To enable the benets of a truly condition-based maintenance philosophy to be realised,
robust, accurate and reliable algorithms, which provide maintenance personnel with
the necessary information to make informed maintenance decisions, will be key. This
thesis focuses on the development of such algorithms, with a focus on semiconductor
manufacturing and wind turbines.
An introduction to condition-based maintenance is presented which reviews dierent
types of maintenance philosophies and describes the potential benets which a condition-
based maintenance philosophy will deliver to operators of critical plant and machinery.
The issues and challenges involved in developing condition-based maintenance solutions
are discussed and a review of previous approaches and techniques in fault diagnostics
and prognostics is presented.
The development of a condition monitoring system for dry vacuum pumps used in semi-
conductor manufacturing is presented. A notable feature is that upstream process mea-
surements from the wafer processing chamber were incorporated in the development of a
solution. In general, semiconductor manufacturers do not make such information avail-
able and this study identies the benets of information sharing in the development of
condition monitoring solutions, within the semiconductor manufacturing domain. The
developed solution provides maintenance personnel with the ability to identify, quantify,
track and predict the remaining useful life of pumps suering from degradation caused
by pumping large volumes of corrosive
A comprehensive condition monitoring solution for thermal abatement systems is also
presented. As part of this work, a multiple model particle ltering algorithm for prog-
nostics is developed and tested. The capabilities of the proposed prognostic solution for
addressing the uncertainty challenges in predicting the remaining useful life of abatement
systems, subject to uncertain future operating loads and conditions, is demonstrated.
Finally, a condition monitoring algorithm for the main bearing on large utility scale
wind turbines is developed. The developed solution exploits data collected by onboard
supervisory control and data acquisition (SCADA) systems in wind turbines. As a
result, the developed solution can be integrated into existing monitoring systems, at no
additional cost. The potential for the application of multiple model particle ltering
algorithm to wind turbine prognostics is also demonstrated.
Repository Staff Only(login required)
||Item control page