Sparse data inference for point process failure models incorporating multiple maintenance effects

Kearney, J

Abstract

The primary scenario within repairable system reliability estimation investigated is that of a
single failure mode the likelihood of which occurring is supposed to be affected by
maintenance activities of varying effect and degree. Since the structural composition of the
systems considered are unknown, the models developed are simplifications premised either
on a mechanistic conception of a maintenance action (the Proportional Renewal Model) or
by empirically representing the effect of the maintenance action by the transference of a
subset of system components from being in an unmaintained state to a maintained state
with the reverse process determined by some decay process (Maintenance Decay Model).
Maintenance actions are classified either as 'corrective' (CM) if undertaken in response to
failure or as 'preventive' (PM) if elective. The datasets analysed in this work - collected in
the petrochemical industry over a number of years - are typically sparse and contain
observations of a number PM types. The interactions of different maintenance types on a
single failure mode (one type of CM) are investigated and related to the problem of
maintenance scheduling optimisation. Given the complexity of the models and the sparse
nature of reliability data, statistical methods to assess the level of confidence in the model
parameter required to incorporate diverse maintenance effects are compared with
particular focus given to Bayesian methods of statistical inference which have the advantage
of being able to incorporate the use of prior knowledge in the estimation procedure.

Citation

Kearney, J. Sparse data inference for point process failure models incorporating multiple maintenance effects. (Thesis). Salford : University of Salford

• Thesis Type: Thesis
• Deposit Date: Oct 3, 2012
• Award Date: Jan 1, 2011