Track 1 Session 9
8:00 to 9:10 a.m. Friday April 13, 2007

Comparison of Different Statistical Methods to Analyze Failures and Degradations

Components that are important for power delivery can be checked through non-destructive testing. Preventive maintenance tasks that avoid failures can also be conducted. Related historical experience data and expertise are available for reliability estimation. Several statistical modeling choices must be made: frequentist or Bayesian inference, taking into account influence variables if available (e.g. pressure, temperature), modeling of preventive maintenance effects on reliability, etc.

This presentation aims at showing and measuring the interest of using methods that allow a modeling which is as close as possible to what is known from the field. It is illustrated by two examples taken from the energy industry.

Key Words: Maintenance Efficiency, Statistics, Predictive Reliability, Non-Destructive Testing, Decision to Replace