This paper deals with the parameter estimation of lifetime distributions from available field data. The focus is on extending known methodology such that it is applicable in industrial practice and to put this in a tool which can be used also by statistically unskilled users. In the given case the further aim is to do reliability analysis of car components, but the methodology is not restricted to this application domain. The tool covers a number of distributions relevant for reliability analysis including a particular “bathtub” distribution able to represent the behaviour over different timescales. Furthermore, it is able to deal with large data volumes, clustered and suspended data as typically encountered in practice. The parameter estimation is based on adapted versions of linear regression and maximumlikelihood-estimation. Numerical issues are taken care of, local and global optimisation, and interval arithmetics is used. Methods for the computation of confidence intervals are also provided.
[1]
Jochen Heistermann.
Unterstützung Genetischer Algorithmen mittels paralleler Architekturen
,
1994
.
[2]
Ulrich W. Kulisch,et al.
Numerical Toolbox for Verified Computing I
,
1993
.
[3]
Bayerischen Milchwirtschaft,et al.
Statistik
,
2002,
Jugend Medien Schutz-Report.
[4]
Robert B. Abernethy,et al.
The new Weibull handbook
,
1993
.
[5]
J. Miller.
Numerical Analysis
,
1966,
Nature.
[6]
Gordon Johnston,et al.
Statistical Models and Methods for Lifetime Data
,
2003,
Technometrics.
[7]
K. A. Semendyayev,et al.
Handbook of mathematics
,
1985
.
[8]
William J. Thompson,et al.
Atlas for Computing Mathematical Functions: An Illustrated Guide for Practitioners with Programs in Fortran 90 and Mathematica
,
1997
.
[9]
R. Steele.
Optimization
,
2005
.
[10]
CATALINA M. LLADÓ,et al.
Splitting Techniques for Interval Parameters in Performance Models
,
2000
.