Deterioration processes

  • Mohamed Abdel-Hameed

Abstract

In reliability studies, the question of assessing the behavior of the failure rate of a given device often arises. In practice, it is assumed that the life length of the device has a certain distribution, such as exponential, Weibull, gamma, or that its life length belongs to a given family of distributions, such as increasing failure rate, increasing failure rate average. Based on field data collected about the failure times of identical devices optimal estimates of the failure rate are obtained and hypothesis testing for the parameters of the assumed distribution function are carried out. In many cases collecting enough data for sound statistical conclusions to be drawn is not possible either because of prohibitive cost or insufficient time available to observe all failure times of items on test. Even if enough data can be collected, the validity of the inference procedures are questionable due to the sometimes unfounded but necessary assumptions that must be imposed concerning the distribution function of the failure time. One way to avoid the above difficulties is to examine the failure mechanism of the given device and thus determine in a proper fashion the form of its distribution function. In this paper we concern ourselves with deterioration models, and we discuss three such models. In the first model a device is subject to damage and wear. The damage is assumed to be an increasing strong Markov pure jump process and the wear occurs at a constant rate. This amounts to saying that damage occurs because of shocks and that the times and magnitudes of shocks form a Poisson random measure on R+ × (0,∞) and the rate of wear is a constant a > 0. We call this model Pure Jump Damage Process With Drift. The second model differs from the first in one regard: between shocks the device wears at a rate which is equal to the damage accumulated right before the occurrence of the shock. For example, between the first and second shock the device fails at a constant rate which is equal to the left hand limit of the damage level at the time of occurrence of the first shock. We call this model Deterioration Processes With Wear Depending On Damage Level. The third model differs drastically from the first two. In this model, the deterioration process is assumed to be a Markov additive process (X,Z). The process X = (Xt) describes the state of the environment, and the increasing process Z = (Zt) describes the accumulated deterioration the device suffers. The process Z has conditionally independent increments given the paths of the environment process. In all three models, the device is assumed to have a threshold and it fails once the accumulated deterioration exceeds or is equal to the threshold. We study life distribution properties of such devices and the effect of the parameters of the deterioration process on the failure rate.

Keywords

Failure Rate Failure Time Damage Function Wear Process Deterioration Process 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Abdel-Hameed, M.S. (1984a). Life distribution properties of devices subject to a Levy wear process. Math. Oper. Research 9, 606–614.MathSciNetMATHCrossRefGoogle Scholar
  2. Abdel-Hameed, M.S. (1984b). Life distribution properties of devices subject to a pure jump damage process. J.Appl.Prob. 21, 816–825.MathSciNetMATHCrossRefGoogle Scholar
  3. Cinlar, E. (1972). Markov additive processes, II. Z.Wahrscheinlichkeitstheorie Verw. Geb. 24, 94–121.MathSciNetGoogle Scholar
  4. Cinlar, E. (1975). Levy systems of Markov additive processes. Z.Wahrscheinlichkeitstheorie Verw.Geb. 31, 175–185.MathSciNetMATHCrossRefGoogle Scholar
  5. Cinlar, E. (1976). Shocks and wear models and additive processes. Proceedings of the Conference on Theory and Applications of Reliability with Emphasis on Bayesian and Nonparametric Models, Edited by C.P. Tsokos and I.N. Shimi, Academic Press, 193-214.Google Scholar
  6. Cinlar, E. and J. Jacod (1981). Representation of semimartingales Markov processes in terms of Wiener processes and Poisson random measure. Seminar on Stochastic Processes, Edited by E. Cinlar, K.L. Chung and R.K. Getoor, Birkhauser, Boston, 159–242.Google Scholar
  7. Feldman, R.M. (1977). Optimal replacement for a system governed by Markov additive shock processes. Ann. Prob. 5, 413–429.MATHCrossRefGoogle Scholar
  8. Hollander, M. and F. Proschan (1972). Testing whether new is better than used. Ann. Math. Statist. 43, 1136–1146.MathSciNetMATHCrossRefGoogle Scholar
  9. Karlin, S. (1969). Total Positivity. Standford University Press, Standford.Google Scholar
  10. Marshall, A.W. and F. Proschan (1965). Maximum likelihood estimation for distributions with monotone failure rate. Ann. Math. Statist. 36, 69–77.MathSciNetMATHCrossRefGoogle Scholar
  11. Marshall, A.W. and F. Proschan (1972). Classes of distributions applicable in replacement, with renewal theory implications. Proceedings of the 6th Berkeley Symposium on Mathematical Statistics and Probability Vol.1, Edited by J. Neyman, and E.L. Scott, University of California Press, 395-415.Google Scholar

Copyright information

© Springer Science+Business Media New York 1986

Authors and Affiliations

  • Mohamed Abdel-Hameed
    • 1
  1. 1.Department of MathematicsUniversity of North Carolina and Kuwait UniversityUSA

Personalised recommendations