International Journal of Reliability, Quality and Safety Engineering

A Bayesian continuous reliability growth model is presented. It is assumed that the development phase of a product consists of m stages. In each stage, the failure mechanism of the product follows a competing risks model with two specific failure modes: inherent and assignable-cause. The hazard rate for each mode is time-invariant within one stage. Under the assumption that modifications of the product improve its reliability, we assign a reasonable joint prior distribution for the hazard rates. Then Bayesian analysis is carried out using this prior distribution. It turns out that the posterior pdf of the hazard rates of interest is just a weighted average of pdf's which have the same form as the prior pdf. A numerical example is given for illustration.

The competing risk model involving two Weibull distributions is characterized by four independent parameters. The shapes of the density and failure rate functions for the model depend on the parameter values. In this paper a complete parametric characterization of these shapes is presented.

In the randomly-censored model, we define Y = min(X, T) and Z = I_{{X < T}}, where X is the life length, and T is the random censoring time which is independent of X. Couple (Y, Z) is observed. Sufficient conditions are found to ensure that the Maximum-Likelihood Estimators (MLE) are strongly consistent. Application is made to usual life distributions.

A reliability-based design (RBD) of a mixed series–parallel system with deteriorative components for minimal life cycle cost is presented in this paper. Two formulations are presented based on the type of preventive maintenance modeling. These formulations incorporate the effects of preventive maintenance and minimal repairs to adjust the system failure rate. Genetic Algorithms are used to obtain an optimal system design. The economic life of the designed system is also evaluated. The results have practical applications in the area of computer integrated manufacturing where a system must perform below a given failure rate.

A methodology for reliability and safety analysis of a certain class of repairable fault-tolerant systems is presented. The analysis leads to a closed-form approximation of the probability of an absorbing state in a state transition diagram. This expression can provide insight into the relationship between various system parameters and system reliability and/or safety. The approximation technique is based on the combination of results from the analysis of several failure mechanisms, each studied by itself, into an expression for the approximate reliability of a system. The resulting approximation error can be analyzed in order to evaluate whether or not the approximation is useful in a given situation.

The applicability of the fatigue-life distribution of Birnbaum–Saunders for modeling the components of partially prestressed concrete girder bridges is argued by the agreement of the calculated distribution with a large data set comprised of many small censored samples, each put into standardized form, of the fatigue-lives of bridge components. Also, the Weibull and lognormal distributions are compared by the appropriateness of the assumptions in their derivation to the cumulative damage to bridge components.

The maximum likelihood estimates of the parameters of the fatigue lives of the prestressing steel, mild steel, and concrete are found from randomly censored samples, for several censored data-sets under various stress regimes. The parameters are postulated to be log-linear functions of the maximum stress and stress regimes. Weighted least-square regression is applied to determine the unknown coefficients. To compensate for differences in the precision of the estimates obtained under each stress regime the variances are calculated by the bootstrap statistical method. This procedure gives an estimated distribution of the fatigue life with its parameters functions of the stress regime which is applicable under virtually any realistic condition.