Reliability Engineering最新文献

The probability models of reciprocals of normal, Gamma and Weibull variates are called inverted normal, inverted Gamma and inverted Weibull distributions respectively. These distributions arise naturally in the study of certain stochastic processes causing wear-out failures of mechanical systems.^1–4 In this paper characteristic features of the hazard functions based on this inverted class of distributions are discussed and their possible uses are explored. Hazard functions and mean residual life of inverted normal, inverted Gamma and inverted Weibull are compared with the normal, Gamma and Weibull hazards.

The event-tree analyses of the Angra 1 nuclear power plant probabilistic safety assessment were performed by means of the explicit method of event-tree modelling of shared-equipment dependencies, from which some insights were drawn. At first, we discuss in this paper the effects of collapsing equal impact vectors (from the quantification point of view also). The second feature discussed is related to a critical analysis of the dependency matrix concept in view of a counterpart to it—namely, the GO methodology in the context of modelling intersystem dependencies. Finally, we discuss the possibility of using our computerized event-tree analysis methodology for assessing the unavailability of the involved standby safety systems, most of which are recognized to be noncoherent structures.

Examination of data on software failure reveals that models which assume continuous reliability growth are not accurate. Simulation models incorporating more realistic assumptions have given results in closer accord with the data. Both the simulation models and the real data can be fitted by a simple logistic model.

A new parametric model for common cause failure quantification in reliability analysis is described. Although formally similar to the Multiple Greek Letter model, its parameters are based on event frequencies, not on component failure frequencies. The model is defined in terms of its inputs from event data evaluation and its outputs which correspond to fault tree basic events.

Fatigue failure occurs when some dominant crack or cracks in a component extend to a critical level under the application of cyclic loading. By using the theory of stochastic processes Birnbaum and Saunders proposed a probability model to characterize the time (i.e., the number of cycles) required to propagate a fatigue crack past a critical value. The model is phenomenologically quite sound and provides a probabilistic interpretation of Miner's rule. In statistical literature a thorough treatment of the model is missing. For example, no work has been reported about the renewal and related functions of this model. This paper presents: (i) a summary of some known characteristics of the model; (ii) parameter estimation methods, and K-S test statistics for the model validation; (iii) the nature of hazard function in terms of the coefficient of life variation; (iv) the renewal function, renewal rate function and variance of number of renewals in graphical form; and (v) a comparison of a typical set of various functions.

In structural system reliability analysis it is usually assumed that the elastic properties of the members are deterministic while the loading parameters and/or member resistances are assumed to be stochastic in nature. Theoretically, all potential modes of failure must be considered in an estimation of system failure probability. In reality, however, the probability of occurrence of many of these failure modes is so low, compared with the structural system failure probability, that they make only an insignificant contribution, and hence can be neglected. Thus it is usually sufficient to consider only the so-called dominant failure modes in a system reliability calculation.

For rigid frames composed of ideal plastic or elastic-plastic material the dominant failure modes of interest in the reliability assessment are drawn from the set of all plastic collapse mechanisms. Several empirical procedures have been previously developed to enumerate these dominant mechanisms with varying degree of success. Although in principle it is possible to enumerate all dominant modes relative to some criteria, in practice no method has yet proved to be applicable to large frames, and the validity of the identified mechanisms is often uncertain.

This paper proposes a modification of the existing truncated enumeration method (TEM) applied to plastic frames. An example is demonstrated, and the results compared to other methods recently developed. The results show that almost all the dominant mechanisms may be successfully found. The proposed procedure can be systematically programmed and may be applied to estimate reliability of large structural systems.

A comprehensive study of Subsea Blowout Preventer (BOP) performance in the North Sea between 1978 and 1986 has been carried out to identify BOP reliability problems. Reliability of the BOP as a barrier against blowouts has been assessed together with rig downtime caused by BOP failures and malfunctions. An overall conclusion from the study is that the BOP reliability has improved significantly during the last few years.

The existing fault tree technique is static, whereas the proposed technique using the Markovian process can treat the fault tree dynamically.

By using the Markovian process, it is possible to model the dynamic features of the existing fault tree and to handle the dependencies on the state of the system. This conbination allows detailed consideration of component maintenance, which is normally not considered in the on/off logic of the fault tree.

The Markovian process is based on the probabilistic models. It is also characterized by the state and the time, so the system, which is composed of a number of basic events, can be described at any time by specifying its state at that time.

In the Markovian approach for fault tree, the concept of the supercomponent is introduced in order to reduce the number of system states and the size of the transition matrix. Now, a number of basic events are considered to be one component in the Markovian process.

Using the proposed dynamic fault tree analysis, a sample calculation is performed. As a result, the unavailability is much less than the value for the static fault tree analysis. Namely, the conservatism of the current analysis is excluded in this paper. The dynamic behavior of each system state and of the overall system is well analyzed. The interactions between the supercomponent tested and the supercomponent not tested are dynamically analyzed, too.

In conclusion, by using the Markovian process and the concept of the supercomponent, the size of transition matrix is reduced, and especially, the effect of the tested supercomponent on the system is dynamically analyzed.