Common-cause Failure Model Research Articles

Discrete Predictive Analysis in Probabilistic Safety Assessment

This paper presents methods for predicting future numbers of component failures for probabilistic safety assessments (PSAs). The research is motivated and illustrated by discrete failure data from the nuclear industry, including failure counts for emergency diesel generators, pumps, and motor operated valves. Failure counts are modeled with Poisson and binomial distributions. Multiple-failure environments create extra problems for predictive inference, and are a primary focus of this paper. Common cause failures (CCFs), in particular, refer to the simultaneous failure of system components due to an external event. CCF prediction is investigated, and approximate inference methods are derived for various CCF models.

Coupling Model: A Common-Cause-Failure Model with Consideration of Interpretation Uncertainties

The quantification of common-cause failures (CCFs) is often connected with uncertainties in how to interpret observed CCF events and with how far they are applicable to the specific group of components in question. A method has been developed that allows consideration of these kinds of uncertainties on the basis of a modification of the Binomial-Failure-Rate model. The quantification of interpretation uncertainties by means of interpretation alternatives is discussed as well as their effects on the estimation of the coupling parameter of the underlying CCF model. The estimation of the coupling parameter under consideration of the aforementioned uncertainties is performed by a Bayesian approach. To facilitate the specification of interpretation uncertainties, a default proposal of the interpretation vector is automatically generated on the basis of component fault states gained by expert judgment. Modification of the default vector is possible depending on engineering judgment of technical or operational differences between the observed and the target group of components.

Common-cause failure models, data, quantification

Common-cause failure models, data, quantification

A knowledge-based multi-dimension discrete common cause failure model

Common cause failure (CCF) is analyzed as a manifestation of the probabilistic characteristics of component failure rate/probability stemming from stochastic environment load. CCF revolved concepts, such as ‘root cause’ and ‘coupling mechanism’ are interpreted mathematically from the viewpoint of random environment load bringing about failure dependency. Opinions about ‘inherent CCF’ and ‘additional CCF’, ‘absolute CCF’ and ‘relative CCF’ are presented and discussed. An easy-to-use CCF model is developed through multi-dimension environment load-component strength interference analysis and knowledge based parameter discretization. Owing to its strict statistical foundation, such a model has the ability of estimating component failure rate/probability and common cause failure rates/probabilities consistently, dealing with low redundancy system CCF and high redundancy system CCF uniformly, and predicting high multiplicity failure rate/probability based on low multiplicity failure data satisfactorily.

Risk follow-up by probabilistic safety assessment—experience from a Finnish pilot study

Risk follow-up by probabilistic safety assessment (PSA) provides a systematic method to analyze incidents. Events can be evaluated from the safety point of view to get feedback from operating experience, for the identification of risk contributors and for the verification of PSA models. This paper is concerned with the risk follow-up methodology, which is based on a marked point process framework. This framework provides a theoretically rigorous method for retrospective risk evaluations. Basic concepts for the modelling and an evaluation of the operating history by PSA are presented, and a Finnish pilot study on risk follow-up is summarized. We recommend the ‘total memory approach’ for the assessment of the unavailability of the standby safety systems. According to our experience, accurate evaluations with time-dependent component models are not necessarily needed but simplifying approximations can be used. The modelling of common cause failures remains a problem in practice, however, because there are not enough data to distinguish them properly.

The theory and quantification of common cause shock events for redundant standby systems

Abstract System unavailability equations are obtained for redundant standby safety systems with alternative testing strategies. The model is based on time-related general multiple failure rates rather than probabilities per demand. Common cause failures (CCF) of all multiplicities are accounted for. The results indicate that while staggering the tests is advantageous to most systems, extra testing rules may be beneficial in addition for certain types of redundancy. The results for common parallel redundancy systems are used to obtain proper quantitative values for the basic events of general fault tree models of large systems. Several parametric CCF models, their hidden assumptions and mutual relationships are critically evaluated. A certain class of component-caused CCFs is introduced. It turns out that most commonly used models are valid only for the subset of external shocks that are independent of the number of components in a system. Estimation methods are presented for multiple failure rates and other model parameters. Coupling equations between model parameters cause correlations between the parameters and lead to constrained estimation. Numerical examples are used to illustrate the methods and to demonstrate the significance of various aspects of the models.

Reliability and uncertainty analysis of hardware failures of a programmable electronic system

Abstract This paper presents a methodology for evaluating the performance of Programmable Electronic Systems (PES) used for safety applications. The most common PESs used in the industry are identified. Markov modeling techniques are used to develop the reliability model. The major aspects of the analysis address the random hardware failures, the uncertainty associated with these failures, a methodology to propagate these uncertainties in the Markov model, and modeling of common cause failures. The elements of this methodology are applied to an example using a Triple Redundant PES without inter-processor communication. The performance of the PES is quantified in terms of its reliability, probability to fail safe, and probability to fail dangerous within a mission time. The effect of model input parameters (component failure rates, diagnostic coverage), their uncertainties and common cause failures on the performance of the PES is evaluated.

Experiences from dependent failure analysis in nordic countries

This paper provides a review of recent experiences from dependent failure analysis in Nordic countries. A brief summary of the present status of the treatment of explicitly modelled dependencies is given. However, the main emphasis is on analysis of common cause failures (CCFs). In this context insights from the Nordic CCF Data Benchmark Exercise and from qualitative and quantitative retrospective studies are covered. Specific findings concern the impact of alternative methods, data and assumptions in modelling of common cause failures on the results of current PSAs, and use of systematic procedures for search and quantification of CCF contributions. Based on these experiences, priorities for future research have been set. Thus the ongoing activities concentrate on proper consideration of plant-specific defensive measures against CCFs, on generation of representative CCF data which take into account efficiency of such defences, and on quantification of CCF contributions in systems with an ultra high level of redundancy.

Treating Data Uncertainties in Common-Cause Failure Analysis

Uncertainties in the estimation of parameters for common-cause failure models arise not only because of the small number of common-cause failure events but also because recorded events may not be r...

Basic aspects of stochastic reliability analysis for redundancy systems

Much confusion has been created by trying to establish common cause failure (CCF) as an extra phenomenon which has to be treated with extra methods in reliability and data analysis. This paper takes another approach which can be roughly described by the statement that dependent failure is the basic phenomenon, while ‘independent failure’ refers to a special limiting case, namely the perfectly homogeneous population. This approach is motivated by examples demonstrating that common causes do not lead to dependent failure, so far as physical dependencies like shared components are excluded, and that stochastic dependencies are not related to common causes. The possibility to select more than one failure behaviour from an inhomogeneous population is identified as an additional random process which creates stochastic dependence. However, this source of randomness is usually treated in the deterministic limit, which destroys dependence and hence yields incorrect multiple failure frequencies for redundancy structures, thus creating the need for applying corrective CCF models.

A shock model for common-cause failures

Abstract This paper deals with the modeling of common-cause failures in redundant systems. Two of the well-known models for common-cause failures, the β-factor and the binomial failure rate (BFR) model, are discussed and evaluated. Applicability and shortcomings of these models are pinpointed. Further, a new parametric model for common-cause failures, here denoted the random probability shock (RPS) model, is suggested. The main feature of this model is the ability to model various degrees of dependence between the components of the system in a rather straightforward way. The β-factor and BFR models are special cases of this new model.

A systematic procedure for the incorporation of common cause events into risk and reliability models

Common cause events are an important class of dependent events with respect to their contribution to system unavailability and to plant risk. Unfortunately, these events have not been treated with any kind of consistency in applied risk studies over the past decade. Many probabilistic risk assessments (PRA) have not included these events at all, and those that have did not employ the kind of systematic procedures that are needed to achieve consistency, accuracy, and credibility in this area of PRA methodology. In this paper, the authors report on the progress recently made in the development of a systematic approach for incorporating common cause events into applied risk and reliability evaluations. This approach takes advantage of experience from recently completed PRAs and is the result of a project, sponsored by the Electric Power Research Institute (EPRI), in which procedures for dependent events analysis are being developed. Described in this paper is a general framework for system-level common cause failure (CCF) analysis and its application to a three-train auxiliary feedwater system. Within this general framework, three parametric CCF models are compared, including the basic parameter (BP), multiple Greek letter (MGL), and binominal failure rate (BFR) models. Pitfalls of not following the recommended procedure are discussed, and some old issues, such as the benefits of redundancy and diversity, are reexamined.

Unavailability of redundant diesel generators in nuclear power plants

Analytical expressions are derived for the reliability, mean time between failure and availability systems of arbitrary redundant levels. A common-cause failure model is introduced to account for the impact of such failure modes on the failure of units in a redundant system. An approximate evaluation method of common-cause failure rates is provided, based on a Poisson model. Special considerations are given to diesel generators (DG) used for emergency AC power supply in commercial nuclear power plants. Applying the results to compute availabilities of DG units in existing plants showed little improvements in availabilities in using more than three DG units per plant. The results are given in a form that can directly use operational and life test data on failure rates and repair times.