Failure-free operation

In order to ensure virtually failure-free operation, a pohcy of changing this spring at 50,000 cycles of operation might be adopted. 

The majority of inhibited pol3uner coatings serve as structural elements of metal parts. The cost-effectiveness of coatings is a result of the extended time of failure-free operation of the equipment, the raised quality of manufactured goods and the reduction of expense for the protection of equipment against corrosion. 

SOFTWARE RELIABILITY is defined as the probability for failure-free operation of a program for a specified time under a specified set of operating conditions. It is one of the key attributes when discussing software quality. Software quality may be divided into quality aspects in many ways, but mostly software reliability is viewed as one of the key attributes of software quality. 

As stated in opening, software reliability can be defined as the probability of failure-free operation of a computer program in a specified environment for a specified time. This definition is straightforward, but, when the reliability is expressed in this way, it is hard to interpret. 

If we can express the probability of the elements failure-free operation by the parameters of the division of causal variable intervals between the failures, the total probability of the failure-free operation can be expressed by the summary statistic parameters determined from the values of the total amount of generated data (MONOSI, M. PALUCH, B.). 

The KLT-40S power plant [13] was developed on the basis of a standard KLT-40 type nuclear propulsion plant that has the experience of more than 250 reactor-years of failure-free operation. Components of the original plant have been modernized to increase plant reliability, to extend its service life and to improve the conditions of maintenance. The design of safety systems is based on safety regulations for marine reactors and was updated to meet the requirements of the Russian Regulatory Authority - GAN RF - for nuclear power plants. 

Perrow represents a rather pessimistic view of the operators and maintenance personnel s ability to maintain a safe state in high-technology systems involving major accident risks. Experience shows, however, that some organisations are able to sustain failure-free operation of such systems. Studies of so-called high-reliability organisations (HRO) have identified three aspects characterising such systems in particular (LaPorte and Consolini, 1991 Reason, 1997 Rijpma, 1997). 

In the rest of this paper. Section 2 examines how to use a claimed probability of fault-freeness Pp towards claims of actual interest, namely probability of failure-free operation (reliability) over prolonged operation (possibly a system s lifetime). Section 3 discusses how to integrate via Bayesian inference evidence from failure-free operation to improve the claimed reliability it shows how to avoid the crucial difficulty of choosing a prior distribution and obtain predictions that are guaranteed to be conservative (not to err on the side of optimism). Section 4 positions our contribution with respect to other past and ongoing work on related approaches. Last, Section 5 examines how a claim of a certain Pp can be supported, and addresses the crucial issue that absolute certainty of fault-freeness can never bee achieved, even for a product that is indeed fault-free. The last section discusses the value of the reported results and future work. 

We can describe the operation of Bayesian inference as reducing the values of the probability density function more for those values of pfd that are less likely to be true in view of the observed failure-free operation. Thus seeing no failures reduces the values of the probabihty density function for high values of pfd, and shifts probability mass towards the origin (towards pfd = 0). 

These considerations highlight another important point a prior that is pessimistic in terms of the reliability it implies may produce optimistic inference. Here, moving QN closer to 1 implies, before failure-free operation, pessimism a system Hkely to fail in few demands from the start of operation. But then observing it not faihng over even few demands then logically makes it very likely to have 0 pfd (optimism). Which prior distributions will produce pessimistic posteriors depends both on which posterior prediction we wish to minimise (e.g. posterior rehabihty for demands vs posterior probability of fault-freeness) and on the specific observations (here, the number of failure-free demands). It would thus be wrong to take from the worst-case posterior distribution we obtain here any measure different from posterior rehabihty for tfut demands, e.g. a certain percentile, or a posterior probabihty of fault-freeness, and believe it to be a conservative value for use in further claims about this system. 

This relies on (i) using the process evidence to support a claim of probability of fault-freeness (ii) applying Bayesian inference from the observation of failure-free operation and (iii) given strong uncertainty about the prior distributions to use, applying worst-case reasoning. 

Mean time to repair. The mean time between the occurrence of a failure and the return to normal failure-free operation after a corrective action. This time also includes the time required for failure detection, failure search and re-starting the system. 

---