Abstract:
It is well known that the conventional reliability analysis using probabilities has been found to be inadequate to handle uncertainty of failure data and modeling. To overcome this problem the concept of fuzzy probability has been used in the evaluation of reliability of systems. It has been amply- demonstrated that the fuzzy set theory can be conveniently used for system reliability evaluation, particularly when there exists uncertainty in the failure data. In reality, even though one may use the best data collection procedures, failure data uncertainty always exists. In the presented work a concept of possibility of failure i.e. fuzzy set defined on probability space is used to evaluate system reliability. The notion of the possibility of failure is more predictive than that of probability of failure, the latter is the limiting case of the former.
In the presented approach a fuzzy-set analysis is made for various system structures viz, fault tree, event trees etc. such analysis can not be made by hand calculations due to complexity of trees. Hence a computer algorithm is developed for each case. Also some approach of neural network is described for the reliability analysis. A feed-forward recursive neural network is used to perform the reliability analysis
