Statistical Monitoring of Failures - Methods and Use

The paper deals with the statistical view of failure rate detection. We discuss various probability models and methods of estimating the observed parameter that arise from them. The discussion of robustness of used methodologies and their partial independence of the chosen model is also a part of the paper Firstly, we chose the interval estimation of the failure rate. Further values applicable in the energetic experience are derived from it. One of them is the tolerance interval, i.e. the interval in which the future number of failures will lie with a given probability. Next is the estimation of the most probable number of failures in near future. These are rather important data for the planning of future maintenance. The failure rate, as a probability parameter has double determination a descriptive one, which describes the technical state of a group of appliances, and a predictive one, which models a future expectable number of failures and defects. In the paper two approaches are confronted and interconnected. The first one is exact from the mathematical-statistical point of view and accordingly the (practical) second one that follows the maximal applicability and availability of the obtained results. This second approach is accepted in the assumed structure of the necessary data. The methodologies published in this paper are motivated also with analogies to methods of statistical quality management. This problem has been solved within the project MPO 2A - 2TP1/051 named "Zvysent spolehlivosti a bezpecnosti elektrickych siti" (in English: "Reliability and Safety Enhancement of the Electrical Power System").