極値統計を活用した構造信頼性評価に関する検討

抄録

In recent years, the technologies to measure working load and stress applied to the machinery systems, which are the kind of the design factors of them, by using ICT have been developed. In order to increase the reliability of machinery systems, it is effective to apply structural reliability analysis to predict the failure probability of machinery systems because the design factors have stochastic natures and structural reliability analysis could take into account uncertainty of design factors. In order to improve prediction accuracy of failure probability, it is important to know the edge shape of probability density function(PDF) of design factors because the edge shape affects the accuracy of integral calculation of failure probability. We proposed a method to evaluate the edge shape of PDF by applying peak over threshold method, a kind of extreme value theory, to measured data of design factors. The edge shape of PDF could be estimated directly by the developed method even if the mathematical expression of PDF is unknown. We applied the developed method to data analysis of wind velocity, which is one of the design factors of wind turbine, and estimated the edge shape of PDF for wind velocity above 28 m / sec. It was confirmed that the gap between estimated PDF and the data accumulated over 50 years decreased by 64 % compared to conventional technology.