Application of Reliability and Fiber Probabilistic Strength Distribution Concepts to Composite Vessel Burst Strength Design

Abstract

With the expanding use of composite vessels, including space and tactical applications, accurate prediction of performance becomes increasingly important. A criti cal parameter in performance prediction is the burst strength as measured by the delivered fiber ultimate strain-to-failure. This paper discusses a composite pressure vessel design approach that is based on reliability and probabilistic fiber strength distribution concepts.1 The method is based on the fiber strain-strength interference reliability theory. A substan tial part of this paper examines how fiber strength distribution parameters generated by tow tensile tests of impregnated carbon fiber strands can be used in pressure vessel strength design. The fiber statistical strength distribution is analyzed using the Weibull distribution function. It is shown that the bimodal Weibull distribution function fits the impregnated fiber strand strength distribution data much better than the single mode Weibull function. This is particularly important because the bimodal function gave a much more accurate representation of the lower tail of this distribution. According to the strain- strength interference theory, the lower tail of the fiber strength distribution is the most sig nificant in pressure vessel strength reliability calculations. The impregnated fiber strand strength distribution agrees well with the strength distribu tion of 4-in. biaxial composite tubes manufactured from the same carbon fiber lot. Good agreement was also shown with the strength distribution of pooled fiber lot tow tensile data and hydroburst strength distribution of full-scale composite pressure vessel manufactured from these lots. In this comparison, the A-basis fiber strength allowable calculated from the strand data was slightly lower (3%) than those calculated from the full-scale pressure vessel strength data, whereas, the B-basis allowables calculated from the strand and the full-scale pressure vessel strength data were the same. However, because A-basis and/or B-basis allowables require many data points from many material lots, the use of fiber lot qualification data is advantageous because it meets these requirements.