Statistical Analysis of Fatigue Test Data

The offshore environment contains many sources of cyclic loading. Standard design S-N curves, such as those in DNVGL-RP-C203, are usually assigned to ensure a particular design life can be achieved for a particular set of anticipated loading conditions. Girth welds are often the ‘weak link’ in terms of fatigue strength and so it is important to show that girth welds made using new procedures for new projects that are intended to be used in fatigue sensitive risers or flowlines do indeed have the required fatigue performance. Alternatively, designers of new subsea connectors, used for example in tendons for tension leg platforms, mooring applications or well-heads which will experience cyclic loading in service, also wish to verify the fatigue performance of their new designs. Often operators require contractors to carry out resonance fatigue tests on representative girth welds in order to show that girth welds made using new procedures qualify to the required design S-N curve. Operators and contractors must then interpret the results, which is not necessarily straightforward if the fatigue lives are lower than expected. Many factors influence a component’s fatigue strength so there is usually scatter in results obtained when a number of fatigue tests are carried out on real, production standard components. This scatter means that it is important first to carry out the right number of tests in order to obtain a reasonable understanding of the component’s fatigue strength, and then to interpret the fatigue test results properly. A working knowledge of statistics is necessary for both specifying the test programme and interpreting the test results and there is often confusion over various aspects of test specification and interpretation. This paper describes relevant statistical concepts in a way that is accessible to non-experts and that can be used, practically, by designers. The paper illustrates the statistical analysis of test data with examples of the ‘target life’ approach (that is now included in BS7608:2014 + A1) and the equivalent approach in DNVGL-RP-C203, which uses the stress modification factor. It gives practical examples to designers of a pragmatic method that can be used when specifying test programmes and interpreting the results obtained from tests carried out during qualification programmes, which for example, aim to determine whether girth welds made using a new procedure qualify to a particular design curve. It will help designers who are tasked with specifying test programmes to choose a reasonable number of test specimens and stress ranges, and to understand the outcome when results have been obtained.