Using statistical and uncertainty analyses in design, applied to a tokamak central solenoid

An uncertainty analysis combines the resulting prediction limits with estimated effects of a number of factors, of which the most important are: the statistical variability and limited number of specimens, the largest undetected crack length, the plate-to-plate variability, and the allowance for oversights. The total effect of these (and many more minor) effects is to reduce the allowable stress, for chosen odds against fracture of 10000 to 1, to about 2/3 of the central statistical value. The analysis highlights the factors needing further study and illustrates the value of statistical and uncertainty analyses, which should be combined with cost-benefit and fault-free analyses to complement code-based design.<>