Many strength and load related variables such as fatigue strength, yield strength, alternating stress and mean stress are not in general normally distributed as often assumed. In this paper, non-normal distributions of these variables are treated by AFOSM(Advanced First Order Second Moment) method to obtain reliability for the probabilistic constraints. It is based on suitable transformations of the non-normal variables to equivalent normal variables. Then the reliability based optimal design in the form of POD(Parametric Optimal Design) can be utilized and naturally extended to accommodate very general practical problems. The validity of using AFOSM method is shown by the Monte Carlo simulation.
The algorithm is programmed and added to the existing reliability based optimal design program. Several test problems from trusses and frames are taken for distributions such as lognormal and Weibull. The results are compared with one another. Depending on the skewness in the lognormal and the Weibull distributions, for example, the optimal design by the normal distribution assumption can be either conservative or nonconservative and hence care need be exercised.