A METHOD FOR ESTIMATING A PROBABILISTIC DESIGN FACTOR

Edward E. Osakue, Lucky Anetor

Abstract: A probabilistic method for determining a design factor is presented based on the lognormal probability density function. Design parameters are characterized by mean values and coefficients of variation (covs). The design capacity model variability is estimated using first order Taylor series sensitivity analysis. The method provides a lower bound and an upper bound estimate for a design factor which can be used to initiate a design task without an arbitrarily chosen “safety factor”. Two design cases of static bending and axial tensile fatigue are considered. In first case, the range of the design factor is 1.485 to 1.620 at a reliability level of 99.9%. In the second case, the range of the design factor is 1.631 to 1.739 at a reliability level of 99%. The higher values of the design factor in the second case compared to the first, even at lower reliability level are attributable to the greater variability of design parameters for the second case. The illustrated two cases are stress-based problems but the method is not limited to stress-based design. It can be applied to other serviceability criteria such as buckling, lateral deflection, torsional deformation, critical frequency, etc. What is required is the formulation of a design capacity model for a failure mode that is related to an appropriate serviceability requirement. Because a probabilistic design factor serves the function of a “safety factor” in deterministic design, it follows that deterministic design is converted to probabilistic design by its use. This is an easier, faster and less costly approach to probabilistic design than classical methods.

Keywords: Lognormal, Reliability, Variability, Normal Variate, Design Factor, Over-design Factor

DOI: https://doi.org/10.15623/ijret.2017.0608018