Observations in Reliability Analysis of a 1-D Consolidation Problem

This paper presents the results of a reliability analysis for a one-dimensional consolidation problem. The peculiarity of this problem is that there are two local solutions for the design points, depending on the initial point of the search algorithm in the first-order reliability method, probably due to the non-differentiability of the performance function. One of the local design points is the global solution of the first-order reliability method, while the other is a fake one. Unfortunately, it turns out to be relatively easy to find the fake solution in certain scenarios. This phenomenon of multiple design points is studied and documented in detail in this paper. The analysis results by using other reliability methods are also presented for comparison. Finally, recommendations are given for the suitable reliability methods for this particular example. The conclusion of this paper suggests that cautions should be used when implementing any design-point based reliability method to problems with non-differentiable performance functions.