Abstract
A partial factor format involves multiplying a nominal resistance such as the mean bearing capacity by a resistance factor or dividing a characteristic strength such as the lower 5% undrained shear strength by a partial factor. The purpose of a partial factor format is to implement reliability-based design without requiring the engineer to perform reliability analysis. Hence, the performance of a partial factor format must be measured by its ability to produce designs achieving a desired target reliability index within an acceptable error margin. It is an article of faith that it is possible to achieve an approximately consistent reliability index by recommending a single numerical value for a single resistance factor. This practice is widely adopted in the form of the Load and Resistance Factor Design (LRFD) approach. Based on the examples studied in this paper (pile installed in homogeneous clay and pile installed in clay and sand), it is demonstrated numerically that the deviation from the target reliability index can be so large, particularly on the unconservative side, that the purpose of reliability calibration becomes nearly meaningless. In other words, the partial factor format may not perform significantly better than an uncalibrated format in achieving the prescribed target reliability index. In our opinion, it is difficult to provide a compelling reason to engineers to go through the hassle of changing a code format in this case. If a design guide is insistent on recommending a single numerical value in association with a partial factor format, it is more sensible to recommend a constant quantile value, rather than a constant resistance factor.