Probabilistic Models for Structural Performance of Rounded Dovetail Joints.

This paper presents probabilistic models for the structural performance of rounded dovetail joints. The models are developed with a Bayesian technique, which implies that the model uncertainty is explicitly characterized by random variables. The Bayesian approach also promotes model updating when new test results become available in the future. Practical insight is gained from the modeling process, which includes a novel search for influential parameters, and from the subsequent probabilistic analysis with the models. The models are based on 80 tests of single and double dovetail joints with varying geometric parameters, specifically the flange angle and the dovetail height. A significant effort was made to record variables that conceivably influence the performance of this type of joint, including a series of material parameters: tension strength perpendicular to grain, shear strength parallel to grain, moisture content, density, growth ring density, and growth ring orientation. This paper explores the significance of each parameter and proposes models that include the most significant parameters while retaining a measure of the model uncertainty. In contrast to most models used in structural design, the probabilistic models presented herein are unbiased and suitable for future reliability-based calibration of code equations. [ABSTRACT FROM AUTHOR]