Influence of Experimental Scatter on the Analysis and Extrapolation of Stress-Rupture Data

摘要

A great deal of research has been performed in the area of analysis and extrapolation of stress rupture data for more than three decades. Most of this work has concentrated on development of clever model forms for describing behavior in a simple mathematical way. However, little direct consideration has been given to the importance of inherent data scatter for such analyses. This paper presents a direct discussion of that topic, including an illustration of the impact of data scatter for several computer simulated data sets for which ''true'' behavior and the ''true'' amount of scatter are known by definition. The results clearly indicate that less attention should be paid to development of clever model forms for fitting data, and more should be paid to the use of sound, valid statistical techniques for the fitting of whatever models are used. (ERA citation 11:036551)