Proposal of Standard Regression Method for S-N Curve Based on Fatigue Test Data for Metallic Materials

Abstract

It is very important for the reliability fatigue design of mechanical structures to draw the best fitted curve of an S-N data set. This paper describes a new non-linear regression method for adopting some S-N models for fatigue data regarding metallic materials. The proposed method has an advantage of the best curve fitting for the various types of the S-N data in the fatigue strength database, which was established by the JSMS Committee on Fatigue. This new method is based essentially on the conventional Gauss-Newton method. Furthermore, an additional parameter is introduced in order to achieve high iterative stability for various types of data. Accurate results regarding the curve fitting were obtained using the proposed regression method.