A novel method for analysis of fatigue life measurements based on modified Shepard method

• Fatigue life measurements in replicated tests are analyzed using modified Shepard method.
• The simplicity and accuracy of modified Shepard method are compared with Bayesian statistics.
• Tikhonov stabilization method reduces the influences of measurement noise without significantly distorting final results.
• Tensile surface residual stresses, introduced by polishing procedure, effectively decreases the fatigue life of specimens.

Fatigue is one of the major reasons of engineering parts failure. Still performing experiments is an effective way for determination of fatigue life. Due to inherent randomness of fatigue data, Bayesian statistics is widely used in the literature. In the present study, modified Shepard method is used to determine the best estimate for fatigue life in replicated tests and results are verified by Bayesian statistical approach. Linear regression method based on ASTM standard is utilized to generate fatigue S–N curve, as well as Tikhonov stabilization method is shown to effectively reduce the influences of measurement noise.