Statistical Characteristics of Fatigue Life under Random Loading with an Exponential Distribution : for Mild Steel and Aluminum Alloy A5083P-0

Abstract

Many stractural members and mechanical components are generally subjected to variable-amplitude random-sequence loading, or so-called service loadings. At present, in order to predict their fatigue strength, the most reliable method is considered to be random loading fatigue tests simulating real service loadings. In such a case, it is important to examine the statistical characteristics of fatigue life, stress-strain relations under random loading conditions, and so on. In this paper, the following experimental programs were carried out in order to study the above items. The materials used were mild steel and aluminum alloy A5083P-0. Both constant and random loading fatigue tests were performed under load-controlled, axial, fully reversed conditions. The frequency distribution of stress amplitudes was chosen to be exponential. The probability distribution and scatter band of both fatigue life and cumulative damage sum were studied. For aluminum alloy, the life range was approximately 3×10^4 cycles and 5×10^5 cycles; for mild steel, 3×10^4 cycles and 10^6 cycles. About 30 plain specimens were tested for each life range. From these tests the following results were obtained. (1) It was found that the probability distribution of fatigue life for all the above test conditions conformed approximately to normal or logarithmic normal distribution. This was found by both plotting the results on the probability paper and by the x^2 test. (2) Using the F test, it was shown that, for aluminum alloy, the scatter band of fatigue life for random loading was nearly equal to that of constant loading, while for mild steel, it was considerably narrower. (3) It was also shown that the scatter band of fatigue life for aluminum alloy was a little wider than that of mild steel, under both constant and random loading conditions. (4) The probability distribution of cumulative damage sum can be expressed as a normal distribution. (5) Stress-strain relations during random loading cycles are considerably different than those during constant loading.