Selection of Method for Fatigue Life Prediction Under Random Loading

Abstract Solder fatigue continues to be one the primary reliability concerns in electronic packages. Three basic factors are necessary to cause fatigue failure. These are (i) a stress of sufficiently high value, (ii) varying amplitude stress, and (iii) a number of cycles of applied stress. Fatigue failure of a solder joint may also occur under cyclic thermal loading. The goal of this project is to compare the experimental results of fatigue testing with the various theoretical approaches (Rainflow counting method, Steinberg’s 3 band approach, and Rayleigh’s method) and come up with a formulation that would predict solder fatigue failure under random vibration loading. Experiments have been performed to collect data. Initially various components having frequency range from 20–2000Hz were combined at random phase to produce a random profile. Rainflow cycle counting was done on 125 random profiles using an algorithm developed by downing (1982). The basic assumption made in this algorithm is that the strain-time history starts with either the maximum peak or the minimum valley. Further each cycle was segregated into one of the three bands depending on the value of the range of the cycle. It was found that the results matched very closely with Gaussian distribution curve. The first phase of testing was to observe the behavior of aluminum and soldered copper (forming a lap joint) specimens under completely reversed sinusoidal loading in order to evaluate S-N curve. Three different stress levels were chosen and all experiments were carried out at a frequency of 20Hz. The second phase of testing consisted of observing the behavior of both the samples under semi-random loading. Random waveform was generated through a waveform function generator and was externally fed to the MTS machine. The present paper presents the result of experiments along with the applicability to the three approaches to prediction.