Fatigue behavior of the fiber reinforced plastic laminates with scarf joints

The fatigue behavior and fatigue damage accumulation in fiber reinforced plastic (FRP) laminates with scarf joints are studied. Careful observations about the failure mode of FRPs with scarf joints are made. Experiments are conducted on FRPs with scarf joints and fatigue parameters needed for calculation of fatigue lives are obtained. All of the specimens used for this study were made with 24 oz. (5:4) E-Glass woven roving (WR) with 510A vinyl ester resin. Tests were conducted under both constant amplitude as well as the variable amplitude loadings. These tests were stress controlled at five different stress levels, all with zero mean value. The constant amplitude tests were performed at four different stress levels and their results were analyzed through two different methods. Their S/N curves were analyzed through both the log-log approach as well as the semi-log approach. The variable amplitude loading tests were performed under the zero-mean, Gaussian, and narrow-band stochastic process with the probability distributions that had Rayleigh distributed extremas, all, at the same root mean square (rms) stress value of 4 ksi. The fatigue parameters obtained from linear regression analysis of the constant amplitude tests results, through both methods, were then used to predict fatigue lives based on the Palmgren-Miner hypothesis. The parameters obtained through both methods were also used in conjunction with the Rayleigh Probability Distribution Function in order to analytically predict the fatigue life of these specimens under variable amplitude loading. The analytical results predicted for both approaches are then compared with the geometric mean of experimental results of the variable amplitude tests. Comparison of the results rendered the S/N curve obtained through the semi-log approach as the most appropriate method for finding the fatigue life of the FRP laminate specimens with scarf joints investigated in this study.; The results obtained through the semi-log S/N curve approach were less than 7.5% higher than the geometric mean of the number of cycles to failure obtained through the experiments, whereas the expected number of cycles to failure obtained through the log-log S/N curve approach was more than 50% higher than those obtained experimentally.