Observations and Analysis of Fatigue Crack Growth from Cold Expanded Holes

Abstract

The split-sleeve cold expansion process of Fatigue Technology Inc. (FTI) has been used for over 35 years by the aircraft industry to improve the fatigue life of structures by inducing compressive residual stresses around holes. Current methods of predicting the fatigue life and crack growth from cold expanded holes demonstrated a lack correlation between the predicted and actual crack growth and have also been shown to be very sensitive to possible errors. Since a good correlation of the predicted and real crack growth is essential to an accurate and efficient damage tolerance program, the applicability and effectiveness of the current method is questionable. A preliminary three-dimensional investigation into the crack mechanics indicated that the crack front shape of a cold expanded hole was distinctly different than crack growth from a non cold expanded hole. A series of fatigue tests of 7075-T651 open hole dogbone specimens were run to investigate the crack front shape. The results showed that the crack front shape was significantly different from a cold expanded hole than a non cold expanded hole and varied some with thickness. The distinct P shaped crack front shape was investigated further by growing a crack from a cold expanded hole in a 3-D Finite Element Analysis (FEA) Model. At this stage in the development of the model, CTOA was used as an indicator of relative speed of crack growth in various directions rather than associating it with an absolute crack growth rate. The model showed a similar crack front shape via the simple CTOA criterion as the experiments due to the complex thru thickness residual stress distribution. This thru thickness distribution was the cause for the P shaped crack and the faster crack growth observed on the entry side than the exit side surface of the cold expanded specimens. Because the crack front shape of a cold expanded hole is different than a non cold expanded hole, any prediction method is better served by accurately modeling the real crack growth rather than correlating the life with a model of a physical different crack shape with a slower crack growth rate.