PREDICTION OF THE FIRST SPINNING CYLINDER TEST USING DUCTILE DAMAGE THEORY

Using the standard Rousselier ductile damage theory, a prediction is made of crack growth in the First Spinning Cylinder Test carried out by AEA Technology, Risley. Only data from small‐scale mechanical tests and metallography have been used. The prediction has been “blind” in the sense that, although we were aware of the observed high J‐R ‐curve (hitherto defying detailed interpretation by any of the standard fracture mechanics techniques), we restrained ourselves from any attempt to use the detailed results of the test itself in making the prediction. Indeed, we were in some confusion about the reported J‐R‐curves until a final meeting when our predictions were discussed with the experimental workers. Furthermore, crack growth versus rotation speed was predicted before sight of the experimental results. Originally, reasonable estimates of the size scale determining the averaging involved in the damage theory gave predictions of crack growth versus rotation speed which agreed very well with the cylinder test, except near initiation, where the apparent increase in initiation toughness was not predicted. These predictions involved the parameter J only through the use of the compact tension specimen results in fitting the damage theory parameters and were quite independent of any complications which might arise from the use of J when centrifugal or body forces are present. The J‐R ‐curve predictions were not quite so good, but were still well above the small specimen results. Again, there was discrepancy near initiation. Subsequently, we have learned that metallography which we were led to believe applied to a section perpendicular to the line of the mean crack front in fact referred to a section parallel to the mean crack plane. A rough correction using limited additional data improves our “blind” prediction slightly, but other limitations of the available small‐scale data mean that, had some suggestions in the literature been adopted, a result more in accord with that expected from the small‐scale tests might well have been obtained. It is concluded that further development of the modelling and the associated large‐ and small‐scale testing is required. Some comments on this further development and the potential of the method as a predictive engineering tool for the transfer of results from specimens to structures are made.