Failure load prediction for fiber-reinforced composites based on acoustic emission

In the design and quality control of fiber-reinforced structures, testing on coupon level and structure level are frequently carried out. In order to accept or reject a final product or material charge, means of quality control are carried out. In safety relevant structures, this is often based on holding a certain proof load. Acoustic emission is already used for the monitoring during proof load testing, but is only used for simple accept/reject diagnosis. For the accepted components typically no assessment is made for the expected residual capacity. We propose an acoustic emission based approach able to perform prediction of the ultimate strength values and to evaluate the materials present stress exposure while being tested. We base our approach on accepted acoustic emission measures, such as the Felicity ratio or the Shelby ratio to assess the structural integrity. Using a combination of an artificial neural network to predict the materials present stress exposure and a simple linear extrapolation we are able to predict the failure strength within the margin of prediction error for all test cases studied. The approach is benchmarked for three types of specimens, systematically changing test volume and load condition. We used tensile tests on fiber-reinforced thermoplastic tape samples, classical tensile test samples and bearing strength samples, all made from the same material.