Neural Network Prediction of Failure Load in High Strength Composites Using Acoustic Emission Method 1

Abstract- The objective of this paper was to predict the failure load of carbon/epoxy composite test specimens using an online acoustic emission (AE) monitoring and artificial neural networks (ANN). The test specimens were Carbon/epoxy rings made of carbon T700 fibers and Epoxy resins these rings were tested in BISS 300KN Servo-hydraulic (UTM) Universal Testing Machine with the help of split disk test fixtures to ensure uniform distribution of loads on the ring and fixing AE sensors on the specimen at discrete locations. A series of 24 carbon/epoxy rings were monitored with an acoustic emission (AE) system, while loading them up to failure. AE signals emitted due to different failure modes in tensile specimens were recorded. Amplitude, duration, energy, counts, etc., were the effective parameters to classify the different failure modes in composites, viz., matrix crazing, fiber cut, and delamination, with several subcategories such as matrix splitting, fiber/matrix debonding, fiber pullout, etc. A Multi-layer Back propagation neural network was generated to predict the failure load of tensile specimens. The network was trained with the amplitude distribution data of AE collected up to 50%, 60%, and 70 % of failure loads, respectively along with their slope of cumulative amplitude distribution plot. 10 specimens were in the training set with their corresponding failure loads. The trained network was able to predict failure loads of remaining 14 specimens within the acceptable error tolerance. The results were compared, and we found that the network trained with 60 % data having better prediction performance