Artificial neural network based prediction of ultimate strength of composite tensile specimen using acoustic emission RMS data

Abstract

Acoustic Emission Testing AET has become a recognized Non-Destructive Test NDT method commonly used to detect and locate faults in mechanically loaded structures and components Acoustic Emission AE can provide comprehensive information on the origination of a discontinuity newline(flaw) in a stressed component and also provides information pertaining to the development of this flaw as the component is subjected to continue or repetitive stress The intensity of an AE signal detected by a sensor is considerably lower than the intensity that would have been observed in the newlineclose proximity of the source This is due to attenuation AE spreads from its source in a plate like material its amplitude decays by 30% every time it doubles its distance from the source In three-dimensional structures the signal decays on the order of 50% The failure characterization will be changed according to AE parameters such as amplitude count RMS energy and duration The sensors are fixed at different places on the specimens For every hit the sensing amplitude by the sensors at different places will be different because of distance between the sensor and source So the definition of failure mode is not accurate for this prediction work But the RMS value is same at each sensor for every source So the predicted value will be accurate Due to attenuation the RMS values is selected for the prediction of the ultimate failure load AE signal collected from the composite tensile specimen during proof testing were interject with an Artificial Neural Network ANN and the ultimate failure strength was predicted Three batch of composite tensile specimens were prepared In the first batch 14 number of Glass/Epoxy tensile specimens were fabricated and tensile test was conducted up to its failure AE data were collected during the tensile test The RMS values of AE are collected up to 70% of its failure load were used to predict the ultimate failure strength A back propagation neural network was generated to predict the failure load Out of 14 specimens 10 specimen s