Thesis Defense Announcement

Herff College of Engineering announces the Final Thesis Defense of

Rowan Lumb

for the Degree of Master of Science

April 1, 2019 at 11:00 AM in Engineering Science Building, Room 317

Advisor: Dr. Steve Wayne

Evaluating Fatigue Damage Information in Acoustic Emission Data

ABSTRACT: This study examines the acoustic emission (AE) data obtained during static tension and fatigue loading for 4340 steel and 7075 aluminum with the aim of characterizing the AE fatigue damage information. Scanning electron microscopy was used to examine fracture surfaces and revealed the presence of damage in the form of voids, microcracks, slip, debonding of second precipitates, intergranular and transgranular fracture. Analysis of AE data using Andrew's plot from initial static tension testing shows that AE data is grouped by damage mechanism. In this study AE data is collected using a new experimental loading procedure based on the Kaiser effect and Dunegan corollary. The new test protocol interrupts a continuous cyclic fatigue test with sample unloading to zero at predetermined intervals. Upon reloading back to the fatigue test cycles, a simple uniaxial 'ramp' occurs, which induces uniaxial elastic strain in the test specimen. This study therefore examines the AE data obtained during continuous fatigue and ramp loading for 4340 steel and 7075 aluminum. Fatigue damage information is assessed using three methods: traditional AE analysis, information entropy, and a supervised neural network. Results from the 4340 testing showed that traditional AE analysis using the total energy parameter had a low correlation to cycles under dynamic loading with a Spearman correlation coefficient of 0.419, and no correlation for 7075 Al. The information entropy parameter shows a distinct delineation between elastic and plastic strain in static tension tests. For 4340, no correlation to cycles is observed, but for 7075 Al the information entropy parameter has low correlation with a Spearman correlation coefficient of 0.511. The neural network was assessed to be 50.2 ± 18.0% accurate in predicting cycles under dynamic loading for 4340 steel and to be 55.0 ± 15.5% accurate in predicting cycles under dynamic loading for 7075 Al. The results of this study indicates that damage information from a fatigued 4340 steel or 7075 aluminum alloy is contained within the collected AE data, yet challenges remain with regard to using such information for predicting the remaining life of components subjected to fatigue loading.