Acoustic emission nondestructive testing has been used for real-time monitoring of complex structures. All of the structures were made of materials at least 0.070 inch thick. The purpose of this research was to demonstrate the feasibility of using neural networks to classify acoustic emission signals gathered from a pressure vessel made of 2024-T3 aluminum 0.040 inches thick, i.e. thin aluminum sheet. AE waveforms were recorded during fatigue cycling of one pressure vessel using a wide band transducer and a digital oscilloscope connected to a computer. The source for each signal was determined using two narrow band transducers and a LOCAN-AT data acquisition system. The power spectrum was calculated for each waveform. A Kohonen self-organizing map (SOM) was used to cluster the spectra. The network clustered the data on a two-dimensional feature space according to the source of the signal. A total of 3,600 power spectra were used to train the neural network, and 1,800 were used to test the network. Initially there was overlap between the clusters on the two-dimensional feature space; however, this was found to be due to human error. The SOM itself correctly classified all of the signals.
","abstract_html":"<p>Acoustic emission nondestructive testing has been used for real-time monitoring of complex structures. All of the structures were made of materials at least 0.070 inch thick. The purpose of this research was to demonstrate the feasibility of using neural networks to classify acoustic emission signals gathered from a pressure vessel made of 2024-T3 aluminum 0.040 inches thick, i.e. thin aluminum sheet. AE waveforms were recorded during fatigue cycling of one pressure vessel using a wide band transducer and a digital oscilloscope connected to a computer. The source for each signal was determined using two narrow band transducers and a LOCAN-AT data acquisition system. The power spectrum was calculated for each waveform. A Kohonen self-organizing map (SOM) was used to cluster the spectra. The network clustered the data on a two-dimensional feature space according to the source of the signal. A total of 3,600 power spectra were used to train the neural network, and 1,800 were used to test the network. Initially there was overlap between the clusters on the two-dimensional feature space; however, this was found to be due to human error. The SOM itself correctly classified all of the signals.</p>","abstract_has_math":false,"creators":["Thornton, Weldon Paul"],"institution":null,"degree_name":"Master of Science in Aerospace Engineering","degree_level":"Thesis - Open Access","degree_discipline":"Aerospace Engineering","degree_department":null,"school":null,"contributors":["Eric v. K. Hill","Frank Radosta","Sathya Gangadharan"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":1995,"date_issued":"1995-04-01T08:00:00Z","date_published":"1995-04-01T08:00:00Z","updated_at":"2026-07-27T19:25:16Z","subjects":["Acoustic emission","nondestructive testing","pressure vessel","aluminum","Aerospace Engineering","Structures and Materials"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://commons.erau.edu/db-theses/198","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Eric v. K. Hill","Frank Radosta","Sathya Gangadharan"]},{"key":"dc:creator","label":"Author","values":["Thornton, Weldon Paul"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"thesis:degree_discipline","label":"Discipline","values":["Aerospace Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis - Open Access"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science in Aerospace Engineering"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Acoustic emission","nondestructive testing","pressure vessel","aluminum","Aerospace Engineering","Structures and Materials"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://commons.erau.edu/db-theses/198"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Acoustic emission nondestructive testing has been used for real-time monitoring of complex structures. All of the structures were made of materials at least 0.070 inch thick. The purpose of this research was to demonstrate the feasibility of using neural networks to classify acoustic emission signals gathered from a pressure vessel made of 2024-T3 aluminum 0.040 inches thick, i.e. thin aluminum sheet. AE waveforms were recorded during fatigue cycling of one pressure vessel using a wide band transducer and a digital oscilloscope connected to a computer. The source for each signal was determined using two narrow band transducers and a LOCAN-AT data acquisition system. The power spectrum was calculated for each waveform. A Kohonen self-organizing map (SOM) was used to cluster the spectra. The network clustered the data on a two-dimensional feature space according to the source of the signal. A total of 3,600 power spectra were used to train the neural network, and 1,800 were used to test the network. Initially there was overlap between the clusters on the two-dimensional feature space; however, this was found to be due to human error. The SOM itself correctly classified all of the signals.
"]},{"key":"dc:title","label":"Title","values":["Classification of Acoustic Emission Signals from an Aluminum Pressure Vessel Using a Self-Organizing Map"]}]}],"canonical_facts":{"dc:contributor":["Eric v. K. Hill","Frank Radosta","Sathya Gangadharan"],"dc:creator":["Thornton, Weldon Paul"],"dc:description.abstract":["Acoustic emission nondestructive testing has been used for real-time monitoring of complex structures. All of the structures were made of materials at least 0.070 inch thick. The purpose of this research was to demonstrate the feasibility of using neural networks to classify acoustic emission signals gathered from a pressure vessel made of 2024-T3 aluminum 0.040 inches thick, i.e. thin aluminum sheet. AE waveforms were recorded during fatigue cycling of one pressure vessel using a wide band transducer and a digital oscilloscope connected to a computer. The source for each signal was determined using two narrow band transducers and a LOCAN-AT data acquisition system. The power spectrum was calculated for each waveform. A Kohonen self-organizing map (SOM) was used to cluster the spectra. The network clustered the data on a two-dimensional feature space according to the source of the signal. A total of 3,600 power spectra were used to train the neural network, and 1,800 were used to test the network. Initially there was overlap between the clusters on the two-dimensional feature space; however, this was found to be due to human error. The SOM itself correctly classified all of the signals.
"],"dc:identifier":["https://commons.erau.edu/db-theses/198"],"dc:subject":["Acoustic emission","nondestructive testing","pressure vessel","aluminum","Aerospace Engineering","Structures and Materials"],"dc:title":["Classification of Acoustic Emission Signals from an Aluminum Pressure Vessel Using a Self-Organizing Map"],"thesis:degree_discipline":["Aerospace Engineering"],"thesis:degree_level":["Thesis - Open Access"],"thesis:degree_name":["Master of Science in Aerospace Engineering"]},"updated_at":"2026-07-27T19:25:16Z"}