In this work, the ultimate strength of aluminum/silicon carbide (Al/SiC) composites was predicted by using acoustic emission (AE) parameters through artificial neural network (ANN) analysis. With this aim, a series of fourteen Al/SiC tensile samples were loaded up to the failure to investigate the amplitude distribution of AE events detected during loading. A back propagation ANN was prepared to correlate the amplitude values generated during loading up to 60% of known ultimate strength with ultimate failure strength of the samples. Three individual neural networks generated with parameters like hits, the Felicity ratio and rise angle were trained towards anticipating the ultimate strength value, which was predicted within the worst case error of -3.51 %, -4.73 %, and -2.73 %, respectively. The failure prediction accuracy by using rise angle as input was found to be slightly better, although the three neural networks all proved effective.
Neural network prediction of aluminum-silicon carbide tensile strength from acoustic emission rise angle data / Christopher, C. Mahil Loo; Sasikumar, T.; Santulli, C.*; Fragassa, C.. - In: FME TRANSACTIONS. - ISSN 1451-2092. - ELETTRONICO. - 46:2(2018), pp. 253-258. [10.5937/fmet1802253M]
Neural network prediction of aluminum-silicon carbide tensile strength from acoustic emission rise angle data
Fragassa, C.
2018
Abstract
In this work, the ultimate strength of aluminum/silicon carbide (Al/SiC) composites was predicted by using acoustic emission (AE) parameters through artificial neural network (ANN) analysis. With this aim, a series of fourteen Al/SiC tensile samples were loaded up to the failure to investigate the amplitude distribution of AE events detected during loading. A back propagation ANN was prepared to correlate the amplitude values generated during loading up to 60% of known ultimate strength with ultimate failure strength of the samples. Three individual neural networks generated with parameters like hits, the Felicity ratio and rise angle were trained towards anticipating the ultimate strength value, which was predicted within the worst case error of -3.51 %, -4.73 %, and -2.73 %, respectively. The failure prediction accuracy by using rise angle as input was found to be slightly better, although the three neural networks all proved effective.File | Dimensione | Formato | |
---|---|---|---|
130_FME, Santulli, Neural network prediction of aluminum–silicon carbide tensile strength.pdf
accesso aperto
Descrizione: 130
Tipo:
Versione (PDF) editoriale
Licenza:
Licenza per Accesso Aperto. Creative Commons Attribuzione (CCBY)
Dimensione
989.08 kB
Formato
Adobe PDF
|
989.08 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.