| About this Abstract |
| Meeting |
TMS Specialty Congress 2024
|
| Symposium
|
Accelerating Discovery for Mechanical Behavior of Materials 2024
|
| Presentation Title |
Predicting Charpy Toughness From Fractographic Images |
| Author(s) |
Nathan Bianco, Kaitlynn Fitzgerald, Dale Cillessen, Nathan Brown, Jay Carroll, Kimberly Bassett, Brad Boyce |
| On-Site Speaker (Planned) |
Nathan Bianco |
| Abstract Scope |
Evaluating the fracture toughness of failed in-service parts without physical material testing can accelerate failure analysis. In this study, additively manufactured Charpy bar samples were produced over a wide range of process conditions. The Charpy V-Notch toughness was measured on over 400 samples alongside corresponding optical images of the fracture surface. A series of convolution neural network models were trained to determine a correlation between these fracture toughness values and the corresponding fractographic images. In addition to analyzing the native images at 1800 x 1800 pixels, the images were downsampled to various levels to evaluate the tradeoff between image fidelity, predictivity, and computational efficiency. The models successfully predicted the fracture toughness for a diverse range of fractographic images while also identifying interpretable physical characteristics associated with changes in toughness, such as: porosity, shear lips, and fracture surface edges. This work illustrates a machine learning approach to facilitate failure analysis. |
| Proceedings Inclusion? |
Definite: Other |