Grade Level at Time of Presentation

Sophomore

Major

Mechanical Engineering

Institution

Western Kentucky University

KY House District #

24

KY Senate District #

5

Department

Dept. of Science and Engineering

Abstract

We present the preliminary results on using a deep learning neural network to predict a metal sample failure based on a set of images obtained with a Scanning Electron Microscope.

Various metal alloy samples were prepared according to ASTM E8/E8M-11 standards for a tensile test. Each sample was prepared for circle grid analysis and then stressed on a tensile machine. Stress and strain values were obtained for each position along the sample by measuring dimensions of each elongated circle. Increasing stress and strain values were found closer to the breakage of the sample with low values found at the holding positions of the sample. Thus, each sample provided a scale of stress/strain values. The correlated stress and strain values were used to determine a failure likelihood for the sample at each position.

Multiple SEM images of the failed steel sample were taken at various locations. SEM images were taken using, Large-Chamber SEM (LC-SEM) at the NOVA Center, WKU. All SEM images were tagged with previously obtained stress and strain values for each circle, correlating position of the circle and position of the image. Using obtained set of SEM images, we trained neural network to classify SEM images based on their stress/strain values. The predicted values were used to analyze the failure likelihood of each sample.

Share

COinS
 

Prediction of Metal Sample Failure from Scanning Electron Microscope images using Deep Learning Neural Network

We present the preliminary results on using a deep learning neural network to predict a metal sample failure based on a set of images obtained with a Scanning Electron Microscope.

Various metal alloy samples were prepared according to ASTM E8/E8M-11 standards for a tensile test. Each sample was prepared for circle grid analysis and then stressed on a tensile machine. Stress and strain values were obtained for each position along the sample by measuring dimensions of each elongated circle. Increasing stress and strain values were found closer to the breakage of the sample with low values found at the holding positions of the sample. Thus, each sample provided a scale of stress/strain values. The correlated stress and strain values were used to determine a failure likelihood for the sample at each position.

Multiple SEM images of the failed steel sample were taken at various locations. SEM images were taken using, Large-Chamber SEM (LC-SEM) at the NOVA Center, WKU. All SEM images were tagged with previously obtained stress and strain values for each circle, correlating position of the circle and position of the image. Using obtained set of SEM images, we trained neural network to classify SEM images based on their stress/strain values. The predicted values were used to analyze the failure likelihood of each sample.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.