Deep Neural Network as a Tool to Classify and Identify the 316L and AZ31BMg Metal Surface Morphology: An Empirical Study

Pushpanjali Bhat1,#

Tanmay Shukla2,#

Nithesh Naik3

Daniel Korir4

Princy Randhawa5

Antony V Samrot6

Ramya S7,Email

Salmataj S A8,Email

1Department of Chemistry, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, 576104, Karnataka, India
2Department of Quantitative Biomedical Science, Dartmouth College, Hanover 03766, New Hampshire, USA.
3Department of Mechanical and Industrial Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal  576104, Karnataka, India.
4Department of Chemistry and Biochemistry, East Texas Baptist University, Marshall Texas 75650 USA.
5Department of Mechatronics, Manipal University Jaipur, Jaipur 303007, Rajasthan, India.
6Faculty of Medicine, MAHSA University, Jenjarom Selangor  42610, Malaysia.
7Department of Electronics and Communications Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal 576104, Karnataka, India.
8Department of Biotechnology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal  576104, Karnataka, India.

#These authors have contributed equally to this work.

Abstract

Identifying the severity of corrosion is crucial in physical and biological sciences. Developing a highly accurate deep learning model capable of classifying corrosion severity across a wide metal surface, even with limited training data, represents a significant advancement over traditional investigation methods. Unlike traditional approaches such as electrochemical measurements that assess only provided inspection areas, deep learning models can classify the entire metal surface. In the realm of biomaterials, this corrosion identification approach aids researchers in selecting appropriate materials for body implants, reducing the impact of corroded metal reactions within the implanted body. This research advocates for an objective and automated examination of metal surfaces, employing convolutional neural networks to classify corrosion intensity based on scanning electron microscope (SEM) images. Despite the limited number of samples from electrochemical laboratories, the deep learning model provides valuable insights across the entire metal plate surface, effectively distinguishing between different corrosion states. Electrochemical measurements were implemented to see the corrosion such as electrochemical impedance spectroscopy (EIS), potentiodynamic polarisation (PDP) techniques. Generative Adversarial Network (GAN) is implemented to generate synthetic images. SEM images were obtained to evaluate the changes at microlevel and CNN was used to classify the images with an efficiency of 92.7%. 

Deep Neural Network as a Tool to Classify and Identify the 316L and AZ31BMg Metal Surface Morphology: An Empirical Study