Abstract
Scanning Electron Microscopy (SEM) image processing and metallurgical analysis are critical for gaining a deep understanding of the structures, properties, and performance of metals. Since metals exhibit complex microstructures composed of grains, phases, and defects that significantly influence their mechanical properties, it is important for researchers to visualize and characterize these intricate structures at a microscopic level, providing valuable insights into the material’s behavior. Although advanced SEM platforms and tools have been widely applied for research and development of novel materials, novel algorithms and image processing techniques are needed to automatically and efficiently analyze a large number of SEM images.
This research paper presents the development of machine learning (ML) and artificial intelligence (AI)-enhanced SEM image processing techniques for the characterization of microstructures in metals. Our long-term goal is to establish an adaptive, data-driven framework for extracting optimal information from individual SEM images, ultimately working towards the creation of a fully automated image processing and analysis methodology. Various data preparation processes, such as image filtering and quality assessment methods, will be investigated for feature identification, extraction, and localization in the SEM images. Then, the extracted information will be used to identify and quantify targeted physical properties and deposition attributes by denoising and image segmentation techniques. Our approach encompasses a comprehensive range of AI/ML techniques, including deep learning, data analytics, and pattern recognition, amalgamated to streamline the image processing workflow. The developed algorithm and software will be able to automatically analyze images and delve into the intricate aspects of physical properties, microstructures, and grain structures, thereby facilitating comprehensive property characterization. The research outcomes will be beneficial for the characterization of 3D printed metals, such as stainless steels, and will be further extended to other alloys in the future.
