pySDAS - an AI tool for SDAS detection

Source code: https://github.com/JunOu/pySDAS

Background

Secondary Dendrite Arm Spacing (SDAS) is one of the most important microstructural features in the dendritic solidification of alloys, due to its significant correlation with multiple critical properties of materials. However, manual measurement of SDAS is labor intensive. pySDAS is a software tool designed to automatically detect SDAS from optical microscopy (OM) images. This tool is aimed to be used by both research and industrial users, and thus a simple Graphic User Interface is developed, allowing easy use.

Work flow

pySDAS has a workflow as shown in the image below. The input file, an OM image, is firstly pre-processed, which includes a number of tasks such as dimension calibration, scale bar removal, and brightness and contrast adjustment. Then the features necessary for the next step, including area fraction of eutectic and alpha phases and alpha/edge interface length, are extracted from the image using a number of image processing techniques. These features are then used by the pre-trained machine learning model to detect the SDAS in the OM image. In addition, another tool is also written for convenient manual measurement of the SDAS.

Machine learning model training

This tool uses the OM microstructure of a A356 aluminum alloy casting as an example. Users may add their own material to this tool, while re-training the machine learning model and making a slight revision to the code of the tool. The code, written with Jupyter Notebook, for training machine learning model is also included in this repositories - see ml_model_train_paper_v1.ipynb.

Example/Tutorial