MACHINE LEARNING APPLICATION IN QUALITY ASSURANCE IN ADDITIVE MANUFACTURING

Additive Manufacturing (AM), particularly Cold Spray Additive Manufacturing (CSAM), is gaining momentum as a cutting-edge technology
capable of producing complex and high-performance components without melting the feedstock material. However, ensuring consistent
quality, reliability, and performance across CSAM parts remains a critical challenge due to the intricate interactions between process
parameters, feedstock properties, and resulting microstructures. This research proposes a comprehensive investigation into the
application of machine learning (ML) techniques for enhancing quality assurance throughout the CSAM workflow—from powder feedstock
characterization to prediction of final mechanical and corrosion properties. The study is structured into six key stages, each targeting a
specific component of the CSAM process. First, ML models, including Support Vector Machines (SVM), Random Forest (RF), will be trained to
predict the flowability of copper powder based on physical, and chemical properties such as size distribution, oxidation level, and moisture
content. Second, ML models will be employed for powder morphology analysis from SEM images to classify powder flowability in CSAM.
Third, the project focuses on porosity prediction using a dataset collected from the literature, where 14 critical process parameters will
serve as input for various ML regression models to estimate porosity levels accurately. Subsequently, this research addresses the
prediction of mechanical and corrosion behavior using a multimodal dataset combining SEM, EBSD, and electrochemical test results. Deep
learning models integrated with GANs will be developed to augment data and improve robustness. The use of other data augmentation
like SMOTE and ADASYN will also help overcome the issue of limited experimental datasets.