Jagger Vicente, “Developing Machine Learning Models to Predict Mechanical Properties in Graphene Reinforced Aluminum Metal Matrix Composites”
Mentor: Pradeep Rohatgi, Materials Science & Engineering
Poster #9

The addition of graphene into aluminum-based composites has attracted significant attention in recent years due to the potential for improving mechanical properties of the Graphene reinforced Aluminum matrix composites (GAMMCs). The objective of this research is to establish accurate predictive models that can expedite the evaluation of these properties, which is crucial for the design and optimization of GAMMCs. The growing importance of these materials across various industries necessitates the development of effective and reliable prediction models. We are developing machine learning (ML) models to predict mechanical properties, specifically the ultimate tensile strength (UTS), yield strength (YS), and hardness of GAMMCs. The dataset used in this study consists of a wide range of aluminum-graphene composite compositions, each characterized by varying Graphene type, Graphene content, processing methods, and Aluminum alloy composition. Experimental data on mechanical properties of GAMMCs have been collected from the most recent research papers in this area of research. The data is analyzed using several ML algorithms with cross validation: linear regression, artificial neural network (ANN), K nearest neighbor (KNN), support vector machine (SVM), gradient boosting machine (GBM), and random forest (RF). Grid search is used to find and tune the hyperparameters of the ML models. The results should demonstrate the feasibility of employing machine learning models to predict the mechanical properties of aluminum-graphene composites with high accuracy, offering insight into the intricate relationship between material composition, processing conditions, and mechanical behavior. As of now, the machine learning models have found a correlation of 0.061 between graphene content (wt%) and UTS. We believe the dataset is not discriminative enough to predict mechanical properties. By collecting more data and restricting the number of samples to a specific type of aluminum, we expect the correlation to increase.