This project explores alternative methods to monitoring and predicting dynamic loads of underwater shells. There is a pressing need for quality health monitoring systems for underwater structures due to the catastrophic nature of unplanned failure. Although other autonomous methods exist, neural networks excel in picking up features that might otherwise be missed through direct programming and mechanical methods. Machine learning with neural networks has already shown greatly improved results over the current state of the art in foundation crack detection. In our project, we develop finite element simulations of shell structures with which we extract “ground truth data”. This data is used to develop machine learning algorithms, and train models capable of analyzing the structural integrity of a shell structure, and further predicting probable failure.