Data-driven prediction and analysis of chaotic origami dynamics

Abstract: Advances in machine learning have revolutionized capabilities in applications ranging from natural language processing to marketing to health care. Here, we demonstrate the efficacy of machine learning in predicting chaotic behavior in complex nonlinear mechanical systems. Specifically, we use quasi-recurrent neural networks to predict extremely chaotic time series data obtained from multistable origami systems. Additionally, while machine learning is often viewed as a "black box", in this study we conduct hidden layer analysis to understand how the neural network can process not only periodic, but also chaotic data in an accurate manner. Also, our approach shows its effectiveness in characterizing and predicting chaotic dynamics in a noisy environment of vibrations without relying on a mathematical model of origami systems. Therefore, our method is fully data-driven and has the potential to be used for complex scenarios, such as the nonlinear dynamics of thin-walled structures and biological membrane systems.