Integrating Deep Learning in Rolling Process Design for Nanocomposites

Abstract

The rolling process is considered to be an important step in the production of advanced materials such as nanocomposites, which require the use of some desirable mechanical properties such as tensile strength and hardness. Here in this work, we provide a novel method, for the first time, the neural network (NN) model is embedded into a rolling processing design, which cultivates a reliable and high-precision model capable of predicting the strength of nanocomposites. Our research serves to bridge an important gap that exists in the current intellectual property: robust, flexible tools for predicting nanocomposite strength under a variety of rolling parameters. This work provides a model-based solution to predict fully charged points using an artificial NN by taking advantage of the properties stated above, aiming to predict more accurately and less computationally cheaply. The conceptual model based on the proposed NN training to experimental and simulated data governed nonlinear interactions between parameters such as rolling temperature, number of passes, and initial thickness of sample. The findings showed for the first time that it was possible to forecast strength at a level previously considered state-of-the-art, which could lead to an improvement in real-time process optimization and cost savings. The invention of this new way to incorporate NNs into the design of a rolling process therefore represents a significant advancement in making high-quality nanocomposites more efficiently, accurately, and at a larger scale.