Machine learning applications in polymer composites

The primary interest in numerous research problems in both polymer composites and Machine Learning (ML) is to develop predictive models for one or more variables of interest using relevant independent variables, or inputs. However, these two fields have often adopted different approaches, where modeling of composite behavior is often based on physics-based models and phenomenological theories. These physical models are more precise and robust, but often suffer from restricted predictive capability since they are confined to a specific set of conditions. ML models, on the other hand, can be more efficient during the design phase as they allow managing massive and high dimensional data sets to extract the best fit or a predictive behavior for the application at hand.In this context, material scientists would benefit from understanding and implementing some of the powerful ML methods, in order to predict or characterize a behavior of interest of a polymer composite. In this paper, we present a general methodology aimed at employing supervised machine learning models for predicting the properties of polymer composites, including thermo-mechanical properties, environmental effects such as moisture saturation level, durability, or other important behavior, based on the composite constituents, manufacturing processes, relevant process parameters, and expected life-span of the composite product.The primary interest in numerous research problems in both polymer composites and Machine Learning (ML) is to develop predictive models for one or more variables of interest using relevant independent variables, or inputs. However, these two fields have often adopted different approaches, where modeling of composite behavior is often based on physics-based models and phenomenological theories. These physical models are more precise and robust, but often suffer from restricted predictive capability since they are confined to a specific set of conditions. ML models, on the other hand, can be more efficient during the design phase as they allow managing massive and high dimensional data sets to extract the best fit or a predictive behavior for the application at hand.In this context, material scientists would benefit from understanding and implementing some of the powerful ML methods, in order to predict or characterize a behavior of interest of a polymer composite. In this paper, we present a general method...