Reptation-guided deep-learning for polymer discovery from rheological measurement
발표자
()
초록
내용
The use of machine learning to predict rheological properties of polymers has great potential to accelerate the design of novel materials. Here, we have suggested the analogy between the double reptation model and the deep neural network model. The double reptation theory itself can be the special case of the neural network model; the linear activation function and the identical weights for the two-layers are the characteristics of the double reptation model. The identical weight in the double reptation model is related with the molecular weight distribution. We showed that the deep neural network model is available to determine, the molecular weight distribution, entanglement molecular weight (plateau modulus), and monomeric friction factors from the experimental rheological data without any additional information when the machine has been trained properly. Overall, a noteworthy conceptual improvement in the determination of major factors that determine the rheological behavior has been achieved via application of machine-learning methods.