Skip to Main Content
Polymer fibers are finding increasing applications in commercial optical communication systems. Polymer optical fibers, with specified desirable consumer characteristics, can be computationally designed. Through the use of an extensive structure - property correlation database, properties of polymers can be predicted by a Neural Network. In this paper we are focusing on glass transition temperature (Tg) that influences a desired outcome in polymeric optical fibers. Performance of such fibers can be optimized by engineering a polymer to exhibit a lower refractive index and Tg. This paper compares and discusses a neural network model and a linear model that have been developed to correlate Tg and repeating units of polymers. A comprehensive neural network model with 28 descriptors was developed to predict T values of 6 g randomly selected polymers from a database containing 71 polymers. The network was trained with the remaining 65 polymers and had an average training RMSE of 17 K (R2 = 0.95) and prediction average error of 17 K (R2 =0.85) based on 10-time experiments. A linear regression model developed for comparison had an average error of 32 K (R2 = 0.81).