Deep learning techniques for polycaprolactone molecular weight prediction via enzymatic polymerization process

Abstract

Polycaprolactone (PCL) is a widely used polyester in biomedical field conventionally synthesized under harsh conditions such as high reaction temperature, long reaction time, organic solvents. To address these drawbacks, in this study, PCL is synthesized via enzymatic polymerization using Candida Antarctica lipase B (CALB) as it does not produce toxic residues and operates under mild conditions. Prediction of polymer molecular weight based on the effects of reaction parameters such as temperature and reaction time is critical in understanding the enzymatic polymerization process. Hence in this study feedforward artificial neural network (FANN) and adaptive neural fuzzy inference system (ANFIS) models were used to stimulate the process. Predictions were done with both models by changing training algorithms and sub-parameters in FANN model and number and type of membership functions in ANFIS model. Comparison of results between FANN and ANFIS proved that ANFIS model had better prediction. In ANFIS model, GAUSS GAUSS2 with [8,4] membership function set was the most effective membership functions to develop model for training with low MAE, MAPE, MSE and RMSE values being 0.01872, 0.00012, 0.001226 and 0.035011, respectively. It can achieve with high overall R2 at value of 1 and 99.99% for accuracy of validation.