Data-Driven Radial Basis Function Approach to Evaluate the Effect of Process Variables on Solar-Assisted Degradation of Starch-Plastics Composite in an Extruder Reactor

Abstract

This study investigated the effect of process variables on the biodegradation of starch-plastics composite. The extrusion process was carried out with low-density polyethylene (LDPE)-starch blends in varied extrusion temperature and extruder speed and starch content. The parametric analysis based on the three-dimensional plots revealed a non-linear relationship between the input parameters and output. The datasets obtained from the extrusion process was employed for data-driven modeling using radial basis function, a machine learning algorithm. The radial basis function was trained using the backpropagation rule resulting in the prediction of the tensile strength, elongation to break, and yield point of the LDPEstarch composite. The robustness of the radial basis function in modeling the process is evident from the R 2 of 0.804, 0.855, and 0.766 obtained to predict tensile strength, elongation to break, and yield point, respectively, with minima prediction errors. The extrusion temperature, extruder speed, and starch content significantly influenced the predicted tensile strength, elongation to break, and yield point. However, the extrusion temperature was found to have the most significant effect. This study can be employed in understanding the appropriate conditions of parameters required to obtain a degradable material based on its mechanical properties.