Optimization using ANN Surrogates with Optimal Topology and Sample Size

Industrial scale process modelling and optimiza\udtion of long chain branched polymer reaction \udnetwork is currently an area of extensive research owing to the advantages and growing popularity of \udbranched polymers. The highly complex nature of these reaction networks \udrequires\uda large set of stiff \udordinary\uddifferential equations\udto model them mathematically with adequate precision and accuracy. In \udsuch a scenario, where execution time of model is expensive, the idea of making the online optimization \udand control of these processes seems to be a near impossib\udle task. Catering to these problems in the \udongoing research, the authors presented a novel work where the kinetic model of long chain branched \udpoly vinyl acetate has been utilized to find the optimum processing con\udditions of operation using Sobol\udsequence \udbased \udANN \udas meta models in a fast and highly efficient manner. The article presents a novel \udgeneric algorithm, which not only disables the heuristic approach of designing the \udANN\udarchitecture but \udalso allows the computationally expensive first principle m\udodel to determine the configuration of the \udANN\udwhich can emulate it with maximum accuracy along with the size of training samples required. The \uduse of\udsuch a fast and efficient Sobol\udbased ANN as surrogate model obtained by the proposed algorithm \udm\udakes the\udoptimization process 10\udtimes \udfaster \udas compared to a case where optimization is carried out \udwith \udthe expensive \udfirst principle model