Intelligent optimization strategy of process parameters for PET industrial fiber production based on data-driver

Abstract In response to the challenges of the traditional PET industrial fiber development process, where production parameters typically require extensive trial and error and performance testing, leading to prolonged development cycles and difficulties in applying mathematical modeling for process simulations, a data-driven artificial intelligence (AI) strategy for process parameters configuration based on real production processes of PET industrial fibers is introduced. By utilizing a residual neural network, production process parameters are precisely configured according to the property indicators of PET industrial fiber, achieving a configuration determination coefficient ( R 2 ) of 0.98. Further optimization of these parameters using a particle swarm optimization algorithm results in an R 2 value exceeding 0.99. This approach circumvents complex mathematical modeling, enabling rapid configuration of process parameters through a data-driven AI algorithm based on product property indicators, thus significantly reducing the development cycle and cost of PET industrial fibers.