Scholarlink

<p>The copolymerization of ethylene and 1-butene in the presence of hydrogen using a dual-site metallocene catalyst was optimized to produce copolymers with pre-designed bimodal molecular weight distributions (MWDs). This optimization employed artificial neural networks (ANNs) featuring both forward and inverse models. A rigorous training phase was conducted for the ANN models using a dataset derived from Monte Carlo simulations, followed by validation and testing procedures. The forward model accurately predicted bimodal distributions obtained through the Monte Carlo method based on initial copolymerization conditions, which included concentration ratios of ethylene to 1-butene and ethylene to hydrogen, while keeping co-catalyst concentration and copolymerization temperature constant. High alignment between predicted and simulated MWD was confirmed through weight fraction comparisons across multiple concentration ratios. Additionally, the inverse model effectively estimated initial copolymerization conditions using weight fraction data for specific chain lengths from bimodal MWD diagrams. As a result, the initial copolymerization conditions in the Monte Carlo simulations were successfully optimized through the integration of ANNs, leading to the generation of pre-designed bimodal distributions. The results demonstrated that HDPE synthesized under different conditions exhibited distinct properties: Case (1) produced higher crystallinity and density with lower comonomer incorporation, while Case (3) resulted in higher molecular weight but lower crystallinity. Case (2) displayed intermediate properties, resembling a bimodal distribution with similar peak heights. This study highlighted the efficacy of integrating Monte Carlo and ANN techniques for precise control over MWD, providing a robust framework for tailoring HDPE properties to enhance performance across diverse applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Decoding the predesigned HDPE synthesis recipe: utilizing the power of ANN and Monte Carlo for tailored molecular weight distribution

  • 期刊论文

  • 发表日期：

    2025年4月15日

查看全文
  • Ramin Bairami Habashi,
  • Mohammad Najafi,
  • Reza Zarghami,
  • Alireza Sabzevari

摘要

The copolymerization of ethylene and 1-butene in the presence of hydrogen using a dual-site metallocene catalyst was optimized to produce copolymers with pre-designed bimodal molecular weight distributions (MWDs). This optimization employed artificial neural networks (ANNs) featuring both forward and inverse models. A rigorous training phase was conducted for the ANN models using a dataset derived from Monte Carlo simulations, followed by validation and testing procedures. The forward model accurately predicted bimodal distributions obtained through the Monte Carlo method based on initial copolymerization conditions, which included concentration ratios of ethylene to 1-butene and ethylene to hydrogen, while keeping co-catalyst concentration and copolymerization temperature constant. High alignment between predicted and simulated MWD was confirmed through weight fraction comparisons across multiple concentration ratios. Additionally, the inverse model effectively estimated initial copolymerization conditions using weight fraction data for specific chain lengths from bimodal MWD diagrams. As a result, the initial copolymerization conditions in the Monte Carlo simulations were successfully optimized through the integration of ANNs, leading to the generation of pre-designed bimodal distributions. The results demonstrated that HDPE synthesized under different conditions exhibited distinct properties: Case (1) produced higher crystallinity and density with lower comonomer incorporation, while Case (3) resulted in higher molecular weight but lower crystallinity. Case (2) displayed intermediate properties, resembling a bimodal distribution with similar peak heights. This study highlighted the efficacy of integrating Monte Carlo and ANN techniques for precise control over MWD, providing a robust framework for tailoring HDPE properties to enhance performance across diverse applications.