基于免疫算法的聚烯烃分子量分布去卷积分析

Deconvolution analysis of polyolefin molecular weight distribution based on immune algorithm

聚烯烃分子量分布(MWD)的去卷积分析是聚合反应动力学参数估计的重要步骤,对烯烃聚合过程建模、模拟和优化至关重要。针对经典MWD去卷积分析方法中传统算法需要较精准的初始值否则无法准确获得每个活性中心反应动力学参数的问题,提出了全局搜索的免疫算法(IA),用于多活性位Ziegler-Natta催化剂制备的聚烯烃MWD的去卷积分析。该算法通过随机产生初始种群,利用免疫操作算子克服一般算法易陷于局部最优的缺点,实现全局寻优。结果表明,在无需给定初始值的条件下,该算法能够有效地计算出催化剂活性位个数、每个活性位所占比例以及所生成聚合物的数均分子量。最后,将免疫算法的求解结果与Levenberg-Marquardt(L-M)算法结合,有效地提高了去卷积分析的精度。

Deconvolution analysis of polyolefin molecular weight distribution(MWD)is a significant step for the estimation of polymerization kinetic parameters which are of great importance for modeling,simulation and optimization of olefin polymerization processes.A global searching immune algorithm(IA)was proposed for the deconvolution analysis of polyolefin MWD produced by multisite Ziegler-Natta catalyst because the conventional algorithm cannot accurately obtain the kinetic parameters of each active site without a precise initial value.The algorithm overcomes falling into local convergence of common method and realizes the global optimization by immune operators.The results show that the algorithm can effectively calculate the number of catalyst active site,the mass fraction and the number average molecular weight of the polymer produced on each active site without giving an initial value.Then,IA combining with Levenberg-Marquardt(L-M)algorithm further improves the accuracy of deconvolution analysis.