Abstract:

In industrial polymerization processes of Innovene and Chisso for polypropylene, agglomeration of polymer particles in horizontal stirred bed reactors (HSBR) has negative impacts on the reactor operation efficiency. A method for agglomeration detection based on attractor comparison by using acoustic emission (AE) sensors is presented. The method, which belongs to time domain analysis, is based on the comparison of time series of AE signals acquired form a normal state with time series measured during operation of the reactor. The nature of the method is whether two time series are generated by the same mechanism. A statistical characteristic S is used to test the null hypothesis that two multidimensional probability distributions are identical. When S <3, it means that the operation state now is normal, if S >3, it indicates with more than 95% confidence that agglomeration in the reactor has taken place. The results from lab-scale experiments showed that the method was sensitive to small changes in the particle-size distribution when the AE sensors were located at the optimal position and the original signals had been pre-processed, for examples, de-noising and normalization, etc . Thus, the agglomeration in HSBR could be quickly and exactly detected online by the acoustic emission technique.

摘要：

在卧式搅拌床反应器丙烯聚合工艺中，结块是严重影响装置长期稳定运行和产品质量的隐患之一。本文以时域分析的角度，提出了一种基于吸引子比较法的声发射结块检测技术。该方法通过比较当前状态与正常状态的声发射信号，并借助统计量S值对两个时间序列的分布进行检验，从而判断两者是否由同一机理产生。当S<3时，则表明当前状态为正常；当S>3时，则表明反应器内有95%的可能性发生了结块。冷模实验结果表明，在合理选择传感器位置的前提下，通过对原始声发射信号进行消噪、归一化等预处理后，吸引子比较法能够灵敏地检测出卧式搅拌床反应器内粒径分布的细微变化，继而预警结块的发生。