管式聚合反应温度分布控制方法的研究

该文利用B样条神经网络建立了管式聚合反应温度分布函数的数学模型,为分布参数系统的建模提供了新方法。研究结果表明,神经网络的组合可以很好的解决分布函数这个高维数据集合的建模问题。该文以期望的温度分相与实测温度分布的误差集合为控制指标,通过最优化方法求解,得到了最优控制序列。以实验室聚合反应器为背景,对上述方法进行了仿真实验,并针对存在噪声和施加干扰的环境,提出了自适应的控制策略,获得了预期的研究结果。

Study of Temperature Distribution Control in Tubular Polymerization Reaction

The dynamic mathematic models of temperature distribution in tubular polymerization reaction have been developed utilizing combinative B-spline neural networks, that provide a new modeling approach for a kind of distributed systems. The researches have demonstrated that the combinative neural networks are feasible in modeling high-dimension data distribution. Taking the error sets between desired temperature distribution and sampled ones as the criterion of control, an optimal control sequence is deduced and applied to a series of simulative polymerization experiments. Aimed at conditions with noises and disturbances, the adaptive control strategies are put forward, and expected results have been obtained.