支持向量机在产品成分估计中的应用研究

Research on Application of Support Vector Machine in Product Component Estimation

化工产品成分的在线分析对于实施有效的过程控制非常重要,但是传统的方法都存在明显的缺陷,软测量技术为解决这一问题提供了新的途径。采用支持向量机对萃取精馏塔工艺过程中产品成分进行估计,给出了建模过程以及参数选取的原则与方法。与神经网络和多元回归方法相比,支持向量机能更好地克服噪声数据对建模的影响,泛化性能好。支持向量机在软测量系统中的深入发展应用,为化工过程的先进控制与优化运行提供了良好的基础。

Online analysis of chemical product components is vital to carry out effective process control. However, there are obvious limitations in traditional solutions respectively. So a new methodology for solving the problem was presented by using soft sensing technology. The product component in an extraction distillation tower was evaluated by adopting support vector machine. The modeling process as well as the principle and the method of parameters selection were presented, in comparison with neural networks and multiple regression, the influence of noise to modeling can be overcome by using support vector machine, and it has good generalization performance. Thus a solid foundation of advanced process control anti optimal operation of chemical processes have been provided by further development and application of support vector machine in soft sensing.