化工过程动态监控中的RBF神经网络方法研究

通过分析测量数据预测过程发展趋势，进而对过程实行监控，早已成为国内外学者所关心的热点课题．根据 ＲＢＦ神经网络训练速度快的特点，提出将其用于化工过程的动态监控．第一级网络用于预测未来一时间段内的有关状态量，第二级网络根据预测结果判断是否将会发生事故．为了在有限样本条件下取得较可靠的监控效果，提出了改进ＲＢＦ神经网络插值性能的措施，并提出对第二级网络输出结果进行变换以准确确定事故可能性的方法，以上方法被用于蒸馏塔开工过程的动态监控，结果令人满意．

A Study on the RBF Neural Network Method Used for the Dynamic Monitoring of Chemical Processes

Predicating the developing trends of a process and monitoring it on the basis of the analysis of measured data has long been a subject drawing wide attention of scholars at home and abroad. Considering the high training speed of RBF neural networks, a method based on a two-stage RBF neural network is proposed for the dynamic monitoring of chemical processes. The first stage is used to predicate the future variable values in the coming time, the second is used to forecast the faults. For the purpose of achieving reliable monitoring effects with limited samples, some measures are proposed to improve the interoperation performance of the RBF neural network, together with a transformation acting on the output of the second stage to determine the possibilities of the faults more accurately. They were applied to the dynamic monitoring for a distillation tower. The results showed a great success.