化工故障诊断中基于遗传算法优化的SOM网络

针对自组织映射神经网络(SOM)启发式训练算法中的缺陷,采用遗传算法SOM中的权重失真指数(LWDI),形成基于遗传算法优化的SOM(GA-SOM)训练算法。将GA-SOM算法应用于化工过程故障诊断,以某工厂甲醇合成反应器故障数据样本为研究对象,研究结果表明,对比基本SOM算法,GA-SOM算法对故障数据能够得到较优的分类辨识结果,且该算法实现简单,便于工程应用,对甲醇合成生产中的故障诊断有非常显著的指导作用。

Genetic Algorithm Based SOM for Fault Diagnosis in Chemical Process

Direct optimization of "locally weighted distortion index" by Genetic Algorithm(GA) is substituted for Kohonen's heuristic-based training algorithm in SOM.The new GA-SOM algorithm is applied practically to identify and classify fault data of methanol synthesis reactor.The emulational experimental results show this algorithm can deal with complicated data,obtain better classification result and identify fault's type correctly than basic SOM algorithms,at the same time its realization is fit for scientific calculation and engineering application.And classification result can more greatly direct the optimization of methanol synthesis reactor parameters and yield monitoring.