Fault Classification and Identification in Chemical Processes Based on Advanced SOM Algorithm

SOM algorithm is applied to fault data classification and identification in chemical processes.Direct optimization of 'locally weighted distortion index' by particle swarm optimizer(PSO) algorithm is substituted for Kohonen's heuristic-based training algorithm in SOM.A practical application of the new PSO-SOM algorithm in identifying and classifying fault data of methanol synthesis reactor is provided.The emulational experimental results show this algorithm can deal with more complicated data,obtain better classification results and identify fault's type more correctly than basic SOM algorithms.At the same time its realization is easier and simpler and is fit for scientific calculation and engineering application.Classification results can more greatly direct the optimization of methanol synthesis reactor parameters and yield monitoring.