Steam soft-sensing for dyeing process via FCM-based multiple models

Abstract

Aimed to the measuring problem of steam consumption in Dyeing process, a multiple neural network soft sensing modeling of Dyeing steam consumption based on adaptive fuzzy C-means clustering(FCM)is presented. The method is used for separating a whole real-time training data set into several clusters with different centers, and the clustering centers can been modified by an adaptive fuzzy clustering algorithm. Each sub-set is trained by radial base function networks (RBFN), then combining the outputs of sub-models to obtain the finial result. This method has been evaluated by a soft sensing modeling of steam consumption in Dyeing process and a practical case study. The results demonstrate that the method has significant improvement in model prediction accuracy and robustness and a good online measurement capability.