Manufacturing process control through integration of neural networks and fuzzy model

Abstract

Artificial neural networks (ANNs) and fuzzy logic have been widely applied in many areas. This research is trying to discuss the integration of these two technologies. Three fuzzy models are utilized to update dynamically the training parameters in order to speed up the training. In addition, a fuzzy model is proposed which is self-organizing and self-adjusting, and able to learn from experience. In a self-organizing and self-adjusting fuzzy model (SOSAFM), the inputs and outputs are partitioned by Kohonen's feature mapping and the premise and consequence parameters are updated through an error backpropagation (EBP)-type learning algorithm. Physical experiments for manufacturing process control are implemented to evaluate the proposed methods. The results showed that updating the training parameters by using fuzzy models can accelerate the training speed. Moreover, SOSAFM is better than the multiple regression and artificial neural network both in speed and accuracy for the purpose of multi-sensor integration.