Control Chart Pattern Recognition Using Associated Rules and Optimized Classifier

Control chart pattern (CCP) recognition is an important issue in statistical process control because unnatural control chart patterns exhibited on control charts can be associated with specific causes that adversely affect the manufacturing processes. In this paper, a hybrid method is introduced to CCPs recognition. In the proposed method, radial basis function neural network (RBFNN) is used as intelligent classifier and shape features are used as effective features. In each pattern recognition problem, the dimension of input data has vital role in classifier performance. Therefore, in the proposed method, the association rule (AR) technique is used to select the effective features and remove the redundant features. Also, in RBF neural network, free parameters such as spread and number of radial basis function have high effect on network performance. Therefore in the proposed bee’s algorithm (BA) is used to select the best values of these parameters. The proposed method is tested on real world data and the obtained results show that the proposed method has excellent recognition accuracy

Keywords:

Spread , RBF , BA , CCP , Association rules