Recognition method for control chart patterns based on improved sequential forward selection and extreme learning machine

Control chart is one of the important statistical process control tools. Abnormal situations and the potential quality problems in the production process can be judged and revealed according to the state of control chart. Thus the recognition of control chart is of great importance. To improve patterns recognition performance of control chart, a new method based on improved sequential forward selection (ISFS) and extreme learning machine (ELM) was presented. Firstly the 13 time domain features were extracted from control chart; secondly, the improved sequential forward selection method was used to select the features to reduce the relevance and redundancy between features and improve recognition rate; finally, ELM was adopted to identify control chart. Experimental results show that the proposed method can achieve a significant classification performance with accuracy of 98.7%, providing a new method for the control chart recognition.