Feature-Based Control Chart Pattern Recognition

Accurate monitoring and control of the manufacturing processes has become very important In today's dynamic world due to rapid Increase In demand for highly precise end products. For every process, there is a target and the goal of statistical process control (SPC) is to ensure that the process mean lies nearer to the target value without inflation o f process variability. Control charts are widely used to identify the situations when control actions will be needed for the manufacturing processes. Control charts usually exhibit one of the eight basic types of patterns. Identification of these patterns leads to more focused diagnosis and significantly minimizes the effort towards troubleshooting. Pham and Wani presented a feature-based heuristic approach for control chart pattern (CCP) recognition, which has been very appealing to the shop-floor people, because In this approach, the practitioners can clearly understand how a particular pattern has been identified by the use o f relevant features. The heuristics proposed by Pham and Wani can only differentiate six types of CCPs based on extraction of nine features. In this paper, a new set of nine features is proposed and the heuristics for CCP recognition based on these features is also presented that can efficiently differentiate all the eight basic types of CCPs. The features are chosen such that the need for human intervention for their extraction is eliminated and thus the CCP recognition is totally automatic. The proposed feature-based CCP recognition approach can be applicable to any process operating on the known target value and will be considerably robust against the variation of true process standard deviation from its estimated value. Thus it has enough potential for use in real-time process control applications.