Control chart pattern recognition using K-MICA clustering and neural networks

Area of Science:

• Industrial Engineering
• Artificial Intelligence
• Statistical Process Control

Background:

• Manufacturing processes increasingly require automated detection of abnormal control chart patterns (CCPs).
• Existing methods may lack the accuracy and efficiency needed for real-time industrial applications.

Purpose of the Study:

• To develop and evaluate a novel hybrid intelligent method (HIM) for the automatic recognition of common control chart patterns (CCPs).
• To identify the most effective neural network classifier for CCP recognition within the proposed HIM.

Main Methods:

• A hybrid intelligent method (HIM) combining a modified imperialist competitive algorithm (MICA) with K-means clustering for data preprocessing.
• A classifier module employing various neural networks (MLP, PNN, RBFN) to determine pattern membership based on Euclidean distance from clusters.
• Experimental evaluation to select the optimal neural network for CCP recognition.

Main Results:

• The proposed HIM achieved a high recognition accuracy of approximately 99.65% for control chart patterns.
• The study successfully identified the most effective neural network architecture for the classification task within the HIM.
• The hybrid approach demonstrated superior performance in distinguishing between normal and various abnormal CCPs.

Conclusions:

• The novel hybrid intelligent method (HIM) offers a highly accurate and effective solution for automatic control chart pattern recognition.
• The integration of MICA-K-means clustering and advanced neural networks provides a robust framework for statistical process control.
• This method has significant potential for enhancing quality control and reducing defects in manufacturing processes.