Classification of DC micro spot welding quality using fuzzy ARTMAP on acoustic emission monitoring

This paper presents a fuzzy ARTMAP to classify quality of nugget formation in DC micro spot welding process using online extraction of AE parameters. The fuzzy ARTMAP is proposed to classify the quality of nugget formation into one of three levels: "weak", "good", and "excessive". It is chosen over feedforward neural networks due to its appealing properties. These include automatic selection of network structure, convergence properties and its online learning. An experiment was conducted to show its performance. Peel and metallographic tests plus spatter exploding observation were used to identify the quality of nugget formation. The result shows that the approach performs well. In addition, the performance can be enhanced greatly if spatter exploding observation is used in combination with the AE parameters.