Inspection of IC leadframes using an unsupervised neural network

In this paper, the authors study the use of an unsupervised neural network to perform the inspection of IC leadframes. The network used is the learning by experience (LBE) type. They show the steps of varying the tolerance of acceptance whenever the network envisages some parts which cannot be classified. The Euclidean distance is used as the similarity measure. Here, two different types of unsupervised neural networks, namely adaptive resonance theory (ART2) and LBE, are compared. Experimental results on the classification of some patterns in a leadframe are also included.