Principal Component Analysis Based Kullback- Leibler Divergence for Die Cracks Detection

Die cracks are a vital issue that directly influences the quality of chip assemblies. In this paper, we focus on detecting die cracks using principal component analysis (PCA) and Kullback-Leibler (K-L) divergence. Our method involves data fusion, including three steps: 1) apply PCA to convert high-dimensional data to low-dimensional data; 2) obtain the frequency distribution histograms of the transformed data and fit them; 3) use K-L Divergence based state index to quantitatively evaluate die cracks. Our method works very well with real-life data. Die cracks are identified according to die cracks data showing skewed distribution, while normal data have Gaussian distribution. Moreover, the proposed state index could successfully detect die cracks.