Radial Basis Function Neural Networks for LED Wafer Defect Inspection

Wafer defect inspection is an important process before die packaging because a good yield ratio is key index to earn benefit in semiconductor manufacturing. Conventional wafer inspection was usually performed by human visual judgment. A large number of people visually examine wafers and hand-mark the defective regions. As a result, potential misjudgment may be introduced due to human fatigue. Besides, traditional method bring out a considerable personnel cost. In order to solve these shortcomings, our research intends to develop an automatic inspection system, which recognizes defective patterns automatically. The radial basis function (RBF) neural network was adopted for inspection processing. Actual data obtained from a wafer fabrication facility in Taiwan were used in experiments. The results show the proposed RBF neural network successfully identifies the defective dies on LED wafers images with good performance.