The rapid growth in printed circuit board production has led to the growing importance of microdrilling. Owing to the trend for high-density circuits on printed circuit boards, the diameter of a microdrill has to be much finer than ever before. The inspection for microdrills becomes more difficult and the detection of defects is harder. An automated visual inspection scheme for detecting the major defects of microdrills is proposed. The colour images of microdrills are first derived, and the boundary of the lip relief plane is extracted. A robust K-curvature corner detection algorithm is then used to detect the corners on the blade boundaries and to split boundaries into segments. The least-squares linear regression method is used to fit the segments into linear equations. With the fitted equations, the intersections on the boundaries of two facets are located and their distances (gap or overlap) are measured. In addition, the defects of taper and flare are identified by comparing the angle measurements with predefined specifications. Two types of microdrills are used to verify the proposed inspection process. Experimental results show that the proposed scheme reliably achieves precise inspection.

Producing silicon wafers is a complex and important process for semiconductor manufacturers. Optimally utilizing each wafer to overcome the quartz shortage is tantamount to achieving maximum total profit. This makes the post-slicing inspection task a crucial step. This paper found that determining post-sliced wafer reuse is equal to measuring the maximum inscribed circle (MIC). This can be considered a nonlinear optimization problem. Due to silicon wafer flatness and fragility, we propose a machine vision-based inspection approach that uses a heuristic method to facilitate solving this MIC problem. The proposed method incorporates the Hooke-Jeeves pattern search with particle swarm optimization (PSO) to achieve fast convergence and quality solution. An experimental design was conducted to first verify the feasibility and identify the feasible control parameters for both PSO and the proposed HJ-PSO. Thirty defective wafers were then used to validate the proposed scheme. The experimental results reveal that the proposed scheme is effective and efficient for wafer post-slice inspection.

Silicon wafer is the most crucial material in the semiconductor manufacturing industry. Owing to limited resources, the reclamation of monitor and dummy wafers for reuse can dramatically lower the cost, and become a competitive edge in this industry. However, defects such as void, scratches, particles, and contamination are found on the surfaces of the reclaimed wafers. Most of the reclaimed wafers with the asymmetric distribution of the defects, known as the “good (G)” reclaimed wafers, can be re-polished if their defects are not irreversible and if their thicknesses are sufficient for re-polishing. Currently, the “no good (NG)” reclaimed wafers must be first screened by experienced human inspectors to determine their re-usability through defect mapping. This screening task is tedious, time-consuming, and unreliable. This study presents a deep-learning-based reclaimed wafers defect classification approach. Three neural networks, multilayer perceptron (MLP), convolutional neural network (CNN) and Residual Network (ResNet), are adopted and compared for classification. These networks analyze the pattern of defect mapping and determine not only the reclaimed wafers are suitable for re-polishing but also where the defect categories belong. The open source TensorFlow library was used to train the MLP, CNN, and ResNet networks using collected wafer images as input data. Based on the experimental results, we found that the system applying CNN networks with a proper design of kernels and structures gave fast and superior performance in identifying defective wafers owing to its deep learning capability, and the ResNet averagely exhibited excellent accuracy, while the large-scale MLP networks also acquired good results with proper network structures.