Support Vector Regression in the Analysis of Soft-Pad Grinding of Wire-Sawn Silicon Wafers

摘要

Silicon wafers, the substrates for more than 90% of integrated circuits, are widely used in semiconductor industry. A series of processes are required to manufacture high-quality silicon wafers. Surface grinding is one of the processes used to flatten wire-sawn wafers. A major issue in grinding of wire-sawn wafers is the reduction and elimination of wire-sawing induced waviness. Several approaches have been proposed to address this issue, among which soft-pad grinding is regarded as the most promising approach. But how to predict and control the waviness reduction based on certain system conditions remains to be a major problem since many factors influence the waviness reduction during soft-pad grinding. In this paper, support vector regression (SVR), a powerful tool for solving multidimensional regression problems, is applied to address this problem. It is demonstrated that SVR can be an effective approach in predicting the waviness reduction in grinding of wire-sawn silicon wafers. In addition, this paper presents the effects of five factors (pad material and geometrical properties, and grinding process parameters) on waviness reduction. The work in this paper shows that SVR is a suitable tool for the modeling and analysis of manufacturing processes where large dimensionality is involved, but only limited data is available.