Detection and Prediction of Chipping in Wafer Grinding Based on Dicing Signal

Abstract

Simple regression cannot wholly analyze large-scale wafer backside wall chipping because the wafer grinding process encounters many problems, such as collected data missing, data showing a non-linear distribution, and correlated hidden parameters lost. The objective of this study is to propose a novel approach to solving this problem. First, this study uses time series, random forest, importance analysis, and correlation analysis to analyze the signals of wafer grinding to screen out key grinding parameters. Then, we use PCA and Barnes-Hut t-SNE to reduce the dimensionality of the key grinding parameters and compare their corresponding heat maps to find out which dimensionality reduction method is more sensitive to the chipping phenomenon. Finally, this study imported the more sensitive dimensionality reduction data into the Data Driven-Bidirectional LSTM (DD-BLSTM) model for training and predicting the wafer chipping. It can adjust the key grinding parameters in time to reduce the occurrence of large-scale wafer chipping and can effectively improve the degree of deterioration of the grinding blade. As a result, the blades can initially grind three pieces of the wafers without replacement and successfully expand to more than eight pieces of the wafer. The accuracy of wafer chipping prediction using DD-BLSTM with Barnes-Hut t-SNE dimensionality reduction can achieve 93.14%.

Ensemble Meta-Learning-Based Robust Chipping Prediction for Wafer Dicing

“…In wafer dicing, people encounter severe problems, such as loss of collected data, nonlinear data distribution, and loss of related hidden parameters. Our previous work [3] found that data sets collected from the wafer dicing process can reveal new information (hidden features) about the wafer representation, especially in wafer coverage areas smaller than 30 for a single wafer. In machine #1, the coverage area of wafer chipping ratio of 10%, 10~15%, 15~20%, and 20~30% (i.e., four groups of data sets) is equal to the number of samples in the single kerf dicing process from the beginning to the replacement of the wafer.…”

“…Our previous study [3] proposed dimensionality reduction with linear PCA [14,15] and nonlinear t-SNE [16,17]. Two of them are shown below.…”

“…Our previous paper [3] go backward into the LSTM model (denoted Backward LSTM) to learn the relationship between future data and present data. Finally, we merge the predicted results of the forward and backward outputs by weighted averaging or summing.…”

“…In our previous work [3], we collected data sheets related to wafer dicing from a semiconductor company in Kaohsiung, Taiwan. The data sheets showed the dicing position coordinates of 112 wafers; the coordinates of each dicing position marked the corresponding kerf, cutline, and channel number.…”

“…In our previous study [3], we proposed several methods to check dicing signals generated from the machine during wafer dicing that can analyze fab machines to see…”