A unified outlier detection method for trace data analysis in semiconductor manufacturing

Abstract

Process trace data (PTD) is an important data type in semiconductor manufacturing and consists of a huge amount of time series data collected from many different sensors during all manufacturing processing steps. PTD contains abundant information about the tool status and thus can be used for improving tool stability and tool matching. For this aim, some industrial applications have been developed on the basis of PTD analysis, but the existence of outliers adversely affects the effectiveness of these applications. Due to the complexities of PTD and the resultant outlier patterns, this paper proposes a unified outlier detection method which takes advantages of data complexity reduction using entropy and the cumulative sum (CUSUM) method. A novel algorithm is developed to meet the practical requirements of PTD analysis by taking into account of the related domain knowledge, and its effectiveness is demonstrated by using real process trace data sets.