Big data analytic for multivariate fault detection and classification in semiconductor manufacturing

Abstract

Nowadays, there are more attentions on cost control and yield enhancement in the semiconductor industry. Many manufacturers have the ability to collect the physical data called Status Variables Identification (SVID) by sensors embedded in the advanced machines during the manufacturing process. To maintain the competitive advantages, process monitoring and quick response to yield problem are pivotal in detecting the cause of the faults with the help of the sensor data. To state the physical nature of certain SVID, we usually transform SVID into Fault Detection and Classification parameters (FDC parameters) using statistical indicators. The data containing FDC parameters is called FDC data. This study aims to develop a multivariate analysis model to find out the crucial factors which may lead to process excursion among a large amount of FDC data. We proposed a 2-phase multivariate analysis framework: (1) the Least Absolute Shrinkage and Selection Operator (LASSO) is applied for key operation screening. (2) And Random Forest (RF) is used to rank the FDC parameters based on the key operations. Based on the results, domain engineers can quickly take actions responding to low yield problems.