Abstract
LCD has been the most developing industry in Taiwan in the latest decades and is facing not only quality and cost problem when production increase and enlargement of panel size but also the fierce competitions come from other countries. Array process in LCD manufacturing is similar to semiconductor manufacturing process, where multivariable analysis has been applied on the monitoring and fault detection and has numbers of successful cases regarding process diagnosis and control. Splash is fine melt particles formed in the sputtering process which reduce quality of products. And more than one factor lead to this defect. Analysis of residual gas is used to attribute the root factor of splash. First, it is necessary to decrease the dimensions of such a considerable dataset in batch process. Variables are chosen based on operation recipe or the physical meaning and separated to mean part, time scaling and residual. Obvious decay trends are discovered in time scaling, which is also regards as an important factor causing splash. A key residual gas V13 largely increases after preventive maintenance (PM) of equipment and lead to splash. Moreover, when V13 is back to a stable state, other two residual gases with high value are found leading to a high probability of splash. Besides, a multivariable analysis is applied on the residuals to examine that no incorrect operation causes splash.