Time-Series Forecast-Based Endpoint Thickness Compensation for Thinning of Sapphire Wafer

Abstract

The grinder consults measurement modules to control the endpoint thickness of the sapphire wafer thinning process, but these modules have shortcomings. We propose a forecast-based endpoint thickness and online error compensation approach for grinding hard, brittle material. We leverage forecasts to improve grinding efficiency, and address the shortcomings of conventional probe gauges, which can be used only when grinding is paused, as well as those of contact gauges, which require correction from time to time due to wear. We construct a multi-signal time-series forecast model and compare the prediction performance using features from various signals. We implement the forecast model in a vertical grinder for actual grinding to evaluate the effectiveness of the proposed approach. The results show that online error compensation reduces contact times for the contact gauge and maintains precise thickness, greatly facilitating the wafer thinning process. The proposed measurement approach applies not only to wafer thinning but can also be extended to other grinding processes.