Stüven, OleOleStüvenMoenck, KenoKenoMoenckFlehmke, MalteMalteFlehmkeDege, JanJanDegeSchüppstuhl, ThorstenThorstenSchüppstuhl2026-10-232026-10-23202659th CIRP Conference on Manufacturing Systems, CIRP CMS 2026https://hdl.handle.net/11420/65422Drilling is one of the most frequently used manufacturing methods. In aircraft manufacturing, common passenger planes require a very large number of drill holes. Verifying whether these holes meet specifications is a labor-intensive task that could benefit from automation. However, due to the large number of holes, in-process measurement is time-consuming and can delay production, rendering it infeasible. An alternative approach is to predict hole quality using process data collected during drilling. Previous studies have demonstrated the feasibility of extracting valuable information about manufacturing processes, tools, and products from such data. However, only single features of the drill holes are predicted. In this work, we propose a method to predict the hole geometry based on force and torque time series data recorded during the drilling process. The resulting point cloud can then be used to predict multiple features or serve as the basis for a more advanced quality control process that is not solely based on verifying whether a few geometric parameters fall within specification.en2212-82712026493498Elsevierhttps://creativecommons.org/licenses/by-nc-nd/4.0/deep learningdrillinghole quality predictionmanufacturingTechnology::621: Applied PhysicsTechnology::629: Other Branches::629.1: Aviation::629.13: Aviation EngineeringComputer Science, Information and General Works::006: Special computer methodsConference Paper10.1016/j.procir.2026.03.25110.15480/882.18989