US 11,829,124 B2
Method for predicting occurrence of tool processing event and virtual metrology application and computer program product thereof
Fan-Tien Cheng, Tainan (TW); Yu-Ming Hsieh, Kaohsiung (TW); and Jing-Wen Lu, Tainan (TW)
Assigned to NATIONAL CHENG KUNG UNIVERSITY, Tainan (TW)
Filed by NATIONAL CHENG KUNG UNIVERSITY, Tainan (TW)
Filed on Dec. 24, 2020, as Appl. No. 17/134,133.
Claims priority of application No. 109118836 (TW), filed on Jun. 4, 2020.
Prior Publication US 2021/0382464 A1, Dec. 9, 2021
Int. Cl. G05B 19/418 (2006.01); G06F 18/22 (2023.01); G06F 18/243 (2023.01); G06F 18/214 (2023.01); G06F 18/2413 (2023.01)
CPC G05B 19/41875 (2013.01) [G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06F 18/2413 (2023.01); G06F 18/24323 (2023.01); G05B 2219/32194 (2013.01); G05B 2219/45031 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method for predicting an occurrence of a tool processing event, the method comprising:
obtaining a plurality of sets of historical process data, wherein the sets of historical process data are used or generated by a production tool when a plurality of historical workpieces are processed by the production tool, and the sets of historical process data are one-to-one corresponding to the sets of historical workpieces;
obtaining a plurality of historical processing event index values used for indicating if a processing event occurred when the production tool processed each of the historical workpieces, wherein the historical processing event index values are one-to-one corresponding to the sets of historical process data, and the historical processing event index values and the sets of historical process data respectively form a plurality of sets of model-building data;
performing a model-building operation, comprising:
building a classification model by using the sets of model-building data in accordance with a classification algorithm, wherein the classification model comprises a plurality of decision trees; and
building a reliance index model by using probabilities of the decision trees; and
performing a conjecturing operation, comprising:
obtaining at least one set of process data, wherein the at least one set of process data is used or generated by the production tool when at least one workpiece is processed;
inputting the at least one set of process data into the classification model, thereby obtaining at least one event predicted value used for indicating if the processing event occurs when the production tool is processing each of the at least one workpiece; and
using the reliance index model to compute a reliance index value of each of the at least one event predicted value for indicating a reliance level of each of the at least one event predicted value;
wherein the at least one event predicted value is used to actually adjust the production tool.