Method and system for virtual metrology in semiconductor manufacturing

A semiconductor product manufacturing system, comprising:

• a manufacturing execution system (MES) configured to control a plurality of production equipments to process a plurality of wafers, during which time the MES generates process data;
• a fault detection and classification (FDC) system configured to receive real time data from a plurality of sensors each coupled to one of the plurality of production equipments, wherein the FDC includes a sensor data pre-processor mechanism configured to pre-process the real time data;
• a statistical process control (SPC) system configured to perform statistical process control on metrology data produced by a plurality of physical metrology tools when the plurality of production equipments process the plurality of wafers;
• an artificial neural network configured to process the pre-processed real time data received from the FDC system and results of the statistical process control received from the SPC system to produce and train a control model based on a training model, wherein the control model is configured to generate prediction data based on a weight approach which is automatically provided to the training model of the artificial neural network;
• an advanced equipment control (AEC) system configured to compare the prediction data with the metrology data and, if a difference between the prediction data and the metrology data exists, provide the prediction data to the SPC system, thereby further training the control model;
• and
• an advance processing control (APC) system configured to use the prediction data to adjust control of the plurality of production equipments.

A semiconductor product manufacturing system, comprising:

• a manufacturing execution system (MES) configured to control a plurality of production equipments to process a plurality of wafers, during which time the MES generates process data;
• a fault detection and classification (FDC) system configured to receive real time data from a plurality of sensors each coupled to one of the plurality of production equipments, wherein the FDC includes a sensor data pre-processor mechanism configured to pre-process the real time data;
• a statistical process control (SPC) system configured to perform statistical process control on a history of metrology data produced by a plurality of physical metrology tools when the plurality of production equipments process the plurality of test wafers, wherein results of the statistical process control are provided to the artificial neural network to train the control model;
• and
• an artificial neural network configured to process the pre-processed real time data received from the FDC system and results of the statistical process control received from the SPC system to produce and train a control model based on a training model, wherein the control model is configured to generate prediction data based on a weight approach which is automatically provided to the training model of the artificial neural network.