Virtual metrology apparatus, virtual metrology method, and virtual metrology program

A virtual metrology apparatus, comprising:

• a memory;
• and
• one or more processing circuits coupled to the memory and configured to:
• acquire a time series data group measured in association with processing of a first target object in a predetermined processing unit of a manufacturing process;
• and
• cause a plurality of network sections to process the acquired time series data group, cause a connection section to consolidate output data produced by the plurality of network sections, and obtain, as inferred inspection data of a resultant object obtained upon processing the first target object, virtual metrology data output from the connection section,
• wherein the plurality of network sections and the connection section are trained by machine learning, the machine learning ensuring that virtual metrology data output from the connection section upon using the plurality of network sections to process a time series data group acquired in advance approaches inspection data measured from a resultant object obtained upon processing a second target object in a predetermined processing unit of a manufacturing process,
• wherein each of the plurality of network sections includes a plurality of layers, and processes the acquired time series data group in mini-batches, and
• wherein a last layer of the plurality of layers includes a pooling part, and the pooling part performs a GAP (global average pooling) process for each of the mini-batches, thereby causing an output from each of the plurality of network sections for a given one of the mini-batches to be an average obtained for the given one of the mini-batches.

A virtual metrology method, comprising:

• acquiring a time series data group measured in association with processing of a target object in a predetermined processing unit of a manufacturing process;
• and
• training, in a training unit including a plurality of network sections configured to process the acquired time series data group and a connection section configured to consolidate output data produced by the plurality of network sections to generate virtual metrology data, the plurality of network sections and the connection section by machine learning, the machine learning ensuring that the virtual metrology data output from the connection section approaches inspection data measured from a resultant object obtained upon processing the target object in the predetermined processing unit of the manufacturing process,
• wherein each of the plurality of network sections includes a plurality of layers, and processes the acquired time series data group in mini-batches, and
• wherein a last layer of the plurality of layers includes a pooling part, and the pooling part performs a GAP (global average pooling) process for each of the mini-batches, thereby causing an output from each of the plurality of network sections for a given one of the mini-batches to be an average obtained for the given one of the mini-batches.