IntLearn ML – Machine Learning for Process Corrections

Learns from equipment process records, finds how defects such as bubbles follow temperatures and gaps, and predicts setpoints with a Transformer-based model.

Machine learning software that learns from the process data your equipment records, finds the cause of defects and predicts the next setpoints

Find the cause of defects in process data and set the line

IntLearn ML reads the process records and defect records that build up in equipment databases together, finds which process values drive the defects and predicts the next setpoints.

It shows which inputs drive the result and keeps training results as reports.

• Connect: Loads process records and defect records from the equipment database, and values from the PLC.
• Train: Trains five model types, a Transformer-based one among them, on the same data and compares them.
• Explain: Shows each input's contribution and the error on screen and in HTML and Excel reports.
• Apply: People confirm and apply the predicted setpoints, and the line moves on to autonomous control.
• Bubble correlation: Sorts the bubbles on the bubble map by type and weighs their correlation with the drying temperature and the coating gap, time lag included, to find which value adds bubbles, filtering out correlations that only come from trends moving together.
• Setpoint prediction: Predicts the next setpoints with a Transformer-based model. Autonomous control, in which the line adjusts itself, is being implemented; today people confirm and change the setpoints.