FUTURE MANUFACTURING
Producing workpiece variants correctly – selecting the right gripping program
THOMAS HAHN AND MARTINA RISTESKA
When producing different workpieces in a production cell , companies often use different robot-assisted gripping programs . Depending on the production task , the operator decides which gripper or machining program to use . If the operator chooses the wrong one , production errors occur , resulting in scrap . An artificial intelligence-based solution assists in selecting the right program .
Within an electric motor plant , 18 different variants of workpieces are processed in the production cell . The robot picks up the delivered workpieces and feeds them to the machining station . In the machining station , the one of the five different clamping devices that fits the workpiece must be attached .
To implement these requirements five different robot gripping programs were developed . The operator of the production cell decides which of these robots gripping and machining programs is to be used for a specific production task . If the operator selects a wrong robot gripping program or incorrect machining program , the workpieces cannot be gripped safely and cannot be fed correctly to the to the machine . Thus , errors in the production cell and scrap will happen . In addition , feeding a workpiece incorrectly to the machining station can cause that the entire production cell will stop .
Limits of machine learning models
A machine learning model trained with image data was developed that automatically recognizes the type of workpieces being produced and selects the correct robot gripping program . A particular challenge for the training process was the low availability of representative image data . The lack of representative training data is a typical phenomenon in industrial artificial intelligence ( AI ) applications .
Sources : Siemens AG
Gripper in action : AI-supported , the system selects the right robot gripping and machining program .
30