FANUC's new AI functions that utilize machine learning and deep learning
FA: AI Servo Monitor :
A mechanical breakdown caused by sudden malfunction of the spindle axes or feed?axes of a machine tool could lead to major problems such as a long-term suspension of a machining line. In order to prevent this, it is necessary to detect signs of anomaly in the spindle axes or feed axes before a malfunction occurs.
FANUC and PFN have developed a new AI function called AI Servo Monitor which collects control data of feed axes and spindle axes of machines with high-speed sampling. It applies deep learning to the collected data and shows the anomaly score based on the current state of the machine components.
AI Servo Monitor trains a model using torque data from motors as input while the machine is operating normally. The trained model has extracted features of the torque data and can represent its normal state. During the machine’s actual operation, AI Servo Monitor takes the torque data as input and compares it with the normal state to calculate and display the anomaly score. By monitoring this, machine operators can observe a symptom indicative of malfunction in the feed axes or spindle axes as they work with the machine tool.
AI Servo Monitor notifies operators before failure relating the feed axes or spindle axes, allowing for maintenance to be performed. This will contribute to the improved availability of machines.
Robot: AI Error Proofing:
FANUC and Preferred Networks, Inc. (PFN) has introduced a new AI Error Proofing function designed for part inspection using machine learning technology.
FANUC robots equipped with the new function can check and determine whether a part is good or bad, based on example images of good and bad parts. AI Error Proofing does not require an external PC because it is implemented directly on the FANUC robot controller as part of FANUC’s integrated vision system - iRVision. The new function is ideal for various inspection processes in manufacturing, such as checking presence of assembled parts or weld nuts, and part orientation verification.
For conventional machine vision to check whether a part is present or not, it would make decisions based on being able to or not being able to detect a pre-taught part’s shape and position. However, this method is often affected by spatter or soot on parts, halation of images due to reflection of metal surfaces, which could lead to false results and require skilled expertise to optimize vision settings.
FANUC’s AI Error Proofing does not try to detect a part’s shape or position, but utilizes machine learning to determine whether the image itself is good or not, enabling a much more robust inspection against fluctuations of the environment or halation. It also allows for higher accuracy inspections by providing several to a few dozen image data sets and teaching it which should pass and which should fail - all without having to do detailed vision parameter tuning.