A number of experimental studies on burr formation in face milling operations have been pursued. They usually focus on machining parameters such as depth of cut, leed rate, spindle speed and in-plane exit angle. But it if difficult to set the correlation between burrs and the parameters on burr when such parameters are considered at the same time. Therefore, in this paper, acoustic emission (AE) is considered as an alternate way to predict burr types regardless of machining conditions. AE signals during face milling were sampled and processed, then fed into an artificial neural network to classify burr types \"on-line\".\".uot;.