This project addresses machine cutting of threads, a key step in reliably manufacturing many products. The research intent is improving the process for enhanced accuracy. This will be accomplished by using nonlinear piezoactuators controlled by neural nets to obtain 5 arc minutes accuracy in positioning a cutting tool. Specifically, in Phase I, neural networks will learn to control a highly nonlinear set of piezoactuators as simulated on a computer, whereas in Phase II, a machine tool will be instrumented to demonstrate the concept in hardware. Tool positioning information will be supplied by a laser measurement device already available. There are many types of metal cutting operations where this technology could be applied to reduce waste and improve quality. The utility of the neural network is apparent since it can learn to compensate for many repeatable variables in the machining operation and also compensate for tool wear.
