Crankpin non-circular grinding progress error forecast and compensation based on RBF-NN

Crank shaft non-circular grinding is a new method of crank shaft processing; it uses X-C two axes synchro-motion method to grind the crank journal and crankpins on one time clamp with excellent flexibility. Different angle of the crank shaft has different stiffness and also the grinding force is always varying in machining process, which will affect the roundness of the crank pin. If machining crankpin by non-circular grinding method just according to the theory equation without compensation, the roundness is hard to be assured. Combined the motion model with compensation, this paper uses RBF neural networks to predict the errors at different angle of the crankshaft. This method is applied to the numerical control machining compensation of the H405BF machine tool. Experiment and simulation results show: using RBF neural networks can predict the errors in machining process comparative exactly, which solves the difficult problem of error compensation in crank pin non-circular grinding process and also assures the quality of crank pin grinding.