Using general regression neural network to determine profile of roller

The measured data from the profile of mill roll forms a shape curve which containing high frequency noise. In this paper, we use polynomial approximation and general regression neural network (GRNN) respectively to determine shape curve of roller. We not only show the good characters of GRNN through comparing the merits and disadvantages of these two methods, but also show that GRNN can also be used to catch local peak on the surface of mill roll. The simulation proves the effectiveness of the GRNN in practical use.