Research on thermal error compensation technology based on wavelet neural networks during abrasive machining

Based on the synthetic analysis of thermal error source on grinding machine and the time domain locality property of wavelet transform, a new WNN (Wavelet neural networks) method is proposed to fulfil the thermal error modelling of grinding machine. The structure and algorithm of the WNN is also proposed then. Finally, hardware system of thermal error compensation is presented, and it is tested during abrasive machining. The experimental results shows that WNN has a quicker convergence speed compared with other neural networks, which improved the machining accuracy of grinding machine effectively.