Identification Of Feeding Speed Fluctuation In NC Machining Based On Wavelet Transform And BP Neural Network

In the process of NC machining,inevitably there is a fluctuation in the feed speed,and the fluctuation will reduce the processing quality.Therefore,analyzing the speed fluctuation have great significance for improving the processing quality.In this paper,a method of speed fluctuation identification based on wavelet transform and BP neural network is proposed.It aims to accurately search the fluctuant region in machining feed speed,and lays a foundation for speed optimization.This paper proposes three kinds of selection criteria of wavelet basis,and designs a comprehensive indicator to combine their advantages and balance their conflicts.The Pearson coefficients between the wavelet reconstruction signal and the original signal is introduced to be the selection criteria for the wavelet decomposition layer.This paper proposes the selection criteria of feature extraction frequency band based on energy coefficients and a method of wave region partitioning based on wavelet coefficient amplitude and position index.Wavelet coefficient energy and length of the fluctuating region are determined as characteristic parameters to represent the speed fluctuation.This paper establishes a BP neural network that is suitable for the identification of machining speed fluctuations,and optimizes the network parameters and structure based on the accuracy of identification.This paper proposes a normalized method of the characteristic parameters and a boundary processing method of the fluctuant region to improve the identification effect.In addition,a speed optimization method is introduced to optimize the speed fluctuations in impeller machining.