Abstract: A reliability evaluation method based on grey neural network modeling is put forward to tool reliability evaluation of missing data in failure sample.The vibration signal and tool wear is obtained by online experiment.Wavelet packet decomposition,time domain statistics and correlation analysis are used to extract significant features.The obtained feature is wavelet energy entropy,seventh band and ninth band wavelet energy,seventh band mean,RMS amplitude,RMS value,standard deviation.Building grey neural network mode.Take 7 significant features as input,take tool wear as output to train network model.Enter the random sample data,to obtain the tool wear data.The pseudo failure life is determined by the preset failure threshold.At last,to establish and complete the Weibull distribution model of reliability assessment.Tool simulation experiment results show that grey neural network model has higher prediction accuracy compared with the grey forecasting model.

Key words: reliability evaluation, grey neural network, the Weibull distribution, features, tool wear, prediction