Research On BTA Drill Bit Wear Monitoring Technology Based On LSTM Recurrent Neural Network

As a common tool in deep-hole drilling machine,staggered-tooth BTA drill bit completes deep-hole machining of parts by cutting cutter teeth and self-guiding guide strip.The drilling process will inevitably cause wear and tear of cutter teeth and guide strip on the drill bit.The wear of drill bit will not only affect the accuracy and quality of the workpiece,but also increase the processing cost and reduce the production efficiency.So it has practical significance and practical value to study the monitoring technology of drill bit wear in deep hole drilling.According to the particularity of deep-hole drilling,a deep-hole drilling experimental platforrm for collecting the torque signal of spindle motor is established,and the wear monitoring technology of BTA drill bit based on LSTM recurrent neural network is systematically studied.The relationship between the torque signal of spindle motor and its eigenvalues and the wear of BTA drill bit during deep hole drilling is studied in time domain and frequency domain respectively.It is found that there is a strong correlation between the statistical characteristics of the torque signal in time domain and the drill bit wear;the power spectral density of the torque signal in frequency domain also shows a strong correlation with the drill bit wear,and the statistical characteristics in frequency domain show a basically consistent trend with the drill bit wear history.The multi-resolution characteristics of wavelet transform and the basic principle of Mallat algorithm are discussed.Daubechies 4 orthogonal wavelet is used to analyze the torque signal and power spectrum of the spindle motor.The wavelet decomposition signal,its statistical characteristics and the variation law of the depth while drilling are studied.From the low-frequency wavelet decomposition coefficients and the reconstructed poxwer spectrum envelope of the torque signal of the spindle motor,the features with strong correlation with the drill bit wear are obtained,which lays a foundation for the recognition of the drill bit wear state.For BTA drill bit in the condition of two kinds of wear pattern of the characteristics of distribution of overlapping,LSTM is proposed to solve the thought of drill bit wear condition monitoring,the system of the LSTM in drill bit wear monitoring is studied the basic principle and algorithm implementation process,proposed the BTA drill bit wear monitoring method based on the LSTM,respectively,using the time domain,time domain and frequency domain characteristics,the characteristics of wavelet decomposition and multiple class feature fusion sample set corresponding LSTM model is set up,and to establish a good model for training and testing,the actual results showed that:When input vectors are time-domain features,time-domain and frequency-domain features and multi-feature fusion samples,the classification effect of corresponding models can well meet the actual monitoring requirements.When the input vector is the wavelet decomposition feature sample,the classification effect of the corresponding model can also meet the actual monitoring requirements.