The prediction of remaining tool life under complex working conditions has become the key to guarantee the machining quality and machining efficiency. In this paper, MCWDCNN is constructed as a milling cutter wear state identification model, and the vibration data of plane milling is used for model validation. For plane milling data, the envelope spectral information of the two signal processes in one cycle of forward milling and reverse milling is used as the input data of the model dual-channel, and the tool life prediction model under complex working conditions is constructed through the BiLSTM and attention mechanism, and the vibration signal sample data set is constructed by using the tool health factor and the RMS as the labels, respectively, and is inputted into the tool life prediction model, and the result has a higher The results have a high degree of fit, which verifies the effectiveness and generalizability of the tool life prediction model.