Tool life prediction is of great significance for improving the machining accuracy of workpieces and the processing efficiency of production. The data distribution of tool monitoring signals under the same working conditions and models is inconsistent, resulting in the limited effect of historical life prediction models on tool life prediction. This paper proposes a dynamic tool life prediction method based on the deep convolutional neural network (DCNN). Firstly, the DCNN is used to mine the degradation trend characteristics of historical tool monitoring signals, a tool life prediction model is built, and an attention mechanism is added to weight the DCNN output to strengthen the tool life characteristics learn to improve the accuracy of life prediction. Then, by detecting the inconsistency of the tool monitoring signal data distribution based on the KL divergence, iteration is performed on the basis of the existing tool life prediction model. Finally, the iterated model is used to perform tool life prediction again. The method better reflects the influence of the actual machining process of the tool on the tool life. Taking the milling data set as an example, the effectiveness of the method is verified.