To keep quality machining against tool wear or breakage, an effective approach to tool wear status recognition is proposed, leveraging deep learning and data fusion. The data fusion method fuses time-series data from multiple sensors into two-dimensional images, utilizing a triangular matrix of angle summation. It performs the data fusion without redundancy and effectively preserves the time relationship between multi-sensor data. The deep residual convolution network facilitated with attention mechanism is employed to extract deep features from the image and to effectively recognize tool wear types. The network takes advantages of the attention mechanism in selecting the information in both the channel domain and the spatial domain, and combines the residual block to deepen the network so as to extract the deep features. Finally, a cutting experiment is carried out for verification of the proposed method, in which the model is trained with multi-sensor data. The findings indicate that the accuracy for recognition of tool wear types is as high as 93.40%.