Chatter will affect machining accuracy, production efficiency, tool wear and workers' health. In order to avoid chatter early, a grinding chatter online monitoring model based on multisensor fusion information and hybrid deep neural network is proposed. First, the grinding experiment of acoustic emission (AE), force and displacement multichannel signal acquisition are carried out. Then, the grinding process is divided into five stages: air cut; stable; slight chatter; severe chatter; and severe chatter with beat effect, the correlation between sensor signals and classic evaluation indicators is analyzed. Next, a hybrid deep neural network model is established, and the feature classification ability, testing accuracy, sensitivity and generalization ability of the model are studied. Finally, the proposed model is applied to microstructured grinding wheel to further verify the generalization ability and chatter prediction ability of the model. The results indicate that our approach can predict the occurrence of flutter 0.15-0.45 s in advance.