•A hybrid predictive maintenance method for CNC machine tools driven by Digital Twin model and Digital Twin data is proposed.•Digital Twin model is built in multi-domain and reflects the actual working conditions; Digital Twin data is gathered by different types of sensors and used for data-driven remaining useful life prediction model; then the system observation value and theoretical derivation value are fused by particle filtering algorithm. The validity and accuracy of the proposed method are verified by cutting tool life prediction of CNC machine tools.•This method enables the Digital Twin model and data to be better integrated and applied, thus can give out a more accurate and intelligent result than before. As a typical manufacturing equipment, CNC machine tool (CNCMT) is the mother machine of industry. Fault of CNCMT might cause the loss of precision and affect the production if troubleshooting is not timely. Therefore, the reliability of CNCMT has a big significance. Predictive maintenance is an effective method to avoid faults and casualties. Due to less consideration of the status variety and consistency of CNCMT in its life cycle, current methods cannot achieve accurate, timely and intelligent results. To realize reliable predictive maintenance of CNCMT, a hybrid approach driven by Digital Twin (DT) is studied. This approach is DT model-based and DT data-driven hybrid. With the proposed framework, a hybrid predictive maintenance algorithm based on DT model and DT data is researched. At last, a case study on cutting tool life prediction is conducted. The result shows that the proposed method is feasible and more accurate than single approach.