A digital twin-based predictive maintenance framework for automation manufacturing unit(DT-PMAMU) was proposed to improve the operational reliability,safety and economy of complex equipment represented by automation manufacturing unit(AMU).General composition structure and fault characteristics of AMU were analyzed,the key factors of DT-PMAMU were established.A multi-layered predictive maintenance framework consisting of physical equipment layer,the internet of things data collection layer,digital twin data layer,virtual equipment layer and predictive maintenance service layer was constructed.The practical application of an AMU for milling proves that the framework achieves good application effect,which can help relevant personnel to quickly establish a predictive maintenance system,and effectively reduce the complexity of the AMU to achieve intelligent operation and maintenance tasks.