This paper aims to propose a digital twin-driven (DTD) approach that consists of the machining data (MD) in twin data (TD), design of MD acquisition methodology, construction of intelligent algorithm, real-virtual data interaction analysis and fusion technology, which improvs the predictability and management of on-line quality control of marine diesel engine (MDE) critical parts. Firstly, this paper introduces the theoretical framework of DTD on-line quality control in machining process. Secondly, we construct the process of DTD on-line quality control and introduce the digital twin model of on-line quality control based on TD-driven; the operation of data-driven quality on-line control based on digital twin including description and modeling of MD; acquisition of MD based on digital twin; TD-driven on-line tool life prediction and data fusion on-line machining parameters optimization methods. Finally, a case study is applied to validate the accuracy and availability of the DTD approach. The proposed approach provides a new way for the on-line quality control of MDE critical parts in machining process.