Predictive maintenance is a technique for creating a more sustainable, safe, and profitable industry. One of the key challenges for creating predictive maintenance systems is the lack of failure data, as the machine is frequently repaired before failure. Digital Twins provide a real-time representation of the physical machine and generate data, such as asset degradation, which the predictive maintenance algorithm can use. Since 2018, scientific literature on the utilization of Digital Twins for predictive maintenance has accelerated, indicating the need for a thorough review. Hydraulic system has been the mainstream choice in large engineering equipment due to its smooth transmission, large bearing capacity, and small volume. However, because of the tightness and invisibility in hydraulic equipment, it is difficult to check and predict its faults. Common fault diagnosis and maintenance methods for the hydraulic system can be divided into two types: a signal analysis based on the mathematical model and a machine learning algorithm based on artificial intelligence. The first method can only diagnose specific faults based on the mathematical model, which is not universal, and the second one must rely on abundant history fault data, which is impossible to obtain in the early running stage. In order to address these questions, a digital twin framework is proposed which combines the virtual model with the real part to solve practical problems. As a concrete realization form of a five-dimension digital twin model, this framework provides a more feasible solution mode for fault diagnosis in the hydraulic system. Meanwhile, it expands the functions of faults prediction and digital model.