As a means to achieve optimal maintenance, the digital twin is expected to be applied to equipment diagnostic technology and remaining life prediction of machines. Since the digital twin can reproduce assumed trouble data by simulation, predictive maintenance can be performed by predicting the life of equipment using data-driven models constructed with the results or AI analysis learned by the Hybrid. This paper evaluates the reliability of physical models in a digital twin based monitoring method for reciprocating compressors. Most problems in reciprocating compressors are reported as wear and tear of crosshead-pin, piston-ring, rider-ring, etc. However, there are cases of unexpected damage, and it is often difficult to find and solve the causes of such problems. Therefore, by creating a physical model of a reciprocating compressor using Ansys motion, it is possible to generate vibration data that is close to reality by creating many examples in a virtual space, thus enabling data-driven monitoring. In this verification, the simulation results obtained from the physical model were used to represent the vibration characteristics during operation by verifying them with acceleration data from an experimental machine under conditions assuming normal conditions, suggesting the effectiveness of the digital twin monitoring method.