Natural gas compressor units, as the core equipment in natural gas transportations, play a crucial role in daily industrial production. Meanwhile, with the increase of equipment operating time, the performance of each component will inevitably decline and even fail in severe cases. Apparently, the unit simulation model, also known as the mechanism model with the characteristic maps, lays foundation for remotely monitoring the conditions of this rotating machine and giving the suggestions for equipment maintenance. And it should be accurate and adaptive according to the performance degradation. To achieve this goal, some thermodynamic parameters and their data samples are decided to represent this model with the Grey Relational Analysis and the optimal Latin Hypercube Sampling. Then, a data-driven model based on Gaussian process is established and the Bayesian calibration method is applied to obtain the model uncertainty, which means the degradation uncertainty in this study. This approach can realize the uncertainty quantification and the result can be adopted to judge and correct the characteristic map in the simulation model. A case study is implemented on an industrial compressor unit to demonstrate this method’s effectiveness. The corrected model outputs are compared with the actual data to acquire the model accuracy and their variance is 0.993, which points out their similarities. This approach avoids the disadvantages of the traditional model correction methods such as low utilization rate of operational data, non-use of prior information, and long fitting time. It makes full use of prior information and numerous data. Besides, it can reduce the diagnostic error, provide new possibility for the real-time monitor and promote the development of health management in the industrial equipment.