As we learned in the previous chapter, we can develop a digital twin either by using a data-driven approach or by using a physics-based mathematical formula. A data-driven approach tries to infer a statistical or mathematical model from data using advanced technologies, such as ML or DL. A physics-based approach tries to emulate the numerical model of the equipment. Both of these methods have their own pro and cons. While the data-driven approach is good for general purposes and can be applied to different equipment models, it is not always able to make decisions about new behavior that has not been acquired from the training data. The physics-based model is very specific to each asset, but it might provide inaccurate results due to the degradation of the equipment or uncalibrated data. The best thing to do is to use a combination of both approaches.