In engine research and development there are often different engine parameters that produce similar effects on the endpoint results. As a consequence, an optimum for a single response regression model can be reached by using many different engine parameter settings. Although multiple responses can be used together with engine design constraints to obtain a unique optimum, interactions between the engine parameters makes these regression models imperfect. When calibrating modern engines, a large number of model parameters need to be set, including compensation parameters for model imperfections. Therefore simpler, more robust, physically based models are beneficial both for powertrain calibration and prediction performance. In this study, we present a new approach for DoE called Intrinsic Design of Experiments (IDoE) that is based on the concept of intrinsic variables instead of engine parameters. The aim is to adapt the DoE to the phenomena of interest and hence create models capable to better capture the underlying physics. The proposed methodology is illustrated for a heavy-duty diesel engine with Variable Valve Timing (VVT). The engine parameters were: boost pressure, IVC timing and EGR. Their intrinsic counterparts, derived from simplified thermodynamic models, comprise gas density and temperature at the start of injection, and oxygen concentration. The injection system parameters including the Fuel Injection Pressure (FIP) and Start of Injection (SOI) were also varied in the DoE. Regression models were made using both engine parameters and intrinsic variables to predict indicated specific fuel consumption (ISFC). The resulting regression coefficients were compared and analyzed. The results show that the regression models based on IDoE needed fewer interaction terms and more quadratic terms to achieve the same fit. Therefore, employing IDoE facilitates the optimization process. Furthermore, the usage of intrinsic physical properties makes the models using intrinsic variables more general and hence suitable for a robust powertrain optimization.