In recent years, engine control systems have become more and more complex because of the growing pressure to develop technical innovations due to social pressures such as global warming and the depletion of fossil fuels. On the other hand, products must be launched on the market in a timely manner and at low cost. For these reasons, calibration processes have become more sophisticated. It is possible to improve the efficiency of calibration by making good use of models, and a calibration process that incorporates models is called model based calibration (MBC). MBC is a valid means of reducing the number of measurement points to some extent by statistical engine modeling and design of experiment (DoE) methodology which places measurement points in order to maximize modeling accuracy. However, it is still necessary to spend much time carrying out boundary detection testing before DoE. Especially for diesel engine calibration, boundary detection measurement is very time consuming because there are large numbers of calibration parameters. Boundary detection that maximizes the efficiency of DoE is indispensable for accurate engine modeling.This paper proposes a new methodology for reducing the steady state measurement time by predicting steady state values during on-line measurement, and introduces an application example of boundary detection measurement. As a result, this methodology is capable of shortening the measurement time while maintaining measurement accuracy.