In order to cope with the increasing and conflicting demands in terms of pollution abatement, drivability and fuel economy, modern combustion engines have become complex systems with many actuators. However, this increase complicates the generation of mathematical models, necessary for optimization, control and on-board diagnosis of the engine. Models representing engines or engine subsystems (e.g. engine airpath, emission formation, etc.) usually contain several inputs and are nonlinear. Hence, if models are generated by means of data based identification, application of Design of Experiment (DoE) approaches is essential to increase the usable information in the data and moreover to reduce measurement costs. This article presents a general discussion and experimental results for different engine applications, namely an iterative nonlinear method to model particulate matter emissions of a passenger car Diesel engine, a backward DoE method which omits poor data points in order to improve the estimation quality and a smart utilization of DoE for direct controller parameterization.