Initially, relevant a priori knowledge is collected and a set of independent variables (e.g., physical quantities) is determined which defines the investigated system (parameterization of the system). Next, one or more mathematical model structures are proposed that are able to predict the system’s behavior (forward modeling). A precondition is that the models are structurally identifiable, i.e., all model parameters can be identified uniquely. In the third step, the predictive quality of the models is improved by adjusting the intrinsic model parameters to the actual measurements (parameter estimation). Subsequently, the validity of its structure and parameter estimates is tested against new experimental data. When the descriptive quality of the model is insufficient, an adaptation of the model structure and/or parameter estimates is needed. As such, model structure selection and parameter estimation are combined in an iterative model building cycle. Finally, a model is obtained that agrees with reality as closely as possible.