Model based analysis has the potential to facilitate maintenance of complex real-time systems, as it allows for impact analysis with respect to the systems' temporal behavior. Model based analysis of temporal behavior of a legacy real-time system has also the potential to support migration toward component based system. However, since most software systems today have been developed in a traditional, code oriented manner, sufficiently detailed models are typically not available. To apply model based analysis on these systems, models have to be extracted from their implementation and observed run-time behavior. This requires methods for model validation. The paper proposes a novel method for model validation and presents a framework for evaluation of model validation methods, which will be used to evaluate the proposed method. The method is targeting temporal models extracted from complex real-time systems.