One of today's challenges in the field of modeling and simulation is to model increasingly larger and more complex systems. Complex models take long to develop and incur high costs. With the advances in data collection technologies and more popular use of computer-aided systems, more data has become available in many organizations. This often allows for and requires a certain degree of automation in modeling. The research presented in this dissertation studied how to automatically generate simulation models. The method proposed uses domain specific model components as building blocks for model generation and applies graph transformation based algorithms to compose large simulation models according to the existing data. The method has been applied practically in the domain of light-rail transportation.