In operations management, the benefit of simulating manufacturing processes with data-driven models has been proven in scenario-based capacity and performance analytics. The availability of data is typically not a barrier anymore, as process parameters can be accessed and modelled relatively easily, however, the system logic representation and extraction has remained challenging. In this paper, a systematic method is presented to build prediction models for a complex manufacturing system that extracts not only the process parameters, but also the routing and operating logic. The approach combines network analytics and statistical modelling techniques to automate the model building and scenario analytics.