The development of mechatronic systems increasingly focuses on dependability aspects as reliability and functionalsafety. Model-based development of mechatronic systems enables for extensive dependability analysis at early designphases. During operation phase, state-of-the-art condition monitoring algorithms as well as methods of advanced controltheory ensure an effective dependability investigation. The models used for development of mechatronic systems includesubstantial information about the architecture, behavior and dependability of the investigated system. Based on thesemodels, a digital twin for use in predictive maintenance can be set up and combined with condition monitoring data ofthe real system. The integration of those models into a digital twin offers extensive potentials and supports its setup. Thepaper at hand describes and illustrates the integration of these models and shows the potentials that arise from the use ofdigital twins for predictive maintenance.