Modern development of complex embedded systems utilizes models to describe multiple different views on the same system. Consistency between these models is essential to successful development but ensuring it is in current practice often a manual effort. In this research project, we aim to develop a methodology that helps developers to maintain consistency in industrial model-based development projects by identifying inconsistencies throughout the development and maintenance of the system. For such support to be applicable in industrial practice, it should fit in with current development, i.e., should be able to identify inconsistencies between models expressed in different modeling languages and created in different modeling tools. Furthermore, the required user interaction to defining consistency checks should be minimal. This paper sketches an approach meeting these requirements, initial results towards it and discusses future research plans towards a doctoral dissertation.