I am working on creating a traceability model for a model transformation approach, which defines trace-links between the elements of the UML model (source model) and the elements of LQN Perfromance model (target model).
You have to qualified models contents in terms of architecture layer (enterprise, systems, platforms) and knowledge management (supporting rationales).
On that basis the traceability of transformations can be supported by built-in mechanisms.
I made something similar a few year a go by using the concept of weaving models, thus by means of a traceability model I created (using ATL) a weaving model, this is a model which stores trace links in order to know which elements of the source model corresponds with the elements of the target model, the visualization tool exist and it cab be downloaded from the eclipse modelign project, was done by Marcos Didonet del Fabro. There are several examples of this. Best.