In the electricity production field, optimization of power plants maintenance operations is a major issue. In order to insure the longest power plants availability, and to avoid incidents, some of these operations have to be simulated with computers before they are performed. To do this, we need a C.A.D. model which describes the actual state of the power plant : such a model is called a As Built Model.
Industrial CAD models (e.g. power plant models) are made of primitive shapes such as planes, spheres, cylinders, cones and tori. This kind of description carries a high level information, and really fits the industrial sites components (pipes, walls, ...)

Laser scans in power plants generate point clouds that are used to keep As Built C.A.D. models up do date. The current process which is used for model reconstruction from point cloud data is not automated in practice, thus it takes a lot of time and is tedious. We aim to speed-up this process by automating it. Given the fact there are existing C.A.D. models that almost fit the point clouds, we propose to use these models to guide the point cloud reconstruction.