Orthognathic surgery is used to treat misaligned jaws in adults by repositioning them. Intervention planning has to take into account different clinical, anatomical, functional, and aesthetic parameters to determine the optimal position. Current planning systems usually present 3D surface models obtained from a CT or Cone-Beam scan of the patient. The surgeon then interactively positions the upper and lower jaw. Thus, the surgeon can manually optimize anatomical aspects but has to consider functional and aesthetic effects simultaneously, which may be error-prone. We are developing a computer-assisted planning system which generates an optimized position for both jaws based on different analyses (cephalometric, plaster model, photostat) of the head, using a gradient descent algorithm. For this purpose, landmarks are interactively identified on a 3D surface representation. The system is developed as a plugin for MITK utilizing a knowledge base realized in the sematic web standard RDFS, which is queried with SPARQL requests. In a preliminary evaluation with five different cases we compare the automatically generated planning proposal with the planning results of a maxillofacial expert (ground truth). Good general agreement is observed, although more research for the identification and development of 3D cephalometric analyses is needed.