There is no general consensus on how to decide if a particular design violates a model quality. In fact, we find in literature some defects described textually, detecting these design defects is usually a difficult problem. Deciding which object suffer from one defect depends heavily on the interpretation of each analyst. Experts often need to minimize design defects in software systems to improve the design quality. In this paper we propose a design defect detection approach based on object oriented metrics. We generate, using gradual rules, detection rules for each design defect at model level. We aim to extract, for each design defects, the correlation of co-variation of object oriented metrics. They are then modeled in a standard way, using the proposed UML profile for design defect modeling. We experiment our approach on 16 design defects using 32 object oriented metrics.