Software product line engineering is a promising paradigm for developing software intensive systems. Among their proven benefits are reduced time to market, better asset reuse and improved software quality. To achieve this, the collection of products of the product line are specified by means of product line models. Feature Models (FMs) are a common notation to represent product lines that express the set of feature combinations that software products can have. Experience shows that these models can have defects. Defects in FMs be inherited to the products configured from these models. Consequently, defects must be early identified and corrected. Several works reported in scientific literature, deal with identification of defects in FMs. However, only few of these proposals are able to explain how to fix defects, and only some corrections are suggested. This paper proposes a new method to detect all possible corrections from a defective product line model. The originality of the contribution is that corrections can be found when the method systematically eliminates dependencies from the FMs. The proposed method was applied on 78 distinct FMs with sizes up to 120 dependencies. Evaluation indicates that the method proposed in this paper scale up, is accurate, and sometimes useful in real scenarios.