We demonstrate how Residual-Based Fault Detection can be improved by means of Genetic-Fuzzy Systems (GFSs). Thus, the performance of a pure Data-Driven Fault Detection System, which relies on system identification models, is improved using models created by Genetic-Fuzzy Systems. The evolutionary approach is used in the cases where a deterministic training of the fuzzy systems is not able to produce good results. As such, when the deterministic optimization algorithm is trapped in local optima, GFSs are used in order to improve (fine tune) the non-global solutions using built-in genetic operators that are able to help converged solutions escape from their locality. The results are presented by means of Fault Detection Curves (FDC) -inspired by Receiver Operating Characteristic (ROC) curves-and show how, even when considering a Fault Detection (FD) system with good detection capabilities, the introduction of new, genetically evolved, fuzzy models still produces an important improvement, reflected by higher Areas Under the Curve (AUC).