This paper presented tries to improve finishing accuracy through monitoring the lateral gap-width by an adaptive fuzzy logic controller. Since the complexities and stochastic properties inherited in EDM processes, it seems promising to introduce soft computing algorithm in such a process control application. The adaptive fuzzy logic system (AFLS) composed of a fuzzy logic system and a nearest neighborhood-clustering algorithm, which adjusts not only the parameters of fuzzy system but also its own structure. A detailed material removal mechanism was established through the concept of discharge-angle for analysis on the machining processes. Formation of lateral gap width was analyzed through material removal swept by the discharge angle. Combined with discharge power and other operation conditions, a feasible gap-width estimation model was established. And an adaptive control based on the AFLS was proposed to improve the conventional servo-voltage strategy in both the machining accuracy and the stability aspects. Experiment results showed that the proposed estimation model is effective, and the proposed control strategy help significantly in improving the geometric accuracy of Wire-EDM’s finishing process.