Recently, in order to increase the autonomy of machines, the knowledge engineering is being applied widely in the field of the electrical machining. The fuzzy control is known well as one of the most effective ways for the knowledge engineering to maintain the stability of the non-deterministic machining such as die-sinking EDM [Int. J. Mach. Tools Manuf. 43 (13) (2003) 1287; J. Mater. Process. Technol. 129 (1–3) (2002) 45; Modern Mach. Shop 70 (7) (1997) 262; IEEE Trans. Automatic Control 35 (2) (1990) 148; Modern Mach. Shop 66 (5) (1993) 130]. Even the fuzzy control is powerful, the ability of the machine would never grow as long as the fuzzy rules are fixed. This investigation was carried out to examine the validities of a self-adjusting fuzzy control on die-sinking EDM. The stability of machining could be calculated properly by using a simplified fuzzy inference with only two input signals such as a frequency of short circuit and a frequency of arcing. And by controlling the jump heights and the approaching movement of the tool electrode according to the calculated stability, a remarkable improvement in machining speed and the maximum depth of cut was achieved.