Based on the fuzzy measuring of fuzzy rule sets，a scheme system used for fault diagnosis is proposed in this paper. The extraction of fuzzy rules involves two steps: (1) the whole space of the trained data is adaptively divided into subspaces in the form of hypersphere. It is probable to solve the problems of recognition in the high dimension space. (2) A fuzzy rule in each subspace is generated and confidence of each fuzzy rule calculated. We extend the conception of the traditional fuzzy rule for the purpose of resolving the contadictions among the generated fuzzy rules. The simulation research has shown that the fuzzy measuring of the fuzzy rule sets can be used for the fault diagnosis of the liquid rocket engine.