By analyzing the fault features of the auto-verification line for electricity meters, this paper develops a fault diagnosis expert system based on fuzzy theory. The structure of the expert system is introduced which has a knowledge base designed on fuzzy relationship matrix. According to the working conditions, this paper designs a fuzzy membership function to describe the symptoms. Two level thresholds are used in fuzzy evaluation of the results produced by the weighted average model. A practical example shows an acceptable result which proves that this expert system is effective.