In order to control the quality of the energy meter manufacturing process, BP neural network is introduced as the classification tool of the abnormal topic pattern set. The longest common subsequence algorithm and the central time series algorithm are used to measure the similarity and fault features of the abnormal topic pattern set. Finally, the correlation between the seven typical fault characteristics with high frequency is analyzed to determine the cause of the increase in the defective rate. The results show that this method can focus on finding the cause of the fault and has important guiding significance for improving the quality of the energy meter.