Combining the aerospace terminology standardization with deep learning， and exerting the powerful semantic representation ability of deep learning in text modeling， we propose a BERT based aerospace terminology standardization method. First， the text similarity calculation method and the sequence standard method for natural language processing tasks are introduced. Second， for a certain amount of space text original word data， the candidate standard words are obtained by using Jaccard similarity calculation. Finally， the prediction standard words are obtained by constructing the pre-training model of BERT. The experimental results show that BERT has strong advantages in natural language processing tasks， and the accuracy rate on the evaluation data set based on multiple terminology standards reaches 89.93%. The proposed method can improve the standardization level and effect of aerospace terms.