Aiming to identify fault issues in a large amount of flight test and operation data for reliability indicator calculation and evaluation, a text classification method based on convolutional neural networks (CNN) in deep learning was proposed. By collecting a large amount of manually classified flight test and operation problem texts as the experimental datasets and performing corresponding preprocessing, the Word2Vec model was used to train the problem description text into word vectors, and a TextCNN model was constructed for training to complete the classification of problem texts. Finally, the experiments show that the classification method for flight test and operation problem based on TextCNN model can provide reference for the automated classification of flight test and operation problems.