基于梯度提升树的飞机机身对接状态识别

为了实时监控飞机机身的对接过程，针对机身对接数据没有标注和样本不平衡的特点，提出基于梯度提升树（GBDT）的机身对接状态识别方法. 通过定位器及定位器上的载荷传感器，实时获取机身对接过程中的位移和载荷数据. 结合飞机部件对接的工艺流程对历史对接数据进行状态标注，提出准确、高效的对接状态自动标注方法. 在经过标注的对接数据上训练基于GBDT的机身对接状态识别模型，通过该模型可以获得各个特征的重要性. 与长短期记忆网络（LSTM）、卷积神经网络（CNN）以及一些传统机器学习方法相比，该方法对接状态识别的宏F₁（macro_F₁）指标高达0.998，能够精准地识别每一种对接状态且训练速度较快.

State recognition for fuselage join based on gradient boosting tree

A state recognition method for fuselage join based on gradient boosting decision tree (GBDT) was proposed by considering the practical conditions of the lack of label and sample imbalance of fuselage joining data in order to monitor the process of fuselage join in real time. The displacement and the load data were acquired in real time through the positioners and the load sensors during the process of fuselage join. The joining state of historical data was labeled based on the process of airliner component join, and an accurate and efficient automatic labeling method for fuselage joining state was proposed. The state recognition model for fuselage join based on GBDT was trained through the labeled data, from which the importance of each feature was obtained. The macro_F₁ for joining state recognition of the proposed method was as high as 0.998, compared with the latest deep learning methods such as long short-term memory (LSTM), convolutional neural network (CNN) and some traditional machine learning methods. Each joining state was accurately recognized, and the model training process was more efficient.