A Regression Analysis and Prediction Method for Assembly Error of Aviation Product Components

There are many factors that would affect the assembly error in aircraft assembly process, and the coupling rules are complex among those factors. The existing numerical methods cannot effectively analyze the assembly error, resulting in poor assembly accuracy and long development cycle. This paper aims at the small sample size of aircraft assembly site data. By measuring and analyzing the data on the assembly sites, the error modeling method and information feedback mode based on assembly site data were established. On this basis, an assembly tolerance prediction method based on artificial intelligence is proposed. Then, taking typical aircraft parts assembly as an example, the error prediction results of the proposed method in this paper and the traditional neural network method were compared. The prediction accuracy of this paper's method (0.003 mm) is obviously higher than that of the latter (0.011 mm), which verifies the effectiveness and reliability of the proposed method. Finally, the predicted assembly error probability was used to guide the adjustment of the assembly tolerance scheme, thereby improving the assembly accuracy and shortening the assembly cycle.