Identification and precise optimization of key assembly error links for complex aviation components driven by mechanism and data fusion model

Abstract

As assembling complex aviation products, due to factors such as part deformation under loads, numerous process parameters, and complex error transmission path, the effective identification and optimization of key error links that affecting assembly accuracy significantly is challenging. In this paper, a mechanism and data fusion method for solving this problem was proposed. Firstly, the geometric-physical coupling relationship among composite thin-walled parts and the entire locating/clamping/joining/rebounding operations was analyzed. Then with the actual error information, the Jacobian-torsor matrix that representing error accumulation relationship was modified, and assembly error was calculated with the mechanism model. Secondly, with actual data processing solution to obtain the deviation of theoretical calculation results, the fusion model of integrating mechanism and data analysis results was proposed for predicting the final assembly accuracy. Subsequently, with massive data samples from the fusion model, the Sobol method was adopted to gain the global sensitivity coefficients of different error elements, and the key error links could be identified. Thirdly, with the accurate error fusion results, three single tolerance optimization models for the entire production process were established, i.e. manufacturing cost, assembly quality loss and repair cost. Then a weight parameters design method was proposed, which can avoid the conflict phenomena of data imbalance and optimization deviation problems among different goals, and the multi-objective tolerance allocation model was solved with intelligent algorithm. Finally, for the assembly work of wing-box component, key error links that having an obvious impact on the profile gap and step difference accuracy were identified and optimized, and beneficial quality/efficiency results were gained. This research could provide a strong interpretability for assembly accuracy analysis results, and a good applicability to practical assembly site.