Modeling of random field deviation and assembled geometrical accuracy for large thin-walled structures

Purpose The assembly accuracy of large thin-walled structures is hard to be predicted due to the random field distribution of initial deviations. This paper aims to propose a new shape characterization and assembled accuracy analysis method for large thin-walled structures with random continuous field deviation. Design/methodology/approach In this paper, the random field deviation is described by the basic deviation patterns and corresponding deviation coefficients. The ranges of deviation coefficients for random field deviations are derived by the fluctuation ranges of initial deviations. The assembled geometrical accuracy indexes are calculated. A surrogate model for the prediction of assembled geometrical accuracy is constructed based on neural network method. The mapping correlation of part deviation coefficients and geometrical accuracy of assembly is established and the fluctuation of geometrical accuracy for assembled thin-walled structure is obtained. Findings The assembling process of two cylindrical panels is analyzed theoretically and experimentally, by which the method of field deviation characterization and assembled accuracy analysis is verified. The statistical distribution of assembled accuracy with random initial field deviation is obtained. The assembled accuracy indexes can be predicted rapidly with given initial deviation. Originality/value A new characterization method for random field deviation is proposed for large thin-walled structures, and the fluctuation ranges of accuracy indexes for assembly can be obtained efficiently with that of parts, which is the premise of assemblability evaluation and deviation control for large thin-walled structures.