Variation simulation of final aircraft assembly concerning compliance and contact interaction of parts requires the specialized approaches to setting and generation of initial gaps between the joined parts. The initial gap reflects all assembly deviations for the joining parts. In this paper two models for the generation of the initial gap samples are considered: the random field model and the mode-based model. The random field model represents the initial gap as a random field with defined properties. The mode-based model decomposes the initial gap into a series of natural modes of the parts.
Typically for the real assembly processes, there is not enough accessible information about initial gaps. If the measurements are not included in the production process, then the number of measured samples can be small. The goal of this paper is to compare how considered initial gap models cope with the small number of given measurements. The models are examined on the example of wing-to-fuselage assembly process. For the considered process both initial gap models are trained on different amounts of measurements, and generated gaps are applied for the variation simulations. The comparison with measured gaps allows to determine which method is the most suitable during variation simulation with limited measurement data. The performed study shows that the mode-based approach is more accurate for initial gap modeling in case of the small number of available measurements.