This paper presents a novel multi-scale design strategy for 2D plain woven composites considering the manufacturing process. The design methodology integrates the meso-structural geometry of woven fabric composites to maximise the buckling load of composite stiffened panels at the macro-level. A numerical model to simulate the autoclave manufacturing process of woven composites is developed, in which a new geometric modelling method is proposed to construct a more realistic RVE model for 2D plain woven composites accounting for the compaction effects. Deterministic and robust design optimisations are conducted to find solutions with both high performances and high robustness to uncertainties. In particular, this study considers the morphological uncertainties of the meso-scale fabric structure, i.e. yarns’ waviness and porosity incurred during manufacturing and its influence on the structural integrity of the parts. The optimisation results obtained from the deterministic and robust design problems are presented. A parametric analysis is conducted to characterise the sensitivity of the fabric properties to the geometry of yarns and related uncertainties. It is shown that the space between yarns has a significant negative impact on the structural buckling load and negative impact on the cure-induced residual stresses.