Sandwich structures with honeycomb cores are widely used in lightweight engineering due to their exceptional stiffness-to-weight ratio, thermal stability, and damage tolerance. However, their design and optimization remain challenging because the structural response is governed by coupled phenomena acting across multiple length scales, from material microstructure to global structural behaviour. The objective of this study is to develop and demonstrate a multiscale simulation framework for sandwich honeycomb structures that enables weight-efficient design while ensuring structural integrity, thermal stability, and manufacturing feasibility under combined thermo-mechanical loading. The proposed approach integrates micro-, meso-, and macro-level modelling. At the micro-scale, effective properties of composite face sheets and adhesive joints are derived considering anisotropy, curing effects, and residual stresses. At the meso-scale, detailed honeycomb core models account for cell geometry, height, orientation, and material orthotropy. These homogenized properties are then transferred to macro-scale finite element models of full sandwich panels and representative components. The framework is validated through comparison with experimental data and full-scale tests reported in prior developments. The multiscale simulations reveal that local meso-scale parameters of the honeycomb core and face-sheet layup have a decisive influence on global stiffness, stress distribution, and failure margins. Optimization of honeycomb geometry and adhesive application strategy enables significant mass reduction while maintaining or improving load-bearing capacity. The proposed approach demonstrates weight savings of up to 20% compared to baseline designs, with good agreement between numerical predictions and experimental observations. The presented multiscale simulation framework provides an effective and physically consistent tool for the design and optimization of sandwich honeycomb structures. By enabling systematic transitions between length scales, the proposed approach enhances predictive accuracy 141 and supports the design and optimization of lightweight, reliable sandwich components for advanced lightweight engineering applications.