Bi-layered beams are widely used as stimuli-responsive actuators. The most prominent example is the bimetallic strip, which is used as thermostat for converting a temperature change into a mechanical displacement. The same principle is valid for the material wood. With its innate capacity to swell and shrink in an anisotropic manner in response to moisture changes, a specific actuation can be programmed into a wooden bilayer system. Application ideas are e.g. curved wooden elements [Wood et al., 2016], or auto-rotating solar panel supports [Rüggeberg, Burgert, 2015].
These applications require effective prediction models with respect to the design process. Existing prediction models such as the Timoshenko formula [Timoshenko, 1925], valid for the two-dimensional case of thin layers, tend to lose their accuracy in case of up-scaled element dimensions. Yet, upscaling is required for the various applications. Processes such as moisture transport or location and size of adhesive bonds can highly affect the actuation. Furthermore, the natural variability of relevant design parameters such as swelling coefficients and mechanical stiffness represents a considerable challenge in the accurate prediction of actuation. We combine methods of uncertainty quantification with hygro-visco-elastoplastic finite element models for wood [Hassani et al., 2015] in order to predict upper and lower bounds of the actuation results of any initially flat shaped configuration for a given change in moisture content. We will show, how the design process of such robust bi-layered wooden structures can thus be optimized.