The Oscillating Water Column  (OWC) device is a relatively young and underdeveloped technology  for  generating electricity from wave energy. Compared to other common renewable sources such as solar and wind energy, the costs for bottom-standing OWC’s are too high for the technology to be economically competitive. A first step in lowering the cost was the integration of OWC devices into breakwaters. Often, OWC are built quite robustly due to insufficient confidence in the magnitude of the extreme loads. This leaves room for optimisation, which so far has not been looked into in detail. In an attempt to lower the costs further, the possibilities for structural optimisation are investigated in this research. The U-OWC device in Civitavecchia, an OWC with an additional vertical duct, is used as a case study and is modelled in 3D in a one-way coupled hydraulic-structural numerical model. The model contained both a fluid domain, i.e. the wave tank, and a solid domain, i.e. the OWC structure. Coupling the two gives direct feedback on the wave impact in terms of a stress state in the structure. By converting the stresses to sectional forces, a new minimally required wall thickness can be found. Comparing these to the original thicknesses of the three main walls gives the optimisation potential of the OWC structure. Numerical simulations were carried out in operational and design conditions for varying geometries. Based on this, it was found the OWC walls could have been built using 35% less concrete on average. The results also led to a new insight on the structural behaviour. Clear trends were found in where and when to expect critical loads. It was also found that the geometrical changes barely had any influence on these properties. These findings were used to develop a new and simple design method, making a full structural analysis redundant. Removing the structural domain and making use of the found property that water pressures are constant over the transverse width, imply that a 2D model would be sufficient, reducing the model size  drastically.