The structural health monitoring of wind turbine structures represents a critical focus area, given the inherent benefits it offers in terms of predictive maintenance. This proactive approach proves cost-effective in the long run compared to reactive, corrective maintenance strategies. Offshore wind turbine foundations, exposed to diverse environmental conditions from both wind and marine waves, necessitate robust monitoring systems. To detect damage in these structures, an analysis of data collected by accelerometers attached to the structure is conducted. This study introduces a novel semi-supervised damage detection methodology, which was systematically compared against a traditional supervised approach. Additionally, an examination of the impact of the percentage of labeled data in the semi-supervised [1] methodology was undertaken. The proposed methodology was validated through testing on a laboratory-scaled jacket-type wind turbine foundation [2]. The structure was equipped with 8 triaxial accelerometers strategically placed. Vibration response data was acquired by subjecting the structure to a white noise signal with a shaker at four different amplitudes, resulting in 5740 experiments. Of these, 2460 pertained to the healthy structure, and 3280 to the damaged one. The results demonstrated the effectiveness of the developed semi-supervised damage detection methodology.