A structural health monitoring (SHM) system verifies the mechanical state of a structure toensure its proper functioning and determines whether it needs some kind of maintenance. Thus, SHMfor wind turbines (WT) in remote locations, as offshore, is crucial. Offshore wind farms are increasinglyrealized in water depths beyond 30 meters, where lattice support structures are an interesting optionto withstand the severe environmental actions. In particular jackets appear to be a highly competitivesubstructure type with a wide range of applicability, from approximately 25 to 70 meters water depth.With no doubt, structural damage is a significant issue in these structures. Unlike on-shore structures oreven shallow water structures, the access for regular monitoring and repair is not an easy option, in termsof both the cost and the accessibility. In this work, a methodology for the detection and classification ofstructural damages in offshore jacket-type WT is stated. The proposed method relies on the paradigmthat any damage in the structure produces changes in the vibrational response. However, it is assumedthat the only available excitation of the WT is the wind turbulence, so the input excitation is assumedto be unknown. Therefore, using only accelerometer information, a data driven approach for damagedetection is developed. The scheme of the proposed method can be summarized in the following steps:(i) the wind excitation is simulated as a Gaussian white noise and the data coming from the WT iscollected using a set of accelerometers; (ii) the raw data is arranged in matrix form and pre-processedusing mean-centered group-scaling; (iii) principal component analysis (PCA) is selected as a technique toreduce the dimensionality of the data and the computing time of the next step; finally, (iv) the quadratic-kernel support vector machine (SVM) is used as a classifier. The 5-fold cross-validation technique isemployed to estimate the overall accuracy and to avoid overfitting. In order to experimentally validatethe proposed approach, the damage detection strategy is applied to different types of predefined damagein a small-scale structure —an experimental laboratory tower modeling an offshore-fixed jacked-typewind turbine—. The results that have been obtained for these predefined damages are included anddiscussed to demonstrate the reliability of the proposed approach.