Vibration-based structural health monitoring (VSHM) methodologies provide a robust, data-driven, system for damage diagnosis. However, there are a few challenges that are currently being investigated to ensure the systems are more reliable for decision-making. The features selected from the vibration responses are not only sensitive to damage but also to environmental and operational variations (EOV). This paper aims to investigate the use of a principal component analysis (PCA) based system for VSHM. In particular, the aim is to compare different approaches, using the same dataset, to explore the effect that data manipulation has on the damage detection capabilities of such a system when it is corrupted by EOV. The data that was used for this study was first taken from a simulated five degree of freedom spring-mass-damper system and secondly from an in-operation Vestas V27 wind turbine with damaged and undamaged scenarios. The simulated system was subjected to varying temperatures and involved four states; one healthy state followed by three states with increasing damage, represented by the reduction of a spring stiffness. Each combination of data manipulation was compared to determine their performance and limitations on removing EOV for reliable damage diagnosis.