Functional data analysis recently became a widely acceptable remedy for handling massive amounts of data measurable as one or several variables dependent on a certain argument, with insightful low-dimensional representations of this intrinsically infinite-dimensional phenomenon increasingly gaining popularity. While most of these are based on classical dimension reduction methods, like, e.g., functional principle component analysis, they often lack robustness and thus experience difficulties in correctly visualizing abnormal observations. We introduce a novel method for visualization of functional data, focused on anomalies: functional abnormal component analysis (fACA). Extending already existing—for the Euclidean space—counterpart, we need to overcome difficulties of the functional setting. We showcase the applicability of the developed method within an existing taxonomy of anomalies, on real-world data sets.