Mechanical systems operating in noisy environments create a challenging signal processing and monitoring problem especially in real-time. To detect a particular type of subsystem from noisy vibration data, it is necessary to identify signatures or particular features that make it unique. Resonant (modal) frequencies emitted during its normal operation satisfy this constraint. Monitoring structural modes to determine the condition of a system under investigation is essential, especially if it is a critical entity of an operational system. The development of a model-based detection scheme capable of the on-line detection of structural anomalies applying both system identification methods to extract a modal model and state estimation/detection methods for anomaly detection is developed. Two optimal detectors evolve: an innovations-based processor and a reference detection scheme.