Modern wind turbine blades (WTB) are
designed to last 25 years of service under severe environmental conditions. Damage
is likely to occur during this time span and therefore there is a need of reliable
systems to ensure the structure integrity. Vibration-based structural health
monitoring (VSHM) is becoming one of the most commonly used methods to monitor
damage in wind turbine blades. These methods rely on vibration measurements,
from the instrumented WTB, to either operational loads or artificial induced
excitation that are used as an input to statistical algorithms for
decision-making. One important aspect is that sensors do not directly measure
damage, but they provide information on where to extract reliable sensitive
features that might be used to monitor the damage. This presentation will
outline a methodology that lies within a popular strategic for data-driven
VSHM, which consists to compare new observations measured in the WTB against to
a reference state that defines the original (pristine) status of the WTB. The
methodology learns from the past and present towards successful damage
diagnosis and hence to project the learning onto future decisions. To
illustrate the methodology, two case of study are presented; first in a 34m WTB
in a laboratory testing and secondly in 27m WTB in an operation wind turbine. Finally,
an expansion to detecting instability of centrifugal compressors is also
presented.