| Abstract: | In the present paper, Random Forests are used in a
critical and at the same time non trivial problem
concerning the diagnosis of Gas Turbine blading
faults, portraying promising results. Random forestsbased
fault diagnosis is treated as a Pattern
Recognition problem, based on measurements and
feature selection. Two different types of inserting
randomness to the trees are studied, based on different
theoretical assumptions. The classifier is compared
against other Machine Learning algorithms such as
)eural )etworks, Classification and Regression Trees,
)aive Bayes and K-)earest )eighbor. The
performance of the prediction model reaches a level of
97% in terms of precision and recall, improving the
existing state-of-the-art levels achieved by )eural
)etworks by a factor of 1.5%-2%. Furthermore,
emphasis is given on the pre-processing phase, where
feature selection and outliers identification is carried
out, in order to provide the basis of a high
performance automated diagnostic system. The
conclusions derived are of more general interest and
applicability. |
|---|
Greek