In recent decades, condition-based maintenance (CBM) is acknowledged a
cost-effective and widely used maintenance program for engineering
systems. Diagnostics and prognostics are critical components of CBM
responsible for offering information about present and future system
conditions. These two components are respectively an integrated process
covering several aspects that are essential for successful
implementations of diagnostics and prognostics. For diagnostics, data to
be used should be clean and useful. Data cleaning can provide clean data
by removing outliers caused by noise, while feature selection can select
useful characteristic features for fault classification. For
prognostics, noise may appear in condition indicator values. Using such
noisy values may result in unreliable predictions for prognostics. A
method is thus demanded to provide predictions without noise effects.
Support vector machine (SVM), a machine learning method, is recognized
having good generalization ability and an effective tool for
classification and prediction needed by diagnostics and prognostics.
This book explores the potentials of SVM for addressing above problems
in diagnostics and prognostics.