Multivariate Processing of Accelerometric Condition Indicators

An innovative integrated self-learning health monitoring system has been developed and implemented on a fleet of helicopters in actual service. This system improves significantly the efficiency of a previous accelerometric vibrational monitoring tool by means of multivariate third level processing of the accelerometric features. The paper describes the way in which several problems, typical for the monitoring process of a mechanical system, have been treated in this specific case. The applied techniques could be of much more general interest.