FAULT-DETECTION OF HELICOPTER GEARBOXES USING THE MULTIVALUED INFLUENCE MATRIX-METHOD

Fault detection of a helicopter gearbox by pattern classification is d iscussed. The detection system is composed of two components, a quanti zation matrix to flag the measurements, and a multi-valued influence m atrix (MVIM) that represents the behavior of measurements during norma l operation and at fault instances. Both the quantization matrix and t he influence matrix are tuned during a training session so as to minim ize the error in detection. This detection system was applied to vibra tion measurements collected from a helicopter gearbox rest stand durin g accelerated fatigue tests and at various fault instances. The result s indicate that the MVIM method provides accurate detection when the f ull range of faults effects on the measurements are included in the tr aining set. Furthermore, the fixed structure of MVIM allows evaluation of individual measurements. This feature was utilized to select a sub set of measurements crucial to detection.