| Περίληψη: | This work focuses in two main directions of rotor dynamics field, the simulation of rotor
bearing systems and the fault diagnosis. From the serious multiple faults that can appear in a rotor bearing system two of them are the target of current research: the transverse fatigue
crack of a rotor and the radial extended wear in a bearing. The transverse crack is a defect
able to bring a catastrophic failure of the system when the growth (depth) takes high
percentage values relatively to radius of the shaft (i.e. >60%) and the symptoms of crack
presence have been widely investigated during last four decades yielding efficient methods
for the early crack detection. On the other hand the defect of bearing wear is much less
investigated without results connected with wear diagnosis methods. Concerning previous
works in those two defects the current dissertation’s persuasion is firstly to make a proposal in
bearing wear detection, secondly to achieve a method definition able to detect a breathing
transverse crack in a different way from those referred to literature.
For the subject of crack detection, a different crack breathing model is proposed with
emphasis in coupled local compliances definition and their variation during rotation while for
the subject of bearing wear detection, a wear model from the literature is used with emphasis
in rotor bearing system construction in a different way in relation to what up to now is
available in literature. The rotor bearing system construction (simulation) is a matter widely
investigated since early 60’s and some points of the current work try to differ in the way that
the rotor and the fluid film bearings interact in discrete time. The concept of nonlinear fluid film
forces is confronted in this work leaving out the nonlinear stiffness and damping bearing fluid
film coefficients and assuming that during the journal whirling no equilibrium point must be
defined in order to evaluate the future progress of vibration. Towards generality the fluid film
bearings are not defined geometrically as short or long. These two specific geometric
assumptions of short/long bearing appear widely in real machines and yield analytical
expressions of fluid film forces but in current work the finite fluid film bearing is used
demanding the well known finite difference method in order to evaluate the impedance forces,
as many researches have propose.
Both defects are met in a rotor bearing system parted from a continuous rotor and finite fluid
film bearings. An entire chapter is dedicated in the way that Rayleigh equation of rotor motion
incorporates internal damping using exclusively Real number confrontment, and in the way
that fluid film forces react in rotor motion by defining boundary conditions in every discrete
time moment. The definition of boundary conditions in discrete time makes them functions of
the entire system response yielding a nonlinear dynamic system with the resulting time
xx
histories to be characterized from periodicity or quasi-periodicity sometimes depending in the
defects presence.
An extended analysis of time histories of the intact and the defected system is made in order
to invest the symptoms of each defect in magnitudes of time and frequency domain. Timefrequency
analysis is performed using continuous wavelet transform in virtually or really (the
former from simulation, the latter from experiment) acquired time histories in order to extract
the variable coupling phenomenon exclusively due to the breathing crack from the other two
main reasons of coupling, the bearings and the shaft. Vertical response due to crack coupling
is amplified when the crack coupled compliances become larger under an electromagnetic
horizontal excitation in the rotor. This rapid in time variable coupling due to crack is used at
last in order to detect the crack presence. The external excitation is used also in the case of
wear detection since results of time-frequency analysis yield unexpected amplification of
specific harmonics when the wear defect is present.
Both considerations about the corresponding fault detection are tried in a real experimental
system after the observation that response of the current rotor bearing simulation converges
with the response of the physical system in characteristics that are judged important for the
method robustness. The general speculation is that both defects have to be detected without
the need of operation interruption since this cannot be feasible (high cost) in real turbo
machinery plants and in an early growth that coincides with safe machine operation. The
defect growths have to be at least 10% (of radius) for the crack and 20% (of radial clearance)
for bearing wear so as the methods to be efficient.
|
|---|
