Professional Documents
Culture Documents
GT2018-75961
ABSTRACT 1 INTRODUCTION
This paper furthers recent research by these authors. The
starting point is the pre-optimization of solid dampers, which en- Turbine blades do not benefit significantly from material
sures that all dampers bound to misbehave are excluded since the hysteresis and aerodynamic damping. As a result, external
early design stage. The authors now enlarge the scope of their sources of damping, e.g. in the form of dry friction devices [1–7]
investigations to explore those damper configurations selected such as underplatform dampers, are typically added to turbine de-
inside the admissible design area. The purpose of the paper is signs. Being able to 1) predict and 2) tailor the effect the damper
to present a set of criteria apt to select a damper configuration has on the bladed disk is of primary importance in increasing the
which not only avoids unwanted situations, but in addition guar- engine reliability and reducing the cost of operation and the du-
antees high performance under different design conditions. ration and cost of the design process itself.
The analysis starts with the definition of a set of requirements The prediction (point 1)) of the damper effect on the blades has
a high performance damper should meet. In detail the present been thoroughly addressed in the last decades. In detail, the com-
investigation seeks to answer the following questions: putation of the forced response in presence of friction damping
– in the low excitation regime, what is the frequency shift and requires ad-hoc solution techniques [8, 9], reduction methods to
the stiffening effect each damper can provide? decrease the size of the blades’ FE models [7, 10–12], contact
– for increasing excitation levels, which damper will start slip- models to take into account the presence of friction [1, 13–17]
ping sooner? and techniques to calibrate them [16, 18–21].
– in the high excitation regime, which damper provides the The first efforts in the frame of damper optimization (point 2)
maximum dissipation? started as early as 1980 [1] and were limited to the selection of
Like pre-optimization, it does not involve nonlinear Finite Ele- the optimal damper mass. A few years later, however, the first
ment calculations, and unlike existing optimization procedures, qualitative comparisons of cylindrical and wedge damper shapes
is not linked to a specific set of blades the damper may be cou- started to appear [22, 23]. These early works highlighted the
pled to. The numerical prediction of the blade-damper coupled drawbacks of cylindrical and wedge dampers: it was observed
dynamics is here used only for validation purposes. how cylindrical dampers were prone to rolling (i.e. a fatal blow to
The approach on which this paper rests is fully numerical, how- their damping capability), while wedge-shaped dampers tended
ever real contact parameters are taken from extensive experimen- to detach for high excitation levels. The obvious correction is to
tal investigations made possible by those purposely developed use a flat-curved solid damper having one of the two contact sur-
test rigs which are the distinctive mark of the AERMEC Lab of faces in cylindrical form, thus allowing a perfect mating with the
Politecnico di Torino. underplatforms. A notable work is [24], where the curved-flat
damper, already used in the industry since 1985, is mentioned
for the first time in a scientific publication. Flat-curved dampers
∗ Address all correspondence to this author.
opposed to ”partial” or ”micro” slip [16,17]). If, as in this paper, contact elements
based on Coulomb friction [14] are used to model the interface, the numerical 3 Displacements are adequately captured by their first harmonic contribution
equivalent of ”gross slip” sees ALL contact elements used in slip condition. only: this holds both for linear and nonlinear vibration if pre-optimized dampers
are involved [30].
of the Criteria for High Performance procedure outlined in the 1. zero amplitude, in the limit: the damper is fully stuck and the
following section. equivalent stiffness introduced by the damper, here termed
K, is at its maximum (see also Sect. 5.1);
2. threshold amplitude at bilateral gross slip onset: its value,
normalized by the centrifugal force on the damper, is here
5 CRITERIA FOR HIGH PERFORMANCE OF SOLID termed S and further addressed in Sect. 5.2;
DAMPERS 3. very large amplitude: the asymptotic dissipated energy per
unit amplitude is here termed D and further addressed in
As stated above, the CHP procedure is a multi-step process Sect. 5.3.
where each ”step” corresponds to ”key” levels of platform-to-
platform relative motion and to ”key” design parameters: All results are delivered, similarly to the pre-optimization maps,
in the (θL , θR , h/r) space. For each ”step” a set of level FIGURE 5. REPRESENTATION OF THE NUMERICAL MODEL
curves represents the performance of each damper configuration OF A SOLID DAMPER BETWEEN A SET OF PLATFORMS.
in terms of the key parameters K, S and D outlined above.
Further details are given in Fig. 4, where a mind map of the
it is worth noting that only the component of the contact forces
CHP procedure is shown, together with the tools employed,
oriented along the direction of motion will contribute to stiffen-
the damper requirements it helps satisfying and its relation pre-
ing the structure. Therefore, with reference to Fig. 5, in case
optimization.
of IP motion only the vertical component of the contact force
WR will contribute to creating a constraint, while in case of OOP
5.1 Step 1: equivalent stiffness motion, only the horizontal component UR will oppose the plat-
The presence of a damper between a set of blades can be forms’ relative motion 4 . Since θL and θR influence the magni-
represented as shown in Fig. 5, where the presence of friction is tude of the force components, different full stick FRFs should be
taken into account through state of the art contact elements [14], expected for different damper configurations, all other parame-
characterized by a set of calibration parameters: normal and tan- ters (e.g. contact parameters) being equal.
gential contact stiffness kn and kt and friction coefficient µ. If
the excitation level is low the damper remains stuck between the 5.1.1 Objective The purpose of this first step is to track
platforms and the damper-induced equivalent stiffness is maxi- the value of the equivalent stiffness and link it to the damper per-
mum. In this case, the contact element acts as a simple spring
coupling (i.e. no slip).
As stated in Sect. 4, once the mode under investigation is se- 4 In this investigation the components of the right contact force are chosen as
lected, the direction of the platform relative movement is known: a reference, however the reader should notice that, UR = UL while the vertical
components differ only by a constant offset WR = CF −WL .
5.3.2 Tools One possible way of achieving the objec- remain the same.
tive above is to run the quasi-static numerical tool described in 3. The excursion of the relevant component of the contact force
Sect. 5.2.2 for increasing values of w1P in the IP case and of in case of bilateral gross slip (e.g. (WRIP− −WRIP+ ) in the IP
u1P in the OOP case. However, a closer analysis of the damper example) is completely defined by the value of the friction
force equilibrium can lead to a closed form relationship between coefficients and the damper configuration. With reference
the damper angles and the amount of dissipation provided by the to [28] and to Fig. 9a-b, let us consider the damper force
damper in case of large displacements. equilibrium (where inertia forces are neglected) in the gross
This formulation is based onto three fundamental observations. slip condition. The vertical components of the contact forces
can be defined as:
1. Only the contact force components aligned with the plat- tan(θL + arctan(µL ))
forms’ relative displacement contribute to dissipation, e.g. WRIP− = CF ·
tan(θR − arctan(µR )) + tan(θL + arctan(µL ))
WR or WL in the IP case. This statement has been confirmed tan(θL − arctan(µL ))
by numerical observation: in [34] it was shown that the area WRIP+ = CF ·
tan(θR + arctan(µR )) + tan(θL − arctan(µL ))
of platform-to-platform hysteresis cycle (WR vs. wP cycle)
(13)
equals the sum of the platform-to-damper hysteresis cycles
at the contacts (i.e. tangential contact force vs. tangential Similar relations are easily derived for the OOP case,
relative displacement). (UROOP+ and UROOP− ). Equations 12-13 have both been ver-
2. With reference to Fig. 9b, if wP is sufficiently high, then ified against the quasi-static algorithm from Sect. 5.2.2 and
the area of the platform-to-platform hysteresis cycle can be standard solution techniques implemented in FRIDA.
approximated with the area of the rectangular ideal cycle:
En ≈ 2 · w1P · (WRIP− −WRIP+ ) (i.e. the losses due to the pres-
ence of contact springs become negligible for large displace- 5.3.3 Results Once again, in a matter of seconds the
ment values). Energy dissipation for a given amplitude of complete (θL ,θR ) subspace of interest is mapped for energy dis-
displacement, here termed D, can therefore be defined as: sipation per unit displacement DW and DU , thus offering the
En damper designer precious indications of the effectiveness of each
DW = ≈ WRIP− −WRIP+ for the IP case; damper configuration in case of high excitation levels. The cor-
2w1P
(12) responding level curves are shown in Fig. 7 e-f. Values are ex-
En pressed in N (i.e. µJ/µm) and are related to a CF= 100 N so that
DU = 1 ≈ UROOP− −UROOP+ for the OOP case.
2uP the actual dissipation can be linearly scaled for different values
It should be noted that for increasing values of displacement of CF.
En will scale linearly with 2w1P but the force excursion will The reader will notice that, unlike Fig. 7a-d, these level curves