This document summarizes an analysis of a ventilation fan that failed due to fatigue during operation. CFX was used to calculate the unsteady pressure load on the fan blades, and ANSYS was used to create a structural model of the fan. LINFLOW performed an aeroelastic analysis of the fan under clean and dirty conditions. The analysis found that while the clean fan was stable, dirt buildup changed the aeroelastic characteristics and caused vibration amplitudes to exceed the material's fatigue limit, leading to failure.

1. Illustration on How to Study Response
of a Ventilation Fan Subjected to High
Frequency Loading.
In this work the question was, why
Software tools used: did the fan fail in fatigue during
+LINFLOW operation? Did the initial design have
an aeroelastic problem or was the
+ANSYS problem created by the dirt built up
+CFX that was found on the failed
fan blades?
LINFLOW 1.4 1

2. CFX and the Ventilation Fan Analysis
CFX (Navier-Stokes solver) used for calculation of unsteady
pressure around the fan, which is the load exciting the fan.
Flow Outlet
CFX Velocity contours
Rotor and stator details
Flow Inlet
LINFLOW 1.4 2

3. CFX Unsteady Pressure Up-stream the
Fan.
CFX used for calculation of unsteady pressure on the fan,
which is the load exciting the fan.
Time history of unsteady Pressure An FFT analysis gave the frequency
that fan blades see. spectrum of the load that the fan
blades are subjected to.
LINFLOW 1.4 3

4. The ANSYS Structural model of the
Ventilation Fan
ANSYS Structural Model of the Fan
To perform an aeroelastic stability
and response analysis, a structural
dynamics model is needed. In this
case the ANSYS FE-program was used
to develop the needed model.
LINFLOW 1.4 4

5. The ANSYS Structure Dynamics Used by
LINFLOW
ANSYS Modal Results for the Fan
Mode Number Frequency (Hz) Mode Type
0,1 37.75 Fan axel torsion
0,2 57.42 Blades in bending motion
0,3 80.93 Blades in bending motion
0,4 109.7 Blades in cord bending
0,5 136.8 Blades in torsion motion
0,6 181.1 Blades in mix. torsion/bending
0,7 184.2 Blades in mix. torsion/bending
1,1 34.07 Fan hub bending
1,2 34.07 Fan hub bending
1,3 72.82 Blade bending
1,4 72.82 Blade bending
1,5 95.63 Blades in cord bending
1,6 95.63 Blades in cord bending
1,7 112.1 Blades in bending motion
2,1 68.48 Blades in bending motion
2,2 68.48 Blades in bending motion
2,3 80.19 Blades in bending motion
2,4 80.19 Blades in bending motion
2,5 114.1 Blades in cord bending
2,6 114.1 Blades in cord bending
2,7 136.8 Blades in mix. torsion/bending
3,1 71.38 Blades in bending motion
3,2 71.38 Blades in bending motion
3,3 87.99 Blades in bending motion
3,4 87.99 Blades in bending motion
3,5 136.8 Blades in mix. torsion/bending
3,6 136.8 Blades in mix. torsion/bending
3,7 181.12 Blades in mix. torsion/bending
Table 1, Structural eigenvalues for the fan blade model
LINFLOW 1.4 5

6. The LINFLOW Aeroelastic Model of the Fan
The LINFLOW linearized
fluid dynamics model of the
fan include both the fan blade
and a wake model. LINFLOW
uses boundary elements to
descretize the fluid dynamics,
which means that no flow domain
grid is needed.
Flow Conditions in this case:
Pressure = 5530 Pa
Density = 1.2 kg/m3
Angular Velocity = 870 – 920 rpm
Cp/Cv = 1.4
Blade angle of attack 63o
LINFLOW 1.4 6

7. LINFLOW Steady Fluid Dynamics
In LINFLOW the aeroelastics is
studied around some mean steady
flow condition. In this case we check
that the lift on the blade corresponds to
the force driving the flow through the
fan and that the drag on the blade
Corresponds to the moment that
the fan is driven by.
Picture show velocity contours on the surface of the blades..
LINFLOW 1.4 7

8. LINFLOW Aeroelastic Stability Anlysis of the Fan.
Aeroelastic Stability Check of the Fan
Aeroelastic Frequencies a.f.o. Flow Velocity. Damping requirement for neutral stability
for the 5 first modes in table 1.
The calculation show no sign of aeroelastic stability problem
In the RPM range at which the fan is operated.
LINFLOW 1.4 8

9. LINFLOW Response/Spectrum analysis, 1
LINFLOW Aeroelastic Response Analysis of the Fan with
Clean Blades.
Stess calc. location
Real part of the stress field on the model Imaginary part of the stress field on the model
Spectrum analysis with SRSS summation gave 19.4 Mpa stress level in the root of the blade shaft.
Measurements on clean blade gave approx. 20 Mpa stress level at the same location
LINFLOW 1.4 9

10. LINFLOW Response/Spectrum analysis, 2
LINFLOW Response Analysis of the Fan with Dirt build-up
on the blades.
Real part of the stress field on the model Imaginary part of the stress field on the model
Spectrum analysis with SRSS summation gave 32.0 Mpa stress level in the root of the blade shaft.
This is above the 27 Mpa stress level that is the limit above which fatigue problems start to appear.
Conclusion, the reason for failure was that the dirt built-up had changed the aeroelastic charactiristics so
that vibration amplitude due to subjected load increased to levels above the fatigue limit for the material.
LINFLOW 1.4 10