Presented by: Vishaal Bhana Supervisor: Prof. Stephan Heyns
Transcript of Presented by: Vishaal Bhana Supervisor: Prof. Stephan Heyns
1
ONLINE DAMAGE DETECTION ON SHAFTS
USING TORSIONAL AND UNDERSAMPLING
MEASUREMENT TECHNIQUES
Presented by: Vishaal Bhana
Supervisor: Prof. Stephan Heyns
PUBLIC DEFENCE OF MASTERS DEGREE
DATE: 13TH FEBRUARY 2013
2
• Turbogenerator shafts needs to function
continuously
• Shafts subjected to complex torsional
loadings
• Often leads to failure due to propagation of
cracks
• Catastrophic and dangerous
• Failure results in downtime
• High repair costs
Introduction
3
• High costs for repair
• Negative effect on industry and public
• Literature has shown various failure cases
• Continuous online monitoring is essential
Blackouts
Downtime for repair
High strain on remaining generators
Potential failures
Failures High repair
cost
Cyclic loading,
cracknitiation
High repair cost
Cyclic loading,
crack initiation
Introduction
4
Introduction
• Constantly monitor the system
• Changes in the signal are further investigated
• Problem is isolated
• Machine stopped and repaired/replaced if necessary
• Pre-plan maintenance programs/ cost effective
Online monitoring
system
Vibration signature normal?
Machine operating constantly
No
Yes
Maintenance done
Pre-plan maintenance
schedule
Investigate irregularity
• Lateral and torsional vibration have been successful
• More focus on torsional methods recently
• Introduction of fibre-optic sensors and torsional laser vibrometers and
non-contact measurement methods
• Considerable research by Maynard and M. Trethewey with fibre-optic
sensors
• Found damage in shafts and blades
• DIC methods used for strain visualisation by taking discrete images
• Impact tests done by Lall
• Vibration and mode shape measurements by Helfrick
• Strain gauges ideal for measuring high frequency content
• Yam investigated strain and displacement modes
• Various modelling methods implemented (Analytical, Timoshenko
beams, commercial software)
• Different methods used for vibration and specifically torsional vibration
but no comparison
5
Literature study
Introduction
• Turbogenerator shafts needs to function continuously
• Shafts subjected to complex torsional loadings
• Often leads to failure due to propagation of cracks
• Cracks develop in shafts due to fatigue
• Failure results in downtime
6
Investigate the torsional dynamics of a shaft for
undamaged and damaged shaft
Comparison study of different torsional measurement
techniques
Investigation of undersampling for damage detection
Numerical analysis, FEM-model built through use of Patran and
the Nastran solver, Normal modes analysis done
Verify using lumped spring-mass system
Conduct an experimental analysis on a rotor with increasing
damage
Measurements done using Telemetry strain system, fibre optic
sensor, Digital Image Correlation (DIC) system
Objectives
Introduction
• High costs for repair
• Effects on the public
• Continuous online monitoring is essential
• Offline monitoring results in downtime
• Pre-plan maintenance programs
7
System setup
• Two disks fixed onto a shaft
• Coupled to a 3kW DC motor
• Self-aligning bearings
• Free-end on the other side
Introduction
• High costs for repair
• Effects on the public
• Continuous online monitoring is essential
• Offline monitoring results in downtime
• Pre-plan maintenance programs
8
FEM
• Study of the dynamics of the rotor investigating
the natural frequencies
• Sensitivity analysis with damage
• Key parameters
Geometry- Basic structure that would resemble
manufactured setup
Mesh- Ease of modelling shaft and damage
Boundary conditions- representation of motor
connection and bearings
Material Properties
9
FEM- Model Setup
• Modelled as a single unit
• Surface Quad 8 elements that is extruded
• Control over meshing geometry and modelling the damage (non
equivalenced nodes
• Fixed motor end using MPC
• Node BC applied at bearing location (x and y direction)
• Mild steel properties used
Introduction
• Turbogenerator shafts needs to function continuously
• Shafts subjected to complex torsional loadings
• Often leads to failure due to propagation of cracks
• Cracks develop in shafts due to fatigue
• Failure results in downtime
10
Torsional mode Lumped Patran % Difference
1 136.04 133.02 2.27
2 420.17 408.95 2.74
Lumped model
FEM- Convergence and verification
12
FEM- Mode Shape
• 1st mode shows high rotational response at free-end
• Strain comparison showed higher strain response at the motor-end
• Location of test equipment determined
Introduction
• Turbogenerator shafts needs to function continuously
• Shafts subjected to complex torsional loadings
• Often leads to failure due to propagation of cracks
• Cracks develop in shafts due to fatigue
• Failure results in downtime
13
Experimental setup
• Rotor manufactured consisting of a shaft with two discs
• Damage introduced by cutting a slot in shaft
• Rigid coupling to prevent torsional damping
• Overall setup with telemetry, Fibre-optic sensor and DIC system
• Brake setup for calibration
Fibre-optic sensor
Telemetry
Introduction
• High costs for repair
• Effects on the public
• Continuous online monitoring is essential
• Offline monitoring results in downtime
• Pre-plan maintenance programs
14
Techniques
• Fibre-optic sensor Black-White coded tape (Square
wave output)
Changes in widths represent
torsional vibration
• Telemetry Analog strain data
Gauges mounted 180o apart
• DIC 2 cameras create 3D representation
White markers are picked up and
their position analysed
Strobe lights needed for dynamic
testing
Order content
• Harmonics of rotational frequency present
• Causes natural frequency to be hidden
• Constant time Constant angle sampling
• FFT to order domain
• Remove orders
• Order removed frequency domain
15
Processing
16
Undersampling
• Nyquist theorem , if not then aliasing occurs
• DIC strobe lights operate at 15 Hz
max2sf f
sin(2 10 ) sin(2 80 )y t t
Processing
17
• Fan-paper method
• Fold to base-band frequency
• Frequency of interest band-limited
• Fold causes reverse readings on odd folds
• Base-band 4.5Hz
Processing cont’d
19
• 3 test runs conducted
• Runup test, 2 constant speed tests- 1100Rpm 1640Rpm
• 2 min constant speed runs 12 averages taken
• Frequency resolution 0.1Hz
• Recording of all 3 devices simultaneously
• Undamaged and damaged case- 5%-66% damage introduced
• Measure time domain and post-processing in frequency
domain
• Order removal done
• Investigation of undersampling methods
• Waterfall plot for runup test
• View results in frequency domain
Test procedure
20
• Waterfall plot shows structural response at B
• Averaged 2D plot, Response at 120.6Hz
A B
Post Processing-Fibre-optic sensor
21
• Waterfall and averaged 2D plot- Nothing distinct even
after order removal
• Line frequency harmonics
Post Processing- Telemetry
22
• Capture markers at undersampled 9Hz
• Continuous marker capture. Create “components”
• Images form 3D representation of setup with displacement
and location information
Post Processing- DIC
23
• Some images contain no markers- “Blind side”
• Create continuous time signal
• Undersampled FFT- rotational frequency content
• Expected frequency 3.6Hz
Post Processing- DIC
• Geometry- Include brake, coupling and damage modelling
• Material properties- Apply manual updating by adjusting
parameters within its range
24
FEM Update
• Fibre-optic sensor revealed change in natural frequencies with
damage
• Comparison showed that fibre-optic method most ideal
25
Final Results
26
Final Results- FEM Comparison
• Close comparison between FEM and experimental results
• Change in frequency ratio shows good correlation
• Error range from 0-3.2%
• Undersampled fibre-optic sensor
• Revealed changes
• Pre-knowledge required for location
27
Damage % Actual frequency Base-band equivalent Undersampled data
0 120.6 3.6 3.59
5 120.6 3.6 3.55
8.3 120.5 3.5 3.48
19.3 120.1 3.1 3.12
31.6 119.1 2.1 2.13
41.6 118.1 1.1 0.98
53.3 115.7 1.3 1.36
66.6 109.1 1.6 1.13
Final Results- Undersampling
• The fibre-optic sensor proved to be the successful torsional
measuring equipment for online monitoring and damage
detection.
• The strain gauge provided order content but nothing about
dynamics
• Further investigations showed that high excitation was required
• The DIC also revealed order content which was successfully
obtained through undersampling techniques
• Undersampling was successfully used with the fibre-optic sensor
results to obtain changes in the dynamics for a damaged shaft.
28
Conclusions
• Strain modal analysis may be investigated by applying external excitation or
loading on the system to excite the modes of interest
• By creating a setup with a low enough dynamic frequency and having a high
enough strobe light available, one can investigate systems without undersampling
techniques.
• The DIC system in this study only looked at the markers while they were in camera
view. Methods by which the markers may be studied during the full rotation and
thereby giving a better full scale representation of what happens during each
revolution. The implementation of such a setup together with the external loading
for higher strain could prove useful.
• It would be interesting to apply them on a system with an actual crack. Future
analysis could be done by growing a crack onto the shaft.
29
Recommendations