DYNAFLOW · PDF filePipe stress engineers prefer quasi static approach ... -Slug load calculation (manual??) ... • Analysis accuracy is increased if support structure is

-1-Copyright © 2007 by Dynaflow Engineering B.V. All rights reserved

DYNAMIC MODULE OF CAESAR IIIs it of any use??

lectures 2007

DYNAFLOW

30th August 2007


Why is dynamics not used moreCoade: < 5% of Caesar II license holders also makes use of dynamic module

Dynamical effects are overlooked

Dynamical effects are underestimated

Pipe stress engineers prefer quasi static approach• Dynamic load * 2

Pipe stress engineer feels uncomfortable with dynamics


Function of the dynamical module

Facilitate accurate assessment of the dynamical effects • Effects of unsteady loads on stress and load levels in piping

systems.

When is it used??• During design (avoid fatigue, overloading, large displacements)• Control of Vibration problems (development of mitigation

measures)


Piping Incidents due to dynamic loads

Few examples of what might go wrong!!


Collateral Damage due Dynamic Effects fromFlange Failure


Support Damage (I)


Support Damage (II)


Support Damage (II)


Support Damage (III)


More Incidents due to fluid/gas transients (“small bore piping”)


Examples of incidents as a result of fluid transients


Deluge FF System on Jetty whenTested First Time (I)


Deluge FF System on Jetty whenTested First Time (II)


Flange Failure


Dynamic loads are classified based on time history

Sustained dynamic loads• Periodic loads of moderate amplitude and long duration (minutes-

days)- Forces due to pressure oscillations in pump/compressor

suction or discharge systems- Imposed oscillating displacements at pump/compressor nozzle

connections.- Oscillating forces originating from flow fluctuations/instabilities

Transient, intermittent dynamic loads• Relatively large forces of relatively short duration (seconds)

- Slug Loads- Relief Loads- Waterhammer, surge loads- Earthquake


Effect of dynamic loads

Sustained dynamic loads• Vibration with a periodicity equal to the excitation mechanism• Vibration amplitude depends on separation between excitation

frequency and natural frequency of the piping system• Failure mechanism: High cycle fatigue

Transient dynamic loads• All natural mode shapes and natural frequencies of the piping system

are affected.• Response may show large amplitude vibrations of short duration.• Failure mechanism: • Excessive support loads• Accumulation of strain, low or high cycle fatigue


Required data for dynamic analysis

Dynamic model of piping system• Good distribution of lumped masses, additional nodes• Sufficient number of DOF

- For lumped mass (FE) type calculations typically only the first 33% of the modes shapes are accurate.

• Accurate boundary conditions

Magnitude of the excitation forces• Mechanical• Fluid mechanical

- Slug load calculation (manual??)- Acoustical simulation (pulsation study)- Waterhammer (surge) load calculation, simulation or

manual(Measurement results for bench marking the simulations)


Assessment of dynamic results

Material Fatigue data,• ASME B&PV section VIII div 2 appendix 5• API 579• AD Merkblatter• BS-5500• EN-13445


Vibration Assessment conform VDI 3842

Typical Vibration Level Limits conform VDI 3842, Vibrations in Piping Systems

1.0

10.0

100.0

1000.0

1 10 100 1000

Frequency [Hz.]

Vibr

atio

n Ve

loci

ty [m

m/s

] RM

S

design marginal correction danger


Three Options for Dynamic Response Analysisin Caesar II

1

2

3

0


Harmonic analysis

Application• Sustained vibrations (compressors & pumps)

Input• Periodic Loads can be applied at any node in the system• Per load case one excitation frequency• Many load cases are possible

Solution method• Since response frequency equals excitation frequency solution

procedure is quasi static (fast)

Output• Stress, displacement and load amplitude per frequency


Spectrum Analysis

Application• Transient vibrations: slug, waterhammer, relief

Input• Load time histories at many nodes in the system

Calculation method• Load time histories are translated into response spectra (time

phase between different loads is lost)• Natural frequencies and mode shapes are calculated• Response for each mode shape is determined• Mode shape responses are combined into a final system

response

Output • Maximum stress, largest modal contribution • Maximum loads, largest modal contribution• Maximum displacements


Response Spectrum Generation


Response Spectrum Generation

DLF Related to 4 harmonic cycles

DLF may grow > 2 if time history contains only a limited amount of consistent periodicity


Time History Analysis

Application• Transient vibrations: slug, waterhammer, relief

Input• Load time histories at many nodes in the system

Calculation method• Load time histories are maintained (time phase between different loads is

conserved)• Natural frequencies and mode shapes are calculated• Response of each mode shape is determined• Mode shape responses are combined into a final system response

Output • Maximum stress, actual stress at user defined times (snapshots)• Maximum loads, actual load at user defined times• Maximum displacements, actual displacement at user defined times• By combination of several runs a time history of stress, loads

displacements can be composed


Missing Mass correction

Only limited number of mode shapes are extracted and participating in the response

Only a fraction (preferably close to 100%) of the system mass isparticipating

Only a fraction of the total excitation force is participating

The missing force fraction is calculated and applied statically after multiplication by the largest DLF value above the frequency cut-off.

Important when large axial loads are applied (axial mode shapes have relatively high natural frequencies.


Spectrum versus Time History

Spectrum• Frequency domain• Conservative results: only maximum response per mode shape is

calculated and combined in system response• Mode making maximum contribution is identified (advantage for

vibration control)• Only maxima are calculated, time phase is lost• Number of participating modes is finite (missing mass)

Time history• Time domain• More accurate results: time history response per mode is

conserved and time phasing between model maxima is maintained during combination.

• Load, stress time histories (enables fatigue assessment by meansof cumulative damage, counting of cycles)

• Graphical response is possible• Modal info is not available• Calculation is memory intensive (limits simulation duration,

number of participating modes, time step resolution)• Number of participating modes is finite (missing mass)


Quasi-Static Approach

Dynamic Load amplitude (maximum) * DLF=2 as static load

Quasi static approach is simple and fast

Quasi static approach works when there is only one dominant modeshape that is excited

Quasi static approach focuses on loads (dynamic response is not considered), i.e. solution by change/elimination of modes is not possible.


Best Method for Transient (Impact) Loads Analysis

Quasi static• Simple but only if there is one dominant mode shape

Spectrum• Time phase between several impacts on one system is lost (e.g.

slug hitting consecutive elbows)• Conservative but output provides clues for problem solving.

Mode with largest contribution is identified.

Time history• Exact, timing relation between impacts is maintained (slug,

waterhammer)• Clues for diagnosis are less obvious

For transient loads a combination of spectrum and time history runs provides the best opportunities


Protective Measures

Sustained dynamic loads• Control of mechanical natural frequencies of the piping system

in relation to the excitation frequency• Support functions and support stiffness (in general high

stiffness)• Analysis accuracy is increased if support structure is included

in the modelTransient dynamic loads

• Control of support and nozzle loads• Support flexibility is sometimes useful• Elimination of damaging mode shapes

Protective measures for dynamics may be conflicting with statics




Example 1

ProblemHigh vibration level in compressor suction piping

Steps to solutionVibration Measurements, identification of main contributions in frequency domainVerification of acoustical natural frequencies of piping system (acoustical resonance)Verification of mechanical natural frequencies (mechanical resonance)Identification of source of vibration problemModification proposal


Compressor Location


Steel Supporting (I)


Steel Supporting (II)


Compressor Layout


Step 1. Vibration Measurements and Compressor Harmonics

0.00

20.00

40.00

60.00

80.00

100.00

120.00

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

Frequency (Hz)

Am

plitu

de (d

B)

66 Hz 99 Hz49 Hz

33 Hz

83 Hz16 Hz


Intermediate Conclusion from Step 1

Vibrations are at compressor harmonics

Vibrations must be result of:

Acoustical resonanceor

Mechanical resonanceor

High pulsation forces without resonance (compressor bottle sizing problem)


Step 2. Acoustical Natural Frequencies & Compressor Harmonics (Search for acoustical resonance)

0

50

100

150

200

250

10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00

Frequency (Hz)

Ampl

itude

16 Hz

Purple vertical lines represent compressor harmonics


Intermediate Conclusion from Step 2

Maybe near to resonance condition at first compressor harmonic (16.5 Hz.).

No further acoustical resonance

Vibration peak at 16.5 Hz, most probably is due high shaking forces as a result of near resonant condition.

The other vibration peaks must be the result of: Mechanical resonance

orHigh pulsation forces without resonance (compressor bottle sizing problem)


Step 3. Vibration Measurements & Calculated Mech. Natural Frequencies (Search for Mechanical Resonance)

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

Frequency (Hz)

Am

plitu

de (d

B)

66 Hz.33 Hz

83 Hz

Purple vertical lines represent pipe system natural frequencies


Conclusion from Step 3 & Identification Cause of Vibration Problem

Apparently there is mechanical resonance at 33 Hz and 66 Hz and near mechanical resonance at 83 Hz.

No mechanical resonance condition at the first compressor harmonic (16.5 Hz.) and at 49 Hz. and 99 Hz.

The high vibration levels 33 Hz, 66 Hz and 83 Hz are of mechanical nature.

The high vibration level at 16.5 Hz most probably is an acoustical resonance problem.

The high vibration level at 49 Hz and 99 Hz. must be the result of:High pulsation forces without resonance (compressor bottle sizing problem)


Examination of Mechanical Behavior 66 Hz. Mode Shape

Large amplitude movement in suction manifold


Step 4. Modifications

The high vibration levels 33 Hz, 66 Hz and 83 Hz are of mechanical nature and need a mechanical solution

Better supportingImproved support stiffness

The high vibration level at 16.5 Hz is due to acoustical resonance and needs an acoustical solution, I.e. different bottles and/or orifice plates to introduce more damping

The high vibration level at 49 Hz and 99 Hz. are the result of high pulsation forces without resonance and must be resolved by compressor bottle (re)sizing.


“As Built” Supporting Structure of Compressor Manifold


Modified/Improved Supporting Structure ofCompressor Manifold


Modified Structure Implemented & Connected to Attached Piping


Conclusion from Example 1

Compressor vibration problems are of a mixed nature• Part is mechanical• Part is acoustical

Each category requires a different approach and result in different solutions

Not all vibration problems can be solved by mechanical measures.


Example 2

ProblemFailure in cooling pump discharge line (possibly vibration induced)

Steps to solutionNo vibration measurements just visual observationIdentification of excitation mechanism thru fluid simulations (pump trips & start-up and check valve closures)Time history mechanical simulation to verify stress levelsIdentification of source of vibration problemModification proposal


Steady State Volume Flowrate [m3/s]

Model of the System


Pump trips at t=1 seconds,

Pump inertia: 8 kgm2

Time History of Pumps


Check valve closes in 0.5 second

Time History of Valve


Flow thru checkvalve

Time History of Flow through Valve


Maximum Transient Pressure during pump trip [Barg]

Maximum Transient Pressure


Unbalanced Load Time Histories (I)

Load in Newton * 104



Unbalanced Load Time Histories (II)



Unbalanced Load Time Histories (III)



Unbalanced Load Time Histories (IV)


Unbalanced Load Time Histories available in CAESAR


Conclusion from Example 2:

Alternating stress amplitude of 233 MPa results in stress range of 466 MPamay be responsible for LCF


Example 3

ProblemExcessive pressure in injection line.

Steps to solutionAssumed mechanism entrapped (undrained) fluid propelled by gas at gas velocity during start-up hits valve that is cracked openIdentification of source of problem thru simulation


Valve Damage


Model of the System


Pressure Time History at the Ball Valve in Case of Entrapped Gas at the Valve

Valve cracked open at t=0

Pressure in Barg


Maximum Transient Pressure in trapped gas

0

50

100

150

200

250

300

350

400

450

500

0 100 200 300 400 500 600 700 800 900 1000

Trapped gas volume [Liter at atmospheric pressure]

Max

imum

Pre

ssur

e [B

arg]

.

Slug 29 liter, 20 meter Slug 130 liter, 90 meter

Pressure Time History of Entrapped Gas


Conclusions for presentation

Dynamic analysis is an important and sometimes a critical element in integrity analysis

• Many failure modes are to be addressed• Excessive loads are to be handled

CAESAR offers several types of dynamic analyses to assist in thedemonstration of integrity:

1. Harmonic Analysis2. Spectrum Analysis3. Time History Analysis

Solutions to possible problems are often found by introduction of the right supporting and/or supporting steel structure


END

Thank you for your attention

Documents

DYNAFLOW · PDF filePipe stress engineers prefer quasi static approach ... -Slug load calculation (manual??) ... • Analysis accuracy is increased if support structure is