tp151

7/29/2019 tp151

1/9

An Impeller Dynamic RiskAssessment Toolkit

Donald M. SchifferSupervisor, Solid Mechanics

Dresser-Rand CompanyPaul Clark DriveOlean NY 14760

Phone: 716.375.3170Email: Donald_M_Schiffer@Dresser-

Rand.com

Asadullah SyedAnalysis Engineer

Dresser-Rand CompanyPaul Clark DriveOlean NY 14760

Phone: 716.375.3314Email: [email protected]

Two-Sentence Synopsis

A standardized and cost-effective risk assessmenttoolkit is described to comprehensively analyzecentrifugal compressor impellers for stress understeady and dynamic loads. This advanced analysisapproach has been developed using a blend of

automated computer simulation and broadexperience base resulting in greater reliability andprofitability for both client and manufacturer.

7/29/2019 tp151

2/9

2011 Dresser-Rand. All Rights Reserved. 2

Donald M. Schiffer

Donald Schiffer is the Supervisor of SolidMechanics at Dresser-Rand Company, in Olean,New York. He has 28 years experience withturbine and compressor structural design andanalysis. His main focus has been rotatingcomponent dynamics and the development ofintegrated stress design systems for steamturbines and centrifugal compressors.Mr. Schiffer has a BS degree from SyracuseUniversity and a BSME from Clemson University(1976). He is a member of the ASME.

Asadullah SyedAsadullah Syed is an Analysis Engineer withDresser-Rand Company, in Olean, New York. Hisresponsibilities include structural design and

analysis of various rotating and stationaryturbomachinery components. He has experiencewith remaining life assessment of turbomachinerycomponents using fatigue and fracture mechanicstechniques.Mr. Syed holds a B.S. degree (MechanicalEngineering, 1999) from Jawaharlal Nehru

Technological University, in Hyderabad, India,and an M.S. degree (Mechanical Engineering,2001) from The University of Toledo, in Toledo,Ohio. He is a member of the ASME and ASM.

ABSTRACT

A standardized and cost-effective riskassessment toolkit is described in this paper tocomprehensively analyze centrifugal compressorimpellers under steady and dynamic loads. Thefirst step in this process occurs in the proposalstage where the gas loading levels are appraisedusing a preliminary screening computer program.

The application engineers execute the productconfiguration program for each proposal thatinvokes the screening program. All designs are

screened for steady gas bending stress levels foreach client-required operating point. Anadditional dynamic audit is performed if thesestresses are above a given threshold level. Thealternating loads are derived from upstream anddownstream flowpath disturbances. Naturalvibration modes that need to be avoided areevaluated using the SAFE diagram and harmonicresponse analysis, followed by automated results

postprocessing. The overall minimum factor ofsafety is determined for all impeller modellocations and known excitations using a fullyautomated process.

The use of this advanced analysis approach, thathas been developed using a blend of automated

computer simulation and broad experience base,results in greater reliability, profitability, andreduced risk for both client and manufacturer.Practical real-world examples of impellersanalyzed using this approach will be presented inthis paper.

INTRODUCTION

Centrifugal compressor impellers are beingengineered today with higher power densities,increased gas pressures, smaller package sizes,

and the need to minimize initial project costs. Atthe same time, the requirement to maintain andimprove impeller reliability must be preserved. Adynamic assessment of impeller stages is requiredto achieve these objectives. An impeller is a keycomponent of a compressor. It is subjected toinlet and exit flow variations through the stage,and therefore, it must be designed to withstandthe alternating pressure loads due to thesevariations in addition to withstanding steadyloads. While the requirement for reliable androbust impeller design can be addressed with

elaborate engineering analyses and testing,including static/dynamic FEA and transient CFD,it is still subject to time and cost constraints.

These analyses could become cost prohibitive forthe custom compressor manufacturer that designseach machine uniquely for client requirements.

The historical approach is to satisfy steady statestress requirements and consider modalinterferences on special designs or field issues.

The use of an interference or SAFE diagram is anexcellent first step in the dynamic process, butimpellers will often show multiple interferences

between mode shape and excitation shape with noclear method to resolve the severity. Althoughthis will definitely help improve reliability, theabsence of the analytical component dealing withthe evaluation of alternating loading and resultingstress at potentially resonant frequencies remainsan issue. In order to calculate alternating stresslevels, it is essential to estimate the dynamicpressure forcing functions. The loading on an

7/29/2019 tp151

3/9


impeller can be easily influenced by severalaerodynamic flow phenomena such as swirl, wakeformation, flow separation, stall, surge and otherfluctuations. Marshall and Sorokes [1] havedescribed the various flow phenomena causingforced vibrations of centrifugal compressor rotors.

Impeller vibrations are adversely affected by thesame phenomena. Due to the inherent complexityof these phenomena, accurate estimation of thealternating forcing function becomes difficultwithout a high-end transient CFD analysis and/ortesting. Another approach to estimate the forcingfunction has been to use a portion of the steadyloading for the alternating load component. Theturbine industry has long used steady stagehorsepower for estimating the excitation levels.However, this approach must be validatedthrough various channels, including CFD and

field experience, in order to be successfullyapplied to different impeller families withvariable geometry.

IMPELLER STRESS DESIGNCHALLENGES

Steady Stress Considerations

A first key element of any impeller stress designprocess is to ascertain the steady load carryingcapacity of a particular design. This, in turn, is

governed by the properties of materials used forthe impeller. Various approaches have been usedto predict the centrifugal load deflection andstress on impellers including 1D, 2D and peaknodal stress in 3D solid element FEA models.Geise P., et. al. [2] have presented a processwhere impeller overspeed limits are establishedbased on steady state FEA impeller models usingthe peak element stress. The elemental stressapproach based on a standardized mesh has beenshown to provide excellent correlation with testdata.

Modal Analysis

In the subsequent steps of the process, theimpeller natural frequencies are calculated bymeans of a modal analysis. In turbomachinerydesign, the Campbell diagram has been used formany years, but the SAFE diagram [3 and 4]

represents a significant improvement withadditional mode shape information.

The SAFE diagram is used to check if vibrationmodes fall within the speed range of a machine. Ifthey do, the most prudent solution would seem tobe to make modifications to the design so as to

move the natural frequencies out of the speedrange. However, for a complex structure such as aclosed impeller, this usually proves to be achallenge. In most cases, there are several modefrequencies which could be excited by the forcingfrom stationary structures upstream anddownstream of the impeller. In those cases, therisk usually can be minimized by changing thevane count. Although this solution may addressthe problem for a simple case, there are severalcases where a change in the count of stationariesdoes not provide an interference-free situation and

still results with some modes ending up in thespeed range at either the fundamental excitationor at higher harmonics of that excitation. Thissituation then needs to be addressed by evaluatinghow much of a risk any given interference poses.

This can be achieved by conducting a harmonicresponse analysis for the resonant conditions ofconcern and then evaluating the alternating stresslevels at those conditions.

Excitation Estimation

The challenge in doing a harmonic responseanalysis comes from the difficulty in accuratelyestimating the alternating load levels on animpeller for each excitation type. While high-endCFD analysis or testing can be employed toestimate the loading, it is highly unlikely to beused on every impeller being fabricated at amanufacturers facility mainly due to constraintsof time and resources. On the other hand, thestage horsepower can be successfully used toestimate the alternating loads on an impellerwithout additional time or cost to the overall

project. Figure 1 shows the estimation ofalternating load from the steady horsepower loadon a stage.

Figure 1. Alternating Load Estimation

7/29/2019 tp151

4/9


As shown in Figure 1, the derived stimulus iscomputed as a ratio of the alternating load Aand the steady load S. An initial investment inan effort to validate this method with CFDanalyses, testing and experience will providereturns by enabling higher power density impeller

designs and reducing warranty costs by improvingreliability in the long term.

Harmonic Response Analysis

A commercially available FEA code will apply aspecified harmonic loading across a frequencyrange to any structure and predict the resultingdisplacements and stress levels. Of particularinterest are the solutions at each naturalfrequency. Unless a code with cyclic symmetryresponse capability is used, a full 360-degree

model is required. A typical impeller may haveinlet guide vanes (IGV) upstream and adownstream diffuser with low solidity diffuser(LSD) vanes. The analyst must map these loadson the 360-degree FEA model and solve for thealternating stresses at each resonance. The analystor external program must also postprocess theenormous quantity of output data generated bythese solutions. Recent advancements intechnology have resulted in increased computingpower and disk capacities on computers forprocessing and storage of these enormous data.

High capacity disk configurations can be utilizedto store and process the output generated by thesolutions.

TOOLKIT DEVELOPMENT

Historical Results Analyzed

An important step in developing a process for riskassessment was the evaluation of historical data.

This was achieved by collecting vital fieldexperience, including field incidents in order to

assess the details of operation and levels of stressin each case. A large impeller geometry databasewas created to include all the field experiencedata. It should be noted that it is essential to havedata of both successful and failed impellers inorder to relate the loading history to anyacceptability indicating parameter. An automatedapplication was created to execute the evaluationrun on all cases in the database. Once this was

accomplished, the challenge was now to interpretthe results in order to relate to the trends ofsuccess and failure. This was achieved byrelating the trends as a function of gas bendingload on the impeller. The analysis levels werethen established using these trends. Figure 2

shows the dynamic assessment value comparisondue to the analysis performed.

Figure 2. Dynamic Assessment Value

When no detailed advanced analysis is conducted,the historical warranty costs were maximum withthe historically associated analysis costs at aminimum. The implementation of this newtoolkit is expected to significantly lower thewarranty costs while adding a low fraction of the

warranty savings to the dynamic design analysisthat is required as a result of the new process.

Risk Assessment Toolkit

One of the key elements of this toolkit is the riskassessment filter screen which is accessed at theproposal stage in the sales regions of orderquotation. Figure 3 shows the flow of thisprocess.

Figure 3. Risk Assessment Process Flow

The order quotation personnel in the sales regionsare the first important link in this chain. Byhelping identify the gas bending stress levels inthe selected impellers based on aerodynamic datafor that stage, this tool will facilitate applicationengineering personnel in making the rightselection of impellers for a particular applicationat the start of the cycle. As shown in the flowdiagram, all impellers are screened based on inputfrom the aerodynamic performance database and

an impeller geometry database. The designscreening program then calculates the gas bendingstress levels in the impeller based on a worstcase loading situation estimated automaticallybased on various performance parameters. Oncethe stress levels are calculated, a decision must bemade on the acceptability of the stress levels for aparticular design. Once the highly loadedimpellers are identified, they are flagged for

7/29/2019 tp151

5/9


further analysis. This approach facilitates multipledynamic analysis levels based on gas bendingstress, construction method, materials selected,flow coefficient and other design parameters. Asshown in Figure 3, depending on the load leveland other parameters, the future dynamic analysis

direction is chosen: 1) no additional analysis, 2)standard dynamic audit including modal andharmonic response analysis, or 3) high-endtransient CFD analysis or testing followed by aresponse analysis. The results of this earlyassessment are automatically communicated tovarious departments involved through electronicmail notification leading to a streamlined processin the project cycle.

Automated Dynamic Audit

The core process of this toolkit is the standarddynamic audit described in more detail below.Once a notification for performing a riskassessment is received, the next step is to generatea finite element (FE) model for the impeller to beassessed. The model being generated must have amesh density that is adequate for the accuratecomputation of impeller natural frequencies.Impeller modal tests were conducted followed bycorrelation studies to establish acceptable levelsof FE mesh density to be used on models. Figure

4 shows a typical mesh density used for a closedimpeller.

Figure 4. Typical Impeller Mesh Density

The entire analytical process is automated andfollows the flow shown in Figure 5.

Figure 5. Automated Dynamic Audit Process

Each solution routine is conducted by a

commercially available FEA code. The automatedprocess starts with a static solution at themaximum compressor speed with the availableFEA model for an impeller, and the mean stress ateach node in the model in computed. Alternatingpressures resulting from upstream anddownstream vanes are placed on one impellerblade pitch and converted to nodal forces. Oncecomplete, the model is ready for a modal solution

to calculate impeller natural frequencies ofvibration including prestress resulting fromcentrifugal speed and shrink load. The modalsolution will compute impeller naturalfrequencies that are used to generate a SAFEdiagram in order to help identify any interference

with excitations. The single blade nodal forces arecopied to all blades in the 360-degree model forthe nodal diameter pattern required with eachexcitation. The displacement harmonic responseis then computed for all excitations. A responsestress solution then calculates the alternatingstress levels for all the high-risk modes. The meanstresses for each interference are calculated byscaling them from the centrifugal stressespreviously calculated at maximum compressorspeed. Finally, a factor of safety is calculated ateach node in the FE model at each identified

frequency. An additional program feature is theability to consider the weld material properties inthe calculation of the factor of safety. Theminimum factor of safety is summarized for allvibration modes and all excitations within themachine speed range. The model generationprocess is standardized and the entire remainingprocess is fully automated so that operatorintervention is not normally required. Thedynamic audit process has resulted in well over anorder of magnitude reduction in solution effortand detailed dynamic calculations on production

impellers can now be performed efficiently.

CASE STUDIES

Case Study 1

The intent of this case study is to examine thesuccessful application of the risk assessmenttoolkit to a field problem that was resolved byredesign of the impeller before the release of thistoolkit. Several trains with the identicalcompressor using similar impellers were in

service at various locations. An impeller hadexperienced field incidents with two issues. Ascallop-shaped piece broke off at the OD of thedisk and there were crack indications near theblade leading edge. This impeller design (Case A)had 15 blades and 16 IGVs. Figures 6 and 7 showthe typical mode shapes associated with this kindof indication pattern.

7/29/2019 tp151

6/9

7/29/2019 tp151

7/9


high alternating stresses and a low factor ofsafety. Selecting such a speed range would thenresult in a highly reliable impeller design.

CONCLUSION

The purpose of this paper was to present a riskassessment toolkit developed by the authors andin use at their company. The existing staticanalysis approach has been greatly enhanced bythis toolkit and a fatigue-based approach can nowbe used with much reduced effort and time toaddress impeller reliability. Historical incidentswere used to establish levels of analysis requiredas a function of impeller loading. Each impellerchosen by the application engineers is screenedusing the computer program at the proposal stage

and identified for potential risk. The dynamicaudit tool is fully automated to perform modalanalysis and harmonic response analysis, andprovides a consistent approach to calculate afactor of safety for all impellers built by themanufacturer. In summary, the use of this toolkitresults in better understanding of impellers in adynamic environment and provides greaterreliability, profitability, and reduced risk for boththe client and the manufacturer.

NOMENCLATURE

CFD Computational Fluid DynamicsFEA Finite Element AnalysisBLEM Blade Leading Edge ModeSOP Scallop Out of PhaseRPM Rotations per MinuteOD Outer DiameterIGV Inlet Guide VaneLSD Low Solidity DiffuserHP Horsepower

REFERENCES

1. Marshall D. F., Sorokes J. M., 2000, AReview of Aerodynamically Induced ForcesActing on Centrifugal Compressors, andResulting Vibrating Characteristics ofRotors, Proceedings of the Twenty- Ninth

Turbomachinery Symposium,Turbomachinery Laboratory, Texas A & M

University, College Station, Texas, pp 263-281

2. Cameron D. W., Geise P. R., Abbott J.S.,Establishing Overspeed Limits forCentrifugal Compressor Impellers,

http://www.dresser-rand.com/e-tech/turbo.asp.

3. Singh M. P., Vargo J. J., Schiffer D. M., andDello J. D., 1988, Safe Diagram A DesignReliability Tool for Turbine Blading,Proceedings of the Seventeenth

Turbomachinery Symposium,Turbomachinery Laboratory, Texas A & MUniversity, College Station, Texas, pp 93-101.

4. Singh M. P., Thakur B. K., Sullivan W. E.,and Donald G., 2003, ResonanceIdentification for Impellers,Proceedings ofthe Thirty-Second TurbomachinerySymposium,Turbomachinery Laboratory,

Texas A & M University, College Station,Texas, pp 59 -70

ACKNOWLEDGEMENTS

The authors would like to thank Dresser-Rand forallowing the publication of this paper.

7/29/2019 tp151

8/9


FIGURES WITH CAPTIONS

Figure 1: Alternating Load Estimation

Figure 2: Dynamic Assessment Value

Figure 3: Risk Assessment Process Flow

Figure 4: Typical Impeller Mesh Density

Figure 5: Automated Dynamic Audit Process

Figure 6: OD Scallop-Out-of-Phase mode

7/29/2019 tp151

9/9


Figure 7: Blade Leading Edge Mode

Figure 8. Dynamic Audit Results - Case Study 1,Cases A, B and C

Figure 9. Dynamic Audit Results - Case Study 2

Figure 10. Dynamic Audit Results - Case Study 3

Documents

tp151