benchmark April 2012 - NAFEMS · 2018-05-03 · Simulation Knowledge Management in GKN Driveline....

3
sdm 10 benchmark april12 K eeping track of our simulation data and results becomes more challenging every year as the number of files we need to manage grows at an ever increasing rate. Audi (1) estimate that their analysts are generating a hundred and twenty times as many files as a decade ago while model granularity has grown by an order of magnitude. So we are all drowning in data. We can do more and more sophisticated simulations to more closely represent the physics of the problems we are studying. But the bad news is that this is going to create even more data and more complex, interlinked data sets. Simulation data sets are also becoming harder to manage manually. It’s difficult to keep track of what we changed in a complex data set for each iteration that we made. Technology for managing large simulation data sets has been around for a long time. The first COTS SDM platform was put into production by BMW(2) more than a decade ago. So there is now a substantial body of knowledge describing what early adopters have achieved. However the survey carried out by the NAFEMS Simulation Data Management Working Group found that SDM is not widely deployed, with less than half of all companies using any SDM at all. Neither were respondents using other types of systems to manage their simulation data. NAFEMS members cited a lack of evidence demonstrating the benefits of SDM for not adopting this technology, so the NAFEMS SDMWG has developed a White Paper on the Value of SDM to address this issue. The good news is that SDM solutions can off load analysts of many routine data manipulation and data management tasks. Audi and BMW applied SDM to automatically document which input files were used in a particular vehicle simulation. They also automated the extraction of key results, the storage of these results in the SDM system and insertion of the key results into report templates. This has enormously reduced repetitive, tedious and error prone work, enabling engineers to do more engineering. The CAE guy rightly pointed out in last July’s issue of benchmark that it would be really helpful to automatically create both a report and a presentation of a set of results. The core capability of an SDM solution to programmatically access the all simulation data in context enables any type of report or presentation template to be populated for any simulation or set of simulations. Several vendors now offer report generation modules for their SDM solutions. I have avoided implying that reports can be created automatically because it’s definitely necessary for the analyst to add their interpretation of the results. But it’s great for the solution to perform the basic chores of extracting and inserting all the standard sections, curves and even 3D viewables into a report or presentation template. An SDM solution also enables you to annotate information items, for example to add comments to qualify results. This means that it becomes less necessary to generate a report as a document since the results of a simulation and the analysts’ comments can be consulted directly from the database. While the core capability of a PDM system is to manage tree structures of assemblies, parts, documents and data items, the core capability of an SDM solution is to manage complete sets of simulation items and the linkages which connect them. This is known as a graph structure in data-management- speak. For a finite element analysis,

Transcript of benchmark April 2012 - NAFEMS · 2018-05-03 · Simulation Knowledge Management in GKN Driveline....

sdm

10 benchmark april12

Keeping track of oursimulation data and resultsbecomes more challengingevery year as the number offiles we need to manage

grows at an ever increasing rate.Audi (1) estimate that their analystsare generating a hundred andtwenty times as many files as adecade ago while modelgranularity has grown by an orderof magnitude. So we are alldrowning in data. We can do moreand more sophisticated simulationsto more closely represent thephysics of the problems we arestudying. But the bad news is thatthis is going to create even moredata and more complex, interlinkeddata sets. Simulation data sets arealso becoming harder to managemanually. It’s difficult to keep trackof what we changed in a complexdata set for each iteration that wemade.

Technology for managing largesimulation data sets has beenaround for a long time. The firstCOTS SDM platform was put intoproduction by BMW(2) more than adecade ago. So there is now asubstantial body of knowledgedescribing what early adoptershave achieved. However the surveycarried out by the NAFEMSSimulation Data ManagementWorking Group found that SDM is

not widely deployed, with less thanhalf of all companies using anySDM at all. Neither wererespondents using other types ofsystems to manage their simulationdata. NAFEMS members cited alack of evidence demonstrating thebenefits of SDM for not adoptingthis technology, so the NAFEMSSDMWG has developed a WhitePaper on the Value of SDM toaddress this issue.

The good news is that SDMsolutions can off load analysts ofmany routine data manipulationand data management tasks. Audiand BMW applied SDM toautomatically document whichinput files were used in a particularvehicle simulation. They alsoautomated the extraction of keyresults, the storage of these resultsin the SDM system and insertion ofthe key results into reporttemplates. This has enormouslyreduced repetitive, tedious anderror prone work, enablingengineers to do more engineering.The CAE guy rightly pointed out inlast July’s issue of benchmark thatit would be really helpful toautomatically create both a reportand a presentation of a set ofresults. The core capability of anSDM solution to programmaticallyaccess the all simulation data incontext enables any type of report

or presentation template to bepopulated for any simulation or setof simulations. Several vendorsnow offer report generationmodules for their SDM solutions.

I have avoided implying thatreports can be createdautomatically because it’s definitelynecessary for the analyst to addtheir interpretation of the results.But it’s great for the solution toperform the basic chores ofextracting and inserting all thestandard sections, curves and even3D viewables into a report orpresentation template. An SDMsolution also enables you toannotate information items, forexample to add comments toqualify results. This means that itbecomes less necessary to generatea report as a document since theresults of a simulation and theanalysts’ comments can beconsulted directly from thedatabase.

While the core capability of a PDMsystem is to manage tree structuresof assemblies, parts, documentsand data items, the core capabilityof an SDM solution is to managecomplete sets of simulation itemsand the linkages which connectthem. This is known as a graphstructure in data-management-speak. For a finite element analysis,

sdm

11benchmark april12

this graph structure starts withgeometry, loads and materials onthe left, progressing throughmeshing, assembly into an inputdeck, solution, post-processing, keyresults and finally reporting on theright side of the screen. The graphstructure provides traceability andenables you to navigate from inputdata through process steps toresults.

Like PDM solutions, SDM solutionstrack information items andassociated metadata as well as thedata files. Several representationsof the same data can be attachedto, and accessed from aninformation item. For simulationoutputs, in addition to the bulkdata file, these representations caninclude numeric values stored asmetadata, viewable graphs, anExcel file, a lightweight viewable3D file and a compressed version ofthe output file. The original bulkdata file can be archived or evendeleted whilst retaining theinformation derived from thesimulation and the traceability backto the inputs through the graphstructure.

An SDM solution provides thecapability to provide controlledaccess to simulation data sets sothat they can be consulted byanyone with appropriate accessrights, inside or outside theorganisation. This can lead tosignificant time savings in searchingfor and imparting information.Different studies (3) (4) have foundthat analysts spend between 30and 50 % of their time searchingfor or imparting information. So anaccessible and largely self-explanatory mechanism for

searching, navigating and accessingsets of simulation data allowsengineers to do more engineering.

Typical early adopters of SDM areCAE departments in Aerospace andAutomotive OEMs developingcomplex, high performanceproducts which are refinediteratively. Multiple iterations makeit worthwhile to invest in a solutionthat enables automation andreduces the effort to re-runsimulations with modified inputs.Early adopters have been able toretain the Intellectual Property(IP)from their simulation-basedexperimentation just as they retaintheir test results. Audi (1) reportedthat they had accumulated theresults of 500,000 vehiclesimulations by the end of 2010 andcan access both physical andnumeric test results for anysimulation through a web browser.The NAFEMS SDM Business ValueWhite Paper incudes an extensivebibliography of papers describingend user experiences. In addition,the papers from the 2011European SPDM conference arenow available to NAFEMSmembers.

So is SPDM the same as SDM?Growing interest in gettingsimulation data under control hasled to acronym wars around whatis an SDM system. The NAFEMSwhite paper sets out the corecapabilities of an SDM systembased on the experience of earlyadopters. These include thecapability to capture processinformation which is the necessarycomplement to the input data filesto fully define what was done andwhy. A range of CAE process

implementation solutions arewidely used in industry, fromscripting through process definitionsolutions to stochastic iterationengines. The author argues that it’sthe ability to integrate an analyst’spreferred CAE automation solutionand capture the inputs and outputsthat matters, rather than the needto contain a process manager. It’sclear that the capability of an SDMsolution to enable processautomation can yield largeproductivity gains.

The world of simulation is highlyheterogeneous with manyorganisations using tens tohundreds of applications, whereasthe CAD world is now reduced to ahandful of systems. So anothercore SDM capability is to deal witha diverse range of data types andapplications. Peter Bartholomewpresented an interesting paper (5)on the topic of the capabilitiesrequired to manage simulation datato the RAeS in 2009. He pointedout that the capabilities required ofSDM by an analyst investigatingcomplex aerospace phenomena arevery similar to the capabilitiesrequired by an Aerospace OEM.

Since the SDM solutionrequirements are similar from theOEM to the individual consultant,it’s in the OEM’s interest to provideaccess to their SDM system to theirsuppliers to ensure the traceabilityof externally sourced simulationwork. In the case of Audi, half theanalysts registered on the SDMsystem are external third parties. Inaerospace, a 50M€ project led byAirbus called CRESCENDO (6) isinvestigating the implementation ofSDM solutions across the virtual

Figure 1: Display of a Set of Simulation Data in an SDM Solution (Courtesy of MSC Software)

sdm

12 benchmark april12

enterprise. CRESCENDO isdemonstrating the feasibility andvalue of SDM to improve aircraftdevelopment processes. Anotherkey benefit of SDM is the ability toaccess remote simulation resources(7), both physical and human, withlower supervision and collaborationcosts.

If SDM can deliver all thesebenefits, why are relatively feworganisations using SDM today?One reason has been solutioncosts. In the PDM space, capableand affordable PDM solutions likeSmarteam emerged to meet theneeds of small and mediumbusinesses; even open source PLMsolutions are available. Because ofthe complex requirements of anSDM solution and the relativelysmall number of analystsworldwide, compared to CADdesigners for example, fewdedicated SDM solutions have beendeveloped and not all PLM vendorspropose an SDM solution on theirplatform. Some CAE vendors offerdata management solutions tightlyintegrated with their solutions.These latter solutions are attractiveas long as you only need tomanage the data types included inthe integrated SDM solution.Nissan (8), Ford (9) and Petrobras(10) reported significant gains fromsuch specific, mono-vendorimplementations.

Continued vendor investment inOOTB core functionality and value-added capabilities is decreasing thecosts of implementation andincreasing Return On Investment.BMW (11) announced that they areswapping out their customdevelopments for OOTB SDMapplications. This means that otherorganisations will now be able toacquire as configurable

applications, high end SDMcapabilities such as simulationgenerators that previously had tobe built as bespoke developments.Simulation generators are anadvanced, add-on, SDM capabilitythat enables load-cases and data-sets to be organised so thatsimulations can be re-run onmodified input data with minimaleffort. This is extremely valuable foriterative investigations on complexdata-sets or problems likepedestrian impact on cars wheremany possible positions of thepedestrian need to be considered.

SDM is a dynamic and evolvingdomain that can bring real value toanalysts and enterprises alike.NAFEMS has an SDM conferenceplanned for the USA this yearwhere end users will present theirexperiences. The NAFEMS SDMBusiness Value White Paper thatwas distributed in draft form at the2011 SDM conference will bepublished shortly. Watch thisspace.

Mark Norris is an IndustryPrincipal with Infosys, anengineering, simulation & PLMservices provider and systemsintegrator. He has too manyyears’ experience of PLMconsulting and solution designaround Teamcenter, ENOVIA andSimManager. He is the authorof the SDMWG White Paper onthe Business Value of SDM andcan be reached [email protected]

Bibliography[1] REICHENEDER, Josef GRUBER, Karl.

Experience with 7 Years SDM atAudi. NAFEMS SDM Conference2010.

[2] HAGELE, J HANLE, U KROOP, ASTREIT, M KERNER, CSCHLENKRICH, M. The CAE-BenchProject – A Web-based System forData, Documentation andInformation to Improve SimulationProcesses. s.l. : MSC VPDConference, 2000.

[3] KRASTEL, Marcus. SimPDM –Integration of Simulation andComputation in a PDMenvironment. ProSTEP iViPSymposium 2008.

[4] ROBINSON, Mark. An empiricalanalysis of engineers' informationbehaviors. Journal of the AmericanSociety for Information Science andTechnology, Wiley, 2010.

[5] Bartholomew, Peter. The use ofMetadata to maintain the Integrityof the Analysis Process. RAeSCertification and Validation Seminar25 May 2010.

[6] COLEMAN, Peter. Developing theBehavioural Digital Aircraft.Aerodays 2011www.aerodays2011.org.

[7] MINTER, Steve VALINE, Glenn.Simulation KnowledgeManagement in GKN Driveline.NAFEMS Wiesbaden, Germany2009.

[8] KIM, Daniel. ImplementationFrameworks of SDM (SimulationData Management) in NissanTechnical Center, NA. CPDACincinnatti 2011.

[9] MCCOY, Robert. AutomatedRollover Test Simulation Tool. CPDADetroit 2010.

[10] Simulation Data and ProcessManagement with ANSYSEngineering Knowledge Manager atPetrobras. Bhide, S Largura, L C.NAFEMS SPDM conference 2011 :s.n.

[11] SCHLUTER, Wolfgang THAM,Carsten. BMW 2nd GenerationSimulation Data Management –Experience, Infrastructure andMigration. NAFEMS SDMConference Franfurt 2010.

AUTHOR INFORMATIONMark Norris I [email protected]