Assistant Head Engineering Analysis Branch NASA Langley Research Center Hampton, Virginia 23681 |
 |
Aerospace Engineer Engineering Analysis Branch NASA Langley Research Center Hampton, Virginia 23681 |
Assistant Head Engineering Analysis Branch NASA Langley Research Center Hampton, Virginia 23681 |
|
Aerospace Engineer Engineering Analysis Branch NASA Langley Research Center Hampton, Virginia 23681 |
| ASET | Analysis Set in NASTRAN used to represent an Accelerometer Set |
| BDF | Bulk Data File |
| DMAP | Direct Matrix Abstraction Programming |
| DOF |
Degrees-of-Freedom |
| FEM | Finite Element Model |
| PCL | PATRAN Command Language |
| [dx] | Test / Analysis Cross-Orthogonality Matrix |
| [FA] | Mass-normalized Analysis Eigenvectors |
| [FX] | Mass-normalized Test Eigenvectors |
| [MA] | Analytical Mass Matrix |
![Matrix Equation: [Phi]-Transposed [M] [Phi] = [delta]](Image1.gif){kind=small}
![Matrix Equation: [Phi sub X]-Transposed [M sub A] [Phi sub A] = [delta sub X]](Image2.gif){kind=small}
To make TACT an intuitive tool for PATRAN-based analysts, emphasis was placed on making TACT look like an integrated part of PATRAN. Just like other PATRAN tools, the user enters the TACT tools through a pull-down menu from the PATRAN main menu bar. All of the individual utilities in TACT are PATRAN forms; these forms look the same as all the other PATRAN forms and have the same widgets (buttons, slide bars, list boxes, select data boxes) that the PATRAN user is already knowledgeable in manipulating. The TACT on-line help is written in the native PATRAN help documentation format (FrameMaker) and has a hyperlinked table-of-contents that looks like MSC’s on-line PATRAN documentation.
The TACT main menu was designed to be a step-by-step guide for analysts unfamiliar with the correlation process. There are four main sections in the TACT main menu: Pre-Test Modules, Post-Test Modules, ‘View Results’ Modules, and TACT Help. These sections are described in detail in the following sections. The overall organization of the TACT utilities is shown in the software map in Figure 2.
The successful process of correlating a FEM to modal survey test data begins long before the test ever takes place. The analyst will never be successful in achieving correlation of the FEM to the test data if steps are not taken to make sure that the test data is adequate for the analyst’s needs. It is critical that the analyst and test engineer work together to define a set of accelerometers for the test that are physically accessible yet analytically adequate to capture the necessary test results.
After creating a FEM of the test article, the analyst is ready to use the Pre-Test Modules in TACT to analytically evaluate candidate accelerometer locations and orientations for the test. TACT uses NASTRAN to calculate a cross-orthogonality matrix between results from the full FEM and results from a reduced math model that describes the modal behavior that a particular accelerometer set would be capable of capturing. There are five utilities in the Pre-Test Modules: Run Full FEM Job, Create/Modify ASET, Create/Modify Stick Model, Run Full FEM/ASET Job, and Export ASET UFF.
Run Full FEM Job. After the analyst has created and is satisfied with a FEM of the test article, the first step in the Pre-Test process is to "baseline" the full FEM. TACT modifies the full FEM NASTRAN Bulk Data File (BDF) to include the NASTRAN DMAP in Figure 3, and automatically executes NASTRAN; the DMAP causes NASTRAN to punch the full FEM eigenvalue and eigenvector matrices to DMI cards. The DMI cards are included in future analysis BDFs, and additional DMAP is used to compare results for a particular accelerometer set to these full FEM results.
Create / Modify ASET. After the full FEM has been baselined, the analyst is ready to define sets of accelerometers, which are abbreviated in TACT as ‘ASETs’. Analysts can create and modify ASETs using the form shown in Figure 4 by manually entering or ‘screen picking’ nodes they wish to identify as candidate accelerometers. The user can select the degrees-of-freedom (DOF) to be measured at each accelerometer location; the default DOF assignment is a triaxial accelerometer (XYZ). Accelerometer numbers can be manually or automatically assigned, and the user can modify them at any time. External to PATRAN, TACT establishes and manages directories and results file storage for each ASET.
Create / Modify Stick Model. This tool helps the analyst to define a "stick model" which simply "connects the dots" between accelerometers for easy test result visualization. Definition of the stick model can be done using manual data entry or by selecting entities from the viewport. Figure 5 shows a PATRAN viewport of the research wind tunnel model FEM after stick model creation; notice that the accelerometer numbers and DOF are shown in the viewport along with the stick model.
Run Full FEM / ASET Job. After the analyst has created an ASET and a stick model, the next step in the Pre-Test process is to analytically determine the adequacy of the ASET for capturing the desired modes during the modal survey test. The ‘Run Full FEM/ASET Job’ form prepares and submits a NASTRAN deck that uses the DMAP in Figure 6 to calculate frequency differences and a cross-orthogonality matrix between the full FEM results and the ASET results. The analyst can view the results of this analysis using the tools in the ‘View Results’ Modules. The analyst then modifies the ASET and reviews results in an iterative fashion (shown in Figure 2) until an acceptable ASET has been defined.
Export Analysis Data. Once the analyst and test engineer have settled on the accelerometer DOF and locations, the analyst is ready to export the analytical predictions to the test engineer. The Export Analysis Data form exports a single Universal File Format (UFF) file containing accelerometer locations (U15), "stick model" connectivity (U82), reduced analytical mass matrix (U247 & U252), and eigenvalue/vector results (U55), which can be imported into many test manager software codes, such as I-DEAS.
Once the modal survey has been completed, the test engineer will post-process the results and deliver them to the analyst. Assuming the test results are valid, the analyst then begins the process of modifying the FEM so that the analytical results match the test results. Normally, changes to the FEM are made incrementally, and the effect of a particular FEM modification on the correlation is evaluated after each change.
To aid the analyst in performing this evaluation, TACT provides two Post-Test utilities, Import Test Results and Run ASET/Test Job, which are discussed in the sections that follow.
Import Test Results. This TACT utility reads in Universal File Format U15, U55, and U82 data types from a single UFF file, and stores the data for animation and correlation calculation in PATRAN and NASTRAN. TACT has a user-selectable "Auto-transform" import capability that automatically transforms test data to match the analytical model’s units and coordinate system orientation.
Run ASET / Test Job. After the analyst has imported test results, the next step in the Post-Test process is to calculate the correlation between the test results and the current FEM. The ‘Run ASET/Test Job’ form prepares and submits a NASTRAN deck that uses the DMAP in Figure 7 to calculate frequency differences and a cross-orthogonality matrix between the ASET and test results. The analyst can then view the results of this analysis using the tools in the ‘View Results’ Modules. The analyst makes adjustments to the FEM that will improve the correlation, recreates the BDF, and reviews results in an iterative fashion (shown in Figure 2) until acceptable correlation has been achieved.
Any time the analyst wishes to compare ASET modal results to either the full FEM or test results, the same basic capabilities are needed. First, the analyst needs to be able to view and post-process the frequency percent difference and cross-orthogonality calculations. Secondly, the analyst needs to be able to view the mode shapes and be able to visually compare ASET results to either the full FEM or test results. In the first several iterations of trying to determine where the FEM should be changed or updated to better reflect test data, most analysts use information from visual inspection of the mode shapes in conjunction with the correlation calculations. The two ‘View Results' modules in TACT, Mode Pairs & Correlation and Animated Mode Shapes, are intended to provide a core set of tools for performing this function.
Mode Pairs & Correlation. This form, shown in Figure 8 using the low aspect ratio wind tunnel model results, provides the user with a single place to manipulate mode pairs and view the effects on the frequencies and cross-orthogonality results. TACT has an ‘Auto Pair' capability, which calculates which modes form likely mode pairs based on user defined thresholds of frequency percent difference and on- and off-diagonal values in the cross-orthogonality matrix; Figure 8 actually shows mode pairs predicted by TACT’s Auto-Pair function. TACT also has built-in Mode Pair Reports, which are optionally exported text files containing all mode pair results that are easily imported into test report or correlation report documents.
Animated Mode Shapes. The control panel for mode shape animation is shown in Figure 9. The user can display ASET and test mode combinations to the viewport based either on mode pairs or by selecting modes individually. The widgets on the form change depending upon this selection; in Figure 9, the user has selected Mode Pairs as the mode pairing method for display. All animation parameters on the form can be modified while animating or stepping through an animation.
TACT currently offers five different options for displaying animated mode shape results, including synchronized analysis and test results side-by-side, synchronized analysis and test results overlaid, and an animation of the difference between analysis and test results. Analysts might use any combination of these viewing methods to obtain an understanding of the differences between the ASET and test results. Figure 10 shows the viewport for the research wind tunnel model results using the overlay display method where both the ASET and test results are overlaid on the same undeformed stick model. Figure 11 shows the research wind tunnel model results using the adjacent display method where the results are shown next to each other on top of separate undeformed stick models.
TACT comes with an extensive 74-page on-line Reference Manual. This document was developed in FrameMaker, PATRAN’s chosen format for help documents, to ensure that all potential users would have access to the software (the FrameMaker viewer is distributed with PATRAN) and be familiar with its use. The TACT Reference Manual contains a hyperlinked table-of-contents to make it simple for users to locate their topic of interest. In addition to screen shots/explanations of every form and widget available in TACT, the Reference Manual includes installation instructions and system requirements, file structure descriptions, a trouble-shooting guide, and a list of planned enhancements.
To demonstrate the improvements that TACT offers the MSC/PATRAN- and MSC/NASTRAN-based structural analysis community, an evaluation of the modal correlation process using TACT within PATRAN and external third-party modal correlation packages was performed. Three external third-party modal correlation packages were included in this review: Microdyne, Inc.’s DYNAVIEW marketed by Practical Systems and Technology, Inc. (1-410-758-2008), Structural Dynamics Research Corporation’s (1-703-713-0006) I-DEAS Correlation, and LMS International’s (1-578-453-4100) CADA-X Link. All software was assessed based on the following criteria: (1) compatibility/integration with PATRAN/NASTRAN, (2) training required, (3) cost, and (4) improved efficiency over EAB’s current correlation process. Correlation of a NASA LaRC low aspect ratio research wind tunnel model was used as the case history for this study.
For EAB analysts, software must be compatible and integrate well with their primary structural analysis codes, MSC/PATRAN and MSC/NASTRAN. TACT was developed to allow analysts the ability to perform all modal correlation tasks totally within the PATRAN/NASTRAN environment. Third-party packages such as DYNAVIEW, I-DEAS Correlation, and CADA-X Link are somewhat compatible with these analysis codes; however, this integration and compatibility is complicated by data translation and transfer, operation on/in different and sometimes unfamiliar computer platforms/environments, and the inability to perform internal model updating in PATRAN.
Before the development of TACT, modal correlation tasks were performed in EAB using DYNAVIEW. DYNAVIEW is a PC-based modal correlation package that provides the user with the capability to display, animate, and correlate modal data obtained from experimental and analytical sources using two modules – Modal Manager and Correlation Manager. To utilize this software, analytical models and results, generated in EAB on a Silicon Graphics Indigo2 workstation using MSC/PATRAN and MSC/NASTRAN, were transferred via ftp software to the PC. This required the analyst to operate on an entirely different computer platform, physically moving between computers often located in different rooms. Since DYNAVIEW only imports files written in either Enterprise Software Products, Inc.’s Finite Element Modeling and Post-processing (FEMAP) neutral or universal file format, FEMAP was used to convert MSC/NASTRAN analytical models and results to this file format. Both the analytical and test data were then individually brought into the Model Manager. Once DYNAVIEW analytical and test models were created, they were correlated using the Correlation Manager. During the entire modal correlation process, data was transferred four different times using four different software tools. If model updating was necessary, the user returned to the workstation, modified the analytical FEM in PATRAN, and then started the modal correlation process over again.
I-DEAS Correlation and CADA-X Link are robust workstation-based modal correlation packages that offer the analyst a complete set of tools for comparing modal data obtained from experimental and analytical sources. These tools include pre-test accelerometer set selection and evaluation, excitation setup, geometric mapping, visual comparison, MAC, orthogonality, frequency comparison, COMAC, and frequency response synthesis. Both packages have additional products that can be purchased to allow for optimization of accelerometer and excitation locations, design sensitivity, and modal updating – features currently not available in TACT or DYNAVIEW.
Similar to DYNAVIEW, I-DEAS Correlation and CADA-X Link require data translation of MSC/NASTRAN FEM and results; however, unlike DYNAVIEW, data translation is performed within these packages. Before translation, the MSC/NASTRAN deck must be run with a ‘PARAM, POST’ card in the case control section to output the appropriate files for these codes. While CADA-X Link is run on a Silicon Graphics workstation, I-DEAS Correlation must be run on a Hewlett-Packard 45I workstation because of its compatibility with LaRC’s test data acquisition system. This system provides PATRAN users with two challenges: (1) transfer of analytical and test models and results across platforms and (2) operation in a different computer environment. Finally, these packages, like DYNAVIEW, require the user to perform model updating by returning to PATRAN, modifying the analytical FEM, and restarting the modal correlation process over again.
Another criterion in evaluating potential software packages is the amount of training required. At NASA, training and travel budgets have sharply declined over the years, pressuring organizations like EAB to cut their training and travel costs. Sending employees to software training performed either on- or off-site can cost these organizations thousands of dollars. In addition to these costs, there is lost productivity while the employee is training.
Training on TACT and DYNAVIEW is unnecessary. Analysts acquainted with the modal correlation process will find these codes intuitive and easy to use. For analysts unfamiliar with modal correlation, TACT and DYNAVIEW provide step-by-step instruction through their main menu organization and on-line help.
For most PATRAN-based analysts, training on I-DEAS Correlation and CADA-X Link is a necessity. These packages are complex and sometimes operate in unfamiliar computer environments. For example, MSC/PATRAN users unacquainted with I-DEAS will find working with Correlation very difficult; I-DEAS users unfamiliar with MSC/PATRAN will also find working with TACT challenging. Although the basic functionality is the same for PATRAN and I-DEAS, structuring of each code is extremely different.
As with training and travel, EAB has a limited amount of money available for purchasing and maintaining software, making cost an important criterion in assessing potential software packages. In EAB, modal correlation is only one of the many tasks performed; therefore, spending a significant amount of monetary resources on software used for a small percentage of work is unjustified.
The results of a cost evaluation of TACT and the above third-party packages with their necessary modules and translators are presented in Table 1. For each software package, the government fees (as of September 1997) associated with licensing/software development and annual maintenance are listed. It should be noted that these are the fees that NASA is charged by these companies. In some cases, these prices may differ from the fees charged to other companies and government agencies as part of technology exchange agreements that have been established between the two parties.
| Cost Item | TACT | DYNAVIEW | I-DEAS Correlation | CADA-X Link |
| One Paid-Up License/ Software Development | $8,100 (270 hrs) |
$4,400 | $19,000 | $19,300 |
| Annual Maintenance/ Development | $1500 (50 hrs) |
$995 | $2,565 | $2,895 |
The first cost item shown above is, for TACT, the software development cost (also shown in hours), or, for third-party packages, the purchase price for one license. The software development cost of TACT is substantially lower than the price of one license of I-DEAS Correlation or CADA-X Link. There are times when two or more analysts need to simultaneously use modal correlation software. TACT allows users to access the software when needed; however, additional users of third-party packages require additional licenses. For a fairly inexpensive package like DYNAVIEW, purchasing two or more licenses will exceed the cost of developing TACT. It is important to note that since TACT was developed at LaRC, this code is free to other organizations within NASA; organizations outside NASA will eventually have free access to the software through the Langley Software Server (LSS).
The second cost item included in Table 1 is the annual software maintenance fee. This fee encompasses technical support, software upgrades, and error corrections. Higher-priced correlation packages, such as I-DEAS Correlation and CADA-X Link, have higher annual maintenance costs. These higher costs further tap limited software funds.
I-DEAS Correlation and CADA-X Link have additional software products that can be purchased to allow for optimization of accelerometer and excitation locations, design sensitivity analysis, and model updating - features currently not available in TACT or DYNAVIEW. For EAB analysts, these features are nice, but unnecessary; however, if they are ever needed, these features will either have to be incorporated into TACT or correlation packages like I-DEAS Correlation and CADA-X Link purchased.
The final criterion used in the evaluation of potential modal correlation packages is improved efficiency over EAB’s previous correlation process. Prior to the development of TACT, EAB analysts performed modal correlation tasks using DYNAVIEW. Challenges already presented with using third-party software include data transfer/translation, operation on/in different computer platforms/environments, and external model updating with PATRAN. Other inefficiencies MSC/PATRAN and MSC/NASTRAN users encounter when using third-party software are external editing, bookkeeping, computer modeling, and manual analysis-to-test transformation tasks.
A time comparison of the entire modal correlation process for a NASA LaRC low aspect ratio research wind tunnel model was performed using TACT and DYNAVIEW to illustrate how inefficient and time-consuming third-party correlation packages are for MSC/PATRAN and MSC/NASTRAN analysts. The modal correlation process consisted of pre-test accelerometer set selection and evaluation, correlation of the reduced baseline FEM and test, model updating, and correlation of the reduced updated FEM and test. The results of this comparison were that the entire modal correlation process took 20% longer using DYNAVIEW than TACT. With TACT, MSC/PATRAN and MSC/NASTRAN users perform modal correlation tasks in less time because they never have to perform inefficient tasks such as data translation/transfer, operating on/in unfamiliar computer platforms/environments, external model updating/editing, and manual analysis-to-test transformation tasks.
For the PATRAN-based analysis community, TACT is potentially a more attractive set of modal correlation tools over third-party modal correlation packages for many reasons. Many of the third-party packages have substantial costs associated with them, including purchase price, software maintenance, and training. For MSC/PATRAN and MSC/NASTRAN users, either on-the-job or formal training is required to learn the complex third-party software as well as a different computer environment. Even after the user is fluent in the third-party correlation software, the user must frequently perform inefficient and time-consuming tasks such as data transfer and translation into and out of PATRAN, bookkeeping and file organization, and manual analysis-to-test transformation tasks. TACT eliminates all of these issues for the PATRAN-based analyst, since it has no cost, is an intuitive add-on to PATRAN, and automates the mundane, clerical tasks in the correlation process. TACT streamlines the PATRAN-based correlation process, allowing the structural analyst to focus on the more complicated aspects of model updating.
The development of TACT would not have been possible without the support of several additional key people. Thanks go to Dr. John Whitesides, Director of the Joint Institute for the Advancement of Flight Sciences, for mentoring this project in partial fulfillment of the requirements for a Professional Degree Program, Degree of Engineer at George Washington University. Additional thanks go to Michael Lindell, also of the Engineering Analysis Branch, for developing the original NASTRAN DMAP correlation technique, which is the foundation for the modified correlation calculations performed by TACT.
|
COMPILE SEMODES SOUIN=MSCSOU NOREF NOLIST $ First, dump out full FEM eigenvalues ALTER 152 LAMX, ,LAMA/LMAT/-1$ EQUIVX LMAT/FULLEIG/-1$ $ Then, dump out full FEM eigenvectors ALTER 154 EQUIVX PHG/PHGF/-1$ MATPCH FULLEIG,PHGF,,,// $ |
|
ASSIGN OUTPUT4 = `ASET_1st_try.f15', UNIT = 15, DELETE, FORM=FORMATTED ASSIGN OUTPUT4 = `ASET_1st_try.f16', UNIT = 16, DELETE, FORM=FORMATTED SOL 103 TIME 1000 COMPILE SEMODES SOUIN=MSCSOU NOREF NOLIST ALTER 152 DMIIN DMI,DMINDX/FULLEIG,PHGF,,,,,,,,/ $ LAMX, ,LAMA/ASETEIG/-1$ ADD ASETEIG,FULLEIG/EIGDELTA/1./-1./0 $ ADD EIGDELTA,FULLEIG/PCNTDIFF/1./1./2 $ MATPRN PCNTDIFF,,,, // $ OUTPUT4 FULLEIG,,,,//0/15/0/ $ Full FEM eigenvalues OUTPUT4 ASETEIG,,,,//0/15/0/ $ ASET FEM eigenvalues OUTPUT4 PCNTDIFF,,,,//0/15/0/ $ Percent diff between OUTPUT4 ASETEIG,,,,//0/16/0/ $ ASET FEM eigenvalues ALTER 154 SMPYAD PHGF,MGG,PHG,,,/OCHK/3/1/1/1/1/0/0 $ MATPRN OCHK,,,, // $ OUTPUT4 OCHK,,,,//0/15/0/ $ OUTPUT4 PHA,,,,//0/16/0/ $ OUTPUT4 MAA,,,,//0/16/0/ $ |
|
ASSIGN OUTPUT4 = `ASET_1st_try.f16', UNIT = 16, DELETE, FORM=FORMATTED ASSIGN OUTPUT4 = `ASET_1st_try.f17', UNIT = 17, DELETE, FORM=FORMATTED SOL 103 TIME 1000 REDUCED EIGS COMPILE SEMODES SOUIN=MSCSOU NOREF NOLIST ALTER 152 DMIIN DMI,DMINDX/TESTEIG,PHT,,,,,,,,/ $ LAMX, ,LAMA/ASETEIG/-1$ ADD ASETEIG,TESTEIG/EIGDELTA/1./-1./0 $ ADD EIGDELTA,TESTEIG/PCNTDIFF/1./1./2 $ MATPRN PCNTDIFF,,,, // $ OUTPUT4 TESTEIG,,,,//0/17/0/ $ Test eigenvalues OUTPUT4 ASETEIG,,,,//0/17/0/ $ ASET FEM eigenvalues OUTPUT4 PCNTDIFF,,,,//0/17/0/ $ Percent diff between OUTPUT4 ASETEIG,,,,//0/16/0/ $ ASET FEM eigenvalues ALTER 154 SMPYAD PHT,MAA,PHT,,,/MSF2/3/1/1/1/1/0/0 $ CALCULATE MODAL SCALE FACTORS DIAGONAL MSF2/MSF/'SQUARE'/0.5 $ TRNSP PHT/PHTT $ SOLVE MSF,PHTT/PTNT/0 $ FORM MASS NORMALIZED TEST EIGENVECTORS TRNSP PTNT/PHTN $ SMPYAD PHTN,MAA,PHA,,,/OCHK/3/1/1/1/1/0/0 MATPRN OCHK,,,, // $ OUTPUT4 OCHK,,,,//0/17/0/ $ OUTPUT4 PHA,,,,//0/16/0/ $ OUTPUT4 MAA,,,,//0/16/0/ $ |
