AIAA 2001-1190 Thermal Testing of Tow-Placed, Variable ...mln/ltrs-pdfs/NASA-aiaa-2001-1190.pdf ·...

For permission to copy or to republish, contact the copyright owner named on the first page. For AIAA- held copyright, write to AIAA Permissions Department ,

1801 Alexander Bell Dri ve, Suite 500, Reston, VA, 20191-4344.

AIAA 2001-1190Thermal Testing of Tow-Placed,Variable Stiffness Panels

K. Chauncey WuNASA Langley Research CenterHampton, Virginia

Zafer G�rdalVirginia Polytechnic Institute and State UniversityBlacksburg, Virginia

42nd AIAA/ASME/ASCE/AHS/ASCStructures, Structural Dynamics and

Materials Conference & Exhibit16-19 April 2001

Seattle, Washington

1American Institute of Aeronautics and Astronautics

AIAA 2001-1190

THERMAL TESTING OF TOW-PLACED, VARIABLE STIFFNESS PANELS

K. Chauncey Wu*NASA Langley Research Center

Hampton, Virginia 23681

Zafer G�rdal Virginia Polytechnic Institute and State University

Blacksburg, Virginia 24061

Abstract

Commercial systems for precise placement ofpre-preg composite tows are enabling technologythat allows fabrication of advanced compositestructures in which the tows may be precisely laiddown along curvilinear paths within a given ply. Forlaminates with curvilinear tow paths, the fiber ori-entation angle varies continuously throughout thelaminate, and is not required to be straight andparallel in each ply as in conventional compositelaminates. Hence, the stiffness properties vary asa function of location in the laminate, and the as-sociated composite structure is called a ÒvariablestiffnessÓ composite structure.

Previous analytical studies indicate that vari-able stiffness structures have great potential forimproving the structural performance of compositestructures. In the present research, an experimen-tal program has been conducted to evaluate thethermal performance of two variable stiffness pan-els fabricated using the Viper Fiber PlacementSystem developed by Cincinnati Machine. Thesevariable stiffness panels have the same layup, butone panel has overlapping tows and the otherpanel does not. Results of thermal tests of thevariable stiffness panels are presented and com-

pared to results for a baseline cross-ply panel. Is-sues arising from the manufacturing processesused to fabricate the variable stiffness panels arealso presented and discussed.

Introduction

The widespread use of polymer compositematerials in aerospace vehicles has also resultedin the development of highly innovative, ad-vanced composite structures like the V-22 aft fu-selage and the JSF engine inlet duct. Many, if notall, of these complex structures require sophisti-cated manufacturing technology for their fabrica-tion. A significant advancement in computer-numerical-control (CNC) machine tool technologyis the introduction of commercial systems for pre-cise, repeatable placement of pre-preg compositetows.

One example of such an advanced towplacement system is the Viper¤ Fiber PlacementSystem1 (FPS), developed by Cincinnati Machineand shown in Fig. 1. Advanced tow placementsystems like the Viper FPS are enabling technol-ogy that allows fabrication of complex compositestructures in which the fibers in any given ply maybe laid down along curvilinear paths. In these ad-vanced composite structures, the fiber orientationangle is allowed to vary continuously in each plyand throughout the structure, and is not requiredto be straight and parallel in each ply as in conven-tional composite structures. These configurationsare referred to as Òvariable stiffnessÓ or VS struc-

_____________________ *Aerospace Engineer, Vehicle Analysis Branch, MailStop 365. Professor, Departments of Aerospace and OceanEngineering, and Engineering Science and Mechanics,Associate Fellow AIAA.Copyright © 2001 by the American Institute ofAeronautics and Astronautics, Inc. No copyright isasserted in the United States under Title 17, U.S. Code.The U.S. Government has a royalty-free license toexercise all rights under the copyright claimed herein forGovernmental Purposes. All other rights are reservedby the copyright owner.

_____________________

¤ Identification of commercial products and companiesin this paper is used to describe the materials ade-quately. The identification of these commercial prod-ucts does not constitute endorsement, expressed orimplied, of such products by the National Aeronauticsand Space Administration.


tures, because their stiffness properties also varycontinuously over the domain of the structure.

Several analytical studies have been per-formed to predict the structural performance of VSstructures. G�rdal and Olmedo2,3 presented so-lutions for in-plane response and buckling of con-stant-thickness VS panels with various boundaryconditions and aspect ratios. Variable stiffnesslaminate designs were identified whose bucklingloads were up to 80 percent higher than those ofan equal-weight conventional angle-ply laminate.Waldhart 4 defined a manufacturing constraint onfiber radius of curvature for VS panels whose fiberpaths are generated by shifting a reference path inone direction (the method used in Refs. 2 and 3),and also by projecting new fiber paths which areparallel to the reference path. These constraintswere then applied to optimized designs from Refs.2 and 3. Tatting5 applied the VS concept to thedesign of cylindrical shells such as fuselage struc-tures, and showed that the largest improvement inthe bending response of long cylinders wasachieved through a circumferential shell stiffnessvariation. Langley6 evaluated the in-plane re-sponse of symmetric VS laminates with overlapswhich are generated by placement of adjacenttows during the manufacturing process. Theseoverlaps form areas of increased thickness in thelaminate that resemble discrete stiffeners on thepanel surface, and are much more complicated tomodel and analyze than the previous constant-thickness laminates.

These studies indicate that the VS concepthas great potential for improving the structural per-formance of composite structures. However, noexperimental work has been done thus far toevaluate the thermal performance of VS panels.The primary objective of the present work is topresent the results of an experimental study tocompare and evaluate the structural response ofVS panels when subjected to thermal loads. Thesecond objective of this study is to discuss uniqueissues related to the fabrication of the VS panelsusing the Viper FPS.

Advanced tow placement systems

Commercial systems for precise, repeatableplacement of pre-preg tows are now being usedfor fabrication of advanced composite structures.These advanced tow placement systems repre-sent a combination of the best features of auto-

mated tape layer and tape winding systems. Anadvanced tow placement system typically has aself-contained tow placement head with multipledegrees of freedom (DOF), which is thenmounted on a motion base with several transla-tional DOF. A mandrel with an additional rotationalDOF provides a tool surface on which the tows aredeposited.

One widely used advanced tow placementsystem is the Viper FPS, developed and manufac-tured by Cincinnati Machine. The Viper FPS has atow placement head that collimates up to 32 pre-preg tows from individual supply reels. Each sup-ply reel has a separate feed mechanism with thecapability to cut and restart that tow. The tow feedmechanisms also permit the tows to be paid out atdifferent rates, thus allowing the tow placementhead centerline to precisely follow complex, curvi-linear paths. Heaters inside the tow placementhead provide for partial in-situ compaction of thetows, which allows fabrication of parts with con-cave surfaces. The Viper FPS has a total of sevenDOF in the workspace: three rotational DOF onthe tow placement head, two translational and onerotational DOF on the motion base, and one rota-tional DOF on the mandrel.

The Viper FPS or similar systems are requiredto fabricate advanced composite structures likethe VS panels in this study. Thus, they are ena-bling technology which can allow engineers todesign composite structures that have reachedtheir full performance potential through tailoring ofthe fiber orientation angles, both across the plan-form and through the thickness of the structure.

Selection of design for fabrication

The fiber orientation angle of a VS fiber in thisstudy (shown in Fig. 2) is assumed to be a linearsaw-tooth function with specified initial and finalvalues. These values are specified at a given in-terval along a variation direction, and this variationdirection may then be rotated through some angleabout the global axes. Thus, a reference tow pathmay be described by four parameters. These pa-rameters are Qo, the fiber orientation angle at thepanel center, Q1, the fiber orientation angle at aspecified distance a/2 along the variation direc-tion, and F, the rotation angle of the variation di-rection with respect to the global coordinates. Arepresentative fiber orientation angle is shown inFig. 2, and the corresponding reference tow path


is shown in Fig. 3. In this study, additional towpaths are generated by shifting the reference pathalong an axis through the panel center and per-pendicular to the variation direction. A condensednotation for a VS laminate is [F±<Qo|Q1>]ns,where the subscript n indicates an integer numberof repeating VS plies, and s indicates midplanesymmetry.

Studies performed by G�rdal and Olmedo2,3predicted that significant increases in bucklingload are possible for square, simply supported VSpanels. These constant thickness panels areloaded with end displacements applied parallel tothe shift direction, but transverse to the fiber ori-entation angle variation. A series of VS laminatedesigns are identified whose buckling loads arehigher than those of an equal-weight conventionalangle-ply laminate. These designs all have fiberorientation angles that are smaller on the center-line than on the panel edges. This means that thecenter of the panel has a low stiffness in the load-ing direction, with a higher percentage of the totalaxial load applied through end displacement car-ried by the stiff, simply supported panel edges.

A VS laminate with a [0±<0|75>]9s layup hasthe highest buckling load of all the VS designsevaluated in these studies. The buckling load ofthis VS laminate is 79 percent higher than thehighest straight angle-ply laminate buckling load(obtained with an equal-weight ±45 deg. layup). AVS ply from this laminate is shown in Fig. 4, withthe reference tow path shown as a dark line, andthe shifted paths shown as dashed lines. All of thefibers on the panel load axis centerline are ori-ented at 90 deg. to the panel load axis, which isundesirable because these fibers do not contrib-ute to the laminate axial stiffness. This portion ofthe laminate also violates good composites designpractice, which states that laminates should nothave more than four contiguous plies with thesame fiber orientation angle.7 Waldhart4 deter-mined that this design violates a fiber curvatureconstraint that specifies the minimum radius ofcurvature for the tow placement head, and thus isnot manufacturable with current technology.

For the same panel geometry above, Waldhartidentified a manufacturable VS laminate designwith a buckling load which is 44 percent higherthan the buckling load of a ±45 deg. laminate. Thisdesign has a [90±<30|75>]9s layup. This design isclose to the fiber curvature constraint boundary,

and thus may be difficult to manufacture. Reduc-tion of the 75 deg. edge fiber orientation angleshould make fabrication of the VS laminate easier,but will also slightly lower the buckling load. Smallincreases in buckling load are also predictedthrough replacement of the outermost VS plieswith ±45 deg. straight-fiber plies. A modified de-sign with a [±45/(90±<30|60>)4]s layup is chosenfor fabrication with the Viper FPS.8

Variable stiffness panel manufacturing

The Viper FPS is used to fabricate three com-posite panels from AS4/977-3 pre-preg. Eachpanel has a 26 x 24.50 inch planform, with a 24 x24 inch test section located between the me-chanical test boundary supports. The first twopanels both have the same VS layup, but onepanel has overlapping tows and the other paneldoes not. These panels are used to quantify theimprovements in thermal-structural performancedue to the VS layup. The third panel has a con-ventional [±45]5s layup to provide a baseline forcomparing the response of the two VS panels.

The Viper FPS used for fabrication of the VSpanels in this study is configured to apply up to24, 1/8 inch-wide pre-preg tows during each passof the tow placement head. In the first VS panel, all24 tows are applied during each pass of the ViperFPS, which causes significant overlaps betweentows from adjacent passes towards the paneledges. For the second VS panel, fabricated with-out overlaps, the tow cut/restart capability of theViper FPS is used to maintain a constant 20-plythickness across the whole panel while still apply-ing the VS layup.

Exact tow path

As discussed above, the nominal layup of thetwo VS panels is [±45/(90±<30|60>)4]s, which isboth symmetric and balanced. The outer four pliesof these laminates are straight-fiber ±45 deg.plies, and the inner 16 plies are VS plies with afiber orientation angle of ±60 deg. at Y = 0 inches(the panel centerline), and ±30 deg. at Y = ±12inches. The fiber orientation angle varies linearlybetween Y = 0 and ±12 inches, and repeats out-side the ±12 inch limits.

The relationship between the VS fiber orienta-tion angle Q and the panel geometric coordinatesX and Y is given by dY/dX = tan(Q). Since Q is a


linear function of Y, then dX = cot(Qo + mY)dY,where Qo is the fiber orientation angle on the cen-terline (60 deg. in Fig. 2), and m is the slope of thefiber orientation angle variation (2.5 deg./inch for-12 £ Y £ 0 inches, and -2.5 deg./inch for 0 £ Y £12 inches). Integration of this equation (with X = 0at YÊ= 0) gives an analytical expression for the ex-act tow path of a +Q VS ply, which is

X Ym

mY mY mY

mmY mY mY

mEq

oo o

oo o

o

o

( )[tan ( ) ]

tan( ) ln tan( )cos( ) sin( )

±[cot ( ) ]

cot( ) ± ln cot( )sin( ) cos( )

±ln tan( )[tan ( ) ]

( . )

=+

+ +[ ]{ }

++[ ]{ }

[ ]+

11

11

11

2

2

2

QQ Q

QQ Q

Q

Q

All of the tow paths within a ply are nominallyidentical to the exact tow path, but are shifted inboth directions along the X-axis by integer multi-ples of (24 tows x 1/8 inch/tow)/sin(Qo = 60 deg.),or 3.464 inches. For a -Q VS ply, the tow path isreflected about the X-Z plane. As mentioned ear-lier, shifting of the tow paths may result in overlap-ping of the edges of the adjacent passes. Theseoverlaps may assume the appearance of integralstiffeners in the finished panel. The exact stiffenerpattern resulting from tow overlaps is shown in Fig.5 for the test section of the first VS panel, withpredicted laminate thicknesses between 36 and20 plies. This pattern shows step thickness dis-continuities between the stiffened regions whichcould cause local voids and load path variations.

Actual tow path

The actual tow paths which the Viper FPS towplacement head centerline follows during fabrica-tion of the first -Q VS ply (actually the third ply laiddown on the tool surface) are shown in Fig. 6 forthe VS panel without overlaps. Each tow path inthe figure requires a separate pass by the ViperFPS. The tow paths within the test section are es-sentially identical to the reference tow path whichpasses through the origin (shown as a bold line inthe figure), but are shifted by multiples of 3.550inches along the X-axis. The fiber orientation an-gles for this ply are shown in Fig. 7, with lightershades of gray corresponding to smaller magni-tudes.

Since the VS panel with overlaps is laid up ona flat tool surface, the layup is asymmetric, with araised stiffener pattern on the back of the actual

panel (see Fig. 8). The test section is outlined inwhite in the figure. For the VS panel without over-laps, use of the tow drop/add capability may resultin resin-rich pockets or voids whose locations(shown as black dots in Fig. 7) may be predictedfrom manufacturing data provided by CincinnatiMachine. These data include descriptions of thepaths followed by the tow placement head duringpanel fabrication, and which of the 24 tows wereplaced at any given location within each ply.

These manufacturing data are further proc-essed and evaluated to determine correlation be-tween the exact and actual tow paths. Goodagreement is found for the first two VS plies inboth of the VS panels. This correlation is shown inFig. 9, where the exact tow paths generated withEq. 1 (shown as dashed lines) are plotted with theactual tow paths from Fig. 6. The majority of thedifferences between the exact and actual towpaths in the figure are due to the two different X-direction shift increments listed above.

Manufacturing ply shift

Further analyses of the manufacturing datashow that the centerpoint of each subsequent VSply with the same orientation is shifted by integermultiples of 3/16 inch in both the X- and Y-directions. The reference tow paths for the eight-Q VS plies of the VS panel without overlaps areshown in Fig. 10. As described above, the cen-terpoint for ply 3 passes through the origin, andthe centerpoints for the next seven VS plies areeach shifted by an additional (3/16 ´ Ö2) inchesalong the line Y = X. Note that the ply numbersswitch from odd to even as they pass the nominallaminate symmetry plane between plies 10 and 11.The centerpoints of the +Q VS plies are shifted bythe same increments along the line YÊ=Ê-X.

From a manufacturing point of view, the pur-pose of these ply shifts is to avoid collocation ofply drops and seams between successive passeson the panel planform, since they would occur atthe same points within each ply of the same orien-tation. Collocation of ply drops and seams wouldcause local resin-rich regions through the panelthickness, which could degrade the laminatestrength and stiffness. Unfortunately, these plyshifts cause the actual layups for the two VS lami-nates to be both asymmetric and unbalanced.Thus, every point on the VS panel planform has aunique layup, making the structure even more


complicated than originally intended. The refer-ence tow paths for the ±45 deg. plies of the VSpanels do not exhibit the centerpoint shifts de-scribed above.

It would be desirable to incorporate thebenefits of the ply shifts (no collocation of towdrops and edges) in VS panels, while still mini-mizing the detrimental effects (asymmetry and im-balance in the layup). Two possible methods forreducing the impact of the ply shifts are now pro-posed. One solution would be to reduce themagnitude of the ply shift X- and Y-increment from3/16 inch (1.5 tow widths) to 1/16 inch (0.5 towwidths). While this option does allow the tow dropsto be closer together, it also reduces the variationin fiber orientation angles at any given point.

Another scheme, which can also be used withthe one just described, is to alternate the directionof the shift increment. Several options are illus-trated in Fig. 11 for the -Q plies of a VS panel,along with the current scheme used in fabricationof the panels in this study. Zero indicates no shift(where the VS ply is origin-centered as in Fig. 6), 1is a shift of some magnitude in the direction shownin Fig. 9, and -1 is a shift in the opposite direction.These schemes have the same advantage as theone above, but with the disadvantage that the towdrops and voids in several plies are collocated onthe panel planform, but not in adjacent plies.These options should be evaluated with analyticalmethods to predict their effectiveness in reducingthe panel anisotropy.

Panel imperfections

The geometric imperfections of the threecomposite panels are determined from surveysperformed with a Brown and Sharpe 7300 coordi-nate measuring machine (CMM) fitted with a Ren-ishaw PH-9 indexable probe head. The CMM canlocate a point in space with a accuracy of ±0.0003inches in its workspace. Front and back surfacemeasurements are taken at a grid of evenlyspaced locations (1/8 inch for the VS panel withoverlaps, and 1/2 inch for the VS panel withoutoverlaps and baseline panel) over the 24 ´ 24 inchtest sections.

The asymmetric, unbalanced layups intro-duced due to the ply shifts caused the two VSpanels to warp badly after panel consolidation. Aplot of the front (tool) surface geometric imperfec-

tions is shown in Fig. 12 for the VS panel withoverlaps. The maximum amplitude is 0.383 inchesand the minimum amplitude is -0.331 inches, witha root-mean-square (RMS) amplitude of 0.152inches. The front surface imperfections of the VSpanel without overlaps are qualitatively identical tothe ones in Fig. 12. The maximum amplitude is0.417 inches, the minimum amplitude is -0.371inches, and the RMS amplitude is 0.166 inches forthis panel. The surface contours for the two VSpanels are very similar, which leads to the conclu-sion that the anticlastic shapes of the VS panelsare due entirely to the ply shifts, not from thethickness variations caused by the tow overlaps.

The front surface imperfections for the base-line panel have a maximum amplitude of only0.031 inches, and a minimum amplitude of -0.022inches. The RMS amplitude for the baseline panelis 0.013 inches, or 8 percent of the average ampli-tude of the VS panels. The ply shifts, if present,should have no effect on the baseline panel be-cause it is a conventional composite laminatewhose fiber orientation angles are the samethroughout each ply.

Laminate and ply thickness

The measured thicknesses of the two 20-plypanels are computed as the difference of the frontand back surface measurements. The averagethickness of the VS panel without overlaps is0.149 inches, with a standard deviation of 0.0026inches, and the average thickness of the baselinepanel is 0.153 inches, with a standard deviation of0.0019 inches. The average ply thicknesses forthese panels are 0.00745 and 0.00765 inches,respectively. The VS panel without overlapsweighs 5.48 lbs, and the baseline panel weighs5.65 lbs, for an average material density of 0.0579lb/in3.

The measured thicknesses of the VS panelwith overlaps are computed and shown in a con-tour plot in Fig. 13. The maximum thickness of thispanel is 0.289 inches, and the minimum thicknessis 0.140 inches. An analysis of the manufacturingdata for the VS panel with overlaps shows that themaximum laminate thickness is 38 plies, and theminimum thickness is 18 plies. The actual VSpanel with overlaps weighs 6.77 lbs. Multiplicationof the material density, number of plies, differentialarea, and ply thickness gives a differential weightat each point in the panel planform where the


manufacturing layups are defined, and summationgives an estimate of the panel weight. An as-sumed ply thickness of 0.00765 inches results ina predicted weight of 6.78 lbs, which is very closeto the actual weight of the panel.

Fiber orientation angle

Before advanced fiber placement systems canbe used for fabrication of advanced compositestructures, it is necessary to quantify the precisionand accuracy of the fiber orientation angles as thetows are laid down in the workspace. If the towscan be placed in a predictable, repeatable mannerduring the manufacturing process, then the nomi-nal values can be used in the design of VS struc-tures with a high degree of confidence that theactual angles will conform to their desired values. Ifthis is achievable, laminate designers can theninclude the fiber orientation angle throughout thestructure as a design variable, further opening thedesign envelope and leading to even more highlyoptimized composite structures.

The actual fiber orientation angles for the ref-erence tow path in Fig. 6 are computed frommanufacturing data and plotted with their exactvalues in Fig. 14. The differences between theexact and actual ply angles are shown in Fig. 15,with a RMS difference of 0.38 degrees over the24.50 inch width of the VS panel. This leads to theconclusion that the Viper FPS can precisely orientthe composite tows at their desired angles, whichis significant because it means that manufacturerscan precisely and accurately control the fiber ori-entation angle during the fabrication process.

Thermal free-expansion tests

One objective of the present work is to ex-perimentally evaluate the structural response ofthe VS and baseline panels when subjected tothermal loads. For each of the three panels dis-cussed above, several thermal tests of the uncon-strained panel are performed. In these tests, thepanel (without edge and end fixtures) is heatedfrom room temperature to a maximum of 150 de-grees Fahrenheit (deg. F).

Test facility

A capability for thermal testing is developedwhich includes an oven with thermal control. Elec-trical resistance and forced-air heaters are used to

raise the temperature within the enclosure. Theoven is built from plywood, with foam and ceramicboard insulation lining the interior. Perforatedsheet-metal baffles are installed within the oven toensure uniform heating of the panel by the forced-air heater unit. The oven is installed in a 500 klbelectrohydraulic compression test stand, thus al-lowing mechanical loads to be applied to the paneleither with or without the thermal loads. A personalcomputer-based data acquisition system is usedto collect information from the test instrumentationas well as load and stroke data from the test stand.

The panel is supported in the oven by fixturesthat prevent rigid-body motion of the panel whilestill allowing thermal free expansion. The panelrests on two small quartz rollers that support thepanelÕs weight while preventing direct contact be-tween the panel and lower heated platen. Smallquartz cones are attached to the corners of theoven face opposite spring-loaded steel plungersto support the corners of the panel.

Shadow moir� interferometry

The front face of the oven described above isa clear glass pane that allows observations of thepanel using shadow moir� interferometry duringboth thermal and mechanical load tests. Shadowmoir� interferometry is an experimental techniqueused to obtain qualitative and quantitative meas-urements of the panel normal (out of plane) dis-placements. Only qualitative measurements aretaken during the tests in this study, so thermal de-formations of the moir� system are not consideredto be significant.

A grid of thin black lines (50 lines/inch) onphotographic film base is mounted on the glassfront of the oven. A polarized, high-intensity lightsource is used to project shadows from the gridonto the panel inside the oven. As the panel de-forms, the interference patterns (moir�) betweenthe grid and shadows appear to shift. A Hasselblad70 mm camera is used to take still photographs,and a tripod-mounted video camera is used to re-cord video imagery, of the moir� patterns duringthe tests.

Instrumentation

For the thermal test, the panel response ismeasured with thermocouple and strain gage in-strumentation. The thermocouple and strain gage


patterns used for the three panels are shown inFig. 16. Ten thermocouples, arranged as back-to-back pairs, are attached to the panel front andback surfaces. The thermocouples have an accu-racy of ±1.8 deg. F, and are used to measure thepanel surface temperature during heating.

Between 18 and 34 electrical-resistance straingages, arranged in back-to-back pairs, are bondedto the panel surfaces to measure strains from thethermal tests. The strain gages are deployed tomeasure either axial strains, or both axial andtransverse strains at discrete points on the panel.Strain data are taken as functions of temperaturefor gages bonded to Corning ultra-low-expansiontitanium silicate blocks to determine the thermaloutput of the strain gages.9 After the thermal test,the thermal output data are subtracted from thetotal strain collected during the test to obtain thetrue free-expansion thermal strains.

No panel axial displacements are recordedduring the thermal tests. However, four linear vari-able displacement transducers (LVDTÕs) are lo-cated at the corners of the test stand platens tomeasure their relative motion during the thermaltest.

Heating profile

Electrical resistance and forced-air heaters areused to gradually raise the temperature within theoven to a maximum of 150 deg. F. A feedbackcontrol system is used to provide closed-loop,real-time thermal control. Electrical thermocouplesare attached to the heated platens and oven toprovide feedback to the control system, as well asmeasurements of the temperature input, to thedata acquisition system. The temperatures insidethe oven are raised from room temperature to 90deg. F as a step input, then held at that level for 5minutes to allow all temperatures to equalize. Thetemperatures are then increased to 150 deg. F at2 deg. F/minute. To complete the heating cycle,the temperatures are held at 150 deg. F for 20minutes to allow the panel to reach its maximumtemperature. A representative heating profile isshown in Fig. 17.

Test results

Five thermal tests are performed for each ofthe three panels. The first test for each panel isused to fully cure the adhesives used to attach the

strain gages. Typical test results from the latertests are presented in this section. As notedabove, the strain data presented here are cor-rected for the strain gageÕs thermal output causedby the thermal loading.

A shadow moir� photograph of the VS panelwithout overlaps during the free-expansion ther-mal test is shown in Fig. 18. The contours in thisfigure correspond to the anticlastic panel geome-try from Fig. 12. No changes are observed inthese contours during the thermal tests. This isprobably due to the relatively low resolution of theshadow moir� system and the small induced de-formations, and does not mean that the panels donot deform during the thermal test.

VS panel with overlaps

Back-to-back axial and transverse strains at thecenter of the VS panel with overlaps are plottedagainst the average panel temperature in Fig. 19.The location of the gages is indicated by a blackring on the gage pattern in the figure. As ex-pected, the strain response is linear with tempera-ture. The extensional (in-plane) coefficient ofthermal expansion (CTE) of the laminate is equalto the slope of a plot of the average strain as afunction of temperature.

The approximate layup at the panel center(determined from manufacturing data) is[±45/(±58)4]s. This layup has a relatively high axialCTE of 5.06 ´ 10-6 in/in/deg. F, and a low trans-verse CTE of 0.06 ´ 10-6 in/in/deg. F. Thesemeasured CTEÕs are computed using the averageaxial and transverse strains from Fig. 19. A set ofexperimentally determined material properties(shown in Table 1) and the [±45/(±58)4]s panellayup are used as input for a commercial laminateanalysis code.10 Using classical lamination theory(CLT), the laminate analysis code predicts an axialCTE of 7.12 ´ 10-6 in/in/deg. F, and a transverseCTE of -1.64 ´ 10-6 in/in/deg. F for this layup. Themeasured and predicted CTE values are listed andcompared in Table 2.

Back-to-back axial and transverse strain dataare also taken at a point 5 inches to the right of thepanel center on the transverse centerline, and areanalyzed as described above. The 20-ply layup ofthe VS panel with overlaps is approximately[±45/(±48)4]s at this point, which is close to anorthotropic cross-ply layup. However, the meas-


ured axial CTE of this layup (2.96 ´ 10-6 in/in/deg.F) is still significantly higher than the transverseCTE (0.51 ´ 10-6 in/in/deg. F). A comparison withpredicted values is also shown in Table 2.

VS panel without overlaps

Back-to-back axial and transverse strains at thepanel center are plotted against the average paneltemperature in Fig. 20 for the VS panel withoutoverlaps. Analyses of the test data show that thislayup has an axial CTE of 3.86 ´ 10-6 in/in/deg. F,and a transverse CTE of -0.36 ´ 10-6 in/in/deg. F.The 20-ply layup at the panel center is approxi-mately [±45/(±57)4]s. A laminate analysis predictsan axial CTE of 6.78 ´ 10-6 in/in/deg. F, and atransverse CTE of Ð1.58 ´10-6 in/in/deg. F.These measured and predicted CTE values arecompared in Table 2. While the extensional CTEtrends for this panel are similar to those describedabove for the VS panel with overlaps, the bendingstrains, which are proportional to the differencebetween the front and back surface strains, arereversed.

Comparisons of the measured and predictedCTEÕs at two points, located 5 and 10 inches tothe right of the centerpoint along the transversecenterline, are also shown in Table 2. The ap-proximate layups at these points are [±45/(±47)4]sand [±45/(±35)4]s, which are both 20-ply lami-nates. Good correlation is observed between testand predicted CTE values at these points. Notethat a direct comparison may be made betweenthe first point evaluated here and the corre-sponding point described above on the VS panelwith overlaps. However, a direct comparison is notpossible for the second point on this panel, be-cause the corresponding laminate thickness onthe VS panel with overlaps is approximately 0.245inches (32 plies).

Baseline panel

For the baseline panel, back-to-back axial andtransverse strains at the panel center are plottedagainst the average panel temperature in Fig. 21.The strains are linear and approximately equal,which is as expected, since the [±45]5s laminateshould have the same CTE in both the axial andtransverse directions. The measured CTE of thislayup is 1.58 x 10-6 in/in/deg. F, which compareswell with the predicted value of 1.47 ´ 10-6in/in/deg. F. The maximum temperature for the

baseline panel is about 7 deg. F lower than themaximum temperature for the VS panels becausethe heating profile used for the tests of this paneldid not include the 20 minute hold at 150 deg. Fshown in Fig. 17.

Discussion

A plot of the measured and predicted CTEÕs isshown in Fig. 22, with predicted CTE values plot-ted at 1-degree increments of the fiber orientationangle. The correlation between the measured andpredicted CTEÕs for the composite panels rangesfrom excellent (1.5 percent error) to poor (104percent error). The best correlation is obtainedwhere the layups are close to an orthotropic cross-ply laminate (i.e., [±45]5s), but good correlation isalso observed for the [±45/(±35)4]s laminate.Poor correlation is obtained for larger fiber orienta-tion angles.

The correlation between the measured andpredicted CTEÕs is somewhat surprising, since thefiber angles (and CTEÕs) of the VS panels changedramatically across the panel width, and onlyslightly across the panel height due to the manu-facturing ply shifts. In comparison, the CLT-basedlaminate analysis assumes that the laminate is aconventional composite structure with the samefiber angles at every point. It would therefore bereasonable to expect that CLT is unsuited forthermal analysis of VS structures, but the correla-tion in Fig. 22 indicates that this is not the case, atleast for low to intermediate fiber orientation an-gles. Thus, a numerical procedure with an accu-rate fiber orientation model should also be used topredict the thermal performance of the VS lami-nates and model the interactions between theadjacent laminates of varying layups and thick-nesses, an effort which is, at present, outside thescope of this paper.

A qualitative analysis of the bending strains forthe VS panel with overlaps indicates that the out-of-plane deformations become more exaggeratedduring the thermal test. In other words, the panelgeometry at the maximum temperature looks likethe one shown in Fig. 12, but with even larger am-plitudes. Similar analyses for the VS panel withoutoverlaps indicate that, when heated, it tries to re-turn to a flat configuration. This behavior is whatone would expect since the panel was cured on aflat surface at approximately 350 deg. F, and is anindication of the residual stresses induced during


the curing due to the variation of the fiber orienta-tion angle over the planform area of the panel aswell as the unanticipated through-the-thicknessvariation. This behavior is the opposite of what wasobserved for the VS panel with overlaps, whichsuggests that the different deformation patternsmay be due to the raised stiffeners on the back ofthe panel. Quantitative comparisons of the strainsfor the two VS panels also shows that the strainmagnitudes are larger for the VS panel with over-laps, which may also be due to the additional strainfrom the raised stiffeners.

Concluding remarks

Previous analytical studies demonstrate thatthe VS concept has great potential for improvingthe structural performance of composite struc-tures. In this study, the results of experiments tocompare and evaluate the structural response ofVS panels when subjected to thermal loads arepresented and discussed. Comparisons betweenexperimental CTEÕs and results generated withCLT show that the best correlation is achieved forlaminates that are close to a ±45 deg. Layup. Anadditional goal of this study is to identify and dis-cuss several fabrication issues for the VS panels.Differences between the exact and actual towpaths are quantified. The manufacturing ply shift isalso quantified and its effects on the panel hard-ware are discussed. Several options are pre-sented for achieving the same benefits with a re-duced impact on the panel hardware.

Composite materials have revolutionizedaerospace structures in the past three decades.The recent introduction of advanced fiber place-ment systems now allows composite manufactur-ers to precisely and accurately control the fiberorientation angle during fabrication of the struc-ture. This capability now allows designers of com-posites to use the fiber orientation angle as a de-sign variable in their analyses, not only throughouteach ply as with conventional composites, but ateach point within a ply in an advanced compositestructure. The present work represents a first steptowards developing a verified, reliable capabilityfor the analysis and design of such highly tailoredcomposite structures.

References

1. Evans, D. O., Vaniglia, M. M., and Hopkins, P.C.: Fiber Placement Process Study. 34th Inter-

national SAMPE Symposium, Vol. 34, Book 2,May 1989, pp. 1822-1833.

2. G�rdal, Z., and Olmedo, R.: In-Plane Responseof Laminates with Spatially Varying Fiber Orien-tations: Variable Stiffness Concept. AIAA Jour-nal, Vol. 31, No. 4, April 1993, pp. 751-758.

3. Olmedo, R., and G�rdal, Z.: Buckling Re-sponse of Laminates with Spatially Varying Fi-ber Orientations. Proceedings of 34th AIAA/ASME/ASCE/AHS/ASC Structures, StructuralDynamics, and Materials Conference. April 19-21, 1993, La Jolla, CA, pp. 2261-2269. Avail-able as AIAA-93-1567.

4. Waldhart, C., G�rdal, Z., and Ribbens, C.:Analysis of Tow Placed, Parallel Fiber, VariableStiffness Laminates. Proceedings of 37th

AIAA/ASME/ASCE/AHS/ASC Structures,Structural Dynamics, and Materials Confer-ence. April 15-17, 1996, Salt Lake City, UT.Available as AIAA-96-1569.

5. Tatting B. F.: Analysis and Design of VariableStiffness Composite Cylinders. Ph. D. Disserta-tion, Virginia Polytechnic Institute and StateUniversity, 1998.

6. Langley, P. T.: Finite Element Modeling ofTow-Placed Variable-Stiffness CompositeLaminates. M. S. Thesis, Virginia PolytechnicInstitute and State University, 1999.

7. G�rdal, Z., Haftka, R. T., and Hajela, P.: Designand Optimization of Laminated Composite Ma-terials. John Wiley and Sons, Inc., New York,1998.

8. Tatting B. F., and G�rdal, Z.: Design andManufacture of Tow-Placed Variable StiffnessComposite Laminates with ManufacturingConsiderations. Proceedings of 13th U.S.National Congress of Applied Mechanics(USNCAM). June 21-26, 1998, Gainesville,FL.

9. Anon., Strain Gage Thermal Output and GageFactor Variation with Temperature. TN-504-1,Measurements Group, Inc., Raleigh, NC, 1993.

10. Nairn, J. A.: Utah Laminatesª Manual, Version2.0. RSAC Software, Salt Lake City, UT, 1988.


Table 1: Measured ply properties of AS4/977-3 material_________________________________________________________________________________

E1 18.83 x 106 lb/in2

E2 1.34 x 106 lb/in2

G12 0.74 x 106 lb/in2n12 0.36a1 -0.19 x 10-6 in/in/deg. Fa2 19.1 x 10-6 in/in/deg. F

_________________________________________________________________________________

Table 2: Measured and predicted coefficients of thermal expansionfor variable stiffness and baseline panels

_________________________________________________________________________________

Rosette location, in. Gage CTE, 10-6 in/in/deg. F Error,Layup, deg. orientation Test CLT percent

_________________________________________________________________________________VS panel with overlaps(X=0, Y=0) Axial 5.06 7.12 28.9[±45/(±58)4]s Transverse 0.06 -1.64 103.7

(X=0, Y=5) Axial 2.96 2.85 3.9[±45/(±48)4]s Transverse 0.51 0.28 82.1

VS panel without overlaps(X=0, Y=0) Axial 3.86 6.78 43.1[±45/(±57)4]s Transverse -0.31 -1.58 80.4

(X=0, Y=5) Axial 2.53 2.38 6.3[±45/(±47)4]s Transverse 0.66 0.65 1.5

(X=0, Y=10) Axial -0.98 -1.39 29.5[±45/(±35)4]s Transverse 5.04 6.02 16.3

Baseline panel(X=0, Y=0) 1.58 1.47 7.5[±45]5s

_________________________________________________________________________________


Courtesy of Cincinnati Machine

1. Advanced fiber placement system.

Q1

Qo

Variationdirection

Fiberorientationangle

a/2

2. Fiber orientation angle for VS ply.


Variationdirection

Shiftdirection

a/2

Q1

Qo F

Y

XF

3. Reference tow path for VS ply.

Y

X

Shiftdirection

Fiberÿanglevariationdirection

Endÿshorteningdirection

4. [0±<0|75>] VS ply.


20ÿplies

28ÿplies

36ÿplies

5. Exact stiffener pattern for VS panel with overlaps.

X-coordinate,ÿin.

Y-coordinate,ÿin.

Referencetowÿpath

-20

-10

0

10

20

-20 -10 0 10 20

6. Actual tow paths for ply 3 of VS panel without overlaps.


-20 -10 0 10 20-20

-10

0

10

20

Q,ÿdeg-32.59-35.77-38.95-42.14-45.32-48.50-51.68-54.86-58.05ÿÿÿÿTowÿdrop/townotÿplaced

X-coordinate,ÿin.

Y-coordinate,ÿin.

7. Fiber orientation angles for ply 3 of VS panel without overlaps.

8. Back surface of VS panel with overlaps.


X-coordinate,ÿin.

Y-coordinate,ÿin.

-20

-10

0

10

20

-20 -10 0 10 20

ActualÿtowÿpathExactÿtowÿpath

9. Comparison of exact and actual tow paths for VS ply.

X-coordinate,ÿin.

Y-coordinate,ÿin.

-20

-10

0

10

20

-20 -10 0 10 20

PlyÿshiftÿdirectionPlyÿ3

Plyÿ18

Plyÿ9

Plyÿ12

10. Shifted reference tow paths for VS panel without overlaps.


-1.5-0.5

0.51.5

-101

Currentÿplyÿshiftÿscheme

01 3 5 72 4 6

-101 -1

01

-1.5-0.5

0.51.5

-1.5-0.5

0.51.5

5

9

12

16181818

3

7

14

5

9

12

16

3

7

14

5

9

12

16

3

7

14

35

79

121416

18

35

79

121416

18

35

79

121416

18

18

3

7

14

5

9

12

16

11. Current and proposed VS panel ply shift schemes.

-12 -6 0 6 12

Z,ÿin.ÿ-0.299-0.234-0.169-0.104-0.0390.0260.0910.1560.2210.2860.351

X-coordinate,ÿin.

Y-coordinate,ÿin.

-12

-6

0

6

12

12. Measured front surface geometric imperfections for cured VS panel with overlaps


-12 -6 0 6 12

t,ÿin.0.1470.1600.1740.1870.2010.2150.2280.2420.2550.2690.282

X-coordinate,ÿin.

Y-coordinate,ÿin.

-12

-6

0

6

12

13. Measured thicknesses for VS panel with overlaps.

-20 -10 0 10 20Y-coordinate,ÿin.

0

ActualÿangleExactÿangle

Fiberÿorientationÿangle,ÿdeg.

-50

-40

-60

-30

14. Comparison of exact and actual fiber orientation angles for reference tow path.


-20 -10 0 10 20

Y-coordinate,ÿin.

-1

0

2

-2

1

(Exactÿ-ÿactual)ÿfiber

orientationÿangle,ÿdeg.

15. Differences between exact and actual fiber orientation angles for reference tow path.

ThermocouplesBiaxialÿstrainÿgagesAxialÿstrainÿgages

StrainÿgageÿpatternÿforVSÿpanelÿwithÿoverlaps

StrainÿgageÿpatternÿforÿVSÿpanelÿwithoutÿoverlaps

Strainÿgageÿpatternÿforÿbaselineÿpanel

Thermocoupleÿpatternÿforÿallÿpanels

Frontÿofÿpanelsÿshown

Allÿpatternsÿrepeatÿonbackÿofÿpanels

X

Y

16. Instrumentation layout for VS and baseline panels.


75

100

125

150

0 500 1000 1500 2000 2500 3000 3500

Temperature,ÿdeg.ÿF

Time,ÿsec.

Averageÿofÿ5ÿcontrolthermocouplesÿshown

17. Representative heating profile.

18. Shadow moir� photo of VS panel without overlaps.


-200

-100

0

100

200

300

400

500

75 100 125 150 175


Gageÿlocationandÿorientation1ÿ-ÿAxialÿfront2ÿ-ÿAxialÿback3ÿ-ÿTransverseÿfront4ÿ-ÿTransverseÿback

1

2

3

4

RosetteÿlocationStrain,ÿmicrostrain

19. Strain vs. temperature for VS panel with overlaps.

-200

-100

0

100

200

300

400

500

75 100 125 150 175

Strain,ÿmicrostrain



1

2

4

3

Rosetteÿlocation

20. Strain vs. temperature for VS panel without overlaps.


-200

-100

0

100

200

300

400

500

75 100 125 150 175

Strain,ÿmicrostrain


2,ÿ3


Rosetteÿlocation

1,ÿ4

21. Strain vs. temperature for baseline panel.

Q,ÿdeg.

a,ÿ10-6 ÿin./in./deg.ÿF

CLTÿaxialCLTÿtransverseTestÿaxialTestÿtransverseTestÿ[±45]5s

[±45/(±Q)4]sÿlaminate

30 35 40 45 50 55 60-2

0

2

4

6

8

22. Comparison of measured and predicted laminate CTEÕs.

AIAA 2001-1190 Thermal Testing of Tow-Placed, Variable ...mln/ltrs-pdfs/NASA-aiaa-2001-1190.pdf ·...

Documents

Transcript of AIAA 2001-1190 Thermal Testing of Tow-Placed, Variable ...mln/ltrs-pdfs/NASA-aiaa-2001-1190.pdf ·...