DESIGN ANJ3 TESTING OF THERMAbEXPANSION-IKILDED --EPOXY HAT-STIFFENED SZN37ICH PANELS

* D m C. Jegley

NASA Langley Research Center

Abstract

. a Muumm weight configurations for two types of graphite-epoxy hat-stiffened compression-lwded panels fabricated by the thennal-expansion-molding (TEM) ymufacturing process were evaluated analytically and experimentally for designs with load index NiL values ranging from 100 to 800. The two types of panels contain graphite-epoxy face sheets with a foam core and hat stiffeners which are either open or filled with foam. Constraints on the extensional and shear stiffnesses are imposed on the design so the panels will satisfy typical constraints for aircraft wing structures. Optirrral structurally efficient TEM panels are ccrmpared to cammen=ially available aluminum aircraft structures. Predicted load-strain relationships agree well with experimental results. Significant impact damage to the unstiffened face sheet and foam core does not noticably reduce the load carrying ability of the panels but damage to the stiffened face sheet r&ces the failure load by 20% campared to unimpacted panels.

Introduction

Since laminated composite materials have lower density and higher stiffnesses and strengths than aluminum, they can be used to improve the structural efficiency of aircraft camponents. However, panel configurations which are structurally efficient when made f m metals are not necessarily structurally efficient when made from composite materials. The optimum (most structurally efficient) stiffened panel made from laminated composites may be nothing like the optimum stiffened panel made from aluminum. Structures made fram composite materials can be tailored to meet the expected loading conditions in ways that metals cannot. Iaminate stacking sequences can be arranged to make use of differences in stiffnesses and strengths in each direction. New configurations must be evaluated to determine whether they represent more structurally efficient camponents than the typical aircraft structures commercially available today. However, current laminated camposite parts tend to be more expensive to manufacture than metallic parts. Manufaduring processes must be evaluated to determine if they can be used to fabricate cost- effective and structurally efficient graphite-epoxy -ite panels. One process which may provide a way of fabricating such panels is thermal-expansion molding.

Th-1-expansion molding (TEM) can be used to build hat-stiffened panels made of laminated graphite-epoxy plies and foam core. The TEM procedure includes cutting the foam core to appropriate shapes, compressing it by 20-30 percent and placing it in the skin of the panel and/or inside the hat stiffeners. The graphite-epoxy plies are layed around the foam core by hand. The panel is then placed in a tool and heated in an oven. The foam core applies pressure to the

*Aerospace %inert Structural Mechanics Branch. Member A m .

graphite-epoxy layers as it expands during heating. Thermal-expnsion molding may be more cost- effective than conventional methods for manufacturing hat-stiffened panels since it simplifies the tooling and does not require an autoclave. The foam core can be left inside the hat stiffeners in the finished product or it can be ren~~~ed. Foam is easy to work with and relatively inexpensive.

The results of an optimization study and finite element analysis of panels which can be fabricated using the TEM process are presented in this paper. Several test specimens were constructed using this process and then tested by loading the specimens in axial compression. The experimental test results of undamaged and impact damaged panels are presented and campared with analytical results.

Analysis

optimization

?tro panel configurations which can be constructed using the TEM manufacturing process were evaluated by designing optimum structurally efficient panels and camparing their structural efficiency to that of cammercially available alunhum aircraft wing camponents. The computer ccde PASCO (ref. 1) was used to size and evaluate the o p t h panels. Panels were optimized for compressive loading in one direction at load levels of Nx/L = 100, 250,

500, and 800, where Nx represents the applied . .

axial load per panel unit width and L is the panel length. Minimum overall shear and extensional stiffness requirements based on the design load level were also included. These mimimum requirements are discussed in reference 2.

The configurations evaluated consisted 06 qraphite- eqxy laminates containing a nmhr of 0 , t45O and 90 plies. The only constraint on the l w t e was that the outer plies of the panel be f45 plies of a minimum thickness of .0055 inches and that the foam core (inside the skin and &n the hat stiffener) be s~~~ounded by 245 plies of the same minimum thickness. For the optimization process, an integral number of plies were not required since the minimum thicknesses which would satisfy the loading constraints were desired. A minimum thiclmess of .25 inches for the foam layer in the skin was required for manufacturing convenience. Panel dimensions were specified to be 30 inches long and approximately 24 inches wide. All dimensions of the hat stiffeners were variables. Cross sections of the two types of hat-stiffened panels studied are s h m in figure 1. Each design contains a foam core sarktwiched between graphite- epoxy laminates. The configurations are an open hat stiffener (configuration 1) and a foam-filled hat stiffener (configuration 2). To reduce the chance of failures due to transverse shearing in the foam core, which PXCO cannnot predict, a value of 1.5 was used as a safety factor on the buckling mfle with one longitudinal half wave for all load levels considered. Simply supported boundary conditions on the loaded edges are required in

This paper is declared a work of the U S Government and is not sub~ect to copyr~ght protection In the United States 2146

PASCO. Free boundary conditions on the unloaded edges were assumed. The optimization study was perform3 based on the material properties of two graphite-epoxy material systems, Hercules Incorporated AS4 fiber and 3502 resin and Hercules Incorporated Dl6 fiber and American Cyanamid 18081 interleaved resin. Rohacell 71 foam was used for the core. The material properties used are shown in Table 1. Maxhmm allowable extensional and shear strains for the graphite-epoxy layers and manufacturer recammendations of maximum allowable stresses for the foam are shown in Table 1.

Identification of cammercial products and companies in this report is used to describe adequately the materials. The identification of these mercial products does not constitute endorsement, expressed or implied, of such products by the National Aeronautics and Space Administration or the publishers of these conference proceedings.

Finite Element Analysis A finite element analysis using the STAGS camputer code (ref. 3) was cohcted to verify the results obtained from the PASCO analysis using the configuration with foam-filled hat stiffeners. The STAGS model is based on the actual test specimens. The STAGS ard PASCO d e l s differ in that the boundary conditions on the loaded edges must be sinply supported in PASCO but are modeled as clamped in STAGS to more accurately represent the test conditions.

~xperiment

Test sm2cimens

A moderately-loaded hat-stiffened panel made of AS4-3502 and Rohacell 71 foam was designed with foam-filled hat stiffeners based on the PASCO analysis and construct& using the thermal- expimion-molding process. Small changes were made in the optimum design to allm the panel to be built, such as adjusting each layer thickness to contain an intesrdl numkex of plies. The panel was approximately 24.5 inches long- and 21 inch& wide and contained 16 stiffeners of the type s h m in figure lc with a stiffener height of 1 inch, stiffener spacing of 1 inch and a stiffener width of .25 inches. The stacking sequence of the skin was [_+45/Foam/_+45/O7/T45lT with a foam core

thihess of .25 inches. The stacking sequence of the cap of the hat was [-45/O6/k45/O6/45lT. The

[_+45] plies in the skin away from the stiffeners will be referred to herein as the Ifunstiffened face sheetft and the [_+45/07/_+45] plies and stiffeners

will be referred to as the Itstiffened face sheet." The [_+45/07/_+45] plies alone will be referred to as

the Vhicker face sheet. It A panel of the same sta- sequences and stiffener configuration but containing three stiffeners was constructed using the TEM method as a practice panel to familiarize the mnufacturer with the technique and to provide a small specimen to determine the severity of low speed impact damage.

The panel was ultrasonically evaluated before testing by using a bondascope and was determined to have no defective regions. Four test specimens with dimensions 12 inches long and 10 inches wide we& cut f m this panel. Each specimen contained 8 stiffeners. The ends of the specimens were potted in an epoxy materig and ground flat and parallel to assure uniform loading. The specimens

were then instrumented with strain gages. Axial and lateral back-to-back strain gages were placed on the stiffened face sheets, the unstiffened face sheets and on the webs of the hats. Axial strain gages were placed on the cap of the hats.

One of the four specimens had no detectable defects prior to testing. One panel was damaged during machining such that the foam core in the skin was crushed slightly in two areas of approximately two square inches each at the unloaded edges near the lengthwise caterline. ?tro panels were subjected to low-speed impact damage before any axial load was applied at the locations s h m in figure 2. A list of the test specimens and test conditions is presented in Table 2.

Test hrocedure

An axial campressive load was applied at a rate of 3000 to 6000 lb/min to each test specimen until failure. Strain gage data, displacement measurements and moire patterns of out-of-plane displacments at various load levels were recorded during the tests. These measurements are used to evaluate the structural response and failure characteristics of the specimens.

Results and Discussion

Analytical Results

O p t h panels (minimum weight design) were designed for four load levels of axial compression for both panel configurations. The structural efficiency of these optimum panels is shown in figure 3 in the form of a weight index W/AL (where W is the panel weight, A is the panel area and L is the panel length) versus a load index NdL

(where Nx is the campressive stress resultant).

Both configurations yield optimum designs which are more structurally efficient than cammercially available aluminum aircraft camponents for all load levels considered.

For AS4-3502 properties, the structural effigiency klicates that a mininnnn numkex of 245 and

90 plies should be included. Since the minimum th&ckness of the 90' layers was zero, there were no 90 plies in the optimum design. All o p t h panels contain 12 stiffeners. If fewer stiffeners are used (and the total width maintained), the structural efficiency of the panel is reduced (i.e., the mass of the panel designed to carry a specified load is increased by including fewer stiffeners). Which constraints are criticdl at failure is deperdent upon the configuration and the load level. The webs of the hat stiffeners had to be thicker for the panels with the open hat stiffeners to prevent buckling of the webs of the stiffeners. For the lightly- and moderately-loaded panels with filled hat stiffeners, the foam core inside the stiffeners prevented the webs of the stiffeners from buckling so the webs of the stiffeners could be very thin. The extra weight included by filling the stiffeners with a foam core did not reduce the efficiency of the m t efficient panel since the foam core prevented the webs of the hat stiffeners from buckling. The filled hat stiffener configuration is more efficient than the open hat configuration for all load levels (Fig. 3). The o p t h configuration of the panels lncludes a minimum a m t of foam core in the skin and hat stiffeners that are approximately one inch

high with thin w e h and a thick cap. Buckling ncdes involving one arad/or 30 longitudindl half waves were critical. In most cases the werall extensional stiffness and werall shear stiffness constraints were also critical. For the mre heavily-loaded panels, the allowable strain constraints were critical.

The same constraints on dimensions, materials and loading conditions that were used to find optimum hat stiffened panels were used to design optimum unstiffened sandwich panels. The structural efficiency of these unstiffened panels is shown in figure 3 by the dashed line. The structural efficiency of these panels is ampamble to the foam-filled hat stiffened configuration for all load levels considered. However, these optimum designed unstiffened sandwich panels have thick foam cores which may fail in ways not predicted in PASCO. The o p t h unstif fen& sandwich panel which will carry a load of N>(L = 800 has a foam

core which is over 3.5 inches thick. The safety margin included on the buckling mode of one longitudinal half wave may not be adequate to account for failures due to high transverse shear strains when foam cores becaw thick in unstiffened sandwich panels. Therefore, this ccrmparison of stiffened to unstiffened m i c h panels may be misleading.

The optimization study was also coriduW for hat- stiffened panels with apen hat stiffeners using the material properties of IM6-18081 graphite-epoxy tape and Rohacell 71 foam core. The structural efficiency of the optimum panels made of AS4-3502 and of IM6-18081 graphite-epoxy tape for the open hat stiffener configuration is shown in figure 4.

Ihe same constraints were included for the IM6- 18081 panels as for the AS4-3502 panels but since the longitudinal and transverse stiffnesses are higher for IM6-18081 than for AS4-3502, the structural efficiency of the IM6-18081 panels is better than the AS4-3502 panels. The lower shear mAulus of IM6-18081 does not affect the structural efficiency for this loading condition. The critical constraints and failure modes are similar to those of the AS4-3502 panels.

ESrperimental Results

Several kinds of failure events occur in the TEM panels. Failure events which are only indicated by strain gages in a local region of the panel and which do not cause enough out-of-plane displacement to be evident on the moire patterns are local failures which do not cause a change in overall stiffness of the panel. Failure events which are indicated by all strain gages on the panel and which cause a change in the mire pattern are globdl failures which cause a redistribution of load in the panel and change of overall stiffness in the panel. The failure event which causes the panel to be unable to carry additional load is the final failure event. Lccal failure events do not necessarily affect the global failure of the panel and several globdl failure events can occur while the panel continues to carry load. All TEM panels tested experienced several global failure events before final failure.

Undamaqed Panel. Surface strain results indicate that the specimen behaves linearly as the load is increased until the first failure event. Strain -ts taken frum back-to-back pairs of axial

strain gages at the upper left comer and near the center of the panel are shown in figures 5 and 6, respectively. These gages are located on the unstiffened skin and the top of a hat, as Mcated in the figures. T h e first indication of nonlinearity in the load-strain relationship is at a load of approximately 24,000 lb in an area approximately two inches wide in one comer of the panel. The axial gages on the unstiffened face sheet in this region measured no furthur increase in strain as load was increased wkile the gages in this region on the stiffened face sheet continued to meafllre a linear load-strain relationship. Photographs of mire patterns shcw no change in the out-of-plane displacement pattern at a load of 24,000 lb. Strain gages not in this region of the panel were not affected. (e.g., strain gages in the center of the panel shown in figure 6) . This failure went is a local failure of the unstiffened face sheet or of the bond between the foam core and the unstiffened face sheet. This local failure does not appear to propagate.

All strain gages not in this local region indicated a linear load-strain relationship until an applied load of 48,500 lb was reached. At this load level strains measured by all gages indicate a failure event. The load-strain curves for all gages (i.e., those located on the hat stiffeners, the thicker face sheet and the unstiffened face sheet) show a discontinuity in slope at a load of 48,500 lb. This discontinuity in the slope of the load-strain curves for two axial gages at the lengthwise centerline of the panel for each panel tested is shcwn in figure 6. The solid lines represent the strain of the unimpacted panels, and the dashed lines represent the impacted panels. The higher strains recorded by each back-to-back strain gage pair shown are the strains on the unstiffened face sheet. The lower strains are the axial strains on the top of a center stiffener. The photqraphs of the moire patterns taken immediately after this failure event show that an out-of-plane displacement has cccured in a soall area in one comer of the panel, and are shown in figure 7. This area is not the same comer with the local failure. Since all gages are affected, 48,500 lb is taken to be the load of the first global failure event. The first global failure appears to be a delamination between the unstiffened face sheet arfi the foam core or the formation of a transverse crack in the foam, causing the unstif fened face sheet to buckle. This failure could be related to boundary effects. This failure does not appear to propagate immediately as the loading is increased. At this load level the strain calculated by averaging the axial strain measured by the gages along the lengthwise centerline of the panel is .00396. At the initial failure the axial strain decreases by .001 in the unstiffened face sheet and increases by .001 on the top of the hat. All load- strain curves (except in the region of the initial local failure) then become linear again as load is increased until immediately before the final failure.

The sequence of events at final failure is indicated by the mire patterns shown in figure 8. A delamination between the unstiffened face sheet and the foam core or a transverse crack in the foam core occrurs in the comer opposite the first globdl failure at 53,300 lb, as shown in figure 8a. The two delaminations have propagated across the entire panel before a load of 53,400 lb is reached, as shown in figure 8b. The final failure load is

54,400 Ib. At final failure, the regions with the initial failures unload and a previously undamaged part of the panel fails across the entire panel width, though not parallel to the clamped edge. The unstiffened face sheet buckles, delaminates from the foam core, and tne foam core fails. The mire pattern of the panel at failure is shown in figure 8c. In the failed panel, deformations can be seen across the top and center of the panel. Little evidence of the failures in the lmer part of the panel is visible after the final failure. The most pruminent buckles after failure are shown in figure 8d.

CamDarison of Panel with Fabrication Damage and Undamaged Panel. The failure mode of the undamaged panel (panel 1) is similar to that of the panel damaged during fabrication (panel 2). The first global failure events for both panels are failures in one comer of the panel caused by failures associated with the foam core and unstiffened face sheet. The loads at the first global failure events are 48,552 Ib and 48,802 Ib, for panels 1 and 2, respectively. The axial strains at the first global failure event are .00396 and .00390. The loads differ by .5 percent and the strains by 1.5 percent. The fabrication damage had no effect on the first global failure. After the first global failure went, the panels behave differently. The second lawer comer fails in both panels at load levels differing by 6 percent. The upper center of panel 2 fails but it does not fail in panel 1. Propagation across the bottom of the panel occurs at load levels differing by 9 percent and final failure occurs at loads differing by 20

~ t . The final failure mode of both panels is bucklmg of the unstiffened face sheet and cracking of the foam core.

Impact Damaged Panels. To study the effect of impact damage on the behavoir of TEM panels, damage i n d u d by several impact events with different impact speeds were cypared. The impactor was an aluminum sphere of &ameter .5 inches. The sphere was propelled at specified speeds ranging from 50 ft/sec to 350 ft/sec at a small sample panel with three stiffeners to detenrhe what sped to impact the test specimens. Panels were held in place but not axially loaded during impact. All impacts were on the unstiffened face sheet mid-bay between two stiffeners ard/or wer a stiffener. For an impact sped of 50 ft/sec, there was no visible damage. For 100 ft/sec, there was obvious damage to the unstiffened face sheet and foam core but the damage did not go through the foam core to the thicker face sheet. For speeds of 150 and 200 ft/sec the unstiffened face sheet and foam core were seriously damaged but the thicker face sheet was not. For impact speeds W e 200 ft/sec, the impact caused damage in the thicker face sheet as well as the unstiffened face sheet and foam core. For 300 ft/sec, a mid-bay impact causes splintering of the outer plies of the thicker face sheet.

Liqhtlv Impacted Panel. The first case of impact damage examined was when the impact would severely damage the unstiffened face sheet and foam core without damaging the thicker face sheet. The test specimen (panel 3) was impacted at an impact speed of 150 ft/sec (impact energy 2.3 ft-lb) . The panel was impacted on the unstiffened face sheet at two locations, mid-bay between two stiffeners ad. bekird a stiffener, as shown in figure 2, to determine which location repmted the mre severe problem and which location would initiate

failure when the panel was subjected to an axial compressive load.

After being subjected to impact damage, the panel was loaded in axial compression. The lightly impacted panel failed in a manner similar to the undamaged panel. The first out-of-plane displacement to be evident in the mire pattern was near the both loaded edge when the average axial strain along the center of the panel was about ,0034 and a load of 44,573 Ib was applied. This strain level is within 10 percent of the strain at first global failure of the undamaged panels. Same changes in the displacement field near the impact locations could be seen. but these deformat-ions did not seem to affect the panel failures. The strain gages on the unstiffened face sheet near the lccal buckle formed by the first global failure recorded a constant strain after the first failure event. The second global failure, in the form of a small lo& buckle in the unstiffened face sheet, occured across the tcp of the panel, near the loaded edge, at a load of 57,815 Ib. The final failure load was 63,000 Ib. The buckling of the unstiffened face sheet which cccured at final failure crossed the panel parallel to the loaded edges, not at the angle seen in the unimpacted panels. This prcwinent buckle does not pass through either impact site. The second failure is not visible in the moire pattern after the final failure. Even though the impact damage was obvious on visual inspection before the test and it severely damaged the unstiffened face sheet and foam core in a localized region, the impact damage did not noticably decrease the load carrying ability of the panel. Fhotcqraphs of the moire patterns taken soon after the test started and after failure are shown in figures 9a and 9b. No damage due to the impact events can be seen on the stiffened face sheet after panel failure.

Heavily inwacted panel. 'Ihe second case of impact damage examined was when the impact damaged not only the unstiffened face sheet and foam core, but also the thicker face sheet. The test specimen (panel 4) was impacted at an impact speed of 350 ft/sec (impact energy 12.5 ft-lb) . The panel was impacted at two locations on the unstiffened face sheet, wer a stiffener and mid-bay between two stiffeners . The heavily impacted panel failed at the impact locations upon being subjected to axial cxwpressive loading. The first failure was at an average axial strain along the lengthwise centerline of the panel of .0029 and at an applied load of 36,863 Ib. A lo& buckle fonned in the unstiffened face sheet, starting at the location of the mid-bay impact. A pho-ph of this first failure event is shown in figure 10a. The panel continued to carry load and the buckle did not propagate further or cause additional deladnations. The second failure event was at an axial strain of ,004 and load of 41,326 Ib. A buckle fonned in the unstiffened face sheet at the location of the impact behind a stiffener. This buckle progressed across most of the width of the panel. A phatograph of this second failure went is shown in figure lob. The panel continued to carry load up to an axial load of 48,780 Ib when final failure took place. Upon remcrvdl of the load, the buckle representing the first failure went oauld not be seen. The stiffeners did not fail during the final failure but the thicker face sheet failed. A photograph of the failed test spec- subjected to a light load to hold the

buckles open is s h m in figure 1Oc. Iwmination of the failed panel reveals that the foam core in the skin delaminated from the unstif fend face sheet and frum the thicker face sheet. However, few cracks through the thickness of the core are evident.

Comparison of uniimacted and impacted panels. The lightly impacted panel failed essentially the same way as the unimpacted panels. The only apparent effect the impact had on the failure was that the buckle which formed at ths final failure formed across the width of the panel at an anqle of +out 45 degrees from the loaded edge in the unimpacted panels and progressed across the width of the panel parallel to the loaded edge in the impacted panel.

The heavily impacted panel did not fail in the same mode as the other three panels. Failure initiated at the impact locations, not at the clamped edges. The failure loads and strains for this panel were slightly below the failure loads and strains for the first three panels. A comparison of strains and loads of all four panels is shown in figure 11.

The final failure loads of the undamaged, fabrication damaged and lightly impacted panels, as s h m in figure 11, differ by 20%, while the final failure load of the heavily impacted panel is 20% belm these. The initial failure load of the lightly and heavily impacted panels is 10% and 20% below the initial failure load of the unimpaded panels, respectively. However, the failure strains at initial failure, as shown in figure 11, range frcnn .0040 to .0029 as impact speeds range from 0 to 350 ft/seC. An example of the load-strain relationship for the undamaged panel and the impacted panels for two back-to-back axial strain gages located near the center of the panel are s h m in figure 6. The solid lines represent the strain in the unimpacted p e l , and the dashed lines represent the impacted panels. The difference in the load-strain relationship is very small until the first global failure occurs. The amount of decrease in strain recorded in the unstiffened face sheet when the first global failure occurs is approximately .001 in the undamaged panel but is .004 and .006 for the lightly and heavily impacted panels, respectively. The strain gages on the hats recorded the same increase in axial strain at the first global failure for all panels. The unimpacted panel and the lightly impacted panel continue to withstand increased loading after the first global failure but the heavily impacted panel does not.

Impact studies on other sandwich panels and stiffened panels can be used to help evaluate the results of the current impact damaged panels. Impact studies on laminated sandwich panels with thin face sheets and honeycomb cores, presented in reference 4, indicate that damage induced by impact energies as low as 2 ft-lb (lightly impacted) can cause a 30% reduction in the load carrying ability of the panel. In a sandwich panel without stiffeners, the face sheets must carry the load. In a stiffened panel with a smith skin, the stiffeners can carry much of the load, SG d m g e to the unstiffened face sheet has less effect on the failure load or strain. Impact studies on stiffened panels, presented in reference 5, inciicate that damage induced by an impact event on the skin of a blade stiffened m e 1 with impact energy of 9.25 ft-lb can redace the panel failure load by as much as 30%. For the foam-filled hat

stiffened TEM panels tested, impact with impact energy of 12.4 ft-lb reduced the panel's load carrying ability by 10% campared to the unimpacted panel. These results indicate that the foam core in the skin of the panel helps reduce the effects of impact damage. The foam core in the skin of the panel pmtects the stiffened face sheet fmm damage by abyrbing Wct energy. A panel with a foam core m the skm as thin as .25 inches can withstand significant damage with little reduction in load-carrying ability.

Correlation of test to analysis. The load-strain relationship predicted by PASCO and STAGS agree well with each other and with the experimental results up to the first global failure event. These predictions are accurate until a failure mechanism occurs which they cannot predict, such as face sheet wrinkling, shear crimping or failures due to high transverse shear stresses. The strain at the load of the first global failure event is s h m in figure 3 for both analysis methods and three test specimens. The load and strain at the first global failure for the undamaged and fabrication damaged panels were essentially the same. These data lie between the analytic predictions of the load at the axial failure strain of .004. The experimental results at a slightly lower load shown on the figure represent the load at the first global failure of the lightly impacted panel. - Camparing the results of a PASCO analysis of the TEM panel tested and a similar hat-stiffened panel without the unstiffened face sheet and foam core in the skin indicates that the presence of the foam core in the skin supresses a buckling mode of wavelength 1.43 inches (seven longitudinal half waves in the 10 inches between supported edges). Without the foam core in the skin and unstiffened face sheet, the panel would buckle at a load of 30,900 lb, but with the foam core and unstiffened face sheet, the panel would not buckle until a load of 150,000 lb, and the buckling mode would be one longitudirzal half Wave. The test specimen failed at a lmer load because its failure mode was not a buckling mode. The damage caused by the 150 ft/sec impact events is too localized to cause the buckling mode of longitudinal half waves of length 1.43 inches to occur. Therefore, the foam core still prevents the panel frcnn buckling in that mode and the panel fails in the same way as the unimpacted panels. Lccal impact damage to the unstiffened face sheet and core does not contribute to the failure of the panel as long as the stiffened face sheet is not damaged.

The foam core in the skin of the panel increases the bending stiffness of the panel and raises the buckling load. The failure of the panel m y be induced by high transverse shear stresses near the clamped edge. An analysis of sandwich panels which are clamped on the loaded edges and simply supported on the unloaded edges (reference 6) indicates that location of the highest transverse shear stress ryz in the panel coincides with the location of the first global failure of the TEM stiffened panels. Since these test specimens had free unloaded edges, high transverse shear stresses rx, do not develop in the way indicated in reference 6. However, the analysis approach used in reference 6 may be used to predict failure loads and strains based on the transverse shear stresses in the foam.

Panel failure may be predicted by simplifying the problem to a sandwich plate with face sheets the same as the unstiffened face sheet of the TEM panels and core the same as the foam core of the TEM panels. Such a panel, with the same dimensions as the TEM panels, would be expected to buckle at N x = 220 &/in at an axial strain of .0033. Since tne test specimens exhibited first failure at axial strains of .0029-.0039, this estimate is reasonably accurate.

Criteria for local failures due to face sheet wrinkling and shear crimping of sandwich panels with orthotropic cores and isotropic face sheets are given in references 7 and 8, respectively. An explanation of haw results fm these references can be applied to the TEM panels is presented in m i x A.

Application of the results of reference 7 to the TEM panel yields a criticdl strain in the face sheets of .0047 with a wavelength of .12 inches for face sheet wrinkling. The strain at the first global failure event in the unstiffened face sheet (as seen in figure 6) for the unimpacted panel is .0045. Face sheet wrinkling could cause the first global failure event. Application of the results of reference 8 to the TEM panel yields a critical strain of .019 in the face sheets. Since the strain in the tested panel at failure was much lwer than .019, the panel did not fail due to shear crimping.

Concludiw Remarks

Optimization studies indicate that laminated hat- stiffened panels which can be fabricated by the therml-ex&sion-molding process are 20-40% more structurally efficient than cammercially available aluminum a h & panels. These studies also indicate that when subjected to axial compression load, panels constructed of higher stiffness IM6- 18081 graphite-epoxy materials are more structurally efficient than those constructed of A54-3502 graphite-epoxy materials for all load levels considered. In addition, panels with foam- filled hat stiffeners are more efficient than panels with apen hat stiffeners for lcw and moderate loadings. Hat-stif fened panels containing a thin unstiffened face sheet, a foam core in the skin, a stiffened face sheet, and foam-filled hat stiffeners were constru- using the thermal- expansion-mlding precess and subjected to axial compressive loads. Experimental results agree well with analyticdl predictions. Failure loads for unimpacted panels and panels subjected to impact damage $o only one face sheet and the foam core were alnu>st the same. The mode of failure in all cases was separation of the unstiffened face sheet from the foam core. The failure load was reduced by 20% when the panel had impact damage in both face sheets and the foam core. The layer of foam core in the skin of the panel protects the stiffeners from impact damage by absorbing impact energy. A thin layer of foam core in the skin can protect a thicker face sheet and stiffeners from damage and all= the panel to suffer significant impact damage to the skin but lose little load carrying ability.

Acknowledgement

?he test specimens used in this study were constructed under the guidance of Mr. Robert M.

Baum, Materials Division, NASA Langley Research Center.

Face sheet wrinklinc

To apply the results of reference 7 to the TEM panels, the [+45IT face sheet must be assumed to be

isotropic to apply these criteria to predict the failure of a TEM panel. Since the stiffnesses in the axial and transverse directions of a [+45] laminate are equal, these criteria may be used to estimate failure. A [t45/Foam/T45IT laminate with

the thickness of the foam core equal to .25 inches and the thickness of each face sheet equal to .O1 inches is considered. Since the foam core in the TEM panel is isotropic, the criterion for face sheet wrinkling can be simplified slightly.

For a sandwich with an isotropic core, equation 1.8.10 in reference 7 can be expressed as:

where

and

q = t Gc / (tf E ) p = (I-vf2) / 2 (1 + vc) A = (1 + ( 1 - a 2 ) p ) sinh ( c ) + (1 - (1 - a 2 ) p ) 0 = v (1 + vc) / ( 1 - vc)

and t = core

t = face f G = core

v = core C

E = core

E = face f

v = face f

tickness

sheet thickness

shear modulus

Poisson's ratio

Young's modulus

sheet Young's modulus

sheet Poisson's ratio

T = maximum allowable shear stress in core 6 = amplitude of initial imperfection c = 7rt /L

L = half wavelength of sinusoidal wrinkles

The critical stress is obtained when Q is a P

minimum with respect to c.

Shear Crim~ing

The critical strain for shear crimping to occur if no plasticity is permitted is expressed in reference 8 as equation 4.2-7 and becomes:

= h2 Gc/ 2 tf tc Ocrimp (-42) 4. Rhodes, lbrvin D. : Impact Fracture of

Ccanposite Sandwich Structures. Presented at the where t = core tickness ASME/AIAA/SAE 16th Structures, Structural

Dynamics, and Materials Conference, May 1975. t = face sheet thickness f AIAA Paper NO. 75-748. G = core shear modulus

5. Williams, Jerry G. ; Anderson, Melvin S. ; h - t + t

c f Rhodes, Marvin D.; Starnes, James H., Jr.; a113 Straud, W. Jefferson: Qecent Developmmts in

when the core is isotropic and the face sheets are the Design, Test- and Impact-Damage Tolerance eaual thickness. of Stiffened Ccanposite Panels," Fibrous

References

Anderson, Melvin S.; and Struud, W. Jefferson: A Generdl Fanel Sizing Ccanputer Code and Its Application to Ccarrposite Structural Panels. A Collection of Technical Papers - W A S M E 19th Structures, Structural Dynamics and Materials Conference, April 1978, pp. 14-22. AIAA Paper NO. 78-467.

William, Jerry G; and Mikulas, Martin M., Jr.: Analytical and Experimental Study of Structurdlly Efficient Capsite Hat-Stiffened Panels. Presented at the ASME/AIFA/SAE 16th Structures, Structural Dynamics, and Materials Conferenoe, May 1975. AIAA Paper NO. 75-754.

CamDosites in Structural Desisn. Plenum Press, New York, 1980, pp. 259-291.

6. Minguet, Pierre J. : W1ckling of Graphite-Epoxy Sandwich Plates, TEIAC Report 86-16, M.S. Thesis, M.I.T., 1986.

7. Norris, C. B.; Ericksen, W. S.; March, H. W.; Smith, C. B.; and Boller, K. H.: Wrinkling of the Facings of Sanctwich Constructions Subjected to Edgewise Ccmpression, Fl?L Report No. 1810, M a r c 3 1956.

8. Sullins, R. T.; Smith G. W.; and Spier, E. E.: Manual for Structural Stability Analysis of Sandwich Plates and Shells. NASA CR-1457, December 1969.

Almroth, B. 0.; and Brogan, F. A.: The STAGS -ter Code. CR-2950, 1980' Table 1 Material Properties

Material AS4-3502 IM6-18081 Rohacell 71 ~raphite-epoxy ~raphite-epoxy foam

Young's modulus E 6

18.5 x 1 0 psi 6

1 25.0 x 10 psi 13100. psi

Young's modulus E 6

1.64 x 10 psi 6

2 1.70 x 10 psi 13100. psi

shear modulus G 6

. 8 7 x 10 psi 6

12 .60 x 10 psi 4270. psi

Poisson's ratio p 12

.3 .21 ,534

density p ,006 lb/in 3

.006 lb/in 3

,00255 lb/in 3

Failure criteria

tensile strain E t t X? EY

C C compressive strain E E

x' Y shear strain 7

XY tensile stress ot ot

x ' Y C C

compressive stress a XI OY

shear stress r XY

213 psi

-898 psi

+I85 psi -

Table 2 Test Specimens

Specimen Specimen Final failure number damage before load load per unit

loadine. (klb ) width (klb/in) 1 no damage 54.4 4.53

2 fabrication damage 67.0 5.58

3 impact (150 ft/sec) 63.0 5.25

4 impact (350 ft/sec) 48.8 4.06

Open hat Foam-filled hat

a. configuration 1 b. configuration 2

Stiffened face sheet e 6 "

Unstiffened E 2 5 " face sheet t o l l "

c. test specimen

-1k.25"

Figure 1. Cross-sections of hat-stiffened pmels. Figure 2. m t i o n of impact damage.

2 2~ 20 1 o - ~

2- Commercial aircraft Commercial aircraft aluminum wing

aluminum wing compression panels compression panels

1°9 8

WIAL, 7 Iblin.3 6

5

4 Solution method for failed panels

3 0 PASCO 0 STAGS

2 A Experiment

I

100 200 400 600 800 N xlL, l b h 2

Figure 3. Structural efficiency and experimental failure loads of thermal-eXpnSion- molded panels using AS4-3502 properties.

80000

60000

Load, ~b 40000

20000

Figure 5.

Strain gage locations

+u

\Local failure

//

I I I 1 .001 .002 .003 .004 .005 .006

Axial strain

Axial strain in region of local failure of unimpacted panel.

Figure 4.

60000

Load, ~b 40000

20000

Structural efficiency of open hat- stiffened panels using properties of AS4-3502 and IM6-18081 properties.

Strain gage locations -+ First global failure

- (hat) 1 L,L!/

First global failure (face sheet)

/ , Unimpacted panel

--- Impact at 150 Wsec Impact at 350 Wsec

w 0 .001 ,002 .003 .004 .005 .006

Axial strain

Figure 6. Ccorrparison of strain-load relationship for three panels for axial gages near center of panels.

Location of - local failure

First global failure

Figure 7. First global failure of unimpacted panel.

/ d s t global Second global L! Propagation failure failure of failure

a. First and second global failures. b. Propagation of damage.

Figure 8. f ogress ion of failure in unimpacted panel.

Deformations formed at final failure 7

Buckles formed at \ earlier failures

C. Final failure.

Impact d a m a y ,

I

Stiffene r,

Skin separation

Foam failure

- Foam core

d. Side view of failed panel.

Figwe 8 . Concluded.

Failure of - stiffened skin

a. Initial impact damage. b. Failure of panel.

Figure 9. Failure of lightly impacted panel.

lnitial failure

lmpact 2 locations

a. First global failure.

Second 1 failure

b. Second global failure.

Initial failure

- Final failure

c. Final failure.

Figure 10. Fmqression of failure in heavily impacted panel.

Axial strain

0 Final failure 'Ool 1 A initial failure 20000 1 A ~nitial failure L

0 100 200 300 400 0 -

100 200 300 400 Impact speed, ftlsec Impact speed, ftlsec

Figure 11. Comparison of failure load and strain.