NACA TN No, 1820 NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS TECHNICAL NOTE No, 1820 STRENGTH ANALYSIS OF STIFFENED THICK BEAM WEBS By L, Ross Levin and Charles W, Sandlin, Jr. Langley Aeronautical Laboratory Langley Air Force Base, Va. Washington March 1949i AU TECHNICAL NOTE NO. 1820 ‘STRENGTH ANALYSIS OF STIFFENED THICK BRAM WEES By L, Ross Levin and Charles W. Sandlin, Jr. SUMMARY A previously published method for the strength analysis of stiffened ‘beam webs, with particular attention to computing crippling feilure of the uprights, has been revised and extended to aprly to beams with ratios of applied shear to buckling shear loss than 2.5. A comparison of this revised method with the results of tests of thick-web beams 1s presented, ‘The results in this paper concerning the procedures for calculating the critical shear stresses and for predicting forced crippling failure in the uprights gupersede NACA TN No, 1364, Formas and graphs applying tothe parts of the strength-enalysis method which have been revised aro presented. TITRODUCTION Published methods for strength analysis of stiffened beam wobs aro of doubtful accuracy for beams with thick webs, Kuhn and Peterson sugassted in reference 1 that the methods of that paper be limited to beams with ratios of wob depth to web thickness greater than 200 but less than 1500 and ratios of upright to web thickness greater than 0,6, At that timo there were very Little experimental date to check the accuracy of these strength-analysis methods when applied to thicker webe, and the data that vere available indicated some possibility that the strength-analysis methods of refor— ence 1 vould not be satisfactory for thicker webs. ‘The present investigation was undertaken to determine the accuracy ‘that might be expected from the strength-analysis formulas of reference 1 vhon applied to beams with ratios of web depth to web thickness of approrimately 115, As a result of thie investigation cone parts of the method in reference 1 were modified in order to obtain a method of strength analysis which vould be satisfactory for thick-web beams as well as for thin-veb beams similar to those of reference 1. The present paper, therefore, supersedes the sections of reference 1 which give the procedures for calculating the critical shear stresses and for predicting forced crippling failure in the uprights.| ACA TH Wo. 1820 ‘SYMBOLS cross-sectional area, square inches rivet factor (Het area along line of holes ) Gross area along line of hole: Young's modulus, ksi monont of inertia, inches force, Kips etatic moment about noutral axis of parte of cross section as specified by subscript, inches coefficient of edge restraint (see forma (2)) transverse shear force, Kips specing of uprights, inches @istance from median plane of web to centroid of (single) upright, inches depth of bean, inches (see Special Combinations) d4agonal—tension factor ‘thickne: ey ee (used without subscript signifi of wed! centroidal radius of gyration of cross section of upright about axis parallel to web, inches (no sheet should be included) normal stress, ksi shear stress, ket plasticity reduction factor (ratio of oritical shoar stress in the plastic region to the critical shear stress that would be obtained if the material were wholly elastic) Subsoripte: Fr v vrrtghtAGA TN No. 1620 3 w cale or meas ult Special Combination: ?y web calculated orittcal effective gross eection of web axiom measured ultimate internal force in upright, kips clear width between uprights (measured between rivet lines on single uprights, measured between edges of uprights for double uprights), inches clear depth between flanges, inches dopth of beam measured between centroids of flanges, inches length of upright measured between centroids of upright-to- flange rivet patterns, inches ‘theoretical buckling coefficient for rlates with simply supported edges restraint coefficients for edges of sheet: along flanges and upright, respectively (If Gq > hy, eubstitute b, for a, dy for hy, Rg for Rh, and Ry for Ry.) "pasic" allowable stress for forced crippling of uprights (valid for stresses in upright material below proportional Limit tn compression), ket & z flange flexibility factor oa \ > where ( Go + Tete w and I, are mosonte of inertia, about their ovn axis perpendicular to web, of compression flange and tension flange, respectively,4 ACA TN Wo. 1820 TEST SPECIMENS The test specimens were 24S-13 alumimm-alloy beams with a ratio of. web depth to web thickness of approximately 115. The ratios of stiffener spacing to-web depth vere approximately 0.25 and 0.70. Both. single—upright and double-upright beams were tested. Each beam was given a code designation which parallels the designation used in reference 1, For example beam Y-12-1S has the following meaning: V designates tho present series of tests (series I, II, IIT, and IV were published in reference 1) 12 4s the approximate depth of the beam in inches 1. te the mumber of the beam within the series S stands for single uprights (D for double uprights) ‘The nominal dimensions of the beams and the details of the construction are shown in figure 1. The actual properties of each beam are given in table 1. The specimens vere tested as simply supported beams in the jig shown in figure 2, vhich eupported the beams laterally but aid not-restrein the dending of the beam. Tho flanges of the beam vere siipported by closely spaced vertical bars resting on rollers (not visible in the photograph) that allowed each bar to move parallel to the plane of the web as the beam deflected. ‘TEST PROCEDURE Buckling loads for the web were determined by visual observation of the webs and by meacuring the straine in the uprighte with rosistence— typowire strain gages. There should not bo any strain In the uprighte until the critical shear stress 1s reached; however, because the webs had slight Initial eccentricities, sone strain in the uprights usually occurred fas soon as any load vas applied. Tho buckling load vas determined by plotting the measured strain in the uprights against the shear load on the dean; the point at vnich tho load-strain plot for the upright~derarted from a straight line was taken ae an indication of buckling In the wed. The critical shear stress in the web vas computed fron thie buckling load by the formulaNACA TN No. 1820 + Se 4 4 Sie a 3Q; ‘This forms gives the average shear stress in the web according to the engineering theory of bending. RESULTS AND DISCUSSION ‘The results of the investigation are shown in table 2. Experimental ‘buckling loads and failing loads are recorded. The failures were either forced crippling of the uprights or web rupture. Analysis of the present tests by the methods of reference 1 indicated that critical shear stresses and ellovable upright stresses predicted by these methods were not suttictontiy accurate for bone with thin upetente (Ze 2.3) and tatoe webs (2s u5). Methods, which give satisfactory resulté for thickwed ‘beams as well as for beams similar to those of reference 1 (2 0.6 and 200 <2 < 1500], ere discussed for computing the critical shear stresses and the ellovable upright stresses. Critical Shear Stress The formula for the critical shear stress of the web was given in reference 1 as (emda me} @ A plot of this equation for e penel with four simply supported edges (Band Ra qual to 1.0) is shown as figure 3. A comparicon of tho experimental critical shor stresses with the critical shear stresses computed by forma (2), using the restreints R given by the ampirical curves in reference 1, indicated that the values of R given in rofer— once 1 are satisfactory for webs vith double uprigkte, but are too high for webs with single uprights and = <1.3. Values of the restraint cooftictont Ry for single uprights vere computed from the experizontal6 NACA TN No. 1820 critical shear stresses and vere used to establish the curve shown in figure l, The curve for double uprighte shown in figure 4 is the same as the curve given in reference 1. If the critical shear stress computed by formile (2) 1s beyon4 the elastic range of the material, the stress must be corrected for the reduced value of the modulus, In reference 1 critical stresses in the Plastic range were obtained by drawing tangents to the elastic curve fron yy tt S20, ‘Thnae corvoe are stim in rofqynoe 2 for a panel with simply supported edges. In order to obtain the critical shear stress in the plastic range for any other set of edge conditions or any other material, a separate set of curves mst be dravn. Reference 2 presente a method of computing from the stress-strain curve of the material the plasticity reduction factor n, which 1s the ratio of the critical shear stress in tho plastic region to the critical shear stress that would be obtained if the materiel vere wholly elastic; ‘that 12, . Tor( Plastic) = nTop(elastic) If formula (2) 1s substituted for 7,,(elastic), the expression for the critical shear stress in tho plastic region ie von = men 3(2) fa + 3 (a -2)(S)] @) ‘The critical sheer stress in the plastic range my be obtained by computing Tyy/n from forma (3) end then reading To, from figure 5, wich chovs “7, asa function of Top/n. If the critical shear stress computed fron formia (3) is plotted as a function of ,/t, the gurve in the plastic range ie practically a straight Mino and intersscto 2 = 0 at 739 ket. Formula (3), for mactical purposes, gives the same line to that obtained by drawing & tangent to the elastic curve from Tut decause 7,14 16 between 37 kei und 42 kei. (See reference 3.) The caloulated critical shear tresses based on restraint R obtained from figure and the measured critical shear stresses for both single~ upright and double-vpright-beans are shown in table 2. The ratios ofNACA TN No. 1820 1 measured critical stress to calovlated oritical stress vary from 0.77 to 1.21. These calculated values of 7, are probably adequate for the Purpose of determining the @legonal—tension factor xj better result probably cannot be obtained go long as the restraint Ris represented only es @ function of ty/t, because representing R by this function only 18 en extreme simplification of a complex problem. ‘The application of formulas (2) and (3) and the restraint curves of figure 4 to beams with thin uprights may give critical shear stresses lower than those that would be obtained if the preence of the uprights wore disregarded entirely and if T,, were computed for a web bounded by the flanges and the root and tip bays of the beam, This result was obtained because the value of 1... for e panel between two uprights was assumed to be the same as the valus of 7, for an individual panel bounded by edge members of the samo size as the flanges and uprights. Actually, the adjacent panels in the beam have an appreciable effect on ons another. In oan V-12-125 the value of t,,' obtained by disregarding the uprights was higher then the value obtained by assuming that tho uprights éivided the vob into seperate panels. ‘Tho observed 7... was 61 percent greater than the calculated t,, if the veb was assumed to be divided into separate pansle by the uprights, but only 21 percent greater than the calculated T,. if the presence of the uprights vas disregarded entirely. In practice To» must be calculated by both methods for beans with thin uprighta ani the higher value used, because the ratios t/dg end d/h will be different for the two conditions and because no gonoral rules seem to exist that Predict which mothod vould give the higher value of Toy. Forced Crippling Fatlure in Uprights Four types of failure of urrighte are discussed in reference 1; but only one, forced crippling failure of the uprights, vas observed in the present tests. General elastic instability failure of the web end uprights seems to be the only other type of upright failure likely to cour in thick-veb beams. ‘The shear buckles in the web force the buckling of the upright in the leg attached to the web. The amount of the forced crippling (buckling) depends upon the relative sturdiness of the upright and web. In refor— ence 2 formulas for forced orippling vere based on the paremoter x VEy/t- Tf k was less than 0.5, an effective value of k vas used. Use of ‘this parameter in the present test of thickweb beams indicated that tt was not satisfactory. The allovable stresses were too lov if k itself8 NACA TH Wo. 1620 was used and too high if the effective k was used. Much better agreo— nent with the present teste of thick-web beams and about the same agreo— ment for the teste in reference 1 vere obtained by using the parameter ay yk ol ee offoctive value of x. Figure 6 16 a plot ofthe valuos of oy... computed from the Loads on the bonus et-fallure, for all the eingle-upright beams of the present Investigation and for all the singleupright beams shon in figures 22 end 23 of reference 1. (Tho beams shown in figs, 22 and 23 of refer— ence 1 represent about 90 beans tested by four manufacturers and 32 beans tested by ACA.) Tho stresses oy wore computed with the ald of the Face analysts chart of figure 7. Tite chart covers the lov rango of Ay, /at end t/rop that 18 not-shown in the analysis charts of reference 1. Tho points shown in figure 6 are fairly evenly distributed about the average eurve r 2 oo = 3203(%B)3 (4a) The curve recamended for design is given by the formla @ aosi(eait (am) and is about 20 percent below ths average curve, Only two points fall definitely below this design curve. The lowest of these pointe 2 1 e point at et an (tan) 3 8 col ‘rom, 6 point at “vy = 10.6 ket and 15(‘S)3 = 0.63) was commuted tro: the fatling load for one of the manufacturer's beam tests. The NACA constructed and tested a duplicate of the beam tested by the manufacturer. In the NACA investigation a local buckle developed in the outstanding leg of one of the uprights at a load approximately 11 percent above the failing load given by the manufacturer; the NACA beam continued to carry load until the load was about 73 percent above the failing load given by ‘tho manufacturer, At thie load two local buckles Aeveloped in each stiffener and the edges of-the etiffenors started to crack at these buckles. No dotail information about~the behavior of the manufacturer's beams was furnished but-the behavior vas probably similar to that observed in the NACA test; the manufacturer might have interpreted the first bucklo in the upright ae failure end made no furthor.attempt to apply more load.NACA TN No, 1820 ‘The upper curve on figure 6 is 20 percent above the average curve. One of the points in the present series of tests falls above the curve, end most of those from reference 1 that were more than 20 percent above ‘ths average curve when the parameter x \fty/t was used are still above the 20-percent line in figure 6. Figure 8 shows all the Gate nov svatlable for double uprights using 2 z to 3 3 sie stor + (2). tn tna to he een su fr tome uprights te 4? Gg = 27 @) (5a) ‘The formila for the recommended design curve for double uprights is A 33 op = 2k ( ) (x) % None of the tests points for the beams with double uprighte is below the recomended design curve and only ons point 1s more than 20 percent above ‘the average’ curve. In the present tests the ratio of the actual fatling loads to the predicted failing load ranged from 0.90 to 1.14. In reference 1 it vas suggested that the formula for computing the effective area of single uprights bg 1+ fu * might not be satisfactory for thick-web beams, because the simplifying assumptions implied by this forme may not be justified. These implied assumptions are: (a) Tue eccentricity e of the load on the upright 1s constant (b) The ratio e/p 1s not changed appreciably 1f the contribution of the web to the effective cross section of the upright is neglectedw MAGA TN Wo. 1820 Assumption (a) 18 plausible 1f tho uprights are very closely spaced Pecause the web then moves with the uprights (reference 4). Assumption (b) would not-seen to be justified for thick-wob beans; hoover, for low values Of the ratio 7 /Tgp a largo difference in the total effective area of the upright causes only a small change in oy ant, therefore, satisfactory yesulte are obtained. A study of the analysis chart in figure 7 will help to orietn ‘this facts the curves approach a vertical line as tho ratio 1/Tpp decreases: Web Fatlures The average nominal shear stress in the wed was computed by formula (1); the poak value of the nominal shear streas in the web for predicting web rupture was computed by the formilas of reference 1. Critical shear stresses wore computed from formas (2) and (3) by méans of the restraint coefficients given in figure 4. Tae allowable values of the peak shear stress in the web, which are shovn in figure 9, were obtained fron Teference 5. The values are based on tests of long webs subjected to loads approximating pure shear and contain an allowance for the rivet factor; this factor may be included because tests have shown that the ultimate sheer streas on the gross section 1s almost constant in the normal renge of rivet factor (C, > 0.6). In the eix thick-wob doams which failed by web rupture, the ratio of actual failing load to predicted failing load ranged from 0.92 to 1.18. This degree of accuracy 1s aprrorimately the sane as that obtained for the thinner beans ( 200 <2 <1500) atecuased in reference 1, ALL those The mothods of predicting the critical shear stresses, forced crippling failures of the uprights, and rupture of the webs presented are applicable to stiffened boam'vobs with ratios of web depth to wed thickness ‘vetwoen 115 and 1500. The accuracy of these methods is about the seme as ‘thet of the methods presented in NACA TN No, 1364 which were applicable only to boams with ratios of web depth to web thicknesa between 200 and 1500. = Langley Aeronautical Laboratory Netional Advisory Committee for Aeronautics Langley Air Force Base, Va-, December 16, 19h8WACA TN No. 1820 a ‘REFERENCES 1, Kun, Paul, and Peterson, James P.: Strength Analysis of Stiffened Boam Webs. NACA TN No. 1364, igh7. 2, Stowell, Elbridge Z,: Critical Shear Stress of an Infinitely Long Plate in the Plastic Region, NACA TW No. 1681, 1948. 3. Anon.: Strength of Metal Airoreft Elements. ANO-5, Army-Navy-Civil Comittee on Aircraft Design Criteria. Revised ed., Dec. 19425 Amendment 2, Aug. 8, 1946. h, Wagner, Herbert: Flat Sheet Metal Girders with Very Thin Metal Wed. Part I — General Theories and Assumptions, NACA TM No. 604, 1931, Part IT — Shoot Motel Girders with Spars Resistant to Bending, Oblique Uprights — Stiffness, NACA TM No, 605, 1931. Part III — Shoot Metal Girders with Syars Resistant to Bending. The Stress in Uprights — Diagonal Tension Fields. NACA TM No. 606, 1931. 5. Ievin, L. Ross: Ultimate Stresses Developed by 2ST and Alclad 758-7 AuimimumrAlloy Sheet in Incomplete Diagouel Tension. HACK 1 Wo. 1756, 19!me | ay | aa “ 2 “ view 2.38 | 10.88 2.5 : * 3 0.458 0,58 vase | 35 | 50.60 am | deh es | ox a waew | not | 020 am | 2x2xo.0m wna | ate a vats | ansr | so am | Bxgx cone wey | an ” vase | ns | son eo | 2x2x00m wos | oe os vse | aor | ao em | 2x2x 0.0m er | ts a taom | use | s0.0 roo [abe ae | at ast Tae 1.58 | 20.30 1.00 ix i 0605, 2k Lat raego | uss | 20.50 100 | Bxgxona oo | op ast vase | ase | a00 rao | 2x2 one | age Ln rasan | nse | 20.9 rao | Bx 2xo.ons 6 aa Tasies | 158 | 10.50 1.0 bx bx 0.0604 085, Lat Wasagd | 2.58 | 10.90 1.00 hx 2 x 0,067 an Lat vacate | st | soe ra | 22x o.ope a tar view | ass | w.20 nao | Bx Peo, 0 ast OzgT “ON AL YovNSMO 2. TEBE Da AND REIRS : Tait Fa a a ts, Par fone | _+ AT | ome | o [e en | Seems | ee | damn |e Be | TSES | x | cfnm | EE | Aa) ae (a) | (e) fe) | (a) vase [we [aa | oa [wal no [as [ass] oa |ma [rmommaa] om [om vase | cco | aro | an [ore] os | sas | on] eo] ae |rmtenmine| 0 | 0 ALS 6.1 G4 30.2 116 031] 75] 71.0 | Forcod crippling} 98 | .98 race |v {oes | am | aa] ar | am | ae] ro| 0 | nme wining 26 frat view 9.8 23.7 9 8.0 aT LB 9M] TH2 [117-0 Wed 18 | 128 Te 27 3.8 a T0 as 1.23 .0N8 | 78.0 | 78.5 | Forced ortppling) 96 | .96 rae 4.9 65 2h The 6 2.05, 15! 67.0) 3.0 ‘Woo 2.06 | 1.06 acm | ace | sa | oe | mo] me | aa |aw) volne | wm | aa] ss eset (Rees ace afe et sen atest | eee | a eee ore te eeeette fat | i assm| se | ssa {ca fool ans | ase | am] malmo | om | oo) ve rasa} om | a | oe | ms] xo | ae | aol wolme | om | aes] roe Jvieies | 10.2 22.3 12 BO 23 1.98 70 | 67.5 | 45.0 | Moreed ortypling! 2.23 | 1.13 esx} sa | 3a |e | mo| or | am | a0] mo] ar | root erinsin| 220/120 jyaaaks | Us.0 13.2 oh 39.2 25.4 2.80 123] 72.0] 58.0 | Foroet oriypling) 2.02 | 1.02 rseam | oa | or | x9 [ers] ms | aa | 20] ms] eo [ree crimun| ver] ir (free wo fatter. Yor Toros crighltng failure. Opt te the Loot coe Of the predicted loads FL or Py. 47 to tat predicted load (FL oF Fy) vaich correspmiia to the observed type of failure. Sea OegT “ON NL vou etBV ACA TN No. 1820 jy 4 1 4 Ses] { : a = 51$———__—___—_—+| $20.25 _Lie thicknesses of sheet fa yz I { { 1 13g Ele | Lop) AH 58 4 #0.70 Section A-A Section A-A Double uprights Figure |.-Dimensions of test beams. Single uprights Seeeee eet lee a Ir “nen Inn " is . Ors aT pean a ay ot ia ae ia 87 aAGA TN No. 1820 LT Tops ksi 200 400 600 4 ea Figure 3.— Buckling stress simply supported edg Tor for plotes with E=10,600 ks!.18 NACA TN No. 1820 ° | 2 3 Uh | ze Figure 4.-Empirical restraint coefficients for calculating web buckling stress. 24 PEF 5 \6| 8 dl 0 Sd seg eO ge Ter ksi 4 7 Figure 5.-Relationship of to, to ¥ for 24S-T3 aluminum- alloy sheet.~--= 245S-T above r 12 x Av, ——~ |__ Av. (formula(4a)) Figure6.— Stresses in single uprights at failure caused by forced crippling. oegT “ON ME Your or20 48 44 40 36 32 28 24 gy s|¢ .20 08 04 10 "14 8 22 26 30 Tr Ter Figure 7 Diagonal- tension analysis chart. NACA TN No, 1820 34 «3840 32 gz 12 x Av; —~ 324 Av.(formula(Sa)): 6 16 Recommended for—| design (formula (5b) 8 h~lO 12 25 30 70__] 24ST3 0 Oo Bb O 0 75S-T6 4 0 : | | | ° 4 6 lo 12 4 tus See Figure 8.-Stresses in double uprights at failure caused by forced crippling. son RE Yow, ozgt40) rang [ag I poor ere tae a FS Heavy washer} 32 NS 16} si Gurves based on o,,4=62ksi Curves bosed on oy,=72 ksi 5 2 4,6 8 b 9 @ # 6 8 1 (0) 248-73 aluminum alloy. (b) Alclad 75S-76 oluminum alley. S27 Figure 9. Average ultimate shear sttesses in the gross section. zz. som REVOVE ozet