You are on page 1of 6

J. Sound Vib.

(1968) 7 (3), 431436

SOME EXPERIMENTS WITH STIFFENED PLATES UNDER ACOUSTIC EXCITATION


F. J. FAHY Institute of Sound and Vibration Research, University of Southampton, Highfield, Southampton, SO9 5NH, England
AND

R. B. S. WEE
Department of Mechanical Engineering, University of Southampton, Highjield, Southampton, SO9 5NH England? (Received 4 October 1967) Plates which will be subjected to intense acoustic excitation are often stiffened with beams. These are added either for reasons of static stability or to reduce the vibration. Measurements have been made of the average acoustically induced strains in a series of _ik in. mild steel plates to which identical beam stiffeners have been attached in different ways. It has been found that point attachments such as rivets can be preferable to line

attachments such as welds. The magnitude of the benefit gained from attaching the stiffener at discrete points is not explained by existing theoretical vibration analysis.

1. INTRODUCTION Acoustic fatigue has been a major consideration in the design of aircraft structures for more

than a decade. It has recently reared its ugly head in the field of design of the heavier steel structures used in the gas circuits of large nuclear reactors. The spatially distributed, high frequency nature of acoustic pressure excitation makes the problem of design, or later modification, of structures to minimize severe vibration rather different from the more familiar problems associated with mechanically induced harmonic vibration. Consequently measures which have been found to be beneficial in the latter case may not be so with acoustic excitation. A case in point is that of the stiffening of plate or shell-like structures either to increase their static stiffness or to reduce their vibration stresses. Some simple experiments have been made to obtain a broad idea of the effect of different methods of plate-stiffener connection on acoustically induced plate stress [l]. No attempt has yet been made to investigate the effects of variations in stiffener configuration, or in its stiffness relative to that of the plate. However, it is felt that sufficiently large differences have been observed between the stresses in plates stiffened by continuously attached stiffeners and those in which the stiffener was attached at discrete points to justify early publication of these results, without, as yet, their explanation, to bring to the notice of designers the possible beneficial effects of correctly applied stiffeners.
2. APPARATUS AND MEASUREMENTS

Stiffened and unstiffened rectangular 36 in. x 30 in. mild steel flat plates of & in. thickness were clamped to a frame of 2 in. x 2 in. x + in. steel angle section which reinforced the

i Present address: 2 Depot Road, Block 193, Singapore 5, Malaysia.


29 431

432

F. J. FAHY AND R. B. S. WEE

upper edges of a 32 in. x 26 in. x 24 in. steel water tank of O-104 in. thickness. The cistern was surrounded by a 4 in. thick layer of sand contained in an outer wooden box. An intense (148 dB), broadband noise was produced in the box by allowing high pressure air to enter

-a

Araldite

Wetted

Glued

-and bolted

Point

bolted

Unattached

Figure 1. Stiffener configurations.

$in.xl Clomplng

in x4 in

2.

Figure 2. Strain gauge positions. it at the base of one side through a 3 in. gate valve. The air diameter pipe, also at the base of a side. Aerodynamic buffeting Microphone traverses of the volume of the box showed that levels varied by no more than fl dB from the average over the

left the tank through a 6 in.


of the panels was negligible. the +-octave sound pressure frequency range from 380 to

SOME EXPERIMENTS WITH STIFFENED PLATES

433

12,500 Hz, except in the lower region where aerodynamic effects interfered with the sound measurements. Stiffeners of a standard configuration were attached to the plates in different ways. The stiffeners consisted of a grillage of 1 in. x 1 in. x Q in. channel sections welded in the form of 6 in. squares. The stiffener attachment configurations are shown in Figure 1. The plate was held against the unattached stiffener by a small static pressure differential. The ends of the grid were clamped at the boundary together with the plate. Strain gauges were placed in the positions shown in Figure 2 for all panels. One-third-octave strain levels were measured during excitation. Average mean square values were obtained by taking the arithmetic mean of the individual mean square values. Accelerometer measurements were also made to check that the ratio between average strain and average acceleration approximated to the theoretical single mode, or diffuse field, value given by (c2) = (n2)wf CL, where w, is the centre frequency of the filter band and CL is the speed of longitudinal waves in the plate material. Measurements were made in +-octave bands of the total loss factors of each panel in situ by recording the decay of strain following impulsive excitation. Averages were taken over loading and transducer positions.

3. RESULTS The results are presented in Figure 3 in the form of radiation loss factor curves. These have been calculated from a relationship between average mean square acceleration and
-30

-60 d P c- -70 0 H 2 -80 -90


--+--a--- -Glued and bolted stiffener Point bolted stiffener Rivetted stiffener

f octave

band centre

frequency

(Hz)

Figure 3. Radiation loss factors for the set of stiffened plates. average mean square pressure derived in reference 2, and discussed in section 4 below, together with the theoretical relationship between average mean square strain and average mean square acceleration of section 2 above. They are presented in this way so that direct comparison can be made between the present results and those of previously published literature concerning the response of structures to high frequency, diffuse field acoustic excitation [2,5], in which q radhas been derived from experimental results in a similar manner, albeit directly from measured accelerations. Because the factor of proportionality between T,,~ and the average mean square strain per unit average mean square pressure is the same for all plate configurations, apart from

434

F. J. FAHY AND R. B. S. WEE

the individual mechanical loss factors, as shown in section 4, the curves of l;jrad serve as better indicators of relative merit for the various configurations than the measured strain/pressure curves because they normalize the results to conditions of equal mechanical damping for all configurations. Of course the relationship presented in section 4 can be used to derive strain/pressure values if it is so desired. Figure 4 shows the total loss factors, qt, measured by impulsive excitation and decay. It is interesting to note that the high.frequency damping mechanism proposed by Ungar
I I I I I I I I I I I

0,026

h
Point bolted

.Glued

and bolted

/ --L 0,002 0 I 2 I 3 I 4 I 5 Illll 67691

Unattached

Plain panel

<-....* I 2 I 3 --=x 45

1 102

103 Third octave band centre frequency (Hz)

Figure 4. Measured total loss factors. and Carbonell [7] does not seem to operate here because the rivetted structure, which would be most likely to squeeze air in and out of the beam plate interspace, has relatively low damping. 4. DISCUSSION The theory of the vibration of coupled systems has been developed in references 2,3 and 4. Where the literature considers an acoustic field as one of these systems it is usually assumed that frequencies are sufficiently high to ensure the establishment of an approximately diffuse field with waves incident on the structure at all angles between +7r/2 and --n/2 with equal probability. One feature of such a field is the uniformity of sound pressure distribution over the volume. It would appear that the present experimental field may be considered to correspond to this ideal field for the purpose of using existing theory over most of the frequency range of the measurements. Maidanik [5] deals specifically with the effect of adding stiffeners to an acoustically excited plate. He shows that under the assumption of equipartion of energy between the modes of the structure the relationship between the spectral density of average acceleration of the panel, S,,(w), and the spectral density of the sound field, S,(w), is given by

SOME EXPERIMENTS WITH STIFFENED PLATES

435

where r(w) =: [27r2n,(w)/MP](c/p) and P(W) = qrad/(rlrad+ 71mech) %&I r. Mt, is the mass of = the panel, c i.c,the speed of acoustic waves, p is the density of the air, 7],,d is the loss factor of the panel due to acoustic radiation and q,,,h is the loss factor due to mechanical dissipation and radiation into surrounding structures. In the present cases 7)mec~~%ad and p(w) is given approximately by ~rad/~mech.The modal density of a uniform panel, n,(w), is given by n,(w) = 2/3A/2nhC, where A is the area of the panel and h is the thickness. This relationship had previously been derived for homogeneous panel vibration [2]. It may be used unchanged for the stiffened panels under the assumptions that the modal density of a stiffened panel is equal to the sum of the modal densities of the stiffener, nb(o), and of the panel, n,(w), and also that the ratio of average mean square stiffener vibration velocity to plate velocity is given by < G>K v;> = MP %(U)lMb n,(w). The latter relationship has been derived by Heck1 [6] under the assumption that the typical length, L, of the subpanels formed by the stiffener is related to the flexural wavelengths in the plate and stiffener beams by A,,> L > X,. This relationship holds in the present case for the frequency range 500 Hz cf -c8300 Hz. Lyon and Eichler [3] also obtain this relationship provided that the damping of the stiffener itself is low. The ratio ~,n,(w)/~,,n,(w) is approximately @035 for the tested structures. n*(u) is given by rib(u))= L/2rr(w2 EI/p, A)- /4 where L is the length, I is the second moment of area, ps is the density and A is the crosssectional area of the beam. Maidanik [5] found that the attachment of beams to a plate increased its radiation resistance, R,,, (=Q~,,uJ,MJ, so that the response of the heavier stiffened panel to a diffuse field exceeded that of the unstiffened panel. The explanation of this increase lay in the decoupling of the sub-panels by the stiffeners which destroyed acoustic cancellation by adjacent internode regions of opposite phase at frequencies below the panel critical frequency fc. critical frequency for a uniform panel is given by fc = c2/1+8hC,. In the present case, The fc N 8000 Hz for air at 15C. The ratio of stiffened to unstiffened radiation resistance is given by 1 + PJP, where P, is twice the total length of the stiffeners and P, is the parameter of the panel. In the present case this ratio has the value 5.5. The maximum ratio of stiffened to unstiffened qrad, which was observed with the glued and welded structure at 1000 Hz, agrees well with this figure. Below and above this frequency the ratio falls off, indeed to less than unity, below 630 HZ and above 1600 HZ. The low frequency reduction in radiation resistance exhibited by this type of structure was also observed by Maidanik. He attributed it to the fact that at low frequencies (in the present case below 500 Hz) the panel wavelength exceeds the subpanel dimension L and the stiffener then adds fairly uniform mass and stiffness but does not greatly decouple adjacent sub-panels. The radiation loss factor of the glued and welded panel does indeed fall off rapidly below about 500 Hz. The stiffener configurations other than the glued and bolted structure cannot be considered to act in the same manner because flexural waves would pass, largely unscattered, across the point bolted connections and also it could be observed that panel rotation in a plane normal to the beams was not greatly constrained by the connections of the rivetted structure. Hence it would be expected that these, and the unattached panels would not exhibit the increased radiation resistance discussed above. On this basis it would be expected that the radiation loss factor of these configurations would be similar to that of the plain panel. Figure 3 shows that this is not at all the case and that the average r.m.s. stresses in the attached panels are approximately an order less than those of the glued structure over a wide frequency range. It should be noted here that the difference between the maximum stresses may not be so

436

F. J.

FAHY AND R. B. S. WEE

great because the dynamic stress concentrations at a point attachment could well exceed those at a line attachment [8]. However, it is not likely that this factor would cancel out the great benefit apparent from average strain results. It must again be stressed that the present results have been obtained from measurements on one basic beam-plate system and that no more general conclusion can be drawn than that it is worth considering alternatives to continuously welded stiffeners when dealing with high frequency acoustic excitation. In the absence of theoretical explanations for the present results it would be necessary to investigate the behaviour of any different basic configuration experimentally, especially since mechanical loss factors can vary so much from structure to structure. Of the possible causes of the large discrepancy between theory and experiment it is considered that a deviation from the theoretical ratios of stiffener to plate energy is the most likely. However, modification of the plate wave field, with consequent acoustic decoupling, or severe effects on energy distribution of mechanical dissipation in the platestiffener coupling cannot be ruled out.
5. CONCLUSIONS

Measurements of the response to acoustic excitation of a small number of steel plates stiffened in different ways have indicated that whereas continuous (e.g. fillet-welded) attachment of stiffeners may aggravate plate vibration stresses, local point attachments such as rivets or bolts can greatly reduce these stresses. Different basic plate beam systems will require specific investigation but the present results may offer guidance in this respect.
REFERENCES 1. R. B. S. WEE1967 Report submittedfor the degree of B.Sc. with honours in engineering. Department of Mechanical Engineering, University of Southampton. Acoustically induced vibration of stiffened

panels.
2. R. H. LYONand G. MAIDANIK 1962 J. acoust. Sot. Am. 34, 623. Power flow between coupled

oscillators.
3. E. EICHLER 1965 J. acoust. Sot. Am. 37,995. Thermal circuit approach to vibrations in coupled systems and the noise reduction of a rectangular box. 4. R. H. LYONand E. EICHLER 1964 J. acoust. Sot. Am. 36, 1344. Random vibration of connected structures. 5. G. M~UDANIK 1962 J. acoust. Sot. Am. 34,809. Response of ribbed panels to reverberant acoustic

fields.
6. M. HECKL1961 J. acoust. Sot. Am. 33,640. Wave propagation on beam-plate systems. 7. E. E. UNGARand J. R. CARBONELL 1966 AZAA J. 4, 1385. On panel vibration damping due to structural joints. 8. E. E. UNGAR 1961 J. acoust. Sot. Am. 33, 633. Transmission of plate flexural waves through reinforcing beams : Dynamic strain concentrations.

You might also like