JOURNAL OF THEORETICAL AND APPLIED MECHANICS 43, 4, pp. 763-775, Warsaw 2005 ANALYSIS OF THE INFLUENCE OF ELASTICITY CONSTANTS AND MATERIAL DENSITY ON THE BASE FREQUENCY OF AXI-SYMMETRICAL VIBRATIONS WITH VARIABLE THICKNESS PLATES Maciej Domoradzki Jerzy Jaroszewicz Mechanical Faculty of Suvalki, Technical University of Bialystok e-mail: jerzyj@pb.bialystok.pl Longin Zoryj Mechanical Engineering Faculty Technical University of Lviv, Ukraine In the paper, the influence of Young’s modulus and Poisson ratio as well as mass density of a material on the base frequency of circular plates of the diaphragm type with variable thickness is discussed. In order to solve the boundary-value problem, the Cauchy functionmethod and double Bernstein- Kieropian estimators were applied. An analytical form of Cauchy’s influence function was found and used to construct the characteristic equation in the form of a power series with respect to a frequency parameter. Application of this method allowed a functional dependency of the base frequency on material constants of the plates to be established. The results of calculations forplatesmadeofduraluminandtinwerementionedasexamples.Comparison of the obtained results with those found in scientific literature indicated high accuracy of the method applied therein. Key words: circular plates, variable thickness, boundary-value problem, Cauchy function method 1. Introduction Boundary-value problems of axi-symmetrical vibrations of circular plates fixed around their circumference can be solved exactly for some values of the variable thickness coefficient m and Poissons ratio when a solution to equ- ation (2.2) can be presented with the help of Bessel’s function (cf. Kantham, 1958;Hondkiewič, 1964; Leissa, 1969;Conway, 1958). Thisproblemcanalso be 764 M.Domaradzki et al. solved bymeans of characteristic series and partial discretisation utilizing do- uble sided estimators aswell as Bernsteina-Kieropiana’s tables (Bernštein and Kieropian, 1960). It was exhibited in studies concerning long cylinders and beams (Zoryj and Jaroszewicz, 2000a; Jaroszewicz and Zoryj, 2000) as well as discs (Zoryj and Jaroszewicz, 2000b, 2002; Jaroszewicz, 2000). The the- orem of vibrations of continuous-discrete, linearly elastic systems (based on Cauchy’s influence functions and the characteristic series method), which has been developed in this work, is useful for constructing and studying universal frequency equations (Zoryj, 1982). In order to study such systems, it is neces- sary to solve boundary problems, as a result of which, we obtain appropriate characteristic determinants of multi-parametric problems. By equating a de- terminant to zero, it is possible to calculate its roots bynumerical or analytical means through the use of double sided estimators. For constructing characte- ristic determinants, a form of a general solution of the entry-level differential equation is required,which is rarely known, bymeans of special functions, like for instance, the Bessel functions. That is why approximatemethods aremost commonly used to determine the basic frequencies. Among them one men- tiones: Bubnov-Galerkin’s, Rayleygh-Ritz’s, consecutive approximations and differences or finite element method (Vibracii v tiehnike, 1978). The usage of approximate methods does not, however, address the question of accuracy of solutions. That problem may be resolved either having the exact solution, which is possible only in fewparticular cases (bymeans of Euler’s equation for example) or by supplying double sided estimators, whose difference in value depends on the number of used terms of the characteristic series. The greatest advantage of themethoddeveloped in thiswork is the general form of the power series of the characteristic equation, which gives functional dependency of proper frequencies onmaterials constants of considered plates, where Berstein’s double sided estimators can be easily applied to. 2. Problem formulation Weconsider a clampedcircular plate,with the radius Rflexural rigidity D and thickness described by power functions of the radial coordinate r D=D0 ( r R )m h=h0 ( r R ) m 3 0 0 - to plates of the diaphragm type; m< 0 – disc type plates (Hondkiewič, 1964). We shall analyze the dependence of eigen frequencies on the constants ρ, E and ν. In problem (2.1)-(2.2), a limitation of solutions and their first derivatives with respect to the independent variable r is required (Conway, 1958). 3. Constant thickness plate (m=0) The basic frequency of such a plate can be calculated on the basis of awell known formula.For example, fromanequation found in theworkbyVasylenko (1992) ω1 = γ h0 R2 √ E 12ρ(1−ν2) (3.1) where γ=10.214. Now, wewill compare frequencies of two plates of the same thickness h0 and radius R butmade of different materials (ω1)I (ω1)II = √ EI EII ρII ρI (1−ν2 II ) (1−νI)2 (3.2) 766 M.Domaradzki et al. where I and II correspond to the first and second material. If, for sake of the study, we assume the first one (I) to be duralumin and the other one (II) to be tin, we will obtain (E ρ ) I =2.65 (E ρ )−1 II =7.09 1−ν2II 1−ν2 I ≈ 0.81 0.88 ≈ 0.92 and the corresponding ratio of frequencies (3.2) (ω1)I (ω1)II ≈ √ 2.65 ·7.09 ·0.92=4.16 (3.3) νI ≈ 0.34 νII ≈ 0.44 Remarks • Let the frequency equation be presented in the following form (as a characteristic series) (Zoryj, 2000a) a0−a1pR2+a2(pR2)2− . . .=0 p= ρ h0 D0 ω2 In that case, double sided Bernstein-Keropian’s estimators for the coeffi- cient γ in formula (3.1) should by calculated as follows √ √ √ √ a0 √ a21−2a0a2 <γ< √ √ √ √ 2a0 a1+ √ a21−4a0a2 where a0 =1, a1 =1/96, a2 =1/122880 (cf. Jaroszewicz, 2004), hence 10.204<γ < 10.224 • The simplest lower estimator (γ−) hasbeencalculated fromthe equation a0 − a1pR2 = 0 (cf. Jaroszewicz, 1997), therefore γ− = √ 96, which is lower than the exact value (γ =10.214) by 4%. As we can see, the properties of material (ρ,E,ν) significantly influence the basic frequency of the constant thickness plate. Analysis of the influence of elasticity constants... 767 4. Plate with parabolically variable rigidity Let m=2 (Hondkiewič, 1964; Conway, 1958). In this case, on the basis of (2.4), we obtain L0[u]≡uIV + 6 r uIII + 5+2ν r2 uII + 2ν−1 r3 uI (4.1) Substituting u= rµ in (4.1), we pass to the corresponding algebraic equation µ4+(−2+2ν)µ2 =0 (4.2) with the following roots µ1 = √ 2−2ν µ2 =−µ1 µ3 =µ4 =0 (4.3) Thus we obtain linearly-independent solutions to equation (4.1) for arbitrary ν[ν(00.5)], which are functions (see example) described by Zoryj and Jarosze- wicz (2002) u1 = r µ1 u2 = r −µ1 u3 =1 u4 = lnr (4.4) TheCauchy function corresponding to this example will be determined by the formula K0(r,α) =− µ1 8(1−ν2) (r−µα3+µ−rµα3−µ)− α3 2(1−ν) ln r α (4.5) In the formula, the subscript ”1” at can be omitted, and that is why it will not be used afterwords. In the above case, equation (2.2) will take the following form L0[u]−pr− 4 3u=0 (4.6) That is why the limited solution to equation (4.6) with m=2 can be deter- mined by the S1 and S2 series, which are constructed on the basis of formulas Sj =Sj0+pSj1+p 2Sj2+ . . . p= ρh0 D0 R 4 3ω2 (4.7) and SjK = r ∫ 0 K0(r,τ)τ − 4 3Sj,K−1(τ) dτ K =1,2, . . . j =1,2 (4.8) S10 =1 S20 = τ µ1 where the function K0(r,α) is described by formula (4.5). 768 M.Domaradzki et al. Having determined the first two integrals (4.8) we obtain S11(r)= ar 8 3 a= 34 27(23+9ν) (4.9) and S21(r)= br µ+ 8 3 b= 1 4(1−ν) ( 3 4 ( 2µ+ 8 3 ) − 2 ( µ+ 8 3 )2 ) (4.10) µ= √ 2−2ν So, we can see S1(r)= 1+pS11(r)+ . . . S2(r)= r µ+pS21(r)+ . . . (4.11) Having found the dependencies in forms (4.9)-(4.11), one can determine the simplest lower estimator for the basic frequency of vibrations. In order to accomplish that, we need to determine S′1 = 8 3 par 5 3 S′2 =µr µ−1pb ( µ+ 8 3 ) rµ+ 5 3 ∣ ∣ ∣ ∣ ∣ S1 S2 S′1 S ′ 2 ∣ ∣ ∣ ∣ ∣ =µrµ−1 [ 1+pb ( µ+ 8 3 )1 µ r 8 3 +par 8 3 − 8 3 a µ r 8 3 + . . . ] =0 Hence, considering that r=R, we obtain the first two elements of the charac- teristic series (the frequency equation) of the problem defined by expressions (2.2) and (2.3) 1+pR 8 3 ( b µ+ 8 3 µ +a− 8 3 a µ ) + . . .=0 (4.12) From (4.12), for ν =1/9, = 4/3, we obtain 1+pR 8 3(3b−a)+ . . .=0 (4.13) Considering (4.6) as well as the fact that 3b−a= 33 211 − 33 210 =− 33 211 we under estimate the basic frequency for this case ω− = γ−(ν) ∣ ∣ ∣ ν= 1 9 h0 R2 √ E 12ρ(1−ν2) (4.14) Analysis of the influence of elasticity constants... 769 where γ−(1/9) =8.71 is about 8% smaller than the corresponding exact value γ(1/9)=9.46 (cf. Conway, 1958). For ν =0 (µ= √ 2), on the basis of (4.9)-(4.12), we calculate 1− 8 3 √ 2 =−0.8856181 a ( 1− 8 3 √ 2 ) =−0.0233513 ( 1− 8 3 √ 2 ) b=0.0118247 γ−(0)= 9.3 In the same fashion, for ν = 1/2 (µ = 1) we obtain γ−(0.5) = 7.8, which amounts to about 84% of the value of 9.3. As we can see, for plates characterized by the parabolic variable rigidity of m=2, formula (4.14) differs from (3.1) not only by its multiplier γ−(ν) 7.8<γ−(ν)< 9.3 (4.15) Considering that νI = 0.34, νII = 0.44 we calculate (γI)− ≈ 9, (γII)− ≈ 8 which agrees with (4.15). Hence, the corresponding ratios γ−(νI) γ−(νII) = 9 8 ≈ 1.125 ω−(νI) ω−(νII) = 1.125 ·4.16=4.66 (4.16) Summing it all up, in the case of m=2 for the abovementionedmaterials (I-duralumin, II-tin) the ratio of basic frequencies ωI/ωII increased by 12% as compared to (3.3) for m=0, where the multiplier γ did not depend on ν. Let us notice that formula (4.10) can also be written down as follows S21(r)= 1 4(1−ν) 81x(x+1)2 4(14x+6)(11x+3)2 (4.17) where x∈ (1+ √ 2,∞). In that case µ= x+1 x ν = 1 2x2 (x2−2x−1) (4.18) for ν ∈ [0,1/2], µ∈ [ √ 2,1]. Set forth dependencies (4.17) are extremely useful in calculations. Inparti- cular, it is easy to see that in inequalities (4.15), γ−(ν) changes in amonotone fashion with the Poisson ratio ν. Remarks. The coefficients of the corresponding characteristic series (see re- marks fromSection 3 for case m=0) are as follows (Jaroszewicz, 2004) m=2, ν =1/9 a0 =1 a1 =3 3 ·2−11 a2 =35 ·5−1 ·2−21 770 M.Domaradzki et al. Hence, the double sided Bernstein-Keropian estimators give 9.412<γ (1 9 ) < 9.495 which is in agreement with the simplest lower estimator, see (4.14), (4.15). 5. Disc type plate Let m = −1 (Zoryj and Jaroszewicz, 2002). On the basis of (2.4), we obtain L0[u]≡uIV − 1+ν r2 uII + 2(1+ν) r uI (5.1) and equation (2.2) takes the following form L0[u]−pr− 2 3u=0 p= ρh0 D0 R− 2 3ω2 (5.2) The linearly-independent solutions to equation L0[u] = 0 are 1, r3, rµ1, rµ2 (5.3) where µ1 = 1 2 (3+ √ 5+4ν) µ2 = 1 2 (3− √ 5+4ν) (5.4) Let us notice that all solutions (5.3) are limited for r = 0. That is why it is enough to consider the first two solutions together with their derivatives. The necessary solutions, which correspond to them, are constructed with the aid of formulas (4.7), (4.8), in which K0(r,α) = 1 2 √ 5+4ν (rµ2αµ1 −rµ1αµ2)+ 1 6 (r3−α3) (5.5) and SjK = r ∫ 0 K0(r,τ)τ 2 3Sj,K−1(τ) dτ K =1,2, . . . j =1,2 (5.6) S10 =1 S20 = r 3 Having determined the first two integrals (5.6), we obtain S1(r)= 1+pS11(r)+ . . . S2(r)= r 3+pS21(r)+ . . . (5.7) Analysis of the influence of elasticity constants... 771 where S11(r)= a(ν)r 14 3 a(ν)= 81(1−ν) 140(79−9ν) S21(r)= c(ν)r 23 3 c(ν)= 81(1−ν) 23 ·28(331−9ν) (5.8) Continuing in the same manner as in Section 3 (constant thickness pla- te (m = 0)), we come to the first two elements of characteristic series (the frequency equation) of the problem given by expressions (2.2) and (2.3) for m=−1 3R2+ (5 3 a(ν)− 23 3 c(ν) ) pR 20 3 + . . .=0 After simplification bymeans of R2 reduction, and having considered formula (5.2, we obtain 3+ 1 3 [5a(ν)−23c(ν)] ρh0 D0 R4ω2 =0 (5.9) hence, making use of formulas (5.8), we calculate the coefficient −a1(ν)= 1 3 [5a(ν)−23c(ν)] (5.10) for values ν = 0, νI = 0.34 ≈ 3/9, νII = 0.44 ≈ 4/9 and ν = 0.5 we have respectively 0.0092928, 0.0064991, 0.0055045, 0.0049951 From (5.9), we come to formulas (4.12) and (4.13), then to estimators 17.97<γ−(ν)< 24.5 (5.11) and then to relations γI γII = 21.48 23.35 =0.92 (5.12) So, in the case of m=−1 instead of (4.16), we get ω−(νI) ω−(νII) = 0.92 ·4.16=3.83 (5.13) From (5.13), it follows that the ratio between basic frequencies decreased by 8% as compared to the result obtained from (3.3) for fixed thickness (m=0). It is worth noticing that for m = −1 the exact values of basic frequencies 772 M.Domaradzki et al. are known (Hondkiewič, 1964). Having them applied, instead of (5.12), we obtain a2 ∣ ∣ ∣ ν= 3 9 =0.0000009 a2 ∣ ∣ ∣ ν= 4 9 =0.0000007 a1 ∣ ∣ ∣ ν= 3 9 =0.0064991 a1 ∣ ∣ ∣ ν= 4 9 =0.0055045 (a1(ν) is determined by formulas (5.8), (5.10)). Therefore γI γII = 22.25 24.25 =0.92 (5.14) 6. Conclusions • As far as diaphragm type plates (m> 0) are considered, materials with large values of ν exhibit lower basic frequencies as compared to fixed thickness plates. Likewise, disc type plates (m < 0) have higher basic frequencies. The decrease in the diaphragm type plate of m = 2 with respect to the fixed thickness plate (m=0), is from 9% for ν =0 up to 24% for ν = 0.5. For disc plates of m = −1 we observe an increase in the frequency from 83% for ν =0 and 247% for ν =0.5. • The simplest lower estimators calculated with the first two elements of the series taken into account, allow us to credibly observe the effect of E, ν and ρ on the frequencies of axi-symmetrical vibrations of circular plates, whose thickness or rigidity change along the radius according to ae power function. • In the case of fixed thickness plates (m = 0), the coefficient of basic frequencies γ does not depend on Poisson’s ratio ν. Thus γI/γII =1 for materials having large values of ν ∼ 0.5, as because of the conditions (E/ρ)I > 1, (ρ/E)II < 1 they correspond to smaller frequencies. Selected calculation results for twodifferentmaterials νI ≈ 0.34, νII ≈ 0.44, (E/ρ)I = 2.65, (ρ/E)II = 7.09 and three types of plates are presented as follows: Analysis of the influence of elasticity constants... 773 —Bernstein-Keropian double sided estimators 9.412<γ(1/9)< 9.495 for m=2 10.204<γ < 10.224 for m=0 22.232<γ(4/9)< 22.262 for m=−1 24.233<γ(1/3)< 24.272 for m=−1 — basic frequency parameter ratio γ−(νI) γ−(νII) =        1.125 for m=2 1 for m=0 0.92 for m=−1 — basic frequency ratio (ωI)I (ωI)II =        4.66 for m=2 4.16 for m=0 3.83 for m=−1 — simple under estimate estimators of the basic frequency parameter for Poisson’s ratio 0¬ ν ¬ 0.5 7.8<γ−(ν)< 9.3 for m=2 γ−(ν)= √ 96 for m=0 17.97<γ−(ν)< 24.5 for m=−1 Acknowledgement This workwas financially supported by scientific grant of Technical University of Bialystok No.W/ZWM/1/05. References 1. Bernštein S.A.,KieropianK.K., 1960,Opredelenie chastot kolebanǐı sterzh- nevykh system metodom spektralnoi funkcii, Gosstroiizdat, Moskva 774 M.Domaradzki et al. 2. Conway H.D., 1958, Some special solutions for the flexural vibrations of discs of varying thickness, Ing. Arch., B.26, 6 3. Hondkiewič W.S., 1964, Sobstvennye kolebaniya plastin i obolochek, Kiev, NukowaDumka, p.288 4. KanthamC.L., 1958,Bending and vibrations of elastically restrained circular plates, J. Frank. Inst., 6, 265, p.483 5. LeissaA.W., 1969,Vibration of Plates, NASASP160,WashingtonD.C.,U.S. Governement Printing Office 6. Jaroszewicz J., 2000, Drgania swobodne utwierdzonej płyty kołowej obłożo- nej masami,Prace Naukowe Instytutu Techn. Wojsk Lotniczych, 9, 37-44 7. Jaroszewicz J., Zoryj L.M., 1997, Metody analizy drgań i stateczności kontynualno-dyskretnych układów mechanicznych, Monografia, Białystok 8. Jaroszewicz J., Zoryj L.M., 2000, Investigation of the effect of axial loads on the transverse vibrations of a vertical cantilever with variables parameters, International Applied Mechanics, 36, 9, 1242-1251 9. Jaroszewicz J., Zoryj L.M., Katunin A., 2004, Dwustronne estymato- ry częstości własnych drgań osiowosymetrycznych płyt kołowych o zmiennej grubości, Materiały III Konferencji Naukowo-Praktycznej ”Energia w Nauce i Technice”, Suwałki, 45-56 10. Vasylenko N.V., 1992,Teoriya kolebanǐı, Kiev, Vyshcha Shkola, p.429 11. Vibracii v tekhnike, 1978,Spravochnik,6,Moskva,Mashinostroenie, 1978-1981, Vol 1:Kolebania linijnykh sistem, p.352 12. Zoryj L.M., 1982, Ob universalnykh kharakteristicheskikh uravneniyakh w zadachakh kolebanǐı i ustoichivosti uprugikh sistem,Mekhanika Tverdogo Tela, 6, 155-162 13. Zoryj L.M., Jaroszewicz J., 2000a, An axially symmetric vibrations pro- blemsof longcylinders,Mechanical Engineering, Lviv,4, 5, 33-35(inUkrainian) 14. Zoryj L.M., Jaroszewicz J., 2000b, Influence of concentrated mass on vi- brations of the circular plate,Mechanical Engineering, Lviv, 9, 17-18 (in Ukra- inian) 15. Zoryj L., Jaroszewicz J., 2002, Main frequencies of axial symmetric vi- brations of the thin plates with variable parameters distribution, Mechanical Engineering, Lviv, 5, 37-38 (in Ukrainian) Analysis of the influence of elasticity constants... 775 Analiza wpływu stałych sprężystych i gęstości materiału na częstość podstawową drgań osiowosymetrycznych płyt kołowych o zmiennej grubości Streszczenie W pracy zbadano wpływ modułu sprężystości Younga i liczby Poissona, a także gęstości materiału na częstość podstawową płyt o zmiennej grubości typu diafragmy i dysku.Do rozwiązania zagadnienia brzegowego zastosowanometodę funkcji wpływu Cauchy, najprostszy estimator z niedostatkiem i dwustronne estymatory Bernstejna- Kieropiana. Znaleziono analityczną postać funkcji wpływu Cauchy, z pomocą któ- rej zbudowano równanie charakterystyczne w postaci szeregu potęgowego względem parametru częstotliwości. Zastosowaniemetody pozwoliło wyprowadzić funkcjonalną zależnośćczęstościpodstawowejod stałychmateriałowychwymienionychpłyt.Wcha- rakterzeprzykładuprzytoczonowyniki obliczeńdla płytwykonanychzduraluminium i z cyny. Porównanie wyników obliczeń ze znanymi z literatury potwierdziły wysoką dokładnośćmetody. Manuscript received May 5, 2005; accepted for print June 29, 2005