Download as pdf
Download as pdf
You are on page 1of 17
& INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING, VOL. 38, 2265-2281 (1995) AN EXPLICIT TIME INTEGRATION TECHNIQUE FOR DYNAMIC ANALYSES S. PEZESHK* AND C. V. CAMP* Department of Civil Engineering, The University of Memphis, Memphis, TN 38152, U.S.A. SUMMARY A simple explicit solution technique for problems in structural dynamics, based on a Modified Trapezoidal rule Method (MTM) approximation of the governing ordinary differential equations, is developed. The resulting conditionally stable explicit method (MTM) can be easily implemented and is extremely simple to use. Particular attention is focused herein on the concept of numerical stability of the proposed method for a free-vibrational response of a linear undamped Single-Degree-O-Freedom system (SDOF). To examine the effectiveness, strengths, and limitations of MTM, error analyses for the natural period, the displacement, the velocity and the associated phase angle for a free undamped simple mass-spring system are derived and ‘compared with Modified Euler Method (MEM) and the well-known Newmark Beta Method (NBM). ‘Numerical examples for a SDOF system and a Multi-Degree-Of-Freedom (MDOF) system are presented to illustrate the strengths and the limitations of the proposed method. KEY WORDS: explicit and implicit time integration; stability; trapezoidal rule; structural dynamics INTRODUCTION ‘Many variable-time integration methods are currently employed in transient analysis for both uncoupled and coupled situations. Each of these methods have different levels of accuracy, stability and computational cost. Each of these numerical approaches employs difference equiva- lents to develop recurrence relations which may be used in step-by-step computation to obtain the dynamic response of a structure. In general, the critical parameter in each of these techniques is the largest value of the time step that may be used to provide sufficiently accurate results, as this is directly related to the total computational cost of the analysis. In this paper, a modified version of the trapezoidal rule method in conjunction with a second-order Taylor series approximation is used to solve the equations of motion. The Modified Trapezoidal Rule method (MTM) has several features that make it an attractive approach in dynamic analyses of structural systems. This paper is motivated by recent research by Hahn' and Cromer? in which they discuss a modified version of the Euler method for dynamic structural analysis. Hahn's Modified Euler Method (MEM) is a simple and effective numerical procedure which may be used to solve a variety of Single-Degree-Of-Freedom (SDOF) and Multi-Degree-Of-Freedom (MDOF) prob- lems. In some aspects, MTM is similar to MEM, however, MTM is generally more accurate. The objective of this paper is to present a systematic and fundamental procedure to examine the accuracy and stability of the proposed MTM. In what follows, a formal introduction to MTM ‘Associate Professor CCC 0029-5981/95/132265-17 Received 10 July 1993 © 1995 by John Wiley & Sons, Ltd. Revised 26 August 1994 2266 ‘8. PEZESHK AND C. V. CAMP is given including error analyses. The characteristics of the MTM procedure are investigated in several examples. In addition, the accuracy, the characteristics and the merits of MTM relative to those of Habn’s method! and the well-known Newmark family of beta methods* (NBM) are discussed. MODIFIED TRAPEZOIDAL RULE METHOD (MTM) ‘The free and undamped vibration of a simple mass-spring system may be described by the following linear homogeneous second-order differential equation: Kto%x=0 aw where x is the displacement of the vibrating mass from the equilibrium position, w is the circular frequency, and each dot represents differentiation with respect to time, ‘A discrete solution of the governing equation of motion, equation (1), may be approximated by defining the displacement x,, the velocity %,, and the acceleration %, at some time denoted by the subscript n. The time may be expressed asf, = nAt where Ar is an arbitrary time step. In terms of the defined discrete time variables, equation (1) may be written as &,= 0x, Q) The displacement of the system x,, incorporating a trapezoidal rule for the integration of the velocity, is approximated as wet Leia t See) ® Xt ‘The velocity 2, is represented by a second-order Taylor series feer=(1-}) se st @ where y = w?At?, ‘The modified trapezoidal rule method (MTM) consists of solving equations (2)-(4) at each time step n. The computation is initiated by defining the initial conditions of the system given in terms of the displacement x, and the velocity %,, at t=0, or n=0. With the initial conditions prescribed, equation (2) may be solved for the initial acceleration, %,. The values of the velocity, the displacement and the acceleration at time n + 1 are determined by solving equations (3), (4) and (2) in sequence. The resulting MTM procedure is quite simple and fully explicit. Stability of the MTM ‘The expression for displacement, given in equation (3), may be decremented in time to obtain an expression for x). At + Gann + He) 6) The resulting relationship, defined in equation (5), is subtracted from equation (3) to form a second-order difference equation At int) + ze int Xn Xe Xan = x 6 EXPLICIT TIME INTEGRATION TECHNIQUE 2267 The differences (z, — %,-1) and (,+1 —%,) may-be determined by rewriting equation (4). For example Saar — iyo — 1 + yd 0 Substituting equation (7) into equation (6) gives Kat = yAt TE Giant +) ® The sums (%,- + %,) and (%,—1 + ¥,) in equation (8) may be determined by rewriting equations (3) and (2), respectively. . 2 Bent +e = On nu ) (9) Sant + y= — W2Oya +X) (10) ‘Substituting equations (9) and (10) into equation (8) results in a linear homogeneous difference equation Kata (2+ Wet Xy-1 = 0 (it) Itis interesting to note that the governing difference equation, equation (11), is exactly the same as that of MEM. For convenience, equation (11) may rewritten as Xne2 + (= 24 Vxaer + Xa = 0 (2) A characteristic equation is obtained by assuming a general solution of the form xed 3) and substituting the expression into equation (12) B+(—24+yYA+1=0 (14) The roots of equation (14) may be determined by the quadratic equation A2=—H-24+ D447 - 4y (is) ‘The MTM procedure is numerically stable when the spectral radius, given in equation (15), is less than 1 (ie. [2y,2| <1). The absolute value of 4, is less than 1 when y < 4. In addition, to insure that the roots do not bifurcate, they must remain complex-conjugate. Therefore, the quantity inside the square root should be less than zero p-4y<0 (16) The only real value for y that satisfies equation (16) is yooh? <40 a) Therefore, for a system in which the natural period of vibration is T = 2n/w we can rewrite equation (17) as T ar<= (18) 2268 . PEZESHK AND CV. CAMP Itis also interesting to note that the stability requirement achieved in equation (18) is the same as that of MEM. The roots of the characteristic equation, given in equation (15), in polar notation may be written as Ase, Ase" (19) where i= /—T and p is defined as we tan! (4) 20) 2-7 By substituting the roots of the characteristic equation, defined in equation (19), into equation (13), the general solution for the displacement becomes X_= AM + BY ='Ae™ + Be“ ™ aN) or, in a more convenient form, the real part of the displacement may be written as 2%, = Coos(np) + D sin(ny) 22) where the constants, A, B, C and D are determined from the initial conditions of the system. Displacement and velocity using MTM To investigate the accuracy of MTM, the response of the simple mass-spring system is“ evaluated. Expressions for the displacement and the velocity of the system may be derived from the general solution given in equation (22). The time variable n in the expression for the displacement x,, given in equation (22), may be replaced by n = t,/At n= Coos(n jt) + Dain j2) 3) The velocity %, may be derived in terms of the displacement x, by rewriting equation (4) for the velocity %,+1 and substituting the result into equation (9) written in terms of the velocity x, . +1 — (4 —2y) xn ue em a Equation (24) may be written in a more convenient form as Sy = OXne + Xp (25) where 4 442 "Gaye =a ° ‘Substituting the solution for the displacement, given in equation (23), into the approximation of the velocity ,, equation (25), results in y= (Dlacos y+ 6) ~ aC sing sin (n+ {Clacosn-+)+ aD sing) cos(1i2) en EXPLICIT TIME INTEGRATION TECHNIQUE 2269 ‘The constants C and D in equations (23) and (27) may be determined by applying the initial conditions for both the displacement, x, (¢ = 0) = Xo and the velocity, x, (t = 0) = Vo CX, D= Ye es) In terms of the initial displacement Xo and the initial velocity Yo, the displacement x, is ty) _ [Xo(a cos +b) — Yo] tn’ x= Kocos(i 2) ee in 29) and the velocity %, is . ty _ (a? + b?)Xo + (2bXo — Vo)acos yu bVo (te in = Vo cos (« z) - wing sin( we G0) The solution given by equations (29) and (30) for the free-vibrational problem considered is periodic. Because these functions are periodic, the errors associated with the application of the proposed method’ do not grow as the number of time steps increases. The assessment of the magnitude of error for different types of initial conditions are provided later. COMPARISON OF METHODS ‘Available direct methods can be subdivided into explicit and implicit methods each with distinct advantages and disadvantages. The MEM and NBM with f = 0 (well-known central difference method), and the proposed MTM are explicit methods, while the NBM with B # 0 is an implicit method. In MTM, MEM and NBM (f = 0) the solution at time Ar + 1 is obtained by considering the equilibrium conditions at time t. Such schemes do not require factorization of the stiffness matrix in the step-by-step solution. Hence the method requires no storage of matrices if a diagonal mass matrix is used. In addition, the computational cost per step is much less than implicit methods. However, explicit methods such as MTM and MEM are conditionally stable while many implicit methods are unconditionally stable. The MTM procedure is conditionally stable and requires a time step inversely proportional to the period of the system. Explicit methods compute the displacements at the current time step in terms of displacements, velocities and accelerations at the previous time step. In contrast, an implicit procedure computes the displacements at the current time step in terms’ of current accelerations, The resulting formulation requires an iterative procedure to determine the response for non-linear systems. The iterative nature associated with implicit methods may be avoided by using an incremental approach such as that implemented in Clough and Penzien,* Subbaraj and Dokainish,’ and ‘Hughes.® However, these incremental techniques are based on the use of the concept of the tangent stiffness and may introduce additional errors in the computation of the dynamic response of a non-linear system. An additional advantage of explicit methods, such as MTM, is that they can be easily implemented and used in dynamic analysis of different systems. In the following sections, error analyses are performed for MTM, MEM and NBM by considering the free-vibrational response of a linear and undamped SDOF system. The accuracy and stability of MTM is investigated and compared with MEM and NBM. A comprehensive survey of direct time-integration procedures, including both explicit and implicit methods is presented in a set of companion papers by Dokainish and Subbaraj’ and Subbaraj and Dokainish.* Interested readers are referred to these papers for more information. 2270 S. PEZESHK AND C. V. CAMP ERROR ANALYSIS In order to measure the effectiveness of the MTM solution procedure, expressions for the error in the displacement, the natural period, and the phase angle, as a function of the time step, are developed. The exact solution for the free and undamped free-vibrational motion of ‘a mass-spring system, defined in equation (1), is en 2 Error in the natural period A measure of the error introduced by the MTM approximation is established by comparing the natural period of vibration for the displacement, equation (29), or the velocity with exact solution, equation (30). In this case, the error is calculated as the ratio of the approximate period to the exact period T £ 3) In Figure 1, the ratio 7/Tezs is calculated using the MTM procedure and compared to the both MEM and NBM for various ratios of At/T (see Table I for equations). In all cases, the error in the natural period is negligible for At/T < 0-01. However, in the range 0-01 < At/T'< 0-1 MTM introduces less error than NBM (with f = 3) and performs practically the same as both MEM and NBM (with B = 0). Although the implicit NBM (with f = 4) is unconditionally stable, the results shown in Figure 1 indicate that a small time step is needed to improve accuracy and to avoid period elongation. An estimate for the error introduced in the displacement by the MTM approximation is calculated for two cases, In the first case, the displacement developed using MTM is compared to the exact solution for an initial velocity to be zero, Yo = 0. In the second case, an error estimate for the displacement is developed for an initial displacement to be zero, Xo = 0. Error in the displacement (Vo = 0) For the case where Vo = 0, the MTM displacement given in equation (29) reduces to ty Xo db. bn x= Xocos (ijt) — NCEE sin (ni) 64 or gael 7 can os ( fe ana at °) 5) EXPLICIT TIME INTEGRATION TECHNIQUE 21 1.03] Newmark Beta Method, 6 = } 102 101 ib Modified Trapezoidal Method 0994 Modified Euler Method Newmark Beta Method, = 0) ose: 0.001 6610 6.100 4 Figure 1. Error in the natural period for MTM, MEM and NBM at various values of A/T where the phase angle 0 is given as = tan (eseee*} 06) asing The exact solution for displacement when Vo = 0 is x = Xocos & ‘) (G7) The error in the value of the displacement produced by MTM is measured by two criteria. First, the error in the displacement is calculated as the ratio of the amplitude obtained from MTM to the exact value /a* + 2ab cos p+ 6? (G8) In Figure 2, the ratio defined in equation (38) is computed using the MTM and MEM procedures for various ratios of At/T (see Table I for equations). Error in the amplitude is negligible for At/T < 001 for both MTM and MEM. For 001 < At/T < 0-1 the error for MTM is negligible, however, the error for MEM grows rapidly. For Vo = 0, the NBM gives the exact solution for the displacement amplitude. A second measure of the error in the displacement expression using MTM is the phase angle ‘given in equation (36) In Figure 3, the value of the phase angle is computed using both MTM and MEM for various ratios of At/T (see Table I for equations). Over the range 0-01 < At/T'< 01 the ene SiH 2272 ‘8, PEZESHK AND C. V. CAMP Table I. Summary of Network Beta and modified Euler methods ‘Newmark Beta Method? (NBM), Gamma = + Modified Euler Method" Baas = hy + OS (ALR, + Abin 1) at bide AE — 20 + AeBS in + Salt 1a Xt hee Ae Equations Batt oxes Stability criteria ato? <4 Error in natural period Error in displacement: inital velocity equal to zero x, 2\-12 —s (0 -%) Kee - T caf def ae Penge A otf eA") Error in displacement: ae\2 initial displacement 4) equal to zero 7 Error in velocity: ey initial velocity equal to zero =) error in the phase angle using MTM is negligible. However; it should be noted that the phase angle grows rapidly for MEM and is much greater than MTM. In evaluating the error in displacement for Vo = 0, the NBM (f = 4) and central difference (B= 0) forms of the NBM maintain a displacement amplitude ratio of one; however, different choices of the Newmark parameters can be given different results. Error in the displacement (Xo = 0) For the case where Xo = 0, the MTM displacement given in equation (29) reduces to 1 te noah sn(n 3) 69) The exact solution for displacement when Xo = 0 is Yo. (2 x= sin & ‘) (40) Y ‘oo “L/g jo santa snoURA Ye WAIN PUL WLW 40} 0 = °4 wowaceydsIp am YUM pareroosse a(due oseug “E SUNS » coro 0100 190 S00 7 1000 00 Poway Teprozedery, payrpoW, ro sro seo LAY jo Sonqea snouea ye WAIN PUP WLW 10) 0 = 4 AyDoj% fentaY WE YAH qUoWOEIdsIp ayy Woz -Z SIN * coro 0100 190 560 Poway Teprozedexy, poyIPOW, oot 101 zo sor Powe FIN POTPOWL ot suze 3NDINHDAL NOLLYWOaINI AWLL LOITaxa. 2274 . PEZESHK AND. V. CAMP The ratio of the amplitude obtained from MTM to the exact value is Xn 2x me Tasing “ In Figure 4, the ratio defined in equation (41) is compared to the MEM and the NBM procedures. for various ratios of At/T (see Table I for equations). The error in the amplitude is negligible for At/T < 0-01. However, in the range 0-01 < At/T < 0-1 the MTM approximation is more accurate than either MEM or NBM with B = 0. However, the NBM with f = 4 gives the exact displace- ment amplitude. Again, it should be noted that different choices of the Newmark parameters can give different results. ‘An estimate for the error introduced in the velocity expression by the MTM approximation is calculated for two cases. In the first case, the velocity developed using MTM is compared to the exact solution for an initial velocity Vp =0. In the second case, an error estimate for the displacement when Xo = 0 is calculated. Error in the velocity (Vo = 0) For the case where Vo = , the MTM displacement given in equation (30) reduces to xo asing [a? + 2abcos y+ 6?) sin (« 2) (42) Modified Euler Method ‘Newmark Beta Method, 6 = o Newmark Beta Method, 6 = } Modified Trapezoidal Method ose ‘or ano 100 4 Figure 4, Error inthe displacement with an initial displacement Xo = O for MTM, MEM and NBM at various values of ‘yT EXPLICIT TIME INTEGRATION TECHNIQUE 2215 106, 1095. 104 103 Loy 101 190 Fami o99 ose oor 096] 09s. 094] ‘001 O10 0.100 4 # (Flew 5. Erin the velciy wih an nil wlocky Y= 0 for MM, MEM and NBM at various aero oP ‘The exact value of the velocity is fee Xyorin(7 ‘) 3) ‘The ratio of the velocity calculated from MTM to the exact value is aT fa? + rab osu + 67) (44) Reaer 2a singe In Figure 5, the ratio defined in equation (44) is computed using MTM and compared to the results using the MEM and the NBM procedures for various ratios of At/T. The results indicate that the MTM procedure is more accurate than MEM or NBM with f = 0. However, the MTM approximation is not as accurate as NBM with f = 4. Error in the velocity (Xo = 0) For the case where Xo = 0, the MTM velocity defined in equation (30) reduces to = Vocos( ijt) + Mteeesn tO in( a) co) or Yo Je Tab caT eB cos (nie +) (46) asing 2216 SS PEZESHK AND CV. CAMP where the phase angle-0 is given as om tans {_seoen-e fh 4) asin ‘The exact value for the velocity when Xo = 0 is ‘Qn = = (48 = Vo oxs( T ‘) (48) The ratio of the velocity calculated using MTM to the exact value is identical to that given in ‘equation (38) and is plotted in Figure 3. The phase angle defined in equation (47) is similar to that given for the displacement and shown graphically in Figure 3, except it will have an opposite sign. APPLICATION OF MTM TO DYNAMICS ANALYSES To demonstrate the effectiveness, characteristics and merits of MTM in the analysis of structural dynamics, several application problems are presented. The first example involves a SDOF system and the second example applies MTM to a MDOF system. Dynamic response to harmonic loading The dynamic response of a simple damped mass-spring system subjected to a forced harmonic load is defined as E+ Wor + wx = 07(Xy) sin Ot (49) where x is the displacement, £ isthe critical damping ratio, w is the natural frequency, (X..) is the displacement associated with the maximum static load, and is the frequency of the applied harmonic loading. ‘The dynamic system defined in equation (49) is approximated using MTM by the following discrete equations: = 2Ews — wx + ©(Xy) sin Oe (50) Sav -( ~ Fan kat 61) Xe = + Feat ten (52) Consider the system to be initially at rest, with a natural frequency @ = 2x, a frequency of the harmonic loading © = 2n, and damping ratio = 0-1. The response of the system is approxim- ated using MTM with At/T = 0-01. In Figure 6, the ratio of the displacement x to (Xi) is plotted as a function of time. The results obtained from MTM are in excellent agreement with the exact solution.* Application of MTM for direct integration of MDOF systems The equation of motion for a linear, discrete, MDOF system excited by dynamic loads P(t) can bbe expressed as Mi + Kx 2) (83) EXPLICIT TIME INTEGRATION TECHNIQUE 2277 Figure 6. MTM solution for the response ofa free and undamped simple mass-spring system subjected to a harmonic loading with A1/T = 001 where M and K are the mass matrix and the stiffness matrix of the system, respectively, and x is the displacement vector. ‘An analysis procedure for an undamped free-vibrational MDOF system using MTM is accomplished by direct integration, similar to a SDOF system, as the following: —M™'!Kx, = — Rx, (54) nor = Rik + RAE (55) At Bee het Se thea) (56) where K = M~'K, K, = 1 — 4K (At?), and I = the identity matrix. The correlation between equations (55) and (4) can be expressed by using the modal superposi- tion approach. One can represent the approximations %,, 1, %, and i, as k= LOG D ca) i y 4) (58) it a PL 69) 218 8. PEZESHK AND CV. CAMP where ¢ is the jth natural mode; q,(n) is the associated generalized displacement co-ordinate; and nnis the total number of degrees of freedom. Substituting equations (57)-(59) into equation (55), premultiplying the resulting equation by $;M, and making use of the orthogonality properties of the natural modes (with respect to the mass and stifness matrices) lead to Gn + I= (1 — 407 AP) 4 s(n) + Ge (60) in which TK: of = og (61) is the square of the jth natural frequency. Equation (60) derived from equation (55) is of the form of equation (4) with y = 7 Ar?. Dynamic response of MDOF system When MTM is used to evaluate the response of a multi-degree-of-reedom structure, it is necessary that the period, T, used in equation (18) be interpreted as the period corresponding to the highest natural mode of vibration of the system. This is important because as the period of the higher modes becomes smaller, the accuracy of the results using MTM diminishes. Especially, for systems with a large number of degrees of freedom and where the highest natural period of vibration is very short. Under these conditions, an unconditionally stable procedure, such as Newmark Beta method with f= 4, is preferred. To investigate the error associated with both MTM and MEM, a simple study is conducted. ‘A damped free-vibrational DOSF system is investigated for three different values of A¢/T. The displacement error for various time steps is shown in Figure 7. When At/T = 0-10, the maximum displacement error of both methods approaches unity with slightly more error associated: with MEM compared to MTM (Figure 7(a)). The error decreases as the At/T becomes smaller. Similarly, the same is true for At/T = 0-05 and At/T = 0-01 (see Figures 7(b) and 7(c)). The results demonstrate that in all of these cases MTM is more accurate than MEM. An example of dynamic response of a MDOF system Consider a three-storey frame structure, shown in Figure 8. As a convenience, the physical and vibrational properties of the structure are summarized as 10 0 0 M=| 0 15 0 | (kips?/in) (62) 0 0 2 -1 0 K=600|-1 3 —2| (kipjin) 3) 0 5 145) 31-1} (rad/s) (64) 46-1 EXPLICIT TIME INTEGRATION TECHNIQUE 2279 a @ F Sn ~ Feat + a o” 6 @ 0 2 1 1” ie 20 o os} Yn Keact O 05. “o 2 7 6 € © 2 % ” i” 2 0.10, o@ 00s. 4n—Reaact 0.00 0.0 ou. 0” 2 4” 6 & © i " & " ie " te ” 2 # (seconds) Figure 7. Difference in displacement between the exact solution and the MTM and the MEM approximations; (a) ‘i/T = 010, (b) At/T = 005 and (¢) A/T = O01 10 k- see%/fr PC 15 k- sect AE 2.0 k-seciift [eT ure 8, Idealized properties of a three degree-offreedom structure 2280 Ss. PEZESHK AND CV. CAMP ‘The undamped free vibration of this structure is evaluated for the following arbitrary initial conditions: (05) 0 Xo ={0-4) (in) Vo={9} (in/s) (65) 03, o} where Xo and Vo are the initial displacement and velocity vectors, respectively. Using equations (54)-(56) the free-vibrational motion of each storey is obtained. In Figure 9, the first 5 s of the motion for each storey is shown and compared to the exact solution.* A value of ‘At/T, = 001 (At = 0:004) is selected based on the SDOF error analysis. In general, the time increment is calculated based on the shortest period of the MDOF system. However, the results in Figure 9 show that the MTM approximation agrees with the exact solution at all three degrees of freedom for a time increment based on the largest period of the system. The response of a linear MDOF system may be found by using a mode-superposition procedure,‘ in which MTM is used to solve the uncoupled equations of motion for each mode in terms of the generalized co-ordinates. ts @ os! a a s . &) “ : © cs . + (seconds) Figure 9, Displacement history fora three degree-of freedom structure: (a) first degree of freedom x,,(b) second degree of, freedom x; and (6) third degree of freedom x3 wi) EXPLICIT TIME INTEGRATION TECHNIQUE 2281 SUMMARY AND CONCLUSIONS A simple and effective procedure for computational structural dynamics based on a Modified Trapezoidal rule Method (MTM) is introduced. A systematic and fundamental procedure for stability and accuracy analysis for a SDOF system using the MTM procedure is presented. Based on an analysis of the error in period, amplitude, and phase angle of a free and undamped simple ‘mass-spring system, it is shown that the MTM procedure is more accurate in estimating the response, with practically the same amount of computational cost as the MEM procedure. This particularly true for determining transient responses. The MTM procedure is conditionally stable and requires a time step size to be inversely Proportional to the period of the system. The main advantage of MTM over NBM with B # 0 is its ease of implementation and low computational cost. In addition, from an error analyses of the vibration of a free and undamped mass-spring system, the following conclusions may be drawn: (i the error in period is smaller for MTM compared to NBM for f= 4; (i) the error in the displacement amplitude using MTM with an initial velocity Vo = 0 is negligible; (ii) the error in phase angle using MTM is negligible for At/T < 0-1; (iv) the error in displacement amplitude for an initial displacement Xo = 0 for NBM with f = Sis zero, however, there exist some errors in the NBM procedure with f = 0 which are slightly higher than the error generated by MTM. In general, each error term can be minimized by choosing a smaller At/T with the penalty of additional computational time. REFERENCES 1. G. D, Haha, ‘A modified Euler method for dynamic analyses, Int. j. numer. methods eng, 32, 943-955 (1991). 2. A. Cromer, ‘Stable solutions using the Euler approximation’, Am. J. Phys, 49, 445-459 (198i). 3. N.M, Newmark, A method of computation for structural dynamics, J. Eng. Mech, iv. ASCE, Proceeding paper 2094, ‘85(EM3), 67-94 (1959) R-W. Clough and J. Penzien, Dynamics of Structures, McGraw-Hill, New York, 1975. ind M. A. Dokainish, ‘A survey of direct time-integration methods in computational structural plicit Methods’, Comput. Struct, 32(6), 1371-1386 (1989). 'T,J.R. Hughes, The Finite Element Method, Prentice-Hall, Englewood Cliff, N.J., 1987. ish and K. Subbaraj, ‘A survey of direct time-integration methods in computational structural ‘dynamics—IL. Implicit Methods’, Comput. Struct, 346), 1387-1401 (1989). 8. J.L, Humar, Dynamics of Structures, Prentice-Hall, Englewood Clifs, NJ, 1990. 9. K.J. Bathe, Finite Element Procedures in Engineering Analysis, Prentice-Hall, Englewood Ciiffs, NJ, 1982.

You might also like