MIT - 16.20 Fall, 2002 Unit 22 Vibration of Multi Degree-Of- Freedom Systems Paul A. Lagace, Ph.D. Professor of Aeronautics & Astronautics and Engineering Systems Paul A. Lagace ? 2001 MIT - 16.20 Fall, 2002 Previously saw (in Unit 19) that a multi degree-of-freedom system has the same basic form of the governing equation as a single degree-of-freedom system. The difference is that it is a matrix equation: mq ˙˙ + k q = F (22-1) ~ ~ ~ ~ ~ ~ = matrix So apply the same solution technique as for a single degree-of-freedom system. Thus, first deal wit h … Free Vibration Do this by again setting forces to zero: F = 0 ~ ~ mq ˙˙ + k q = 0 (22-2) ~ ~ ~ ~ ~ Paul A. Lagace ? 2001 Unit 22 - 2 it ω ω MIT - 16.20 Fall, 2002 Again assume a solution which has harmonic motion. It now has multiple components: ω qt () = A e (22-3) ~ ~ where ω are the natural frequencies of the system and: ? M ? ? ? A is a vector of constants = ? A i ? ~ ? ? ? M ? Substituting the assumed solution into the matrix set of governing equations: it ? ? ω 2 mA e + k A e it = 0 ~~ ~~ ~ To be true for all cases: [ k ? ω 2 m ] A = 0 (22-4) ~ ~ ~ ~ This is a standard eigenvalue problem. Either: A = 0 (trivial solution) ~ or Paul A. Lagace ? 2001 Unit 22 - 3 MIT - 16.20 Fall, 2002 The determinant: k ? ω 2 m = 0 (22-5) ~ ~ There will be n eigenvalues for an n degree-of-freedom system. In this case: eigenvalue = natural frequency ? n degree-of-freedom system has n natural frequencies Corresponding to each eigenvalue (natural frequency), there is an … Eigenvector -- Natural Mode ? Place natural frequency ω r into equation (22-4): [ k ? ω r 2 m ] A = ~ 0 ~ ~ ~ ? Since determinant = 0, there is one dependent equation, so one cannot solve explicitly for A. However, one can solve for the ~ relative values of the components of A in terms of (normalized ~ by) one component Paul A. Lagace ? 2001 Unit 22 - 4 MIT - 16.20 Fall, 2002 ? Say divide through by A n : ? M ? ? A A i n ? ? ? [ ~ k ? ω i 2 m ] ? ? = ~ 0 ~ ? M ? ? 1 ? ? ? ? Solve for A i / A n for each ω r ? M ? ? A A i n ? ? ? r Indicates solution ? Call the eigenvector ? ? = φ i () ? M ? ~ for ω r ? 1 ? ? ? ? Do this for each eigenvalue frequency: ω 1 , ω 2 …… . ω n 1 2 φ i () associated mode: φ i () φ i () n ~ ~ ~ Paul A. Lagace ? 2001 Unit 22 - 5 MIT - 16.20 Fall, 2002 For each eigenvalue , the homogeneous solution is: r r r q i hom = φ i () e i ω r t = C 1 φ i () sin ω r t + C 2 φ i () cos ω r t ~ ~ ~ ~ homogeneous Still an undetermined constant in each case (A n ) which can be determined from the Initial Conditions ? Each homogeneous solution physically represents a possible free vibration mode ? Arrange natural frequencies from lowest ( ω 1 ) to highest ( ω n ) ? By superposition, any combinations of these is a valid solution Example: Two mass system (from Unit 19) Figure 22.1 Representation of dual spring-mass system Paul A. Lagace ? 2001 Unit 22 - 6 q q MIT - 16.20 Fall, 2002 The governing equation was: ? m 1 0 ? ? ˙˙ 1 ? ? ( k 1 + k 2 ) ? k 2 ? ? q 1 ? ? F 1 ? ? 0 m 2 ? ? ˙˙ 2 ? ? ? ? = ? ? ? ? ? ? + ? ? ? k 2 k 2 ? ? q 2 ? ? F 2 ? Thus, from equation (22-5): ( k 1 + k 2 ) ? ω 2 m 1 ? k 2 = 0 2 ? k 2 k 2 ? ω m 2 This gives: [ ( k 1 + k 2 ) ? ω 2 m 1 ] [ k 2 ? ω 2 m 2 ] ? k 2 2 = 0 This leads to a quadratic equation in ω 2 . Solving gives two roots ( ω 1 2 and ω 2 2 ) and the natural frequencies are ω 1 and ω 2 Find the associated eigenvectors in terms of A 2 (i.e. normalized by A 2 ) Paul A. Lagace ? 2001 Unit 22 - 7 ?? ?? ?? ?? MIT - 16.20 Fall, 2002 Go back to equation (22-4) and divide through by A 2 : ? ( k 1 + k 2 ) ? ω r 2 m 1 ? k 2 ? ? A 1 ? ? ? ? ? = 0 ? ? k 2 k 2 ? ω r 2 m 2 ? ? 1 ? Normalized constant k ? A 1 = k 1 + k 2 2 ? ω r 2 m 1 for ω r mode Thus the eigenvectors are: ? k 2 ? ? k 2 ? 1 ? k 1 + k 2 ? ω 1 2 m 1 ? φ () ? k 1 + k 2 ? ω 2 2 m 1 ? 2 φ i () = ? ? i = ? ? ~ ~ ? ? ? ? ? 1 ? ? 1 ? For the case of Initial Conditions of 0, the cos term goes away and are left with … r qt () = φ () sin ω r t ~ ~ Physically the modes are: Paul A. Lagace ? 2001 Unit 22 - 8 φφφ φφφ ωωω MIT - 16.20 Fall, 2002 Figure 22.2 Representation of modes of spring-mass system φ (1) ω 1 (lowest frequency) masses move in same direction φ (2) frequency) ωωωω 2 (higher masses move in opposite direction Paul A. Lagace ? 2001 Unit 22 - 9 MIT - 16.20 Fall, 2002 General Rules for discrete systems: ? Can find various modes (without amplitudes) by considering combinations of positive and negative (relative) motion. However, be careful of (-1) factor across entire mode. For example, in two degree-of-freedom case + + + - same mode same mode - - - + ? The more “reversals” in direction, the higher the mode (and the frequency) ? It is harder to excite higher modes This can be better illustrated by considering the vibration of a beam. So look at: Representation of a Beam as a Discrete Mass System Paul A. Lagace ? 2001 Unit 22 - 1 0 MIT - 16.20 Fall, 2002 How? ? Lump mass into discrete locations with constraint that total mass be the same ? Connect masses by rigid connections with rotational springs at each mass ? Stiffnesses of connections are influence coefficients (dependent on locations of point masses) ? Forces applied to point masses So: Figure 22.3 Representation of cantilevered beam as single mass system A becomes for simplest case with torsional spring at joint B where: m = ρ tw l length density thickness width Paul A. Lagace ? 2001 Unit 22 - 1 1 MIT - 16.20 Fall, 2002 Could also put mass at mid-point: Figure 22.4 Representation of cantilevered beam as mid-point mass system ? get a different representation Consider the next complicated representation (simplest multi-mass/degree-of-freedom system) Figure 22.5 Representation of cantilevered beam as dual spring-mass system Paul A. Lagace ? 2001 Unit 22 - 1 2 q q MIT - 16.20 Fall, 2002 each m is one half of total mass of beam for constant cross-section case Use influence coefficients, get C matrix, invert to get K. Resulting ~ ~ equation is: ? m 1 0 ? ? ˙˙ 1 ? ? k 11 k 12 ? ? q 1 ? ? F 1 ? ? ? ? ? + ? ? ? ? = ? ? ? 0 m 2 ? ? ˙˙ 2 ? ? k 21 k 22 ? ? q 2 ? ? F 2 ? Same form as before, so solution takes same form. For initial rest conditions: r qt () = φ () sin ω r t ~ ~ Have two eigenvalues (natural frequencies) and associated eigenvectors (modes) ? Modes have clear physical interpretation here: Paul A. Lagace ? 2001 Unit 22 - 1 3 φφφφφφ ωωω ωωω MIT - 16.20 Fall, 2002 Figure 22.6 Representation of deflection modes of cantilevered beam as dual spring-mass system q 2 (+) φ (1) ω 1 (lowest frequency) φ (2) ω 2 (higher frequency) q 1 (+) q 1 (-) q 2 (+) Paul A. Lagace ? 2001 Unit 22 - 1 4 MIT - 16.20 Fall, 2002 Can extend by dividing beam into more discrete masses --> get better representation with more equations but same basic treatment/approach In considering the modes that result from such an analysis, there is a key finding: Orthogonality Relations It can be shown that the modes of a system are orthogonal. That is: transpose r φ () T m φ ( s ) = 0 (22-6) ~ ~ ~ for r ≠ s If r = s, then a finite value results: r T r φ () m φ () = M r (22-7) ~ ~ ~ some value So the general relation for equations (22-6) and (22-7) can be written as: Paul A. Lagace ? 2001 Unit 22 - 1 5 MIT - 16.20 Fall, 2002 φ () r T m φ ( s ) = δ rs M r (22-8) ~ ~ ~ δ rs is the kronecker delta where: δ rs = 0 for r ≠ s δ rs = 1 for r = s This relation allows the transformation of the governing equation into a special set of equations based on the (normal) nodes … Normal Equations of Motion These resulting equations are uncoupled and thus much easier to solve The starting point is the eigenvectors (modes) and the orthogonality relations One must also note that: φ () r T k φ ( s ) = δ rs M r ω r 2 (22-9) ~ ~ ~ Paul A. Lagace ? 2001 Unit 22 - 1 6 ? ? ? ? MIT - 16.20 Fall, 2002 (can show using equations (22-2) and (22-8) ) Have shown that the homogeneous solution to the general equation: mq ˙˙ + k q = F (22-1) ~ ~ ~ ~ ~ is the sum of the eigenvectors (modes): n r qt i () t i () = ∑ φξ r () (22-10) ~ r = 1 n = number of degrees of freedom Where ξ r (t) is basically a magnitude associated with the mode φ (r) at time t. The ξ r become the “ normalized coordinates ” . Thus: 1 3 ? q 1 ? ? φ 1 () φ ( 2 ) φ () L φ ( n ) ? ? ξ 1 ? 1 1 1 ? ? ? φ 2 2 ? ? ξ ? ? ? ? () M M M ? ? 2 ? ? q 2 ? ? ? ? q 3 ? = ? M M M M ? ? ξ 3 ? ? ? M ? ? M ? ? ? M M M M ? ? ? ? ? n ? q n ? ?? M M M L φ n () ? ? ? ξ n ? Paul A. Lagace ? 2001 Unit 22 - 1 7 MIT - 16.20 Fall, 2002 which can be written as: q = φξ (22-11) ~ ~ ~ Placing (22-11) into (22-1) m φξ ˙˙ + k φξ = F (22-12) ~ ~ ~ ~ ~ ~ ~ Now multiply this equation by the transpose of φ : ~ 2 ? φ () 1 1 φ () LL ? ? 2 ? φ T = ? φ 2 () L ? ~ ? M ? ? ? n ? φ () LL L ? φ T m φ ξ ˙˙ + φ T k φ ξ = φ T F (22-13) ~ ~ ~ ~ ~ ~ ~ ~~ ~ Notice that the terms of φ T m φ and φ T k φ will result in most of the terms ~ ~ ~ ~ ~ ~ being zero due to the orthogonality relation (equation 22-8). Only the diagonal terms will remain. Thus, (22-13) becomes a set of uncoupled equations: (via 22-8 and 22-9) Paul A. Lagace ? 2001 Unit 22 - 1 8 MIT - 16.20 Fall, 2002 ˙˙ ξ M rr + M r ω r 2 ξ r = Ξ r (22-14) r = 1, 2 … n That is: ˙˙ M ξ + M ω 2 ξ = Ξ 11 1 1 1 1 ˙˙ M ξ + M 2 ω 2 2 ξ 2 = Ξ 2 22 ˙˙ M ξ + M ω 2 ξ = Ξ nn n n n n where: r ? φ 1 () ? r φ 2 () L ] m ? φ 2 () ? = Generalized mass M r = [ φ 1 () r ~ ? r ? ? ? of rth mode ? M ? and: Paul A. Lagace ? 2001 Unit 22 - 1 9 MIT - 16.20 Fall, 2002 ? F 1 ? Ξ r = [ φ 1 () r ? ? r φ 2 () L ] ? F 2 ? = Generalized force of rth mode ? ? ? M ? ξ r (t) = normal coordinates The equations have been transformed to normal coordinates and are now uncoupled single degree-of-freedom systems Implication : Each equation can be solved separately The overall solution is then a superposition of the individual solutions (normal nodes) Free Vibration ( Ξ = 0 ) --> solution …use same technique as before ? For any equation r: ξ = a sin ω t + b cos ω t r r r r r Paul A. Lagace ? 2001 Unit 22 - 2 0 MIT - 16.20 Fall, 2002 ? Get a r and b r via transformed Initial Conditions r ξ r () = 1 φ () T mq i () = b r 0 0 ~ M ~ r r ˙ r () = 1 φ () T mq ˙ i () = a r ω r ξ 0 0 ~ M ~ r Notes : ? KEY SIMPLIFICATION is that often only first few (lowest) modes are excited so can solve only first few equations. Can add more modes (equations) to improve solution if needed. ? This is a rigorous treatment -- no approximation made by going to normal coordinates. But , this has all been based on the homogeneous case (free vibration), what about … Paul A. Lagace ? 2001 Unit 22 - 2 1 MIT - 16.20 Fall, 2002 Forced Vibration ( Ξ ≠ 0 ) Response is still made up of the natural modes. Solution is found using the same approach as for a single degree-of-freedom system… ? Break up each generalized force, Ξ r , into a series of impulses ? Use Duhamel ’ s (convolution) integral to get response for each degree of freedom ? Stay in normalized coordinates The solution for any mode will thus look like: 1 t ξ r ( t ) = rr ∫ 0 Ξ r ( τ ) sin ω r ( t ? τ ) d τ M ω and equation (22-11) then gives: n r qt i () t i () = ∑ φξ r () r = 1 Again, use Initial Conditions to get constants Paul A. Lagace ? 2001 Unit 22 - 2 2 ??? MIT - 16.20 Fall, 2002 Exact same procedure as single degree-of-freedom system. Do it multiple times and add up. (Linear ? Superposition) Can therefore represent any system by discrete masses. As more and more discrete points are taken, get a better model of the actual behavior. Taking this to the limit will allow the full representation of the behavior of continuous systems . Paul A. Lagace ? 2001 Unit 22 - 2 3