Numerical

Evaluation of
Dynamic
Response
Chapter 5

8 October 2016

CC Chang / HKUST

given

m!x!i + cx! i + kxi = pi
m!x!i+1 + cx! i+1 + kxi+1 = pi+1
seek

Displacement x

m!x!(t) + cx! (t) + kx(t) = p(t)

Loading p

Time-Domain Analysis
pi
p1p

pi+1

2

Time stepping

x1
x2
t1 t2

time

xi
xi+1
ti tj

time

Time-Domain Analysis
• Interpolation of Excitation Method
• Central Difference Method
• Newmark’s Method
• Wilson-q Method
• Commercial Software

Interpolation of Excitation
m!x!i + cx! i + kxi = pi
t i £ t £ t i +1

Dp i
t
linear pi +
Dt i pi+1
pi

Dt i

Δpi
m!x! t + cx! t + kx t = pi +
τ
Δt i

m!x!i +1 + cx! i +1 + kxi +1 = pi +1

ti

t

Dpi

ti +1

Step response due to pi .Ramp response due to DPi t Dt i .Interpolation of Excitation Dpi m!x! t + cx! t + kx t = pi + t Dt i constant linear ramp Solution of x(t) consists of 3 parts .Free vibration response .

2x C= í +e [( ) sin wD Dt .xwDt ( x 2 sin wD Dt + cos wD Dt ) 1.(1 + ) cos wD Dt ]ý 2 k ï wDt wD Dt wDt ïþ 1 x î 2 ù 1é 2x 2 x -1 2x .xwDt D = ê1 +e ( sin wD Dt + cos wD Dt )ú k êë wDt wD Dt wDt úû .xwDt ( sin wD Dt ) wD ü 2 1 ìï 2x x 2x ï .xwDt 1 .x 1 B = e .Interpolation of Excitation • Displacement x i +1 = A × x i + B × x! i + C × pi + D × pi +1 where A = e .

+ e [( + ) sin wD Dt + cos wD Dt ]ý 2 2 k ï Dt Dt ïþ 1 x D t 1 x î 1 [1 -e -xwDt ( x sin wD Dt + cos wD Dt )] D¢ = kDt 1 .x2 sin wD Dt ) B¢ = e -xwDt (cos wD Dt - x 1.Interpolation of Excitation • Velocity x! i+1 = A¢ × x i + B¢ × x! i + C¢ × pi + D¢ × pi +1 where A ¢ = -e -xwDt ( w 1 .x2 .x 2 sin wD Dt ) ü 1 ìï 1 w x 1 ï xw D t C¢ = í.

Interpolation of Excitation .

Interpolation of Excitation • Note: – Linear system – SDOF system – Accuracy depends on the linear interpolation of loading p .

[ k ]x i ( Dt ) 2 2Dt ( Dt ) 2 2Dt ( Dt ) 2 kˆ unknown Pˆi .x i -1 m +c + kx i = pi 2 2Dt ( Dt ) m c m c 2m [ + ]x i +1 = pi .2x i + x i -1 ï!x! i = i +1 ïî Dt 2 xi x i -1 x i +1 x! i x i +1 .x i -1 ïï ! i 2Dt í x .[ ]x i -1 .x i -1 2Dt Dt Dt t x i +1 .2x i + x i -1 x i +1 .Central Difference Method • Finite difference x m!x! i + cx! i + kx i = pi ìx = x i +1 .

given x 0 and x! 0 p0 .cx! 0 .x i -1 ïï ! i 2Dt í x .2x i + x i -1 ï!x! i = i +1 ïî Dt 2 ( Dt ) 2 x -1 = x 0 .[ k ]x i 2 2 2Dt 2Dt ( Dt ) ( Dt ) At t=0.2 x 0 + x -1 ï!x! 0 = ïî Dt 2 find need To find x1 we need x0 and x-1 m!x! i + cx! i + kx i = pi ìx = x i +1 .kx 0 !x!0 = m ìx = x1 .[ ]x i -1 .Dtx! 0 + !x! 0 2 .Central Difference Method [ m ( Dt ) 2 + c m c 2m ]x i +1 = pi .x -1 ïï ! 0 2 Dt í x1 .

c. p. x 0 and x! 0 p0 .x i -1 x! i = i +1 2Dt x . k.cx! 0 .2x i + x i -1 !x!i = i +1 Dt 2 .kx 0 ( Dt ) 2 x -1 = x 0 .Dtx! 0 + !x! 0 !x!0 = 2 m m c 2m m c a= b=kkˆ = + ( Dt ) 2 2Dt ( Dt ) 2 ( Dt ) 2 2Dt For step i pˆi = pi .Central Difference Method • Summary Given m.a × x i -1 .b × x i pˆ x i +1 = i kˆ x .

given m!x!i + cx! i + kxi = pi m!x!i+1 + cx! i+1 + kxi+1 = pi+1 seek Displacement x m!x!(t) + cx! (t) + kx(t) = p(t) Loading p Time-Domain Analysis pi p1p pi+1 2 Time stepping x1 x2 t1 t2 time xi xi+1 ti tj time .

!x! i +1 !x! i +1 !x! t = !x! i ( !x!i + !x!i +1 ) 2 !x! t = !x!i + !x! i x! i +1 = x! i + x! i Dt ( !x! i + !x! i +1 ) 2 x! i +1 = x! i + x! i xi t Dt ( !x!i + !x!i +1 ) 2 x i +1 = x i + x! i Dt x i +1 = x i + x! i Dt Dt 2 + ( !x! i + !x! i +1 ) 4 t( !x!i +1 .!x!i ) Dt Dt 2 Dt 2 ! ! !x! i +1 + xi + 3 6 xi t .Newmark’s Method Given m!x!i + cx! i + kx i = pi Seek m!x!i+1 + cx! i+1 + kx i+1 = pi+1 Too many unknowns ! Need assumptions.

b are Newmark’s parameters .g ) Dt ]!x! i + ( gDt )!x! i +1 ï í 1 2 2 x = x + D t x + [( b ) D t ] x + ( b D t )!x! i +1 !i !! i i ïî i +1 2 g.Newmark’s Method Given m!x!i + cx! i + kx i = pi Seek m!x!i+1 + cx! i+1 + kx i+1 = pi+1 ìx! i +1 = x! i + [(1 .

Newmark’s Method • Average acceleration 1 1 g= . b= 2 4 ( !x!i + !x!i +1 ) !x! t = 2 !x! i Dt ì x! i +1 = x! i + ( !x! i + !x! i +1 ) ï 2 ï í ï Dt 2 ïî x i +1 = x i + x! i Dt + ( !x!i + !x!i +1 ) !x! i +1 !x! t = ( !x!i + !x!i +1 ) 2 x! i +1 = x! i x! i + Dt ( !x! i + !x! i +1 ) 2 x i +1 = x i + x! i Dt 4 xi Dt 2 + ( !x! i + !x! i +1 ) 4 .

!x!i ) Dt Dt ! ! x = x + ( !x!i + !x!i +1 ) ì i +1 i 2 ï ï í ï Dt 2 Dt 2 !x!i + !x!i +1 ïî x i +1 = x i + x! i Dt + 3 6 !x! i !x! i +1 t( !x!i +1 .!x!i ) !x! t = !x!i + Dt x! i +1 = x! i + x! i Dt ( !x!i + !x!i +1 ) 2 x i +1 = x i + x! i Dt xi Dt 2 Dt 2 !x! i + !x! i +1 + 3 6 .Newmark’s Method • Linear acceleration !x! t = !x!i + 1 1 g= . b= 2 6 t( !x!i +1 .

g )Dt ]!x! + ( gDt )!x! (3) i i +1 ï i +1 i 1 2 !! 2 !! ï ! x = x + D t x + [( b ) D t ] x + ( b D t ) x i +1 (4) i i î i +1 i 2 (2) .x! i = Dt!x! i + ( gDt )D!x! i (6) 1 2 2 !! ! ! ! D x = x x = D t x + D t x + ( b D t )Dx i (7) (4) gives i i +1 i i i 2 .(1) gives mD!x! i + cDx! i + kDx i = Dpi (5) (3) gives Dx! i = x! i +1 .Newmark’s Method Incremental form ì m!x! i + cx! i + kx i = pi (1) ï m!x! + cx! + kx = p (2) i +1 i +1 i +1 i +1 ï í x! = x! + [(1 .

1)c]!x! i bDt b 2b 2b .Newmark’s Method From (6) and (7) D!x! i = 1 2 Dx i - 1 1 !x! i x! i bDt 2b mD!x! i + cDx! i + kDx i = Dpi (5) Dx! i = Dt!x! i + ( gDt ) D!x! i (6) Dx i = Dtx! i + bDt g g g Dx! i = Dx i .) !x! i bDt b 2b 1 2 Dt !x! i + (bDt 2 ) D!x! i (7) 2 (8) (9) (8) & (9) into (5) kˆDx i = Dpˆi g 1 kˆ = k + c+ m 2 bDt bDt m gc m g ˆ ! Dpi = Dpi + ( + ) x i + [ + Dt ( .x! i + Dt (1 .

c. k. £ b £ ) 2 6 4 Dt 2) For each time step i ( i = 0.2.Summary Given m. x 0 and x! 0 Assume g and 1) Initialize 1 1 1 b (g = .1)c 2b 2b Dpˆi = (pi +1 .x! i + Dt (1 . p.kx 0 !x!0 = m g 1 kˆ = k + c+ m 2 bDt bDt m gc a= + bDt b m g b= + Dt ( .1.cx! 0 .) !x! i bDt b 2b D!x! i = 1 .Newmark’s Method .pi ) + ax! i + b!x! i Dx i = Dpˆi / kˆ 1 1 ! !x! i Dx i xi 2 bDt 2b bDt g g g Dx! i = Dx i .…) p0 .

sec / in.Example • Case study: – Given: • Physical properties of an SDF system: 2 m = 0.05 • Half-cycle sine pulse force (shown right) p.253 kip .Newmark’s Method . x = 0.6 .6) t. k = 10 kips / in. sec 0. T = 1 sec ( w = 6.28 rad / sec). kips 10 10 sin(pt / 0.

253 k = 10 c = 0. and b = 2m = 0. Dt .1 sec g = .1)c 2b 2b 2 4 kˆ = k + c + m = 114.kx 0 !x! 0 = 0 =0 m Dt = 0.cx! 0 .5 2 Dt ( Dt ) 4 a= m + 2c = 10.Newmark’s Method .45. b = 2 4 m = 0.4 p .1 p0 .507.2 1.Example • Average acceleration using 1.159 u 0 = 0 u! 0 = 0 p0 = 0 1.kx 0 !x!0 = m g 1 kˆ = k + c+ m 2 bDt bDt m gc a= + bDt b m g b= + Dt ( .1 1.0 Initial calculations: 1 1 Dt = 0.cx! 0 .3 1.

2 Dx i = 2.Dtx! i ) . !x! i +1 = !x! i + D!x! i Dpˆi Dpˆi = 114.1 Dpˆi = Dpi + ax! i + b!x! i = Dpi + 10.1 sec 2.2x! i Dt 4 = ( Dx i .45x! i + 0.1x! i ) .0.Newmark’s Method .2!x! i = 400( Dx i .2x! i = 20Dx i .5 kˆ 2 = Dx i .4 D!x! i 2. x! i +1 = x! i + Dx! i .Example • Average acceleration using Dt = 0.2!x! i 2 ( Dt ) .5 x i +1 = x i + Dx i .0 calculations for each time step 2.507 !x! i 2.3 Dx! i 2.

Example • Average acceleration using Dt = 0.Newmark’s Method .1 sec Results .

b= 2 6 p0 .Newmark’s Method .Example • Linear acceleration using Dt = 0.4 a= 1 1 g= .253 k = 10 c = 0.159 x 0 = 0 x! 0 = 0 p0 = 0 1.3 3 6 kˆ = k + c + m = 166.2 Dt = 0.1 sec 1.68.768.1)c 2b 2b 6 Dt m + 3c = 15. Dt 2 .cx! 0 .1 1.1 p . and b = 3m + c = 0.8 2 Dt ( Dt ) 1.kx 0 !x! 0 = 0 =0 m 1.0 Initial calculations: m = 0.cx! 0 .kx 0 !x!0 = m g 1 kˆ = k + c+ m 2 bDt bDt m gc a= + bDt b m g b= + Dt ( .

0.68x! i + 0.2 Dx i 2.7679 !x! i Dpˆi Dpˆi = = ˆk 166.3 Dx! i 2.1 Dpˆi = Dpi + ax! i + b!x! i = Dpi + 15.3!x! i = 600( Dx i .4 D!x! i 2.3x! i .05!x! i .3!x! i 2 ( Dt ) 2.Dtx! i ) .1 sec 2.0. x! i +1 = x! i + Dx! i .8 3 Dt = Dx i .Example • Linear acceleration using Dt = 0.5 x i +1 = x i + Dx i .0 calculations for each time step 2.Newmark’s Method . Dt 2 6 = ( Dx i .3x! i . !x! i +1 = !x! i + D!x! i .!x! i = 30Dx i .1x! i ) .

Newmark’s Method .Example • Linear acceleration using Dt = 0.1 sec Results .

Newmark’s Method .Stability To keep the algorithm stable Dt 1 £ T p 2 Average acceleration 1 g . b= 2 6 Conditional stable Usually for accuracy Dt < 0. b= 2 4 Unconditional stable Linear acceleration 1 1 g= .2b 1 1 g= .55 T .1T or less! Dt < 0.

f ( x i ) = Df i Dx i @ ( k i )T Dx i Dx i Tangential stiffness function of displacement f Linear ( k i )T f i +1 fi Dx i Df i Nonlinear xi x i +1 x .Newmark’s Method for nonlinear systems EOM of nonlinear system m!x!( t ) + cx! ( t ) + f ( x ) = p( t ) m!x!i (t) + cx! i (t) + f (x i ) = pi m!x!i+1 (t) + cx! i+1 (t) + f (xi+1 ) = pi+1 mD!x! i + cDx! i + Df i = Dpi Df i = f ( x i +1 ) .

!x! i +1 Dx i Correction (minimize the unbalanced force) Original mD!x! i ( t ) + cDx! i + Df i = Dpi Approximate mD!x! i ( t ) + cDx! i + ( k i )T Dx i = Dpi Unbalanced ( j) D R Dx i ( j) = ( k i )T Df i .( k i )T Dx i = DR Newton-Raphson Iteration Dx i t = Dx i + Dx i (1) + Dx i ( 2) + ! . x! i +1.Newmark’s Method for nonlinear systems mD!x!i (t) + cDx! i (t) + (ki )T Dxi = Dpi Newmark’s Scheme x i +1.

p. £ b £ ) 2 6 4 Dt 1) Initialize p0 . x 0 and x! 0 Assume g and 1 1 1 b (g = .Newmark’s Method for nonlinear systems Given m.cx! 0 .1)c 2b 2b . k.(fs )0 !x!0 = m m gc a= + bDt b m g b= + Dt ( . c.

2..2.1.) !x! i bDt b 2b .…) Dpˆi = Dpi + ax! i + b!x! i Dx ( j) DR ( j) = kˆ DR (i) = Dpˆi x i +1(0) = x i g 1 kˆ i = k i + c+ m bDt bDt 2 x i +1( j) = x i +1( j-1) + Dx ( j) j=1.fs( j-1) + ( kˆ T .x i + Dt (1 . .Df ( j) ˆk = k + g c + 1 m (Newton-Raphson iteration) T T 2 bDt bDt D!x! i = 1 Dx i - 1 1 !x! i x! i bDt 2b bDt 2 g g g ! ! Dx i = Dx i . T Df ( j) = fs( j) .Newmark’s Method for nonlinear systems 2) For each time step i ( i = 0.k T )Dx ( j) DR ( j+1) = DR ( j) ..

05 • Force-deformation relation is elastoplastic with x y = 0.Newmark’s Method for nonlinear systems • Case study – Given: • Physical properties of an SDOF system: m = 0. sec 0.6) x.6 .sec 2 / in.253 kip .75 in • Half-cycle sine pulse force f. kips 10 e o 0.75 d 1 10 sin( pt / 0. 10 c 10 1 t.5 b a p. in. kips 7. x = 0.

45.0 Initial calculations: m = 0.Newmark’s Method for nonlinear systems • Average acceleration without iteration using Dt = 0.507.1 4 m + 2c = 10.2 p0 .159 x 0 = 0 x! 0 = 0 p0 = 0 1.3 a= 1.253 k = 10 c = 0.kx 0 !x! 0 = =0 m Dt = 0. Dt .cx! 0 .1 sec 1. and b = 2m = 0.1 1.

5 2 4 kˆ i = k i + c + m = k i + 104.4 ki = 0 for ab and cd . and 2.Newmark’s Method for nonlinear systems • Average acceleration without iteration using Dt = 0.2x! i Dt 2.3 2.5 2 Dt ( Dt ) Dpˆi Dx i = kˆ 2 Dx! i = Dx i .2x! i = 20Dx i .1 Dpˆi = Dpi + ax! i + b!x! i = Dpi + 10.(fs )i +1 !x!i +1 = i +1 .6 x i +1 = x i + Dx i .507 !x! i 2.0 calculations for each time step 2.8 p .cx! i +1 . m 2. x! i +1 = x! i + Dx! i 2.1 sec 2. 2.45x! i + 0.2 ki = k for branches oa. bc and de.7 (fs )i +1 = (fs )i + k i Dx i .

Newmark’s Method for nonlinear systems • Average acceleration without iteration using Dt = 0.1 sec Results .

Newmark’s Method for nonlinear systems • Newton-Raphson iteration using Dt = 0.1 p .cx! 0 .2 Dt = 0.507.kx 0 !x! 0 = 0 =0 m 1.159 x 0 = 0 x! 0 = 0 p0 = 0 1.45.253 k = 10 c = 0.3 a= 4 m + 2c = 10.1 1.0 Initial calculations: m = 0. Dt . and b = 2m = 0.1 sec 1.

Df ( j) g 1 kˆ T = k T + c+ m 2 bDt bDt 2 Dx! i = Dx i . and k i = 0 for ab and cd Df ( j) = fs( j) .Newmark’s Method for nonlinear systems • Newton-Raphson iteration using Dt = 0. bc and de.1 Dpˆi = Dpi + ax! i + b!x! i = Dpi + 10.507 !x! i 2.5 2.6 2.3 2 4 kˆ i = k i + c + m = k i + 104.2!x! i = 400Dx i .5 2 Dt ( Dt ) ( j) DR Dx ( j) = x i +1( j) = x i +1( j-1) + Dx ( j) ˆk T 2.2x! i Dt 4 4 D!x! i = Dx i x! i .0 calculations for each time step 2.40 x! i .fs( j-1) + ( kˆ T .2 ki = k 2. .4 2.2!x! i 2 D t ( Dt ) !x! i +1 = !x! i + D!x! i .1 sec 2.k T )Dx ( j) DR ( j+1) = DR ( j) .2x! i = 20Dx i .7 for branches oa.45x! i + 0.

1 sec Results No iteration Newmark’s Method .for nonlinear systems • Linear acceleration using Dt = 0.

b = Incremental form 2 mD!x! i + cDx! i + kDx i = Dpi !x! Dx! i = Dt!x! i + ( gDt ) D!x! i Dx i = Dtx! i + 1 2 Dt !x! i + (bDt 2 ) D!x! i !x! i +1 2 Linear acceleration and force between ti and ti+q md!x! i + cdx! i + kdx i = dpi 1 dtd!x! i 2 1 1 dx i = dtx! i + dt 2 !x! i + dt 2 d!x! i 2 6 dx! i = dt!x! i + 6 d!x! i D!x! i !x! i Dt ti t i +1 dt = qDt ti+q time .Wilson-q Method Modification of the Newmark’s linear acceleration method 1 1 g = .

x i .pi ) + æç m + 3c ö÷ x! i + æç 3m + c ÷!x! i 2 ø è qDt ø è 3 dt dx! i = dx i .3x! i .Wilson-q Method md!x! i + cdx! i + kdx i = dpi where then dpˆi dx i = kˆ 3 6 kˆ = k + c+ m 2 qDt (qDt ) 6 qDt ö dpˆi = q(pi +1 .!x! i dt 2 1 6 6 ! ! ! ! ! ! ! D x = d!x! i dx i = dx i .3x i i q dt dt 2 .

2.x! i .kx 0 !x!0 = 0 m 3 6 kˆ = k + c+ m 2 qDt (qDt ) 6 m + 3c qDt 1 b = 3m + qDtc 2 a= 2) For each time step i ( i = 0.Wilson-q Method . x 0 and x! 0 Dx i = Dpˆi / kˆ Assume Dt and q (q > 1 normally q = 1. k.4 ) 1) Initialize p .1. p.pi ) + ax! i + b!x! i dpˆi kˆ 3 dt dx! i = dx i . c.cx! 0 .3x! i .!x! i dt 2 6 6 1 !x! i = d!x! i d!x! i = dx i .3!x! i D dt q dt 2 1 Dx! i = Dt!x! i + DtD!x! i 2 1 1 Dx i = Dtx! i + Dt 2 !x! i + Dt 2 D!x! i 2 6 dx i = .Summary Given m.…) dpˆi = q(pi +1 .

83 ´ 107 N/m !x! g .Commercial Software . m/s2 4 2 m 0 -2 -4 0 5 10 time s 15 20 k c 75-story building m = 4.SDOF structure under earthquake excitation 1940 El Centro EQ ground acc.04 ´ 106 N .s/m (1%) k = 5.MATLAB Demonstration .61 ´ 107 kg c = 1.

ground acc. title('Response of a tall building under earthquake'). -k/m -c/m].MATLAB Response of a tall building under earthquake load elcen.4 .83*10^7.C. subplot(2.1. D=[0]. B=[0.acc. plot(time. m').x).4 [x temp]=lsim(A. plot(time. xlabel('time').time). 0 5 10 15 20 0 5 10 time s 15 20 0.61*10^7.04*10^6.-1]. m/s2 clear all.2). C=[1 0]. ylabel('disp. disp.Commercial Software . c=1.1.1). 4 m=4. m 0. k=5.B. ylabel('ground acc.acc). 2 0 -2 -4 A=[0 1.2 -0.D.2 0 -0. subplot(2. m/s^2').

Commercial Software .MATLAB The lsim command operates in the STATE-SPACE domain Original 2nd order m!x! + cx! + kx = p State-Space 1st order ìm!x! + cx! + kx = p í x! = x! î 1 ù ìx ü ìï 0 üï ìx! ü é 0 í ý = ê .k .c ú í ý + í 1 ýp !þ ï ï x î!x! þ êë m î ú îm þ mû output ìx ü y = [1 0]í ý + {0}p îx! þ .

D. [x temp]=lsim(A.time).Commercial Software .B.-acc. Demonstration ! .C.MATLAB ! = AX + BU X Y = CX + DU 1 ù é 0 cú A=ê k êë m m úû ìx ü X=í ý îx! þ ìï 0 üï B= í1ý ïî m ïþ C and D depend on the output desired.

Summary m!x!(t) + cx! (t) + kx(t) = p(t) • 5 methods are introduced – – – – – • • Interpolation of Excitation Method Central Difference Method Newmark’s Method Wilson-q Method Commercial Software Accuracy Extension to multiple degrees and/or nonlinear structures – Newmark’s Method – Wilson-q Method – Commercial Software .