CE 809 - Structural Dynamics

Numerical Evaluation of Dynamic Response – Step-by-step Direct Integration - MDF Systems

Semester – August 2021

Dr. Fawad A. Najam Prof. Dr. Pennung Warnitchai

Department of Structural Engineering Head, Department of Civil and Infrastructure Engineering
NUST Institute of Civil Engineering (NICE) School of Engineering and Technology (SET)
National University of Sciences and Technology (NUST) Asian Institute of Technology (AIT)
H-12 Islamabad, Pakistan Bangkok, Thailand
Cell: 92-334-5192533, Email: [email protected]
STEP-BY- STEP DIRECT INTEGRATION:

The most convenient method for evaluating the dynamic response of non-linear multi-degree-of-freedom
structures.

Similar to the case of a SDF structure, the response time history of a MDF structure is divided into many
short-time increments.

During each time increment, the properties of the structure are assumed unchanged.

After each incremental step, the properties are updated in accordance with the current condition of
deformation.

The method is also applicable to the case of linear MDOF systems. In this case it is not necessary to
update the properties of the system.
Numerical Evaluation of Dynamic Response of MDF Systems

• Linear Systems
• Central Difference Method
• Newmark’s Method
• Constant average acceleration method
• Linear acceleration method
• Nonlinear Systems
• Central Difference Method
• Constant Average Acceleration Method
• Nonlinear Static Analysis

3
INCREMENTAL EQUILIBRIUM EQUATIONS:

In analogy to the case of nonlinear SDOF system, the equations of motion for a non-linear
MDOF system are:

𝒇𝑰 (𝑡) + 𝒇𝑫 (𝑡) + 𝒇𝑺 (𝑡) = 𝑷(𝑡) (1)

The incremental form of the equations of motion are:

∆𝒇𝑰 𝑡 + ∆𝒇𝑫 𝑡 + ∆𝒇𝑺 𝑡 = ∆𝑷(𝑡) (2)

where ∆𝒇𝑰 𝑡 = 𝒇𝑰 𝑡 + ∆𝑡 − 𝒇𝑰 𝑡 = 𝑴∆𝒖(𝑡)

∆𝒇𝑫 𝑡 = 𝒇𝑫 𝑡 + ∆𝑡 − 𝒇𝑫 𝑡 ≅ 𝑪(𝑡) ∆𝒖(𝑡)

∆𝒇𝑺 𝑡 = 𝒇𝑺 𝑡 + ∆𝑡 − 𝒇𝑺 𝑡 ≅ 𝑲 𝑡 ∆𝒖(𝑡) (3)

For single-degree-of-freedom systems:
𝑝 𝑡 + ∆𝑡
𝑝𝑡 Δ𝑝 𝑡 = 𝑝 (𝑡 + Δ𝑡) − 𝑝 𝑡
Δ𝑓𝐼 𝑡 + Δ𝑓𝐷 𝑡 + Δ𝑓𝑠 𝑡 = Δ 𝑝 𝑡
Time
𝑓𝑠 𝑡 𝑡 + Δ𝑡
Slope
𝑑𝑓𝑠
𝑓𝑠 𝑡 + ∆𝑡 Δ𝑓𝑠 𝑡 =
𝑑𝑢 𝑡
Δ𝑓𝑠 𝑡 = 𝑓𝑠 𝑡 + ∆𝑡 − 𝑓𝑠 𝑡
=𝑘 𝑡

𝑓𝑠 𝑡 𝑑𝑓𝑠
Δ𝑢 𝑡 ≅ . Δ𝑢 𝑡
𝑑𝑢 𝑡

≅ 𝑘 𝑡 . Δ𝑢
𝑢
𝑢𝑡 𝑢 𝑡 + ∆𝑡
5
For single-degree-of-freedom systems:
We have introduced the following two approximations:

Δ𝑓𝐷 𝑡 = 𝑐 𝑡 . Δ𝑢 𝑡
𝑑𝑓𝑠
=
𝑑𝑢 𝑡

Δ𝑓𝑠 𝑡 = 𝑘 𝑡 . Δ𝑢 𝑡

They are equivalent to the assumption that the damping force

and restoring forces are linear within 𝑡 and 𝑡 + Δ𝑡.

“Piecewise Linear Approximation of Structural System”

6
For multiple-degree-of-freedom systems:
𝐾11 𝑡 𝐾12 𝑡 ⋯ 𝐾1N 𝑡
𝑲(𝑡) = ⋮ ⋱ ⋮ = 𝐾ij 𝑡 The tangent stiffness
𝐾N1 𝑡 … ⋯ 𝐾NN 𝑡 matrix at time 𝑡

Similarly, 𝑪 𝑡 = [𝐶𝑖𝑗 𝑡 The tangent damping matrix at time 𝑡

𝜕𝑓𝑆𝑖 𝜕𝑓𝐷𝑖
where 𝐾𝑖𝑗 𝑡 = and 𝐶𝑖𝑗 𝑡 = (4)
𝜕𝑢𝑗 𝜕𝑢𝑗
𝑡 𝑡

Substituting Eq. (3) into Eq. (2)

𝑴 ∆𝒖 𝑡 + 𝑪 𝑡 ∆𝒖 𝑡 + 𝑲 𝑡 ∆𝒖 𝑡 = ∆𝑷 𝑡 (5)
 𝑢 𝑡 + Δ𝑢 𝑡
𝑢𝑡 “the acceleration response
“Linear Δ𝑢 𝑡
∆𝑡 𝑢 𝜏 =𝑢 𝑡 + .𝜏 varies linearly during each ………. (a)
Acceleration” ∆𝑡 time increment”.
assumption for
single-degree- 𝑡 𝑡 + ∆𝑡 𝜏
𝜏 𝑢 𝜏 =𝑢 𝑡 + 𝑢 𝜏 𝑑𝜏
of-freedom
𝑢 𝑡 + ∆𝑢 𝑡 0
………. (b)
systems: 𝑢𝑡 Δ𝑢 𝑡 𝜏 2
𝑢 𝜏 = 𝑢 𝑡 + 𝑢 𝑡 .𝜏 + .
∆𝑡 2

∆𝑢 𝜏
𝜏
𝑢𝜏 =𝑢 𝑡 + 𝑢 𝜏 𝑑𝜏
𝑢 𝑡 + ∆𝑢 𝑡 0

𝑢 𝑡 𝜏 2 Δ𝑢 𝑡 𝜏 3 (c)
𝑢 𝜏 = 𝑢 𝑡 + 𝑢 𝑡 .𝜏 + 𝑢 𝑡 . + .
2 ∆𝑡 6
8
∆𝑢 𝜏
“Linear Acceleration” assumption for single-degree-of-freedom systems:
At 𝜏 = ∆𝑡, the above equations for velocity and displacement becomes,

Δ𝑢 𝑡 ∆𝑡 2 ………. (d)
∆𝑢 𝑡 = 𝑢 𝑡 . ∆𝑡 + .
∆𝑡 2
∆𝑡 2 Δ𝑢 𝑡 ∆𝑡 3
∆𝑢 𝑡 = 𝑢 𝑡 . ∆𝑡 + 𝑢 𝑡 . + . ………. (e)
2 ∆𝑡 6

Re-writing the above two equations in terms of Δ𝑢 𝑡 :

6 6 ………. (f)
∆𝑢 𝑡 = 2 . ∆𝑢 𝑡 − . 𝑢 𝑡 − 3 𝑢 𝑡
∆𝑡 ∆𝑡
3 ∆𝑡
Δ𝑢 𝑡 = . ∆𝑢 𝑡 − 3 𝑢 𝑡 − . 𝑢 𝑡 ………. (g)
∆𝑡 2
Equations (f) and (g) are derived from the “linear acceleration assumption”.
9
For multiple-degree-of-freedom systems:
Introducing the assumption of “linear acceleration”, it can be shown that
∆𝑡
∆𝒖 𝑡 = 𝒖 𝑡 ∆𝑡 + ∆𝒖
2
∆𝑡 2 ∆𝑡 2
∆𝒖 𝑡 = 𝒖 𝑡 ∆𝑡 + 𝒖 𝑡 + ∆𝒖 𝑡 (6)
2 6
Rearrange Eq. (6):
6 6
∆𝒖 𝑡 = 2
∆𝒖 𝑡 − 𝒖 𝑡 − 3𝒖 𝑡
∆𝑡 ∆𝑡

3 ∆𝑡
∆𝒖 𝑡 = ∆𝒖 𝑡 − 3 𝒖 𝑡 − 𝒖 𝑡 (7)
∆𝑡 2

Substitute Eq. (7) into Eq. (5): 𝑴 ∆𝒖 𝑡 + 𝑪 𝑡 ∆𝒖 𝑡 + 𝑲 𝑡 ∆𝒖 𝑡 = ∆𝑷 𝑡 (5)

Substitute Eq. (7) into Eq. (5),

𝑲 𝑡 ∆𝒖 𝑡 = ∆𝑷 𝑡 (8)

where
6 3
𝑲𝑡 = 𝑲𝑡 + 2 𝑴 + 𝑪𝑡
∆𝑡 ∆𝑡

6 ∆𝑡
∆𝑷 𝑡 = ∆𝑷 𝑡 + 𝑴 𝒖 𝑡 + 3𝒖 𝑡 + 𝑪 𝑡 3𝒖 𝑡 + 𝒖 𝑡
∆𝑡 2
(9)
Eq. (8) is in the same form as a standard static equation,

hence 𝑲 𝑡 and ∆𝑷 𝑡 may be interpreted as the effective dynamic stiffness and the effective
load increment, respectively.

Two approximations (assumptions) are used in this process:

(1) 𝑪 𝑡 and 𝑲 𝑡 remain constant during {𝑡, 𝑡 + ∆𝑡}

(2) Acceleration 𝒖 varies linearly during {𝑡, 𝑡 + ∆𝑡}

Approximation  errors  accumulations of errors  unstable calculation?

CALCULATIONS FLOW-CHART:
Time = 𝑡 𝒖(𝑡), 𝒖(𝑡) are known

Evaluate 𝐾𝑖𝑗 (𝑡), 𝐶𝑖𝑗 (𝑡) ⟹ 𝑲(𝑡), 𝑪(𝑡)

and 𝒇𝑆 (𝑡) = 𝒇𝑆 (𝒖(𝑡)), 𝒇𝐷 (𝑡) = 𝒇𝐷 (𝒖(𝑡))

Imposing the dynamic

equilibrium condition at time 𝑡
𝒖(𝑡) = 𝑴−1 𝑷(𝑡) − 𝒇𝐷 (𝑡) − 𝒇𝑆 (𝑡)

Formulate 𝑲(𝑡), ∆𝑷(𝑡)

Inverse Matrix ∆𝒖(𝑡) = 𝑲(𝑡)−1 ∆𝑷(𝑡)

3 ∆𝑡
Δ𝒖(𝑡) = ∆𝒖(𝑡) − 3𝒖(𝑡) − 𝒖(𝑡)
∆𝑡 2
𝒖(𝑡 + ∆𝑡) = 𝒖(𝑡) + ∆𝒖(𝑡)
Time = 𝑡 + ∆𝑡
𝒖(𝑡 + ∆𝑡) = 𝒖(𝑡) + ∆𝒖(𝑡)
 ∆𝑡 1 1
In general the analysis is accurate when < ~
𝑇 5 10
∆𝑡 1
But errors will accumulate in an “unstable” manner when >
𝑇 1.8

For an N-DOF system: 𝑇1 > 𝑇2 > 𝑇3 > ⋯ > 𝑇𝑁

To avoid this unstable error’s accumulation, ∆𝑡 must be made short relative to the least
period of vibration contained in the structural system, regardless of whether the higher
modes contribute significantly to the dynamic response or not.
∆𝑡 1 ∆𝑡 1 1
Example: if 𝑇1 = 800 𝑇𝑁 and = then = =
𝑇𝑁 2 𝑇1 2 × 800 1600

1600 time steps are required to complete just one cycle of the 1st mode!
Numerical Damping & Period Elongation: The effects of error accumulation

EXAMPLE: Undamped free vibration of a linear SDOF system. 𝑚𝑢+𝑘𝑢 =0

Oscillation amplitude decays.

Positive (+) artificial damping
(Numerical Damping)

Stable

Oscillation amplitude grows up.

Negative (-) artificial damping
(Numerical Damping)

Unstable
For important modes that contribute significantly to response (usually lower modes),
we want zero (0) or near-zero numerical damping.

For unimportant modes (usually higher modes),

it is better to have positive (+) numerical damping.
(since zero numerical damping is not possible to attain)

We want a calculation algorithm that always produce positive (+) artificial damping regardless
the values of ∆𝑡 𝑇 ∶ Unconditionally stable algorithm
Thank you