DYNA

Dynamic Analysis

SOFiSTiK | 2022
 DYNA
Dynamic Analysis

DYNA Manual, Service Pack 2022-11 Build 218

Copyright © 2023 by SOFiSTiK AG, Nuremberg, Germany.

SOFiSTiK AG

HQ Nuremberg Office Garching

Flataustraße 14 Parkring 2
90411 Nürnberg 85748 Garching bei München
Germany Germany

T +49 (0)911 39901-0 T +49 (0)89 315878-0

F +49(0)911 397904 F +49 (0)89 315878-23
[email protected]
www.sofistik.com

This manual is protected by copyright laws. No part of it may be translated, copied or reproduced, in any form or by
any means, without written permission from SOFiSTiK AG. SOFiSTiK reserves the right to modify or to release
new editions of this manual.

The manual and the program have been thoroughly checked for errors. However, SOFiSTiK does not claim that
either one is completely error free. Errors and omissions are corrected as soon as they are detected.

The user of the program is solely responsible for the applications. We strongly encourage the user to test the
correctness of all calculations at least by random sampling.

Front Cover
Arnulfsteg, Munich Photo: Hans Gössing
 Contents | DYNA

Contents

Contents i

1 Task Description 1-1

2 Theoretical Principles 2-1

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
2.2 Direct Time History Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
2.2.1 Duhamel’s Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
2.2.2 Integration of the Equations of Motion . . . . . . . . . . . . . . . 2-3
2.3 Modal Time History Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
2.3.1 Modal Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
2.3.2 Eigenvalue Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6
2.3.3 Generalized Modal SDOF System . . . . . . . . . . . . . . . . . . 2-7
2.3.4 Computation of the Modal Damping . . . . . . . . . . . . . . . . . 2-8
2.3.5 Solution of the Modal Equations . . . . . . . . . . . . . . . . . . . 2-9
2.4 Modal Steady-State Analysis to Harmonic Excitation . . . . . . . . . . . . . 2-10
2.5 Excitation through a Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
2.5.1 Earthquake Response Spectra . . . . . . . . . . . . . . . . . . . . 2-12
2.5.2 Wind Spectrum Analysis . . . . . . . . . . . . . . . . . . . . . . . . 2-13
2.5.3 Superposition of the Maximum Responses . . . . . . . . . . . . 2-14
2.5.4 Sign of Corresponding Forces . . . . . . . . . . . . . . . . . . . . 2-14
2.6 Dynamic excitation as ground acceleration . . . . . . . . . . . . . . . . . . . 2-16
2.6.1 Effective earthquake forces . . . . . . . . . . . . . . . . . . . . . . 2-16
2.6.2 Uniform Support Excitation . . . . . . . . . . . . . . . . . . . . . . 2-17
2.6.3 Multiple Support Excitation . . . . . . . . . . . . . . . . . . . . . . 2-18
2.6.4 Large Mass and Large Stiffness Method . . . . . . . . . . . . . . 2-20
2.6.5 Baseline Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-21
2.6.6 Dynamic load as ground acceleration . . . . . . . . . . . . . . . . 2-22
2.6.7 Total Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23
2.6.8 Equivalent Modal Loads . . . . . . . . . . . . . . . . . . . . . . . . 2-23
2.6.9 Effective Modal Mass . . . . . . . . . . . . . . . . . . . . . . . . . . 2-24
2.6.10 Base Shear and Base Overturning Moment . . . . . . . . . . . . 2-25
2.7 Storey level definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-28
2.7.1 Computation of the center of rigidity (COR) of a diaphragm
floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-29
2.7.2 Computation of the center of mass (COM) of a storey . . . . . 2-31
2.7.3 Storey results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31
2.8 Kinematic Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31
2.9 Elastic Stiffnesses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31
2.10 Geometric Stiffness and P-delta . . . . . . . . . . . . . . . . . . . . . . . . . . 2-32
2.11 Nonlinear Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-32

SOFiSTiK 2022 i
DYNA | Contents

2.12 FAQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........ 2-33

2.12.1 Manual Checks of the RSA Results . . . . . . . . . ........ 2-33
2.12.2 Modal Participation Factor . . . . . . . . . . . . . . . ........ 2-36
2.12.3 Equivalent Modal Loads . . . . . . . . . . . . . . . . ........ 2-37
2.12.4 Scaling of the Generalized SDOF Properties for a Different
Eigenvector Normalization . . . . . . . . . . . . . . . ........ 2-38

3 Input Description 3-1

3.1 Input Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
3.2 Input Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
3.3 SYST – System Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3
3.4 CTRL – Calculation Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 3-4
3.4.1 SOLV Equation solver . . . . . . . . . . . . . . . . . . . . . . . . . . 3-8
3.4.2 CORE Parallel computation control . . . . . . . . . . . . . . . . . 3-10
3.5 GRP – Selection of Element Groups . . . . . . . . . . . . . . . . . . . . . . . 3-12
3.6 MAT – General Material Properties . . . . . . . . . . . . . . . . . . . . . . . . 3-14
3.7 BMAT – Elastic Support / Interface . . . . . . . . . . . . . . . . . . . . . . . . 3-15
3.7.1 Derivation of bedding constants from reference material . . . . 3-16
3.7.2 Rigid Frictional Interface . . . . . . . . . . . . . . . . . . . . . . . . 3-17
3.8 SMAT – SBFEM - Material Properties . . . . . . . . . . . . . . . . . . . . . . 3-19
3.9 MASS – Lumped Masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-20
3.10 EIGE – Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . . . . 3-22
3.11 MODD – Modal Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-24
3.12 STEP – Parameter of the Step-wise Integration . . . . . . . . . . . . . . . . 3-28
3.12.1 Static Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-28
3.12.2 Steady State Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 3-28
3.12.3 Frequency Dependent Steady State Analysis . . . . . . . . . . . 3-29
3.12.4 Transient Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-29
3.13 LC – Load Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-31
3.14 CONT – Contact and Moving Load Function . . . . . . . . . . . . . . . . . . 3-33
3.15 HIST – Results within Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-35
3.16 EXTR – Evaluation of Max. Internal Forces and Moments . . . . . . . . . 3-38
3.17 ECHO – Extent of Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-43

4 Output Description 4-1

4.1 Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.2 Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.3 General Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.4 Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.5 Natural Frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2
4.6 Load Cases, Functions and Loads . . . . . . . . . . . . . . . . . . . . . . . . 4-2
4.7 Displacements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2
4.8 Internal Forces and Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2
4.9 Time Variations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3

Bibliography 4-5

ii SOFiSTiK 2022
 Task Description | DYNA

1 Task Description

The program DYNA can be used for static and primarily for dynamic analysis of three-
dimensional structures. It can perform the following tasks (Special licenses may be needed):

• Static analysis of load cases acting upon three-dimensional structures, as well as planar
and axisymmetric structures.
• Static analysis of load cases after second-order theory
• Computation of the natural frequencies of three-dimensional structures.
• Computation of the buckling eigenvalues of three-dimensional structures.
• Implicit direct integration of the equations of motion for structures with arbitrary damping
• Explicit direct integration of the nonlinear equations of motion
• Interaction with load trains and wind loading
• Soil structure interaction with the SBFEM
• Integration of the equations of motion by superposition of the mode shapes.
• Steady-state oscillations and excitation through spectra.

The static system is stored in the database after its generation e.g. by the program SOFiMSHA,
SOFiMSHC or SOFiPLUS.

The following elements can be processed by DYNA:

• Point masses (with off diagonal components)

• Three-dimensional prismatic bending beams with haunches, warping torsion and bedding
profiles (piles)
• Truss- and cable elements
• Spring elements, boundary and FLEX elements
• Scaled boundary finite elements (SBFEM)
• Damping elements
• Shell elements
• 3D-solid elements

For the explicit integration not all features are supported. Only the truss, cable, spring and
the 3D volume (BRIC) element are available. All interactions (Wind, Loadtrain, Soil) are not
available. But geometric and material nonlinearity are supported.

The results of the dynamic analysis including the mode shapes are stored in the database as
displacements and stresses with a load case number.

The mode shapes can also be transferred from the database after a calculation with the pro-
gram ASE.

For the purposes of a dynamic analysis, the program may output the maximal and the minimal
of all displacements, velocities or accelerations as well as internal forces and moments, and
eventually the time variation of selected degrees of freedom or internal forces and moments.

SOFiSTiK 2022 1-1

DYNA | Task Description

For speed reasons almost all algorithms follow what is called IN-CORE solutions. The size
of the problem is therefore limited by the amount of available main memory. Modal solutions
transferring the eigenvalues from ASE are not subjected to this limitations.

1-2 SOFiSTiK 2022

 Theoretical Principles | DYNA

2 Theoretical Principles

2.1 Introduction

Equations of Motion
The most basic dynamic system is a an elementary single degree of freedom dynamic system
- the mass-spring-damper system depicted in Fig. 2.1.

(t)
k
k · (t) m · ̈(t)
m p(t) p(t)
c · ̇(t)

Figure 2.1: Mass-spring-damper system

One way of deriving the governing linear equation of motion of this linear single degree of
freedom system (SDOF) is using the D’Alembert’s principle, to assume viscous damping and
writing the balance of forces acting on the mass in motion (Fig. 2.1)

m · ̈(t) + c · ̇(t) + k · (t) = p(t) , (2.1)

where (t), ̇(t) and ̈(t) are displacement, velocity and acceleration, while k , c and m are
stiffness, damping and mass of the SDOF system, respectively. p(t) is the external excitation
force.

Quite often the same equation is reformulated in a so called modal form:

p(t)
̈(t) + 2ξω · ̇(t) + ω2 · (t) = , (2.2)
m
p
where ω = k/ m is the eigenfrequency, while ξ = c/ (2 · m · ω) is the modal damping, which is
also the fraction of the so called critical damping.

Similarly, governing equations of motion of a linear multi degree of freedom (MDOF) system is
given by the following system of coupled 2nd order differential equations:

M · ̈(t) + C · ̇(t) + K · (t) = p(t) (2.3)

where (t), ̇(t) and ̈(t) are the displacement, velocity and acceleration vectors of the MDOF
system, respectively. M, C and K are mass, damping and stiffness matrices of the MDOF
system, respectively. p(t) is the time dependent loading vector.

Mass Damping and Stiffness Matrices

In general, all matrices have the same structure. It is common, however, to diagonalise the
mass and the damping matrix. This is permitted in most cases without large errors, but can not

SOFiSTiK 2022 2-1

DYNA | Theoretical Principles

be applied to rotational masses and kinematic constraints with rotational degrees of freedom.

The consistent mass matrices imposes a larger numerical effort and bigger memory require-
ments, but produce in general more accurate results. However there is another disadvantage.
The violation of the discrete maximum principle may lead in case of very small time steps to
oscillations of the solutions. The times steps in such case must not be smaller that the time
needed by the wave to propagate through a single element.

For these reasons the program provides an input entry for the type of matrix assembly in the
CTRL MCON record.

Dynamic Loading
The loading vector p(t) can be expressed as a sum of np different time-dependent loading vec-
tors pj (t), each of which represented by a product of space vector (load shape, load pattern)
p0,j , which is not a function of time, and a time function gp,j (t)

np
X np
X
p(t) = pj (t) = p0,j · gp,j (t) . (2.4)
j=1 j=1

Solution of the Equations of Motion

To obtain the solution of the governing equations of motion even for a SDOF system is very
demanding task. Only a few problems can be solved analytically. In general, a numerical solu-
tion is necessary. In the following sections some methods of solving the governing equations
of motion will be addressed.

If the number of equations is very large, it might be useful to transform the problem in an other
(modal) vector space. The extra benefit is then obtained if only the lowest significant modes
need to be used.

However, sometimes there are a lot of Eigenvectors not relevant for the response. The general
recommendation is to omit or reduce the masses of those eigenvalues (e.g. the vertical mass
of the slabs with the MASS command) to get rid of those unwanted eigenvalues.

But there exists also a solution technique called Ritz-Analysis as an alternate way. The pro-
cedure is identical to the Lanczos-Eigenvalue analysis, but instead of a general random static
load vector a specific load vector (e.g. acceleration in the global X direction) is used. Thus the
obtained base vectors are more relevant and the number of required vectors might become
much less.

But as the danger exists that essential contributions will be skipped, it is not recommended to
use this type of analysis if the required percentage of modes can be obtained with a classical
eigenvector analysis, which is the accurate solution.

2-2 SOFiSTiK 2022

 Theoretical Principles | DYNA

2.2 Direct Time History Analysis

2.2.1 Duhamel’s Integral

One way to determine the response of a linear SDOF system to an arbitrary excitation force,
whose governing equation is described by Eq. 2.1, is provided by the Duhamel’s integral:
Zt
1
(t) = p(τ) · e−ξω(t−τ) · sin [ωD (t − τ)] · dτ , (2.5)
mωD 0
Æ
where ωD = ω 1 − ξ2 represents the resonance circular frequency of damped vibration.

This solution is limited to linear SDOF systems because it is based on the superposition princi-
ple. It is relatively often used for the solution of the SDOF equation for purposes of determining
the response spectrum. Generally this integral needs to be solved numerically, but the meth-
ods used for its solution are not very efficient [1]. More efficient methods for numerical solution
of the differential equations of motion are presented in the subsequent subsection.

2.2.2 Integration of the Equations of Motion

In the most general case, the equations of motion (Eq. 2.6) need to be solved numerically
by the direct integration of these coupled 2nd order differential equations. In order to do so,
firstly the time is subdivided into discrete time steps t ∈ {t1 , . . . , tn } and then the continuous
dynamic equations are discretized, i.e. written in the the discrete form for the time t+1

M · ̈(t+1 ) + C · ̇(t+1 ) + K · (t+1 ) = p(t+1 ) , (2.6)

where  = 1, . . . , n − 1 and t+1 = t + Δt (time step Δt is usually kept constant). This problem
now becomes a problem of solving the equations of motion for the time t+1 based on the
known solution for the previous discrete times, i.e. t ≤ t , including the initial conditions. As
aforementioned the solution can be achieved by the integration, i.e. by using the following
equations (for simplicity sake we drop the vector notation)
Z Δt
̇(t+1 ) = ̇(t ) + ̈(τ)dτ (2.7a)
0
Z Δt
(t+1 ) = (t ) + ̇(τ)dτ (2.7b)
0

which express the velocity and displacement at the end of the time step Δt = t+1 − t in terms
of the known values at the beginning of the time step ̇(t ) and (t ) with the addition of the
integral quantities. Obviously in order to obtain the solution at the time t+1 we need to assume
how the acceleration ̈(τ) will vary within the time step.

The simplest form is to assume a constant acceleration during each time step (Fig. 2.2, left).
However a more enhanced precision is obtained using a linear approximation for the accelera-
tion (Fig. 2.2, right)
τ
̈(τ) = ̈(t ) + · [ ̈(t+1 ) − ̈(t )] . (2.8)
Δt

According to the Newmark method the following expressions hold for the velocity and the dis-

SOFiSTiK 2022 2-3

DYNA | Theoretical Principles

1 ̈+1 − ̈
̈(τ) = ̈ = (̈ + ̈+1 ) ̈(τ) = ̈ + τ
2 Δt

̈+1 ̈+1
̈ ̈

τ ̈+1 − ̈ τ 2
̇(τ) = ̇ + (̈ + ̈+1 ) ̇(τ) = ̇ + ̈ τ +
2 Δt 2

̇+1 ̇+1
̇ ̇

τ2 τ2 ̈+1 − ̈ τ 3
(τ) =  + ̇ τ + (̈ + ̈+1 ) (τ) =  + ̇ τ + ̈ +
4 2 Δt 6

+1 +1

 

τ τ

Δt Δt
t t+1 t t+1

Figure 2.2: Avarage acceleration (left) and linear acceleration (right) methods

placement at the end of the time interval Δt :

̇ (t+1 ) = ̇(t ) + Δt · [ (1 − δ) · ̈(t ) + δ · ̈ (t+1 )] (2.9a)

 (t+1 ) = (t ) + Δt · ̇(t ) + Δt 2 · [ (1/ 2 − β) · ̈(t ) + β · ̈ (t+1 )] (2.9b)

̈(t)

̈+1
̈
t
t t + (1 − α)Δt t+1 t + θΔt

Δt

Figure 2.3: Integration in time - α and Wilson-θ methods

Then we have the choice between five different possibilities to select how or for which time
t + θ · Δt the equilibrium equation is fulfilled:

• Explicit Integration (θ = 0.0)

As the variation of the acceleration along the time step is constant it is possible to calculate
all values highly efficient explicitly by the mass matrix. The method is well suited to be
parallelized, but it is only stable if the time step is below a certain critical value which is

2-4 SOFiSTiK 2022

 Theoretical Principles | DYNA

approximately equal to the wave speed divided my the minimum mesh size.
• Newmark-Method (θ = 1.0, δ ≥ 0.50, β ≥ 0.25(0.5 + δ)2 )
The default of the parameters has no numerical damping at all. Thus small errors may
enlarge especially for the accelerations. In those cases the value of δ should be enlarged1 .
• Wilson-Theta-Method (θ ≥ 1.37)
This value is a modification of the Newmark method where the numerical damping enlarges
the period to a greater extent, but keeps the amplitudes to a higher accuracy (Abb. 2.3).
In the literature the parameter β is given as α but this has been changed to avoid conflicts
with the next method.
• α -Method Hilber-Hughes-Taylor (θ < 1.0, δ = (1 − 2α)/ 2, β = (1 − α)2 / 4)
This method (Abb. 2.3) has been developed to introduce a numerical damping without
degrading the order of accuracy. It is especially suited for non linear problems. The value
α is taken from the input value θ as α = (θ − 1.0). Thus we have a formal equivalent to the
Crank-Nicholson method (see program HYDRA).
• Modal Analysis
The system of equations to be solved can be significantly simplified if the solution is calcu-
lated in the subspace of a few eigenvectors. This requires knowledge of the eigenvalues
and the eigenvectors, the calculation of which is relatively extensive. But then it is possible
to integrate the linear equations exactly. Nonlinear effects may be treated in a simplified
way if the modes contain the nonlinear displacement possibilities.

The implicit methods lead to a system of equations for the displacements or accelerations at
t + Δt . Compared to the explicit methods, the time step may be chosen considerably larger.
Specific errors (oscillations) may be introduced however by a time step chosen too small with
a consistent mass matrix.

For more info about the integration methods please consult other literature sources (e.g. [2],
[3], [4], [5], etc.).

2.3 Modal Time History Analysis

2.3.1 Modal Decomposition

For a linear system, the displacement vector of an n degrees of freedom system (t) can be
expressed as the superposition of the modal contributions (modal decomposition):
n
X n
X
(t) = k (t) = ϕk · qk (t) =  · q(t) (2.10)
k=1 k=1

where:

k (t) contribution of the k th mode to the displacement vector (t) (k th modal displace-
ment vector),
ϕk k th natural mode shape (eigenvector),
qk (t) generalized modal displacement (coordinate) for mode k (modal coordinate re-
sponse),
matrix of mode vectors,  = ϕ1 · · · ϕn ,
 

1 In the literature parameter δ is often given as γ.

SOFiSTiK 2022 2-5

DYNA | Theoretical Principles

q(t) vector of generalized modal coordinates, q(t) = [q1 (t) · · · qn (t)] T .

In other words, the displacement vector (t) has been expressed as a linear combination of
the time independent eigenvectors ϕk with the time dependent scalars qk (t).

Substituting the Eq. 2.10 into the equations of motion of the MDOF system (Eq. 2.6) and
premultiplying everything with T we obtain the following

T · M ·  · q̈(t) + T · C ·  · q̇(t) + T · K ·  · q(t) = T · p(t) . (2.11)

At the first glance, the Eq. 2.11 is very similar to the Eq. 2.6. However, by a certain choice of
the eigenvectors , the matrices T · M · , T · C ·  and T · K ·  will become diagonal, which
in turn means that the system of COUPLED second order differential equations of motion will
be transformed into a system of UNCOUPLED second order differential equations.

2.3.2 Eigenvalue Problem

One way to determine the eigenvectors  so that the decoupling takes place is to start from
the equations of motion for a freely vibrating undamped system

M · ̈(t) + K · (t) = 0 . (2.12)

The solution of this equation is given by the harmonic function

(t) = ϕ · sin (ωt + φ) . (2.13)

Substituting Eq. 2.13 into Eq. 2.12 we obtain the so-called eigenvalue problem
” —
K − ω2 M ϕ = 0 , (2.14)

in which ω represents the circular eigenfrequency and ϕ is the natural mode shape. λ = ω2 is
called the eigenvalue. This set of n homogeneous algebraic equations has a nontrivial solution
if the following is satisfied
” —
det K − ω2 M = 0 . (2.15)

This equation has n roots {ω1 , . . . , ωn } which represent the circular eigenfrequencies (natural
circular frequencies) of the n modes of vibration. Once the circular eigenfrequency ωk is
known, the corresponding eigenvector ϕk could be determined from the Eq. 2.14.

The computation of the eigenvalues is done either by a simultaneous inverse vector iteration
or directly by the Lanczos method. The algorithms find the eigenforms ϕk and the Eigenvalues
are obtained by the Rayleigh coefficient.

ϕT

· K · ϕ
ω2

= . (2.16)
ϕT

· M · ϕ

One benefit of this approach is that duplicate eigenvalues do not impose any problem for the
solution.

Hint
It is obvious from the eigenvalue problem equation 2.14 that the absolute amplitude of

2-6 SOFiSTiK 2022

 Theoretical Principles | DYNA

(Hint continued...)

the vectors ϕk are not known; only the shape of the vectors are known. In other words,
if a vector ϕk is an eigenvector satisfying the Eq. 2.14, then a vector α · ϕk is also an
eigenvector, where α is a non-zero real scalar value.

A very important property of the eigenvectors is that they satisfy the following orthogonality
conditions

ϕT · M · ϕj = 0 ,  6= j (2.17a)
ϕT · K · ϕj = 0 ,  6= j . (2.17b)

2.3.3 Generalized Modal SDOF System

A very important issue is that in general the important properties of a structure may be de-
scribed by a few eigenforms, generally those with the lowest eigenvalues. We can take this like
a shadow of the full solution to a reduced vector space. So the number n in eq. 2.10 may be
selected with a very low value if no local effects are dominant.

Using the Eqs. 2.17, the Eq. 2.11 can be rewritten, and after singling out the k th row we get

ϕTk · M · ϕk · q̈k (t) + ϕTk · C · ϕk · q̇k (t) + ϕTk · K · ϕk · qk (t) = ϕTk · p(t) , (2.18)

which is a scalar SDOF equation for a mode k . Hence the orthogonality property has enabled
us to uncouple the equations of motion.

Hint
In order to uncouple the equations of motion for a system with damping, it is obvious that
the damping matrix C must be such that T · C ·  is a diagonal matrix. Which form should
C take in order for this to be possible is explained in the Section 2.3.4.

We can simplify the Eq. 2.18 to obtain

mk · q̈k (t) + ck · q̇k (t) + kk · qk (t) = pk (t) (2.19)

where the properties of a this k th -mode generalized SDOF system are:

mk = ϕTk · M · ϕk generalized modal mass for mode k ,

ck = ϕTk · C · ϕk generalized modal damping for mode k ,
kk = ϕTk · K · ϕk generalized modal stiffness for mode k ,
pk (t) = ϕTk · p(t) generalized modal load for mode k .

Hint
In SOFiSTiK eigenvectors ϕk are normalized in such a way that the generalized modal
mass mk is equal to 1 , for each natural mode k , i.e.

mk = ϕTk · M · ϕk = 1 . (2.20)

SOFiSTiK 2022 2-7

DYNA | Theoretical Principles

(Hint continued...)

As a consequence, the generalized modal stiffness kk is given by:

kk = ϕTk · K · ϕk = ω2
k
. (2.21)

Taking Eq. 2.4 into an account, generalized modal load for mode k can be written as
np
X np
X
pk (t) = ϕTk · p(t) = ϕTk · pj (t) = ϕTk · p0,j · gp,j (t) =
j=1 j=1
np np
(2.22)
X X
= p0,kj · gp,j (t) = pkj (t) ,
j=1 j=1

where

pkj (t) = p0,kj · gp,j (t) (2.23)

represents generalized modal load for mode k corresponding to the load function j, while

p0,kj = ϕTk · p0,j (2.24)

is defined as the modal load participation factor 2 for mode k corresponding to the load function
j.

Dividing everything with mk we get

pk (t)
q̈k (t) + 2ξk ωk · q̇k (t) + ω2
k
· qk (t) = , (2.25)
mk

where:

ωk =
p
kk / mk undamped circular eigenfrequency for mode k ,
ξk = ck / (2mk ωk ) viscous modal damping ratio for mode k .

2.3.4 Computation of the Modal Damping

If we have a damping the eigenvalues become complex numbers with an independent imagi-
nary part. However there is no strict need to use complex eigenvalues. We can use the real
eigenvalues in nearly all cases. Formally the description of a system by its eigen forms is just a
transformation into an other vector space. In order for the damping matrix to become diagonal
too, damping could be selected to be proportional to the mass and/or the stiffness matrix:

C=·M+b·K (2.26)

When (2.26) is used, the resulting modal damping ξ (Lehr’s damping factor) is also a diagonal
matrix:
1  δk
 
ξk = · + b · ωk = (2.27)
2 ωk 2π

In a complex system the individual elements may have quite different damping properties.
The proportionality of the damping is then no longer given and the damping matrix C does
2 Modal load participation factors p0,kj = ϕTk · p0,j are outputted in Table 3.34.

2-8 SOFiSTiK 2022

 Theoretical Principles | DYNA

not become a diagonal matrix. One possibility would be to account for this exactly during the
integration of the modal equations. But as the damping is small in most cases and the damping
properties are the most uncertain material properties in the total process, it is possible to use
a simplified diagonal modal damping.

To obtain this, each group of elements may be assigned individual Rayleigh factors A and B. For
an eigenvalue analysis, a modal group damping value D may be used to add for every eigen
frequency a stiffness proportional equivalent damping. Then the complete modal damping
matrix will be calculated, printed if requested and the individual energetic equivalent modal
damping will be calculated or simplified according EN 1998-2 4.1.3 (1) by using the diagonal
entries only for every individual eigen frequency.

The same approach is used for aeroelastic effects.

2.3.5 Solution of the Modal Equations

The SDOF Eq. 2.25 can be solved analytically (if such solution exists) or what is more com-
mon by numerical integration using the Duhamel’s integral or some other numerical integration
technique, as it has been described in Section: Integration of the Equations of Motion. Solving
the SDOF Equations 2.25 means determining the modal coordinate responses qk (t) for each
mode k over time 3 .

Once the solution of the SDOF Equation 2.25 is known, the total displacement response can
be obtained by the back transformation from the modal to physical space using the Eq. 2.10,
which is repeated here for completeness sake
n
X n
X
(t) = k (t) = ϕk · qk (t) . (2.28)
k=1 k=1

In fact, any response quantity R(t) (e.g. cross-sectional forces, stresses, etc.) can be obtained
by the linear combination of the corresponding modal response quantities Rk0 and modal co-
ordinate responses qk (t), i.e.
n
X n
X
R(t) = Rk (t) = Rk0 · qk (t) , (2.29)
k=1 k=1

where Rk (t) represents a k th mode contribution to a response quantity R(t). The modal re-
sponse quantities Rk0 are the quantities which correspond to the eigenvector displacements
ϕk . They are determined by the standard static FEM techniques during the computation of
the eigenvalues and eigenvectors directly from the eigenvector displacements ϕk . This is the
approach which is used in SOFiSTiK.

There is another possibility to determine the response quantities R(t). Once the k th modal
coordinate response qk (t) is known, we can at each given time t compute the internal elastic
forces associated with the k th mode ƒ k (t) as follows

ƒ k (t) = K · k (t) = K · ϕk · qk (t) . (2.30)

Alternatively, by utilizing the eigenvalue problem equation 2.14, i.e. Kϕk = ω2 Mϕk , we can
3 Maximal modal coordinate responses mx |qk (t)| are outputted in Table 3.38.

SOFiSTiK 2022 2-9

DYNA | Theoretical Principles

express these forces in terms of mass matrix

” —
ƒ k (t) = M · ϕk · ω2 · qk (t) . (2.31)

By performing a static analysis at a given time by using these forces as external forces we
obtain Rk (t), and finally by superimposing these using
n
X
R(t) = Rk (t) , (2.32)
k=1

we obtain the total response quantity R(t). The forces given by Eq. 2.30 or 2.31 are also called
as equivalent modal forces (equivalent modal load vector).

The advantage of the direct approach used by SOFiSTiK compared to the equivalent modal
forces approach is that the additional effort to compute the total responses by introducing
new load cases is small. All one needs to do in that case is to compute the modal coordinate
responses qk (t) corresponding to the new load cases and perform the superposition according
to Eq. 2.29.

2.4 Modal Steady-State Analysis to Harmonic Excitation

If a loading is in a form of a harmonic excitation with the amplitude vector p0 and the excitation
circular frequency Ωp , i.e.

p (t) = p0 · sin Ωp · t − φp ,


(2.33)

the solution of the modal generalized SDOF equation given by Eq. 2.19 can be found in an
analytical form. This solution qk (t) comprises of a transient (homogeneous solution, qk,h ) and
a steady-state part (particular solution, qk,p ). In the presence of damping (all real mechanical
systems have damping) the transient part decays with time and, depending on level of damp-
ing, can be damped out after only couple vibration cycles. What persist then is the stead-state
response which has the same frequency as the excitation force but a different phase

qk,p (t) = q0,k,p · sin Ωp · t − φk = qk,st · RD (ηk ) · sin Ωp · t − φk ,

(2.34)

where (see also Eqs. 2.24)

p0,k ϕTk · p0
qk,st = = (2.35a)
k ω2
k
·m
Ωp
ηk = (2.35b)
ωk
q0,k,p 1
RD (ηk ) = =s (2.35c)
qk,st € Š2
1− η2
k
+ (2ξk ηk ) 2
!
2ξk ηk
φk (ηk ) = φp + rctn . (2.35d)
1 − η2
k

qk,st is the static deformation, ηk is the ratio of the loading frequency to the natural frequency
and φk is the phase angle of the k th -mode generalized SDOF system. The ratio of the har-
monic response amplitude q0,k,p and the static displacement qk,st (resulting from the statically
applied force p0,k ) is called the dynamic response magnification factor RD (ηk ). RD (η) and

2-10 SOFiSTiK 2022

 Theoretical Principles | DYNA

φ(η) are show on Fig. 2.4.

RD (η)

ξ = 5%
5

10%
4

25%
2

50%
1
100%

0 η = ωp / ω
1 2 3 4

(a) Dynamic response magnification factor RD (η).

φ(η)
ξ = 5%
180◦
10%

25%
50%
100%
90◦

0 η = ωp / ω
1 2 3 4

(b) Phase angle φ(η).

Figure 2.4: Dynamic response magnification factor and phase angle for a damped system excited
by harmonic force.

These classical response functions have a region below the resonance frequency where the
structure follows the loading with a dynamic magnification factor and a region above the reso-
nance where it is no longer possible for the structure to follow the loading, yielding in a steady
decay of the counter phase response until zero for high frequencies.

These response functions yield the true response including the shift of the resonance peak due
to damping effects. All frequencies used in DYNA are always there for those of the undamped
oscillation.

The transient oscillations may be accounted for based on zero initial conditions by an additional
r-multiple component. The superposition of these oscillations results in a floating effect, which
can be accurately registered by time integration. DYNA can selectively omit this component or
add it if it’s unfavourable.

An accurate calculation of the maximum stressing taking into consideration the phase shift can
be carried out only for the final transient oscillation state by neglecting the transient compo-
nents. if multiple load cases should be combined in this way, the extreme values have to be
searched in the time interval of one periodic response (STEP -n). Thus it is clearly to identify
which loadings are synchronized (like vertical and horizontal out of balance forces) and which

SOFiSTiK 2022 2-11

DYNA | Theoretical Principles

are to be combined based on stochastic. For every single run only one of these methods can
be applied. For the case of multiple machinery a superposition in MAXIMA with SRSS has to
be used for the intermediate results obtained with STEP -n.

Otherwise, and especially if the phase differences are small a statistical superposition as in the
case of spectrum excitation can be carried out.

2.5 Excitation through a Spectra

Dynamic loading which is not deteministic may be accounted for by a spectrum. This is a func-
tion giving any response value based on the frequency for an SDOF. Spectra may be obtained
from a given load-time function or establishes as a power spectrum. They are commonly used
for wind and waves and especially for earthquake response.

2.5.1 Earthquake Response Spectra

If the earthquake ground acceleration function ̈gX (t) is known, the solution of the equation
of motions can be obtained by a numerical integration (see Section 2.2). However, in most
practical applications the seismic excitation is usually represented by the earthquake (elastic
or inelastic) response spectrum. The maximal response of the SDOF is then given by
EQX EQX EQX
mx Dk (t) = SD (Tk , ξk ) = SD,k , (2.36a)
EQX EQX EQX
mx Ak (t) = SpA (Tk , ξk ) = SpA,k , (2.36b)

where SEQX EQX

D (Tk , ξk ) is the spectral relative displacement response and SpA (Tk , ξk ) is the true
acceleration spectral response of the k th -mode generalized SDOF system for the earthquake.

If the damping is small the true acceleration spectra SEQX

pA (Tk , ξk ) may be approximated by the
pseudo-acceleration spectra
EQX EQX
SpA (Tk , ξk ) ≈ ω2
k
· SD (Tk , ξk ) . (2.37)

Stresses and forces in the structure are based on the displacements, thus the pseudo-
acceleration spectra. But measurements are made on the true accelerations. Thus the spectra
to be defined is always expected to be a pseudo-acceleration spectra.

An example of a smooth norm-specific pseudo-acceleration and relative displacement spectra

are shown in the Fig. 2.5.

2-12 SOFiSTiK 2022

 Theoretical Principles | DYNA

SpA (T, ξ)/ g

SB

SA

T
TB TC TD TE

(a) Pseudo-acceleration response spectrum normalized w.r.t. ground acceleration

g
SD (T, ξ)

T
TB TC TD TE

(b) Relative displacement response spectrum

Figure 2.5: Smooth norm-specific response spectra

2.5.2 Wind Spectrum Analysis

For the wind there are various power spectra. Which can be defined with some parameters
in SOFILOAD: RESP. For the structural analysis the main loading is in the frequency range
between 0.1 and 1.0 Hertz. The best known spectrum is that from Karman.

ƒ ·S
σ2
1.0

0.1

0.01

ƒ ·L
0.001 U
0.001 0.01 0.1 1 10

Figure 2.6: Wind spectra according Karmann

For the wind the response is obtained by a background and a resonance response. While the
dynamic response is obtained from a normalized power spectrum, the background contribution
is always assumed to be 1.0, as this is on the safe side and is appropriate if the coherence

SOFiSTiK 2022 2-13

DYNA | Theoretical Principles

effects are introduced by the loading itself.

v
π2 ƒ ·S
u
t
q= 1+ ∗ (2.38)
2δ σ2
ƒ ·S 1 · X + 2 · X 2 + 3 · X 3
= (2.39)
σ2 (1 + b · X c )d
ƒ ƒ ·L ƒ ·z
X= = or (2.40)
ƒm V V

2.5.3 Superposition of the Maximum Responses

The maximum displacements and stresses must be superimposed according to probability
theory. The CQC (Complete Quadratic Combination) method by Wilson, or the Sum of the
Absolute Values or the Square Root of the Sum of Squares (SRSS) can be used for this
purpose. The SRSS method is known to be rather faulty in case of multiple eigenvalues,
but the CQC method is no guarantee for correct results either. It should be especially noted
that the method without modal dampings will yield different values to SRSS only for multiple
eigenvalues.
rX X
R= R · ρj · Rj (2.41a)

3/ 2
8 ξ ξj ξ + ηj · ξj · ηj
Æ
ρj = € Š2 € Š € Š (2.41b)
1 − η2
j
+ 4ξ ξj ηj 1 + η2
j
+ 4 ξ2
+ ξ2
j
η2
j
ωj
ηj = (2.41c)
ω

For the case of shear within a quadratic section it can be easily shown, which error is introduced
by the SRSS Method. The acceleration in X-direction exits the two diagonal Eigenforms with
the same amount of 25 % of the total shear.

Eigenfor m 2

a
= +

Eigenfor m 1 Eigenfor m 1 Eigenfor m 2

Figure 2.7: Shear within a quadratic section

The SRSS-Method yields 35 % for all 4 walls, while the CQC Method will give 50 % for the x-
walls and zero for the y-walls which is the correct value. If the maximum stress at the corner is
required, a different target function is required, the load cases max-My,min-My,max-Mz,min-Mz
do not cover this result.

2.5.4 Sign of Corresponding Forces

For every type of superposition yielding only positive values, the sign of the corresponding

2-14 SOFiSTiK 2022

 Theoretical Principles | DYNA

forces and moments should not be neglected. Although it is quite common to use positive
values for all results, this is not true and uneconomical in most cases.

For example if we look at a plane horizontally loaded framework, the internal forces and mo-
ments vary depending on the sign of the horizontal force, yet in every case it has to be observed
that the sign of the moment and normal force in one of the column are identical, while different
in the other one.

Figure 2.8: Plane framework

If we search for the maximum moment, then the associated normal forces must have different
signs. Exactly this is available from the mode shapes. If we now assemble the maximum
moment of different mode shapes, then we should thus always add the mode shapes only
completely with a global factor. When we intend to add the absolute values, then it is sufficient
that all mode shapes are either added or subtracted according to the sign of the leading force.
So we are replacing the rule of combination
X
SUMj = sj (2.42)


by the general form for the vector of internal forces and for the maximum value of force j:

X 
+1, sj ≥ 0
SUM = ƒ · S ; ƒ = (2.43)
 −1, sj < 0


The same can be used for the method SRSS (Square Root of Sum of Squares) the rule:
vX
u
SUMj = t sj 2 (2.44)


is replaced through
X sj
SUM = ƒ · S ; ƒ  = qP (2.45)
2
  sj

Last but not least the same can be done for the CQC-method. In any case the leading force

SOFiSTiK 2022 2-15

DYNA | Theoretical Principles

value will be positive hence it must be introduced as an alternating load in the final design
superposition.

The same method may be also used for the directional superposition for the most unfavourable
direction. Three load cases with accelerations in orthogonal directions may be analyzed to-
gether in a single DYNA run and combined with the SRSS method. The correct sign of the
corresponding forces will yield correct results, while the method used in other programs with
independent extreme values may yield results considerably to large.

Compared to this the Pseudo-superpositions R + ⊕0.30 · Ry which have to be calculated in

MAXIMA with alternate loading groups is much harder.

2.6 Dynamic excitation as ground acceleration

2.6.1 Effective earthquake forces

Seismic loading does not contribute to the external dynamic loading (i.e. right hand side of the
Eq. 2.6). Earthquake acts as an induced motion on the base of the structure (the part of the
structure in contact with the soil).

t (t)

(t) m · ̈t (t) =

m · ̈g (t) + ̈(t)

 
m

k · (t)
k, c c · ̇(t)

g (t)

Figure 2.9: SDOF subjected to ground motion g (t)

The total (absolute) displacement of the mass t (t) subjected to the earthquake- induced
ground motion g (t) is related to the relative displacement between the ground and the mass
(t) by the following relation (Fig. 2.9)

t (t) = g (t) + (t) . (2.46)

The elastic and damping forces depend solely of the relative motion (t), i.e. the rigid-body
motion g (t) contributes nothing to the internal structural forces. Having this in mind and taking
the Eq. 2.46 into consideration while assuming no external loading is applied to the structure,
the equation of motion of a SDOF system subjected to earthquake excitation can be derived
using the concept of dynamic equilibrium (Fig. 2.9)

m · ̈t (t) + c · ̇(t) + k · (t) = m · ̈g (t) + ̈(t) + c · ̇(t) + k · (t) = 0 ,

 
(2.47)

2-16 SOFiSTiK 2022

 Theoretical Principles | DYNA

where the inertial force is related to the total acceleration of the mass ¨t (t). Since the
earthquake-induced ground accelerations are generally regarded as known g (t), we can rear-
ranging the Eq. 2.47 slightly so that the unknowns remain on the left hand side of the equation

m · ̈(t) + c · ̇(t) + k · (t) = −m · ̈g (t) ≡ peƒ ƒ (t) . (2.48)

In this equation the term on the right hand side, i.e. peƒ ƒ (t) = −m · ̈g (t), represents the
effective earthquake force (or more general effective support excitation force).

Comparing the Eqs. 2.47 and 2.48 we can conclude that the structural response (relative
displacement (t)) caused by the ground acceleration ̈g (t) is equivalent to the response of
the structure with stationary base subjected to an external load equal to −m · ̈g (t) (Fig. 2.10).

t (t) = g (t) + (t) (t)

peƒ ƒ (t) = −m · ̈g (t)

̈g (t) stationary base

Figure 2.10: Effective earthquake force peƒ ƒ (t)

2.6.2 Uniform Support Excitation

Assuming uniform support excitation (all the points where the structure is connected to the
ground undergo identical prescribed motion) we can extending the previously derived concept
from SDOF to MDOF system and we can write

peƒ ƒ (t) = −M · ̈g (t) (2.49)

where peƒ ƒ (t) represents the vector of effective earthquake forces, while ̈g (t) is the ground
acceleration vector.

Choosing a vector basis (X , Y , Z ) we can represent the ground acceleration vector as a linear
combination of these three vectors as follows

̈g (t) = X · ̈gX (t) + Y · ̈gY (t) + Z · ̈gZ (t) , (2.50)

where ̈gX (t), ̈gY (t) and ̈gZ (t) are the three components of the ground acceleration in the
Cartesian coordinate system (X, Y, Z). The vectors X , Y and Z are called the influence
vectors and they represent the displacements of the masses resulting from static application of
a unit displacement in X, Y or Z direction to structural points in contact with the ground (support

SOFiSTiK 2022 2-17

DYNA | Theoretical Principles

points) (see Fig. 2.11).

m1 m2 m3

1 3 5  
1 

 

 

 
2 4 6
 

 


 2 



 

3
 
X =
 4 

 


 

 
Y 

 

5 


 

 
6
 

̈gX (t)

̈gY (t)

1 = 1 3 = 1 5 = 1

  2 = 1 4 = 1 6 = 1  
1  0 

  
 

 
 
 


 
 
 

0 1

 
 
 


 
 
 


 
 
 

1 0
   
X = Y =
 0 

 
  1 

 


 
 
 

   
1 0

 
 
 


 
 
 


 
 
 

0 1
   

gX ≡ 1 gX ≡ 1

gY ≡ 1 gY ≡ 1

Figure 2.11: (top) Frame with two DOFs per node subjected to uniform ground acceleration in X
and Y direction. (bellow) Influence vectors X and Y as static displacements due to gX = 1 and
gY = 1, respectively.

The load which is defined with the command ACCE NO 0 in SOFiLOAD will be interpreted as
the effective support (earthquake) excitation load. For example the input of ACCE TYPE NODA NO
0 AX 1.0 AY 0 AZ 0 will be interpreted as the effective support (earthquake) excitation load
peƒ ƒ ,X (t) = −M · X · ̈gX (t), where the ̈gX (t) is defined with the FUNC command. For more
details about this please refer to SOFILOAD: ACCE.

2.6.3 Multiple Support Excitation

In some cases, e.g. for the structures with large dimensions (bridges, pipelines, tunnels, dams,
etc.) founded on a strongly non-uniform soil, the assumption of uniform seismic excitation on
all structural points in contact with the ground, i.e. disregarding the variation of the seismic
waves in space and time, is sometimes not justified.

In these cases it is necessary to take these effects into consideration by exciting the structure
in multiple supports with different ground acceleration functions ̈g, (t), where  = 1, · · · , ng
and ng is the number of supports.

Under the assumption of linear systems and with some simplifications (neglecting contribution
of the damping and assuming that the mass matrix is predominately diagonal) the effective
force vector for multiple support excitation can be represented in a similar way as in case of

2-18 SOFiSTiK 2022

 Theoretical Principles | DYNA

uniform support excitation (Eqs. 2.49 and 2.50), i.e.

ng
X ng
X
peƒ ƒ (t) = peƒ ƒ , (t) = − M ·   · ̈g, (t) . (2.51)
=1 =1

The only difference to Eqs. 2.49 and 2.50 is that now for each ground support  we have one
influence vector  , whereas in the case of uniform support excitation we had one vector (per
earthquake direction) for all supports.

The influence vector  can be obtained by the static procedure by applying the unit displace-
ment at the supports g, ≡ 1 and determining the resulting structural displacements which
then represent the influence vector  . An example of determining influence vectors is shown
on Fig. 2.12.

 
m1 m2 m3 m4 
 1 


 


 

3 5 7 2 

 
 
1

 

 
3 

 

 

X
 
2 4 6 8 

4


=
5 

 
Y

 


 

 6 

 
 


 

 


 7 



 

8
 
̈g,1 (t) ̈g,2 (t) ̈g,3 (t)

   


 0.553 




 0.255 



 
 
 

   
0.017 −0.010 

 
 
 

 
 
 


 
 
 

0.551 0.257 

 
 
 

 
 
 


 
 
 

−0.017 0.013
   
1 = 2 =
0.533  0.267

  
 


 
 
 


 
 
 

−0.004 0.000

 
 
 


 
 
 


 
 
 

  
 0.264 


 0.524 

 





 
 
 

0.005 −0.002
   

g,1 ≡ 1 g,2 ≡ 1

Figure 2.12: (top) Frame with two DOFs per node subjected to multi support ground acceleration.
(bellow) Influence vectors 1 and 2 as static structural displacements due to support displacements
g,1 = 1 and g,2 = 1, respectively.

For example, assuming that the structure is supported in points 1, 3 and 5 we can define three
load cases with loading in form of a support displacements of those points and once these
load cases are computed statically, the displacement vectors of the structure correspond to
the influence vectors for those support points. For more details about the derivation of the
equations for effective forces in case of multiple support excitation, see for example Chopra
[1], Section 9.7.

Another more general way to realize the multiple support excitation is to use the Large Mass
or Large Stiffness Method, as it is described in the Section 2.6.4. These are also applicable for
a nonlinear analysis.

SOFiSTiK 2022 2-19

DYNA | Theoretical Principles

2.6.4 Large Mass and Large Stiffness Method

It is often the case that the user wants to apply the dynamic loading directly as an enforced (pre-
scribed) motions (accelerations or displacements) and perform the direct time history analysis.
One of the ways to achieve this is to use the Large Mass (LMM) or Large Stiffness Method
(LSM). In these methods an element with large mass or stiffness is placed internally at the
nodes where the accelerations or displacements are to be prescribed. These points are then
excited by the large force which is equal to the product of the large mass with prescribed
accelerations or large stiffness with prescribed displacements.

To illustrate how this numerical approach works, let us imagine that we want to excite the
structure by prescribing the accelerations ̈g, at a degree of freedom . We place a large mass
Mrge at this degree of freedom  and at the same time apply a large force equal to Mrge · ̈g,
to the same DOF
     
m11 ··· m1 ··· m1n 

 ̈ 1





 p1 


 . .. ..  
  .   .. 
 . 
.
  
 . . .   . .
 
 
 

  
     
 m1

··· Mrge ··· mn   ̈  +
· ··· = Mrge · ̈g,  (2.52)
 . .. ..   .  
..
 . .. 
 
   
 . . .   .
 
 
 


 
 
 

   
mn1 ··· mn ··· mnn ̈n pn
   

Choosing now only the equation for the th -DOF and rearranging it so that the term associated
with the Mrge is singled out form the left hand side (LHS)

Mrge · ̈ + (rest oƒ the LHS) = Mrge · ̈g, , (2.53)

and dividing everything with Mrge we obtain

(rest oƒ the LHS)

̈ + = ̈g . (2.54)
Mrge

Assuming that the Mrge is so large that (rest oƒ the LHS) / Mrge << 1 holds, we finally end
up with

̈ ≈ ̈g . (2.55)

To sum up - by prescribing the large mass Mrge on a given DOF and simultaneously applying
a large force equal to Mrge · ̈g , we have practically prescribed the acceleration ̈g to the DOF
in question.

Large Mass Method is activated when the user by the definition of the ACCE TYPE NODA NO
command in SOFiLOAD specifies the node number different than zero for the option NO. DYNA
will then automatically place a large mass at the defined node and apply the large force equal
to the product of this large mass and the defined acceleration function.

The Large Stiffness Method works in the same way only instead of large mass, large stiffness
is used. For this method to work, the user needs to use the SOFiLOAD command NODE TYPE
WXX/WYY ... together with the definition of the displacement function.

LMM and LSM techniques are very useful in the case of the multiple support excitation, es-

2-20 SOFiSTiK 2022

 Theoretical Principles | DYNA

pecially in case of the non-linear systems where the equivalent force method is very difficult
to realize. One should keep in mind that these techniques, unlike effective force methods,
produce absolute accelerations and displacements (rigid body + relative movement).

2.6.5 Baseline Correction

In case where the dynamic loading is defined as the prescribed acceleration of the ground
(for example using the Large Mass Method) the acceleration time-history is often given by the
recorded accelerogram. These acceleograms are obtained from the seismograph, which is
triggered after a some small prescribed level of motion has been reached. This means that the
accelerogram contains an error which when integrated results in a linear error in velocity and
a quadratic error in displacements [6].

To correct these errors a best-fit parabola (baseline correction Δg (t)) can be applied to the
original acceleration function g (t) in order to obtain new, corrected acceleration function ̄g (t)
which, when integrated, will result in zero displacements and zero velocities at a given baseline
time tb (which is usually chosen to be equal to the maximal analysis time)
€ Š
̄g (t) = g (t) + Δg (t) = g (t) + b · t + c · t 2 , (2.56)

where b and c represent the unknown coefficients of the baseline acceleration correction
parabola.

Velocity and displacement time-histories which correspond to the new corrected acceleration
̄g (t) can be obtained by the integration of Eq. 2.56
Zt Zt
b c
 
2 3
̄g (t) = ̄g (τ)dτ = g (τ)dτ + ·t + ·t (2.57a)
0 0 2 3
Zt Zt
b c
 
3 4
̄g (t) = ̄g (τ)dτ = g (τ)dτ + ·t + ·t . (2.57b)
0 0 6 12

The coefficients b and c can now be determined from the system of equations with two un-
knowns which are obtained from the Eqs. 2.57 and the condition that new corrected velocity
and displacement are zero at baseline correction time, i.e. ̄g (tb ) = 0 and ̄g (tb ) = 0
Z tb
b c
· tb2 + · tb3 =− g (t)dt (2.58a)
2 3 0
Z tb Z t b ‚Z t Œ
b c
· tb3 + · tb4 =− g (t)dt = − g (τ)dτ dt . (2.58b)
6 12 0 0 0

The integrals on the right hand side of the Eqs. 2.58 are usually determined by some numerical
integration rule (e.g. trapezoidal).

Baseline correction of an acceleration time-history function can be triggered with the input
of the baseline correction time TBLC in the SOFiLOAD command FUNC (see SOFILOAD:
FUNC).

SOFiSTiK 2022 2-21

DYNA | Theoretical Principles

2.6.6 Dynamic load as ground acceleration

Starting from the equation for the effective earthquake forces (Eq. 2.49) and using Eq 2.22 we
can derive the equations for the generalized modal load in case when the load is given as the
ground acceleration function

pk (t) = −ϕTk · M · X · ̈gX (t) + Y · ̈gY (t) + Z · ̈gZ (t) .

 
(2.59)

Modal equation 2.19 now becomes

EQX EQY EQZ
q̈k (t) + 2ξk ωk · q̇k (t) + ω2
k
· qk (t) = k · ̈gX (t) + k · ̈gY (t) + k · ̈gZ (t) (2.60)

where EQX
k
, EQY
k
and EQZ
k
are modal participation factors for the earthquake in X, Y and Z
direction, respectively. They are given as
EQX EQY EQZ
EQX L EQY L EQZ L
k = k ; k = k ; k = k , (2.61)
mk mk mk

where
EQX EQY EQZ
Lk = −ϕTk · M · X ; Lk = −ϕTk · M · Y ; Lk = −ϕTk · M · Z . (2.62)

Hint
Since in SOFiSTiK mk ≡ 1, we have that k = Lk = −ϕTk · M · .

Hint
mk , Lk and k are dependent on the eigenvector normalization.

Utilizing the superposition principle, we can represent the total solution of the modal equation
2.60 as follows
EQX EQY EQZ
qk (t) = qk (t) + qk (t) + qk (t) , (2.63)

where qEQX
k
(t) is the generalized modal coordinate response due to the earthquake ground
acceleration acting in global X direction, ̈gX (t). This means that we can split the Eq. 2.60
into three equations, each of which corresponding to one direction of the earthquake, and treat
them separately.

Focusing on the earthquake ground motion acting in X direction, we can write the modal equa-
tion of motion as follows
EQX EQX EQX EQX
q̈k (t) + 2ξk ωk · q̇k (t) + ω2
k
· qk (t) = k · ̈gX (t) . (2.64)

Dividing everything with participation factor EQX

k
we can obtain an alternative representation
of the modal equation of motion
EQX EQX EQX
D̈k (t) + 2ξk ωk · Ḋk (t) + ω2
k
· Dk (t) = ̈gX (t) (2.65)

where the relation between the qEQX

k
EQX
(t) and a new generalized modal displacement Dk (t) is

2-22 SOFiSTiK 2022

 Theoretical Principles | DYNA

given by
EQX EQEQX EQX
qk (t) = k · Dk (t) . (2.66)

By solving the equation 2.65 for DEQX

k
(t) using one of the techniques described in Section 2.2
EQX
we can determine qk (t) from Eq. 2.66. 4 The contribution of the k th mode to the displacement
vector due to earthquake in X direction EQX (t) is
EQX EQX EQX EQX
k (t) = ϕk · qk (t) = ϕk · k · Dk (t) . (2.67)

2.6.7 Total Response

The total displacement vector due to earthquake in X direction is given by the superposition of
its modal contributions (modal superposition)
n
EQX
X
EQX (t) = k (t) . (2.68)
k=1

Finally the total displacement vector due to spatial earthquake action is given by the sum of
displacements vectors due to the earthquake in all three directions (spatial superposition)
n ”
EQX EQY EQZ
X —
(t) = EQX (t) + EQY (t) + EQZ (t) = k (t) + k (t) + k (t) . (2.69)
k=1

All other response quantities are determined analogue to the direct method described by
Eq.2.29 given in the Section 2.3.5, i.e.
n ”
EQX EQY EQZ
X —
R(t) = REQX (t) + REQY (t) + REQZ (t) = Rk (t) + Rk (t) + Rk (t)
k=1
n ”
(2.70)
EQX EQX EQY EQY EQZ EQZ
X —
= Rk0 · qk (t) + Rk0 · qk (t) + Rk0 · qk (t) .
k=1

2.6.8 Equivalent Modal Loads

Equivalent modal forces for the mode k due to the earthquake in X direction can be determined
in the same way as it has been described in Section 2.3.5
EQX EQX EQX EQX EQX
ƒk (t) = K · k (t) = K · ϕk · qk (t) = K · ϕk · k · Dk (t) , (2.71)

or with the help of the mass matrix

EQX EQX EQX EQX
ƒk (t) = M · ϕk · ω2
k
· qk (t) = k · M · ϕk · Ak (t) , (2.72)

where
EQX EQX
Ak (t) = ω2
k
· Dk (t) , (2.73)
4q per response spectrum is outputted in Table 3.37.
m,k

SOFiSTiK 2022 2-23

DYNA | Theoretical Principles

denotes the pseudo-acceleration response of the k th mode generalized SDOF system due to
earthquake acceleration in X direction ̈gX (t). 5

2.6.9 Effective Modal Mass

We can also write Eq. 2.72 as follows
EQX EQX EQX
ƒk (t) = Meƒ ƒ · Ak (t) , (2.74)

where
EQX EQX
Meƒ ƒ ,k = k · M · ϕk , (2.75)

represents the effective modal mass vector of the mode k for the earthquake in X direction.

This vector in general contains components of all the nodal DOFs (translational and rotational).
In earthquake engineering it is very useful to have a sum of the translational components (X,
Y and Z ) of the effective modal mass vector, i.e.
EQX EQX EQX
Meƒ ƒ ,k,X = bTX · Meƒ ƒ ,k = k · bTX · M · ϕk (2.76a)
EQX EQX EQX
Meƒ ƒ ,k,Y = bTY · Meƒ ƒ ,k = k · bTY · M · ϕk (2.76b)
EQX EQX EQX
Meƒ ƒ ,k,Z = bTZ · Meƒ ƒ ,k = k · bTZ · M · ϕk , (2.76c)

where MEQX
eƒ ƒ ,k,X
, MEQX
eƒ ƒ ,k,Y
and MEQX
eƒ ƒ ,k,Z
are the effective modal masses 6 of the mode k in X, Y
and Z direction for an earthquake acting in X direction (EQX). The vector bX is the vector which
has ones at places where DOF is a translational DOF in X direction, and zeros elsewhere. 7

Hint
Meƒ ƒ ,k is independent of how the eigenvector is normalized (invariant).

The total effective masses in X, Y and Z direction for an earthquake acting in X direction are
n n n
EQX EQX EQX EQX EQX EQX
X X X
Meƒ ƒ ,X = Meƒ ƒ ,k,X ; Meƒ ƒ ,Y = Meƒ ƒ ,k,Y ; Meƒ ƒ ,Z = Meƒ ƒ ,k,Z . (2.77)
k=1 k=1 k=1

If all the modes of the system are computed, total effective mass must be equal to the total
mass of the structure. But as there might be some masses which are not active because
they are associated to a support, DYNA evaluates two factors. First the above equations
are evaluated for the active masses. The result will be 100 % if all eigen forms have been
calculated. The effective massed will be evaluated based on all mass elements and thus is an
indicator for the quality of the mesh.

It is generally required that all modes which contribute significantly to the total response of the
structure should be taken into account. In some seismic norms this requirement is quantified
5 For the undamped structures the pseudo-acceleration is equal to the total acceleration, i.e. A(t) = D̈(t) +
̈g (t). In reality all structures posses some damping, but this structural damping is usually very small so that the
pseudo-acceleration can be approximated with the total accelerations. The internal resistance structural forces are
proportional to the pseudo-accelerations.
6 A.k.a. Participating mass [5]
7 For the uniform support excitation vectors b , b and b are equal to influence vectors  ,  and  (see
X Y Z X Y Z
Section 2.6).

2-24 SOFiSTiK 2022

 Theoretical Principles | DYNA

by demanding that the total effective mass for the modes taken into account is not less then
certain threshold (i.e. 90% of total mass of the structure). However a standard approach is
to include the "‘Missing Mass"’ as the difference of the sum of all modal loadings to the static
loading R with the spectral value of period 0.0. The enhanced implementation calculates this
for every single degree of freedom based on the participation factors, but a simplified and more
general method based on the percentage of the effective masses is also available.
 X Meƒ ƒ , 
Rm = 1 − · R (2.78)
Mtot,

For a spectral loading with CQC this will become:

rX X
R= R · ρj · Rj + (ZPA · Rm )2 (2.79)

The effective modal masses and total effective masses are not dependent on the earthquake
loading, only on its direction. This means that they are available upon the solution of the
eigenvalue problem. 8

2.6.10 Base Shear and Base Overturning Moment

Sum of all the equivalent modal forces (Eq. 2.74) for mode k in X, Y and Z direction due to
earthquake acting in X direction are (see also Eq. 2.76)
EQX EQX EQX EQX EQX EQX
Vb,k,X (t) = bTX · ƒ k (t) = k · bTX · M · ϕk · Ak (t) = Meƒ ƒ ,k,X · Ak (t) (2.80a)
EQX EQX EQX EQX EQX EQX
Vb,k,Y (t) = bTY · ƒ k (t) = k · bTY · M · ϕk · Ak (t) = Meƒ ƒ ,k,Y · Ak (t) (2.80b)
EQX EQX EQX EQX EQX EQX
Vb,k,Z (t) = bTZ · ƒ k (t) = k · bTZ · M · ϕk · Ak (t) = Meƒ ƒ ,k,Z · Ak (t) , (2.80c)

EQX
where Vb,k,X (t) represents the base shear for mode k in X direction due to earthquake acting
in X direction.

If with heƒ ƒ ,k we denote the height (related to some reference point) at which the resultant force
of the equivalent modal loads is acting (base shear) Vb,k (t), we can then also calculate the
corresponding k th -mode base overturning moments in X, Y and Z direction for an earthquake
acting in X direction
EQX EQX EQX
Mb,k,X (t) = Vb,k,X (t) · heƒ ƒ ,k,X (2.81a)
EQX EQX EQX
Mb,k,Y (t) = Vb,k,Y (t) · heƒ ƒ ,k,Y (2.81b)
EQX EQX EQX
Mb,k,Z (t) = Vb,k,Z (t) · heƒ ƒ ,k,Z . (2.81c)

EQX
heƒ ƒ ,k,X is the effective modal height.

By looking at the formulas 2.80 and 2.81 we can conclude that the k th -mode base shear Vb,k (t)
and base overturning moment Mb,k (t) of a MDOF system is the same as the base shear and
base overturning moment of a SDOF system with a mass equal to effective modal mass Meƒ ƒ ,k
and height equal to effective modal height heƒ ƒ ,k (see Fig. 2.13). This gives a mechanical
meaning to the effective modal mass and the effective modal height.
8 Total effective masses M
eƒ ƒ and effective modal masses Meƒ ƒ ,k in X, Y and Z direction are outputted in Table
3.33.

SOFiSTiK 2022 2-25

DYNA | Theoretical Principles

mn
ƒn,k (t)

Meƒ ƒ ,k

Vb,k (t) mj Meƒ ƒ ,k · Ak (t)

ƒj,k (t)

heƒ ƒ ,k heƒ ƒ ,k
m1
ƒ1,k (t)

Vb,k (t) Vb,k (t)

Mb,k (t) Mb,k (t)

Figure 2.13: (left) Base shear Vb,k (t) and base overturning moment Mb,k (t) (right) SDOF system
with effective modal mass Meƒ ƒ ,k and effective modal height heƒ ƒ ,k

The total base shears in X, Y and Z direction due to an earthquake acting in X direction are
n n n
EQX EQX EQX EQX EQX EQX
X X X
Vb,X (t) = Vb,k,X (t) ; Vb,Y (t) = Vb,k,Y (t) ; Vb,Z (t) = Vb,k,Z (t) , (2.82)
k=1 k=1 k=1

and the corresponding total base overturning moments

n n n
EQX EQX EQX EQX EQX EQX
X X X
Mb,X (t) = Mb,k,X (t) ; Mb,Y (t) = Mb,k,Y (t) ; Mb,Z (t) = Mb,k,Z (t) . (2.83)
k=1 k=1 k=1

Modal Response
The maximal (peak) contribution of the k th -mode to the displacement vector due to earthquake
in X direction EQX
m,k
is

EQX EQX EQX EQX EQX

m,k = mx k (t) = ϕk · qm,k = ϕk · k · SD (Tk , ξk ) . (2.84)

Maximal response quantities of the k th -mode REQX

m,k
can be obtained by using Eqs. 2.29 and
2.70
EQX EQX EQX EQX EQX EQX EQX
Rm,k = mx Rk (t) = Rk0 · qm,k = Rk0 · k · SD (Tk , ξk ) . (2.85)

Equivalent modal loads associated with peak (maximum) response of the k th -mode are (see
Eq. 2.72)
EQX EQX EQX EQX EQX
ƒ m,k = mx ƒ k (t) = M · ϕk · ω2
k
· qmk,k = k · M · ϕk · SpA (Tk , ξk ) . (2.86)

As aforementioned (Section 2.3.5) SOFiSTiK does not work directly with these forces and
hence they are not saved in the database. However, if there is a real need for them, it is

2-26 SOFiSTiK 2022

 Theoretical Principles | DYNA

possible for the user to construct them, as it is explained in the Section 2.12.3.

Similarly base shear forces in X, Y and Z direction associated with the k th -mode peak response
for the earthquake acting in X direction (EQX) are
EQX EQX EQX EQX
Vb,k,X = mx Vb,k,X (t) = k · bTX · M · ϕk · SpA (Tk , ξk ) (2.87a)
EQX EQX EQX EQX
Vb,k,Y = mx Vb,k,Y (t) = k · bTY · M · ϕk · SpA (Tk , ξk ) (2.87b)
EQX EQX EQX EQX
Vb,k,Z = mx Vb,k,Z (t) = k · bTZ · M · ϕk · SpA (Tk , ξk ) , (2.87c)

or written in terms of effective modal masses (see Eq. 2.80)9

EQX EQX EQX
Vb,k,X = Meƒ ƒ ,k,X · SpA (Tk , ξk ) (2.88a)
EQX EQX EQX
Vb,k,Y = Meƒ ƒ ,k,Y · SpA (Tk , ξk ) (2.88b)
EQX EQX EQX
Vb,k,Z = Meƒ ƒ ,k,Z · SpA (Tk , ξk ) . (2.88c)

Base overturning moments associated with the k th -mode peak response are (see Eq. 2.81)10
EQX EQX EQX
Mb,k,X = Vb,k,X · heƒ ƒ ,k,X (2.89a)
EQX EQX EQX
Mb,k,Y = Vb,k,Y · heƒ ƒ ,k,Y (2.89b)
EQX EQX EQX
Mb,k,Z = Vb,k,Z · heƒ ƒ ,k,Z . (2.89c)

Hint
In DYNA base overturning moment is always evaluated relative to the origin of the global
coordinate system.

Sometimes maximum base shear and maximum base overturning moment per mode at some
specified height z are needed as well (see Fig. 2.14 (right)). DYNA can compute these with
the help of the command CTRL BLEV.

Total Response
In Section 2.6.6 (Eqs. 2.68 - 2.70) we saw how in a modal time history analysis (MTHA)
any total response quantity R(t) as a function of time can be determined by the sum of its
modal contributions Rk (t). In the response spectrum analysis (RSA) the things are a bit more
complicated. The results of the RSA analysis are the maximal (peak) response quantities per
mode Rm,k . These values are exact (i.e. equal to those obtained from MTHA). However
since in general maximal (peak) responses for different modes are not reached at same time
instants the total response Rm is no longer a simple sum of the maximal modal responses
values Rm,k . In fact a sum of all the modal response quantities provides an upper bound for
9 Base shear per mode, per direction and per earthquake spectra is outputted in Tables 3.35 and 3.36.
10 Base overturning moment per mode, per direction and per earthquake spectra is outputted in Tables 3.35 and
3.36.

SOFiSTiK 2022 2-27

DYNA | Theoretical Principles

mn mn
ƒm,n,k ƒm,n,k

Vb,k mj mj
ƒm,j,k ƒm,j,k

heƒ ƒ ,k
m1 Vb,k (z)
ƒm,1,k
z

z=0
Vb,k
Mb,k (z)

Mb,k

Figure 2.14: (left) Base shear and base overturning moment for the entire structure (right) Base
shear and base overturning moment above height z

the total response quantity

n n
EQX EQX
X X
REQX
m
= mx R EQX
(t) ≤ mx Rk (t) = Rm,k . (2.90)
k=1 k=1

Obviously one could adopt the sum of absolute values as a representative value for the design,
however in many cases this is a too conservative estimate. In practice one adopts some other
modal combination rule which is less conservative (SRSS, CQC, etc.). We can then write
EQX
” —
n
REQX
m
= MCOMB k=1
R m,k
, (2.91)

where MCOMB denotes that a modal combination rule is used. Which ever rule the user
chooses, it should be clear that the combined value for total response is only an approximation
of the exact total response.

Similarly, once the total response quantities are computed using Eq. 2.91 for each individual
component of ground motion (REQX m
, REQY
m
and REQZ
m
) they must be superimposed by using
some spatial combination rule in order to obtain the total response quantities due to the simul-
taneous action of three components of ground motion Rm
” —
Rm = SCOMB REQX
m
, R EQY
m
, R EQZ
m
, (2.92)

where SCOMB denotes that a spatial combination rule is used (SRSS, 100-30-30, etc.).

Modal and spatial combination of response quantities is discussed in more detail in Sections
2.5.3 and 2.5.3.

2.7 Storey level definition

For assessment of earthquake resistance capability of a multi-storey building, storey level re-
lated design criteria are frequently adopted. Fig. 2.15 shows a multi-storey building and corre-

2-28 SOFiSTiK 2022

 Theoretical Principles | DYNA

sponding storey definitions.

LVL 3
Shear force

h3

drift
LVL 2
Shear force

u
h2

LVL 1
Shear force

ZLVL h1

LVL 0 (base level)

Figure 2.15: Definition of storeys in a multi-storey building

The inter-storey height of a storey level is the vertical distance to the storey level situated below.
The storey level at the very bottom constitutes the base level.

Hint
By definition, the inter-storey height of the base level is always h = 0.

Storey level definition is system related and can, e.g., be accomplished by SOFIMSHC: SLVL.

2.7.1 Computation of the center of rigidity (COR) of a diaphragm floor

The center of rigidity of a rigid diaphragm floor is the center of rotation of the floor under applied
moments. For every force applied in the direction of the COR, the floor will only undergo
translational displacements and no rotations. The other floors can undergo displacements as
well as rotations. The center of rigidity can be calculated by applying three load cases, F and
Fy at an arbitrary point and a Moment Mzz . Writing the equilibrium with respect to the moment,
for the two forces, see Fig. 2.16 (left) we get

− (F · ΔY) = Kzz · φz (2.93a)

Fy · ΔX = Kzz · φzy . (2.93b)

and the equilibrium for the moment Mzz reads as

Mzz = Kzz · φzz , (2.94)

where Kzz denotes the rotational stiffness. Substituting Kzz from Eq. 2.94 into Eqs. 2.93 and
solving for ΔX and ΔY , we obtain the distance of center of rigidity from the point where the
loads are applied. The distance of the center of rigidity from the point where the load is applied

SOFiSTiK 2022 2-29

DYNA | Theoretical Principles

Y Y

U2 2

ΔY1 ΔY2 − ΔY1

F
Mzz

U1 1

ΔY2
Fy
Mzz
ΔY

COR COR

Uy1

Uy2
ΔX

ΔX1ΔX2 − ΔX1

ΔX2

X X

Figure 2.16: Calculation of center of rigidity (left) and calculation of rotations (right)

is calculated with opposite sign as

φzy Mzz
ΔX = − , (2.95a)
φzz Fy
φz Mzz
ΔY = . (2.95b)
φzz F

The rotation of the floor can be calculated by using the small-angle approximation, i.e. φ =
tn (φ).

Alternatively, the center of rigidity can also be calculated by the assumption that under applied
moments the center of rotation of the floor corresponds with the center of rigidity. From the
application of an arbitrary moment, the rotation of the diaphragm can be calculated from the
displacements of two arbitrary points (eg. two nodes)
U2 U1 U2 − U1
φ=− =− =− , (2.96)
ΔY2 ΔY1 ΔY2 − ΔY1

or by using Uy as

Uy2 Uy1 Uy2 − Uy1

φ= = = . (2.97)
ΔX2 ΔX1 ΔX2 − ΔX1

After calculating the rotation from Eq. 2.96 or Eq. 2.97, the distance of the center of rigidity
from the point  where the displacements are evaluated can be computed as
Uy
ΔX = − (2.98a)
φ
U
ΔY = . (2.98b)
φ

2-30 SOFiSTiK 2022

 Theoretical Principles | DYNA

2.7.2 Computation of the center of mass (COM) of a storey

The coordinates of the center of mass of a particular storey are calculated as

(m ·  )
P
X= , (2.99a)
(m )
P

(m · y )
P
Y= P , (2.99b)
(m )
(m · z )
P
Z= P , (2.99c)
(m )

where  , y , z and m are the coordinates and masses of each node  in the storey.

2.7.3 Storey results

Information about the storey results can be found in FEABENCH: Storey-results.

2.8 Kinematic Constraints

Frequent use of condensations is made in dynamic analysis. These require a certain exper-
tise, because the selection of the reference nodes is critical. The same holds for kinematic
constraints too.

While a force at a distance a generates a moment P · , the inertia of an eccentric mass is given
by m · 2 . DYNA takes this effect into account automatically for standard kinematic constraints.
This, however, results in off diagonal mass matrix components, which necessitate the use of
a consistent, i.e. not diagonalised, mass matrix. These do not need not only more memory,
but can lead also to oscillation of solution through violation of the discrete maximum principle
at small time steps that perform disturbingly. But as these matrices are not always acceptable,
the user therefore can switch to the use of a diagonalised mass matrix, which then requires
special care in the description of the constraints. When modeling rigid floor disks one should
place the reference node as close as possible to the gravity or the shear centre in order to get
the most realistic results.

Kinematic constraints increase the band width considerably. The memory capacity can thus be
quickly exceeded in cases of large systems or strongly recursive kinematic constraints.

2.9 Elastic Stiffnesses

DYNA employs very compact formulations of the element stiffnesses. The spring and boundary
elements are in the classical form given and do not distinguish from those ones in programs
ASE or STAR2.

The beam element is a real finite element with a displacement accretion with Hermitian function
of second redundancy (therefore cubical polynomials). That is considered at:

– variable stiffnesses and location of centre of gravity

– variable locations of shear centre
– shear deformations over correction terms
– optional warping force torsion

SOFiSTiK 2022 2-31

DYNA | Theoretical Principles

– elastic bedding for pile elements

– all beam loads from database

The element produces for static load cases results also for internal sections.

The QUAD shell element is available in the basic version (Hughes and/or Bathe-Dvorkin) or
with "‘assumed strains"’. The default version of the QUAD includes the sixth degree of freedom
(drilling perpendicular to the element plane) with a variational principle of the rotational field.

The BRIC element is available in the basic version or with "‘assumed strains"’.

Special provisions are available for so called "‘incompressible elements"’ for the dynamic anal-
ysis of fluids with QUAD or BRIC elements.

Anisotropic material or thickness definitions are not available in DYNA.

2.10 Geometric Stiffness and P-delta

For all elements a primary load case can be read, which can be used for the determination of
the geometric stiffness. The second order theory effects are exact in those cases where the
axial force does not change due to geometric nonlinear effects. For cable elements large de-
formations of this primary load case are also accounted for. For warping torsion there are also
contributions from bending and warping moments which have to be accounted for iteratively.

Thus this approach may be used either for a buckling eigenvalue or for performing a linearized
dynamic or static analysis based on the tangential stiffness. So it exceeds classical P-delta
effects. This approach is faster by magnitudes compared to a general geometric non linear
analysis.

The eigen frequency of member with tension will thereby increase, while those of members
under compression will decrease until they reach the value of zero for the buckling load. How-
ever for cable elements the complete separation of geometric stiffness is not always a good
approach, as this might generate negative eigenvalues in a buckling analysis. On the other
side a buckling factor is defined as the factor of the loading. So it is generally foreseen for
cables to split the prestress in two parts. One part is included in the general stiffness (this
is the value defined with the element itself) and the difference from the actual primary estate
to that general value is then used to form the geometric stiffness for the buckling analysis. If
that general value is not defined and option CTRL PLC does not select otherwise, the primary
estate will be taken as general prestress.

For beam elements there are some extra effects for static load cases available. For analysis
according 2nd order theory a primary load case (in many cases only with a constant normal
force) may be added to the results for a total design case(CTRL PLC W2 3) or omitted to save
incremental results which might be used for a linear superposition (CTRL PLC W2 1 = default)
Further it is possible to calculate the incremental loading for beam elements automatically
(CTRL PLC W2 7).

2.11 Nonlinear Effects

DYNA uses linear Algorithms for performance reasons. Some important non linear effects are
available however:

2-32 SOFiSTiK 2022

 Theoretical Principles | DYNA

– For vehicle-structure interaction a moving contact condition is available

– Spring elements can perform all non linear effects including explicit force-displacement
relations.
There is no extra iteration for the residual forces, thus the change of the stiffness should
not become to large.
These effects are also avaiable for a modal analysis if the used eigen values allow suffi-
cient relative displacements of the spring nodes.
– Beam elements can account for all non linear effects form an interaction of a structure
with a steady-state or transient wind field. This includes galloping and flutter.
– Cable, Truss and QUAD elements may account for relative flow velocities as well
– For BRIC elements all nonlinear material models are available for the explicit integration
scheme

2.12 FAQ

2.12.1 Manual Checks of the RSA Results

In this section manual checks of the extreme and corresponding forces computed using modal
superposition rules during the response spectrum analysis will be illustrated trough an exam-
ple.

The structure is a frame subjected to earthquake loading in global X-direction in form of a

response spectrum (Fig. 2.17). The task is to manually determine extreme value of a moment
My,m and corresponding normal force N and to compare it with the extreme results obtained
from DYNA calculation (Fig. 2.18). SRSS modal combination rule is used. Only first two
eigenmodes are relevant to the overall structural response.

12
X
2 
y
Z
11 z

EQX

Figure 2.17: Frame subjected to response spectrum in X direction: nodes and beams of interest

According to Eq. 2.85 maximal k th -mode moment and normal force can be obtained from

My,m,k = My,0,k · qm,k (2.100a)

Nm,k = N0,k · qm,k , (2.100b)

where My,0,k and N0,k are the moment and normal force which correspond to the eigenvector
displacements of the mode k and which are saved in the corresponding eigenmode load case
(see Fig. 2.19). Maximal modal coordinate responses qm,k can be obtained from the Report

SOFiSTiK 2022 2-33

DYNA | Theoretical Principles

834.6
−280.5

230.6
89.1
My,m 232.4 N

1371.8 280.6

Figure 2.18: Total maximum (extreme) bending moment My,m and corresponding normal force
N computed by DYNA using the SRSS modal combination rule

Browser Table ”Modal Responses” (Table 3.37) and for this example they read

qm,1 = −1.386 · 10−1

qm,2 = +1.168 · 10−3 .

Total maximal moment can, according to Eq. 2.91, be determined by applying the modal
combination rule (SRSS in the example) to the modal moments, i.e.
v
u n
uX
n
My,m = SRSSk=1 My,m,k = t M2
 
y,m,k
. (2.101)
k=1

Normal force N corresponding to the total maximal moment My,m can be computed according
to the rules described in Section 2.5.3, i.e. according to Eq. 2.45
n
X
N= ƒk · Nm,k , (2.102)
k=1

where the factors ƒk are given as

My,m,k
ƒk = . (2.103)
My,m

Finally the results at the ends of the beams 11 and 12 can be determined (Fig. 2.17):

• Beam 11
– Start section (node 1)
Maximal modal moments and normal forces for the modes 1 and 2 (Eqs. 2.100 and
Fig. 2.19):

My,m,1 = (+9893) · (−1.386 · 10−1 ) = −1371.0

My,m,2 = (−46368) · (+1.168 · 10−3 ) = −54.2
Nm,1 = (+2031) · (−1.386 · 10−1 ) = −281.5
Nm,2 = (+11995) · (+1.168 · 10−3 ) = +14.0

2-34 SOFiSTiK 2022

 Theoretical Principles | DYNA

−6019

2023

512
My,0,1 N0,1

2031

9893

(a) Eigenmode 1

−25374

10775

46248
My,0,2 N0,2

11995

−46368

(b) Eigenmode 2

Figure 2.19: Bending moments and normal forces corresponding to eigenvector displacements

Total maximal moment (Eq. 2.101):

Æ
My,m = 1371.02 + 54.22 = +1372.0

Corresponding normal force (Eqs. 2.102 and 2.103):

−1371.0 −54.2
N= · (−281.5) + · (+14.0) = +280.7
1372.0 1372.0

– End section (node 2)

Maximal modal moments and normal forces for the modes 1 and 2:

My,m,1 = (+512) · (−1.386 · 10−1 ) = −71.0

My,m,2 = (+46248) · (+1.168 · 10−3 ) = +54.0
Nm,1 = (+2031) · (−1.386 · 10−1 ) = −281.5
Nm,2 = (+11995) · (+1.168 · 10−3 ) = +14.0

Total maximal moment:

Æ
My,m = 71.02 + 54.02 = +89.2

Corresponding normal force:

−71.0 +54.0
N= · (−281.5) + · (+14.0) = +232.5
89.2 89.2

SOFiSTiK 2022 2-35

DYNA | Theoretical Principles

• Beam 12
– Start section (node 2)
Maximal modal moments and normal forces for the modes 1 and 2:

My,m,1 = (+512) · (−1.386 · 10−1 ) = −71.0

My,m,2 = (+46248) · (+1.168 · 10−3 ) = +54.0
Nm,1 = (+2023) · (−1.386 · 10−1 ) = −280.4
Nm,2 = (+10775) · (+1.168 · 10−3 ) = +12.6

Total maximal moment:

Æ
My,m = 71.02 + 54.02 = +89.2

Corresponding normal force:

−71.0 +54.0
N= · (−280.4) + · (+12.6) = +230.5
89.2 89.2

– End section (node 3)

Maximal modal moments and normal forces for the modes 1 and 2:

My,m,1 = (−6019) · (−1.386 · 10−1 ) = +834.2

My,m,2 = (−25374) · (+1.168 · 10−3 ) = −29.6
Nm,1 = (+2023) · (−1.386 · 10−1 ) = −280.4
Nm,2 = (+10775) · (+1.168 · 10−3 ) = +12.6

Total maximal moment:

Æ
My,m = 71.02 + 54.02 = +834.7

Corresponding normal force:

+834.2 −29.6
N= · (−280.4) + · (+12.6) = −280.9
834.7 834.7

The results obtained by the manual calculation are in agreement with the results obtained by
DYNA (Fig. 2.18).

2.12.2 Modal Participation Factor

Modal participation factor for the earthquake in X direction is defined by (see Eq. 2.61)
EQX
EQX Lk
k = = ϕTk · M · X . (2.104)
mk ≡ 1

DYNA does not output the modal participation factors in the Report Browser, however if the user
has a need to have these values it is possible to get them from the database. If the eigenvalue
analysis has been performed with DYNA, then the participation factors EQXk
, EQY
k
and EQZ
k
are saved in the database in the record LC_CTRL (012/LC) for the corresponding eigenvalue load
case on the positions RX, RY and RZ, respectively. For example the CADINP input

2-36 SOFiSTiK 2022

 Theoretical Principles | DYNA

@KEY LC_CTRL #lc_eig+#k ! lc_eig+k = k-th eigenmode load case

STO#PFx @(RX) ! modal part. factor for eq. in X direction
STO#PFy @(RY) ! modal part. factor for eq. in Y direction
STO#PFz @(RZ) ! modal part. factor for eq. in Z direction

will read the modal participation factors from the CDB record LC_CTRL for the k th -eigenmode
(saved in the load case #lc_eig+#k).

2.12.3 Equivalent Modal Loads

It is often the case that the user wants have the equivalent modal loads of the response spec-
trum analysis as defined by Eq. 2.86 and repeated here for completeness sake
EQX EQX EQX
ƒ m,k = −k · M · ϕk · SpA (Tk , ξk ) . (2.105)

As already explained (see Sections 2.3.5, 2.6.6 and 2.5.1) DYNA does not work directly with
these forces in order to compute the modal responses but it uses the direct method. Hence
these forces are not saved in the database. However it is possible for the user to construct
them. How to achieve this is explained in the following.

During the eigenvalue analysis DYNA will save the following load vector in the corresponding
k th eigenvalue load case

ω2
k
· M · ϕk . (2.106)

Comparing Eqs. 2.105 and 2.106 it is evident that the equivalent modal loads can be con-
structed by applying the factor
EQX EQX
k · SpA (Tk , ξk )
, (2.107)
ω2
k

to the loads from the eigenvalue load case. In the Section 2.12.2 we have seen how to get
EQX
k from database. Values ωk and SEQX pA (Tk , ξk ) can be obtained from the Report Browser
output, or alternatively from the database.

ωk is saved in the CDB record LC_CTRL (012/LC) for the corresponding eigenvalue load case
on the position RPAR and can be, for example, obtained by the following CADINP input

@KEY LC_CTRL #lc_eig+#k ! lc_eig+k = load case of the k-th eigenvalue

STO#w @(RPAR)

Product EQX
k
EQX
·SpA (Tk , ξk ) is saved in the database and it can be obtained from the value LSAF
of the CDB record SLVL_RES for the corresponding response spectrum load case (#lc_rs_x),
level-id (#lvl_id) and the given eigenmode number (#k)

@KEY SLVL_RES #lc_rs_x #lvl_id #k

STO#SPax_PFx @(LSAF)

Now that the load factor is know, all one needs to do is to apply it to the corresponding eigen-
value load case, e.g.

SOFiSTiK 2022 2-37

DYNA | Theoretical Principles

+PROG ASE
HEAD
!...
LC ...
LCC #lc_eig+#k FACT #SPax_PFx/#w**2
!...
END

2.12.4 Scaling of the Generalized SDOF Properties for a Different Eigenvector Nor-
malization
As already mentioned, in SOFiSTiK eigenvectors ϕk are normalized in such a way that the
generalized modal mass mk is equal to 1, for each natural mode k . It is however often the
case that the user wants to determine generalized modal properties for different normalization
of the eigenvectors. The procedure to achieve this is described in the following.

Starting from the known generalized modal properties (mk ≡ 1, ck , kk and pk (t)) and the
eigenvalues and original eigenvectors (ωk and ϕk ) the task is to determine new generalized
modal properties (m̄k ≡ 1, c̄k , k̄k and p̄k (t)) so that they correspond to the new eigenvectors
ϕ̄k . New eigenvectors ϕ̄k and old eigenvectors ϕk must be linearly dependent, i.e. they must
be proportional to each other

ϕ̄k = αk · ϕk , (2.108)

where αk denotes the factor of proportionality (per mode k ). Using this equation and having in
mind the definition of the generalized modal properties, we obtain the following
T
m̄k = ϕ̄k · M · ϕ̄k = αk2 · ϕTk · M · ϕk = αk2 · mk = αk2 · 1.0 = αk2 (2.109a)
T
k̄k = ϕ̄k · K · ϕ̄k = αk2 · ϕTk · K · ϕk = αk2 · kk = αk2 · ω2
k
· mk = αk2 · ω2
k
(2.109b)
T
c̄k = ϕ̄k · C · ϕ̄k = αk2 · ϕTk · C · ϕk = αk2 · ck = αk2 · 2ξk ωk / mk = αk2 · 2ξk ωk (2.109c)
T
p̄k = ϕ̄k · p = αk · ϕTk · p = αk · pk . (2.109d)

So with the known factor of proportionality between the new and original eigenvectors αk , the
new generalized modal properties can be easily obtained from the old ones by using the Eqs.
2.109.

Other properties, like for example Lk or k (see Eqs. 2.62 and 2.61), which depend on eigen-
vector normalization can be similarly transformed, i.e.
T
L̄k = ϕ̄k · M ·  = αk · ϕTk · M ·  = αk · Lk (2.110a)
T
ϕ̄k · M ·  αk · ϕTk · M ·  1
̄k = = = · k . (2.110b)
T
ϕ̄k · M · ϕ̄k αk2 · ϕTk · M · ϕk αk

For a special case where the eigenvectors are scaled in such a way that the maximal vector

2-38 SOFiSTiK 2022

 Theoretical Principles | DYNA

component mx1≤j≤n ϕ̄j,k is equal to 1.0, the proportionality factor αk is given by

1.0
αk =
, (2.111)
mx ϕj,k
1≤j≤n


where mx1≤j≤n ϕj,k represents the maximal component of the original eigenvector for mode
k . This value can be easily obtained by looking (in e.g. WinGRAF, Result Viewer, etc.) at the
maximal displacements of the load case in which the eigenvectors are saved.

SOFiSTiK 2022 2-39

DYNA | Theoretical Principles

2-40 SOFiSTiK 2022

 Input Description | DYNA

3 Input Description

3.1 Input Language

The input is made in the CADINP language (see general manual SOFiSTiK: ’Basics’).

Three categories of units are distinguished:

mm Fixed unit. Input is always required in the specified unit.

[mm] Explicit unit. Input defaults to the specified unit. Alternatively, an explicit as-
signment of a related unit is possible (eg. 2.5[m] ).

[mm] 1011 Implicit unit. Implicit units are categorised semantically and denoted by a cor-
responding identity number (shown in green). Valid categories referring to the
unit ”length” are, for example, geodetic elevation, section length and thickness.
The default unit for each category is defined by the currently active (design code
specific) unit set. This input default can be overridden as described above. The
specified unit in square brackets corresponds to the default for unit set 5 (Eu-
rocodes, NORM UNIT 5).

3.2 Input Records

The following record names are defined:

Record Items
SYST TYPE NCS PROB PHYS CS PLC STAT
CTRL OPT VAL VAL2
GRP NO VAL CS FACS HING RADA RADB
MODD FACP FACM WIND LMAX NCSP
MAT NO E MUE G K GAM GAMA
ALFA EY MXY OAL OAF SPM TITL
BMAT NO C CT CRAC YIEL MUE COH
DIL GAMB REF MREF H
SMAT NO LC EX EY EZ RHOX RHOY
RHOZ ALF BET
MASS NO MX MY MZ MXX MYY MZZ
MXY MXZ MYZ MB
EIGE NEIG TYPE NITE MITE LMIN STOR LC
LCUP
MODD NO D A B PERS
STEP N DT INT A B BET DEL
THE EIGB EIGT EIGS DTF STHE
Table continued on next page.

SOFiSTiK 2022 3-1

DYNA | Input Description

Record Items
LC NO FACT DLX DLY DLZ MODB TITL
CONT TYP REF NR V TMIN LCUV LCUT
LCUR
HIST TYPE FROM TO STEP RESU LCST XREF
YREF ZREF DUMP
EXTR TYPE MAX MIN STYP ACT
ECHO OPT VAL

The records HEAD, END and PAGE are described in the general manual SOFiSTiK: ’Basics’.

A description of each record follows:

3-2 SOFiSTiK 2022

 Input Description | DYNA

3.3 SYST – System Parameters

See also: CTRL, STEP, GRP

SYST

Item Description Unit Default

TYPE Type of System LT REST

REST use existing main system
SECT use subsystem of section SNO
SNO Section number − -
PROB Geometric type of the analysis LT *
LINE linear analysis
TH2 2nd order theory
TH3 3rd order theory
PHYS Physical type of the analysis LT LINE
LINE linear analysis
NONL non linear analysis (all)
NSPR non linear analysis (Spring)
NMAT non linear analysis (Material)
CS Construction stage − -
PLC Primary load case − -
STAT State of analysis LT -
SERV serviceability
ULTI ultimate limit
CALC general nonlinear

The system for the analysis has to exist in the database. DYNA can use also the FE-meshes
of a specific section. This may be selected with SYST SECT nnn, where nnn is the number of
that section. The FE-system of the sections is saved in separate data base in a sub-directory.

The geometric type of the analysis may be linear or according 2nd order theory (small defor-
mations, but stress induced geometric stiffness and 3rd order theory (large deformations but
small strains). The stresses for the geometric stiffness are taken from the primary load case.
Thus without a primary load case the analysis is always linear, with a primary load case the
default is TH2. Option TH3 is currently only available for explicit integration.

The physical type of the analysis may be linear or including the nonlinear properties of the
spring elements and/or the full material non linearity (explicit integration only). The definition of
the state presets the selection of stress strain laws and safety factors according to the INI-file
of the selected design code.

The analysis uses the properties for construction stage CS and the stresses and deformations
according CTRL PLC from load case PLC.

SOFiSTiK 2022 3-3

DYNA | Input Description

3.4 CTRL – Calculation Parameters

See also: ECHO, GRP, MASS, EIGE, MODD, STEP, LC, CONT,
HIST, EXTR
CTRL

Item Description Unit Default

OPT Calculation parameter LT !

VAL Value of the parameter − !

V2 Secondary value of the parameter − -
V3 Secondary value of the parameter − -
V4 Secondary value of the parameter − -
V5 Secondary value of the parameter − -
V6 Secondary value of the parameter − -

CTRL defines parameters of calculation. These are:

PLC Primary load case

When entering a primary load case initial stress stiffness will be included in the
analysis, thus the linearized dynamic analysis is based on the tangential stiffness.
Non-linear stiffness created by AQB will be taken into account.
Details for the procedure may be found in chapter 2.
V2 Usage of primary load case
0= only for geometric stiffness
+1 = displacements for transient analysis will be set (default)
+2 = static beam results will be added
+4 = create incremental beam loading automatically
+7 = all options 1 to 4 together
V3 Usage of primary load case stiffness for beams (default n=7), where:
+1 = axial force
+2 = bending moments
+4 = torsion
V4 Options for geometric stiffness of cables (default 0)
+1 = do not set the general prestress value for cables from the pri-
mary load case for buckling eigenvalues.
+2 = suppress the coordinate update
RLC Result load case for Histories
With input of a result case RLC you will save results to the database after every time
step interval of an direct integration with individual load cases starting at RLC+1.
These results may be used for an animated sequence. Via V2 you may control by
adding the options:
+1 = Displacements (default)
+2 = Results of elements

3-4 SOFiSTiK 2022

 Input Description | DYNA

+4 = Velocities and accelerations

+8 = Loadpatterns for travelling load trains
value = user defined load case number (> 9999)
BEAM Formulation of beam element (Bitpattern), alias: BTYP
+1 = with all sections (static analysis only)
+4 = classical Timoshenko beam
+8 = non-conforming Timoshenko beam (Default)
+12 = classical beam with shear correction factors
CONT Deactivation / activation of the extrapolation of contact point displacement
0= deactivation
1= activation
WARP Warping torsion and lateral torsional buckling
0= no, the 2nd order Saint Venant theory for torsion is applied
1= yes, unless the section is warp free. (CM ≈ 0.) The initial stress stiffness
for lateral torsional buckling is applied.
2= yes, even if CM ≈ 0.0. As the characteristic length becomes zero in that
case, we do not obtain torsional stresses directly at a warping support
condition. Please use this option only for very special cases.
BETA Treatment of buckling length after buckling Eigenvalues
0= save
1= superposition
The estimate of the buckling length will be saved only for those beam elements
where the estimate is less than the limit value LMAX specified in record GRP. It
has to be marked however, that there are many cases not applicable for a buckling
length approach and that the second order analysis will be more suitable in most
cases.
QUAD Formulation of QUAD elements
0= conforming elements with “bbar” correction
1= nonconforming elements
2= four assumed strains + drilling degrees
3= five assumed strains + drilling degrees
BRIC BRIC Formulation of QUAD elements
0= conforming elements with “bbar” correction
1= nonconforming elements
SPRI Formulation of Spring elements (Bitpattern)
+1 = apply eccentricity effects (default)
+2 = account for nonlinear effects
+32 = do not extrapolate displacements in time
+64 = do not extrapolate damping forces in time
MCON Formulation of the mass matrix

SOFiSTiK 2022 3-5

DYNA | Input Description

1= Diagonal mass matrix (default)

2= Consistent mass matrix for system
Consistent translatoric element matrices
(default if such relevant kinematic constraints exist)
3= Consistent element- and system matrices including
rotational masses (default if CTRL WARP 1)
Hint
Rotational masses for torsion are always referred on the shear centre. A con-
straint rotation must be considered therefore with the definition of the cross
section.

CCON Formulation of the damping matrix

(same as MCON, definition however only in very special cases)
HLC Number of interim results for transient analysis of wind histories (only for internalpur-
pose)
SRES Steady-state response
0= Simplified analysis of phases by sign only
1= As above, but account for transient components of
initial conditions (to be used only in very special cases)
2= Exact account of phases within a load case (default) The superposition
of multiple load cases is only possible with methods selected by STYP, if
multiple load cases have to be analyzed synchronously the extreme val-
ues have to be searched with STEP -n for selected time values. If these
intermediate results are then to be combined stochasticly MAXIMA has
to be used.
STYP Superposition of results of spectra and stationary response
MAX The functions are evaluated separately followed by a max/min selection.
ADD All functions are added algebraically.
SUM All functions are added by absolute values.
SRSS The square root of the sum of squares is computed.
The default is dependant on the type of analysis.
V2 Default for EXTR and the evaluation of the resultant shear base values:
ADD All functions are added algebraically.
SUM All functions are added by absolute values.
SRSS The square root of the sum of squares is computed.
CQC Complete Quadratic Combination (default)
V3 Load case to define the sign of the final results:
+n A freely selectable load case with number n
-n the nth Eigenform, may be also defined as n[−] .
BLEV Height ordinate of a layer for which the resultant base shear should be calculated
during the response spectra evaluation (may be defined multiple times). The result-
ing moment is always taken to the reference of the origin of the global coordinate
system to allow the superposition of different levels. Requires the gravity direction

3-6 SOFiSTiK 2022

 Input Description | DYNA

to be aligned with the global Z-coordinate axis (positive or negative).

Hint
This input option has become obsolete; for full functionality, please follow the
approach outlined in section 2.7.

COR Calculation of the center of rigidity for each defined storey level.
With the input COR, the center of rigidity for each defined level is calculated. The
floor is considered as a rigid diaphragm floor and will undergo rigid body motions
in the plane of the defined level. This assumption is only applied internally during
the calculation of the COR and will not be effective in the model for the subsequent
calculations, see subsection 2.7.1.
COM Calculation of the center of mass for each defined storey level.
With the input COM, the center of mass for each defined storey is calculated, see
subsection 2.7.2.

SOFiSTiK 2022 3-7

DYNA | Input Description

3.4.1 SOLV Equation solver

SOLV Description Unit Default

VAL Selection of equation solver − 3

1 Direct Skyline Solver
(Gauss/ Cholesky)
2 Iterative Sparse Solver
3 Direct Sparse LDL Solver
4 Direct Parallel Sparse Solver (PARDISO)

For solving the equation systems of the Finite-Element problem, SOFiSTiK provides a number
of solvers. Which solver is used best depends highly on the type of the system and requires
knowledge of relevant system parameters. Following types are available:

• Direct Skyline Solver (Gauss/Cholesky)

This is the classical solver of the FE-Method. The current implementation works best on
a skyline oriented matrix. The storage needed depends on the internal optimization of the
node numbers and may become quite large for 3D structures.
• Iterative Solver (Conjugate Gradients)
One advantage of the iterative solver lies in its reduced requirements for storage, but it
may also provide reduced computing time compared to the previous two types especially
in case of large volume structures.
• Direct Sparse Solver
These types of solvers correspond to state of technology. A quite efficient version based
on the work of Davis [7] is available as well as a direct parallel solver PARDISO.

The advantage of the direct solvers is especially given in case of multiple right hand sides, as
the effort for solving them is very small compared to the triangulization of the equation system.
Thus they are the first choice for any dynamic analysis or in case of many load cases.

In order to minimize computational effort, the solvers need an optimized sequence of equation
numbers. This optimization step is usually performed during system generation. The programs
SOFIMSHA/C by default always create a sequence which is suitable for the direct sparse solver
(3). The solvers (1) or (2) however require a skyline oriented numbering which may be obtained
using the option (CTRL OPTI 1) or (CTRL OPTI 2) during system generation. The correct
setting will be checked and a warning will be issued in case a correct numbering is not available.

The iterative (CTRL SOLV 2) and the parallel sparse solver (CTRL SOLV 4) can be run in par-
allel providing an additional reduction in computing time. A parallelization basically requires
a license of type ”HISOLV”. More information about parallelization can be found in subsec-
tion 3.4.2 describing the input parameter (CTRL CORE).

The equation solvers are selected using the parameter (CTRL SOLV). The first value defines
the type of the solver, followed by optional additional parameters.

Direct Skyline Solver (Gauss/ Cholesky)

3-8 SOFiSTiK 2022

 Input Description | DYNA

SOLV Description Unit Default

VAL 1 Direct Skyline Solver (Gauss/ Cholesky) − !

No additional parameters are required. However it is mandatory to optimize the equation num-
bers in SOFIMSHA/C using (CTRL OPTI 1) or (CTRL OPTI 2) in order to minimize computation
time as well as storage requirements.

Iterative equation solver

SOLV Description Unit Default

VAL 2 Iterative equation solver − !

V2 Maximum number of iterations − *

V3 Tolerance in numeric digits (5 to 15) − *
V4 Type of preconditioning: − 1
0 Diagonal Scaling (not recommended)
1 Incomplete Cholesky
2 Incomplete Inverse

V5 Threshold value of preconditioning − *

V6 Maximum bandwidth in preconditioning − *

The iterative solver uses a conjugate gradient method in combination with preconditioning. For
the preconditioning, following variants are supported:

• Diagonal scaling (V4=0)

Although this is the fastest method with the least memory requirements, it will need a
considerable high amount of iterations and is therefore not recommended in most cases.
• Incomplete Cholesky (V4=1)
This type of preconditioning performs a partial triangulization of the input matrix. Compared
to a full triangulization with the Cholesky method, the Incomplete Cholesky saves time by
ignoring the so called Fill-In during decomposition.
• Incomplete Inverse (V4=2)
This type of preconditioning is generally inferior to the Cholesky method. This applies to
the convergence-rate as well as the time required for computing the inverses. It shows
however better performance in case of more densely populated matrices (Recommended
threshold V5: 0.01).

For any kind of preconditioning the number of matrix entries taken into account during precon-
ditioning can be reduced either by giving a relative threshold value at V5 or via a maximum
bandwidth size at V6. The optimum choice depends on the type of the structure and may only
be found by some tests.

Hint

SOFiSTiK 2022 3-9

DYNA | Input Description

(Hint continued...)

The correctness of the solution of the iterative solver depends primarily on the tolerance
threshold. Therefore, changing the default setting V3 is not recommended. In any case
the analyst should carry out a proper assessment of the computation results.

Direct Sparse LDL Solver (Default)

SOLV Description Unit Default

VAL 3 Direct Sparse LDL Solver − !

Additional parameters are not required. The mesh generators SOFiMSHA/C generate by de-
fault an equation numbering required for this type of solver which minimizes the so-called Fill
In of the matrix.

PARDISO - direct parallel sparse solver

SOLV Description Unit Default

WERT 4 Direct parallel sparse solver − !

This solver PARDISO uses processor optimized high performance libraries from the Intel Math
Kernel Library MKL. It usually provides the least computing times. It does not require an a
priori optimization of the equation numbers during system generation. Hence, the equation
optimization in SOFiMSHA/C could also be deactivated using (CTRL OPTI 0) in order to save
memory during system generation. On the other hand however, this solver does not allow
reusing the factorized stiffness matrix in other programs. Thus, a usage in combination with
the program ELLA is not possible.

3.4.2 CORE Parallel computation control

CORE Description Unit Default

VAL Number of used threads − *

SOFiSTiK supports parallel computing for selected equation solvers. Additionally, some pro-
grams offer parallel element processing capabilities – independent of the chosen equation
solver (CTRL SOLV).

Activation of parallel computing

By default parallel computing is triggered automatically where it is feasible.

Parallel computing requires corresponding hardware and operation system support. In addi-
tion, availability of an adequate SOFiSTiK license is obligatory.

Hint

3-10 SOFiSTiK 2022

 Input Description | DYNA

(Hint continued...)

Parallel computing requires availability of a HISOLV license (ISOL granule).

Number of available threads for parallel computing

If parallel computing is active, the number of adopted threads is determined as follows (listed
with increasing priority):

a) The software retrieves the information about the number of available physical processor
cores on the system. This number defines the default number of threads that are used
when a parallel computation is activated.
b) This default can be modified via the environment variable SOF_NUM_THREADS, which is
also available as sofistik.def parameter.
c) Finally, an explicit statement CTRL CORE NN (or as relative input CTRL CORE NN[%])
temporarily assigns the number of available threads for the respective run.

Hint
Neither option b) nor option c) state an explicit parallel computation request. The deci-
sion if a parallel computation is triggered, depends on the actual analysis option (parallel
processing must be supported for the specific task) and the availability of an adequate
license. Parallel computing can be suppressed by explitly setting the number of available
threads to 1 (or 0).

Parallel options for equation solvers

License
Solver CTRL SOLV Serial Parallel
Skyline Gauss/ Cholesky 1 – n.a.
Iterativ 2 HISOLV HISOLV
Sparse LDL (default) 3 – n.a.
Sparse Parallel (Pardiso) 4 HISOLV HISOLV

SOFiSTiK 2022 3-11

DYNA | Input Description

3.5 GRP – Selection of Element Groups

See also: ECHO, CTRL, MASS, EIGE, MODD, STEP, LC, CONT,
HIST, EXTR
GRP

Item Description Unit Default

NO Group number − !
VAL Selection LT FULL
OFF do not use
YES use, but do not print
FULL use and print the results
SOIL elements define boundary to halfspace for
SBFEM
CS Number of the construction stage − -
CSDL Number of self weight construction stage − CS
FACS Factor of the group stiffness − 1.0
HING blocked degrees of beam hinges Lt16 -

RADA Mass proportional damping 1/ sec 0.0

RADB Stiffness proportional damping sec 0.0
MODD Modal damping − 0.0

FACP Factor of the primary stresses − 1.0

FACM Factor of the masses of the group − 1.0

WIND Options for Wind-Loading − -

AERO Factor for aeroelastic effects − 1.0
EIGL Estimate of lateral bending frequency (WIND) Hz -
EIGB Estimate of main bending frequency (WIND) Hz -
EIGT Estimate of torsional frequency (WIND) Hz -

LMAX Limiting Slenderness for buckling length − -

NSCP Node number for the scaling point of the SBFEM − *

Method

All elements are used if nothing is input. When there is input, only the specified groups get
activated. This effect has to be especially taken care of, if only some groups get a damping
assigned to. If properties are defined in mutliple secondary groups for the same element, the
definition of the lexically last will win.

The elements of a group can be provided with two damping types. The value RADA represents
an external damping proportional to the mass and thus the excursion (e.g. air or water).The
value RADB represents an internal damping proportional to the stiffness (material damping).

3-12 SOFiSTiK 2022

 Input Description | DYNA

The geometric initial stress stiffness will not be multiplied with the factor RADB in general. Only
for the cable the prestress defined with the element is contributing to the damping.

C = RADA · m + RADB · K
kNsec/ m = 1/ sec · Nsec2 / m + sec · kN/ m

For a eigenvalue analysis it is possible to specify a modal damping for every group. This
value is then converted for every eigen frequency using the stifness matrix to an approximate
equivalent modal damping of every eigen form. A mixture of MODD with RADA or RADB is not
allowed.

The definition of modal damping parameters will have no effect in case of EIGE REST.

More explanations for the damping you will find at MODD.

Aeroelastic effects may be activated if the selected primary load case has a wind field. As
these effects are automatically included in a transient analysis, a definition is only suitable
for a modal analysis. WIND 1 activates aeroelastic stiffness, WIND 2 aeroelastic damping,
WIND 3 both effects. With AERO a factor on these conributions may be defined. The values
EIGL, EIGB and EIGT may be used to specify these values more precisely than with STEP. A
definition may be required to activate the derivativa a little bit earlier.

The description of the half space with the ”Scaled Boundary Element Method” (SBFEM) allows
to define the respective static and dynamic properties of the infinite space accounting for the
radiation damping properties. GRP selects the boundary elements of a 2D Analysis or the
QUAD elements of a 3D analysis defining the boundary of the half space. The local z axis
must show into the direction of the half space. Without definition of NSCP the scaling point will
be located on the upper centre of all soil interface nodes. For more details about SBFEM, see
[8].

SOFiSTiK 2022 3-13

DYNA | Input Description

3.6 MAT – General Material Properties

MAT

Item Description Unit Default

NO Material number − 1
kN/ m2
 
E Elastic modulus 1088 *
MUE Poisson’s ratio (between 0.0 and 0.49) − 0
automatic default for nonlinear materials NMAT
kN/ m2
 
G Shear modulus 1088 *
kN/ m2
 
K Bulk modulus 1088 *
kN/ m3
 
GAM Specific weight 1091 25
kN/ m3
 
GAMA Specific weight under buoyancy 1091 *
ALFA Thermal expansion coefficient [1/ K] 1083 E-5
kN/ m2 1088
 
EY Anisotropic elastic modulus Ey E
MXY Anisotropic Poisson’s ratio m-xy − MUE
OAL Meridian angle of anisotropy about the local x axis deg 0
OAF Descent angle of anisotropy about the local x axis deg 0
SPM Material safety factor − 1.0
TITL Material name Lt32 -

Materials which can be used for SVAL or QUAD and BRIC elements may be defined with the
record MAT and MATE. The number of the material must not be used for other materials.

The differences between the two records are mainly the used dimensions. MATE is analogue to
CONC,STEE etc. (MPa) and has additional strength values, while MAT uses (kN/m2 ) analogue
to NMAT. MAT has older item names for the orthotropic parameters.

3-14 SOFiSTiK 2022

 Input Description | DYNA

3.7 BMAT – Elastic Support / Interface

BMAT

Item Description Unit Default

NO Material number − 1

kN/ m3
 
C Elastic constant normal to surface Cs 1097 0.
kN/ m3
 
CT Elastic constant tangential to surface Ct 1097 0.

kN/ m2
 
CRAC Maximum tensile stress of interface 1089 0.
kN/ m2
 
YIEL Maximum stress of interface 1089 -
MUE Friction coefficient of interface − -
kN/ m2
 
COH Cohesion of interface 1089 -
DIL Dilatancy coefficient − 0.
t/ m2
 
GAMB Equivalent mass distribution 1184 0

TYPE Reference type LT NON

NON Unspecific
RIGD Rigid Frictional Interface
PESS Plane stress condition
PAIN Plane strain condition
HALF Circular bedding at half space surface
CIRC Circular hole in infinite disk
SPHE Spherical hole in infinite space
MREF Number of a reference material − NO
H Reference dimension [m] 1001 !
(thickness H or radius R)
RFAC Interface strength: Strength reduction factor related [m] - 0.0
to reference material’s shear strength

BMAT defines for an existing material (e.g. MATE/CONC/BRIC) properties for elastic support.
For a QUAD element it is thus possible to select the properties of the plate and the soil by a
single material number. The support stiffness may be calculated from the elasticity constants
of MREF, a geometric dimension H and a specific geometry rule. The bedding approach works
according to the subgrade modulus theory (Winkler, Zimmermann/Pasternak). It facilitates the
definition of elastic supports by an engineering trick which, among others, ignores the shear
deformations of the supporting medium. The bedding effect may be attached to beam or plate
elements, but in general it will be used as an own element. (see SPRI, BOUN, BEAM or QUAD
and the more general description of BORE profiles)

The determination of a reasonable value for the foundation modulus often presents consider-

SOFiSTiK 2022 3-15

DYNA | Input Description

able difficulty, since this value depends not only on the material parameters but also on the
geometry and the loading. One must always keep this dependance in mind, when assessing
the accuracy of the results of an analysis using this theory.

The subgrade parameters C and CT will be used for bedding of QUAD elements or for the de-
scription of support or interface conditions. A QUAD element of a slab foundation will thus have
a concrete material and via BMAT the soil properties attached to the same material number.
The value C is than acting in the main direction perpendicular to the QUAD surface in the local
z-direction, while CT is acting in any shear direction in the QUAD plane.

If subgrade parameters are assigned to the material of a geometric edge (GLN), spring ele-
ments will be generated along that edge based on the width and the distance of the support
nodes.

3.7.1 Derivation of bedding constants from reference material

Sometimes it is convenient to derive the subgrade moduli from a reference material’s elasticity
modulus E, shear modulus G and Poisson’s ratio μ[1] by providing a reference dimension H
and the type of support:

• Unspecific:
E G
Cs = Ct = (3.1)
H H
• Planar layer with horizontal constraints e.g. for modeling elastic support by columns and
supporting walls (plane stress condition):
E 1 G
Cs = · Ct = (3.2)
H (1 + μ)(1 − μ) H

• Planar layer with horizontal constraints for settlements of soil strata (plane strain condition):
E (1 − μ) 2·G
Cs = · Ct = (3.3)
H (1 + μ)(1 − 2μ) H

• Equivalent circular disk with radius R on an infinite halfspace:

E 2
Cs = · Ct = Cs (3.4)
R π(1 + μ)(1 − μ)

• Circular hole with radius R in infinite disk with plane strain conditions (bedded pipes or
piles):
E 1
Cs = · Ct = Cs (3.5)
R (1 + μ)(1 − 2μ)

• Spherical hole with radius R in infinite 3D elastic continua:

E 2
Cs = · Ct = Cs (3.6)
R (1 + μ)

Including a dilatancy factor describing the volume change induced by shear deformations, we
have for the bedding stresses the following equations depending on the normal and transverse

3-16 SOFiSTiK 2022

 Input Description | DYNA

displacements:

σ = Cs · (s + DL · t ) τ = Ct · t (3.7)

Non-linear effects are controlled by CRAC, YIEL, MUE and COH:

Cracking: Upon reaching the failure stress, the interface fails in both the axial
and the lateral direction. The failure load is always a tensile stress.
If the bedding reaction is applied to a QUAD element, a deformation
in the direction of the local z-axis will create compressive (negative)
stresses.
Yield load: Upon reaching the yield stress, the principal deformation component
of the interface increases without an increase of the stress.
Friction/cohesion: Defining a friction and/or a cohesion coefficient, the lateral shear
stress can not become larger than:
Friction coefficient * normal stress + Cohesion
Please note, that before reaching this limit the stiff-ness CT will pro-
duce the shear stress only if a deformation is present.

If the principal interface has failed (CRAC), then the lateral bedding acts only if 0.0 has been
entered for both friction- coefficient and cohesion.

The non-linear effects can only be taken into account by a non-linear analysis. The friction is
an effect of the lateral bedding, while all other effects act upon the principal direction.

3.7.2 Rigid Frictional Interface

The Rigid Frictional Interface describes the mechanical behavior of a thin interface layer inbe-
tween two bodies with frictional contasct by adopting a friction law of Coulomb type. For the
interface layer a quasi-rigid behavior is assumed.

A typical field of application is the modeling of soil-structure interaction in geotechnical analy-

ses. The Rigid Frictional Interface Element (I2D) is available in the FEA module TALPA.

Hint
For the Rigid Frictional Interface Element (I2D), an explicit provision of stiffness parame-
ters is normally not required; instead, the program establishes appropriate stiffness val-
ues automatically in a mesh adaptive manner. Explicit provision of stiffness parameters
overrides this mechanism, the provided values are adopted unmodified.

Interface strength is determined from the reference material by means of the strength
reduction factor:

MUE = RFAC · tnφreƒ (3.8a)

COH = RFAC · creƒ (3.8b)
RSS = ƒt,reƒ (3.8c)

Alternatively, a direct input of strength parameters is supported, too.

SOFiSTiK 2022 3-17

DYNA | Input Description

(Hint continued...)

Dilatancy coefficient and equivalent mass distribution have no effect on the interface be-
havior.

3-18 SOFiSTiK 2022

 Input Description | DYNA

3.8 SMAT – SBFEM - Material Properties

SMAT

Item Description Unit Default

NO Material number − 1

LC Characteristic Length m !
EX Coefficient of elasticity in X-direction − 0
EY Coefficient of elasticity in Y-direction − 0
EZ Coefficient of elasticity in Z-direction − 0
RHOX Coefficient of density in X-direction − 0
RHOY Coefficient of density in Y-direction − 0
RHOZ Coefficient of density in Z-direction − 0
ALF Inhomogenity of elasticity − 0
BET Inhomogenity of density − 0

Record SMAT defines dependency of material values for the SBFEM:

|| α |y| α |z| α

–       ™
E = Ereƒ · E + Ey + Ez (3.9)
Lc Lc Lc

 
|| β |y| β |z| β
     
ρ = ρreƒ · ρ + ρy + ρz  (3.10)
Lc Lc Lc

The material number of the soil is taken from the material number of the SBFEM-QUAD-
element in 3D. The material number of the soil is taken from the material number of the
SBFEM-BOUN-eement in 2D. If no material number has been assigned to BOUN element,
material with number 1 is used.

SOFiSTiK 2022 3-19

DYNA | Input Description

3.9 MASS – Lumped Masses

See also: ECHO, CTRL, GRP, EIGE, MODD, STEP, LC, CONT, HIST,
EXTR
MASS

Item Description Unit Default

NO Node number − !
MX Translational mass t 0.
MY Translational mass t MX
MZ Translational mass t MX
MXX Rotational mass tm2 0.
MYY Rotational mass tm2 0.
MZZ Rotational mass tm2 0.
MXY Rotational mass tm2 0.
MXZ Rotational mass tm2 0.
MYZ Rotational mass tm2 0.
MB Rotational mass tm2 0.

LC Load case from the CDB − !

PRZ Percentage of permanent load for mass LT/ − 100
PSI0,PSI1,PSI2 (combination value)
SELE Selection of a load direction LT PG
PG, PXX, PYY, PZZ (translatoric)
M-XX, M-YY, M-ZZ (rotatoric)
ALL (every component on its own)

The masses are additional to the primary masses defined in program SOFIMSHA in the
database. They are maintained over several input sets until they are redefined. They are
not effective as dead load in static load cases as do the primary masses, but the have also
a group number of the last GRP-record (allowing mass proportional damping definitions) but
they are always active. MASS 0 can be used to delete all temporary masses, thus the primary
masses from SOFIMSHA are kept.

A mass acts usually the same in all three coordinate directions and thus, it need to be defined
independently only for special cases. Rotational masses with inclined axis will have off diagonal
masses MXY till MYZ.

The dead weight of the entire structure is always applied in the form of translational masses.
If necessary, rotational masses must be defined separately with MASS or CTRL MCON 3. If the
dead weight of a structure is not to be applied, the dead weight of the material or the cross
section should be input as zero.

MASS can be used also to import nodal loads from the database as masses to DYNA. The load

3-20 SOFiSTiK 2022

 Input Description | DYNA

case number must then be defined at LC. The values MX till MZ, with default value of 1.0, are
then the factors for the individual directions of the mass components which are generated from
the loads in the dead weight direction. If other load directions are to be converted to masses
as well, these directions have to be specified additionally with SELE. The input:

+PROG DYNA
HEAD
!...
MASS LC 12 PSI2
!...
END

creates translational masses from all loads of load case 12 in the direction of the dead weight.
If the load case has an action assigned, the permanent combination value ψ2 will be taken as
factor. By contrast the input

+PROG DYNA
HEAD
!...
MASS LC 12 SELE PZ MX 1.0 0.5 1.0
MASS LC 13 50[%] SELE PY MX 0.0 1.0 0.0
!...
END

creates masses (t) in the x and z direction from all PZ loads (kN) of load case 12. Only half
of the mass is activated in the y direction, however. The second input processes 50% of the
PY-loads of the load case 13 only in Y-direction.

Masses can get also a factor with MASS. For this purpose the literal FACT has to be input for
NO. This can be reasonable particularly for larger systems, where it is favourable to suppress
many low frequencies which are not essential for the analysis. With the input

MASS FACT MZ 0.01

The mass in global Z direction is reduced to one percent only.

SOFiSTiK 2022 3-21

DYNA | Input Description

3.10 EIGE – Eigenvalues and Eigenvectors

See also: ECHO, CTRL, GRP, MASS, MODD, STEP, LC, CONT,
HIST, EXTR
EIGE

Item Description Unit Default

NEIG Number of sought eigenvalues − !

TYPE Method for the eigenvalue computation LT SIMU
REST Eigenvalues already available
SIMU Simultaneous vector iteration
LANC Method of Lanczos
BUCK Buckling (Vector iteration)
BULL Buckling (Lanczos)
NITE Number of iterations − *
Lanczos vectors
MITE Maximum number of iterations and options − *
LMIN Eigenvalue shift 1/ sec2 0
STOR Number of eigenform to be stored − NEIG
LC Load case number of lowest eigenform − 1
OPT Options − *
1 Include Missing Mass

The input of EIGE requests calculation of the eigenvalues and the mode shapes. If the eigen-
vectors have been already computed, one must enter TYPE REST. Special attention must be
paid to this when importing eigenvalues from program ASE.

Eigenvalues and forms may represent dynamic vibration modes or buckling eigenforms. While
the first uses a well defined positive definite mass matrix, the second problem may encounter
indefinite geometric stiffness matrices (negative Eigenvalues) and establish problems. Only
SIMU has provisions for that type of problem. In any case you should start with a few Eigen-
values in those cases.

The mode shapes can be stored in the database similarly to static load cases and can be then
represented graphically as deformed structure. A modal evaluation of forces is possible only
when all required mode shapes have been stored also as stresses or forces of the elements.

With OPT 1 up to three extra load cases are created, containing the rigid body response of
the missing masses. These will be included in the analysis if a uniform ground acceleration is
applied as loading. For a restart the number of the eigenvalues must then match the number
at the evaluation analysis. For other load cases there is a simplified more general method
available via LC BASE AUTO.

The eigenvalue problem can be shifted by one value. This finds application in structures that
are not supported (zero eigenvalue is the smallest value) as well as in checking the number of
eigenvalues by means of a Sturm sequence. The number of skipped eigenvalues is manifested
during the shift by the number of sign changes of the determinant.

3-22 SOFiSTiK 2022

 Input Description | DYNA

The choice of method for the eigenvalue analysis depends on the number of the eigenval-
ues. The simultaneous vector iteration is used in most cases. The number of iterations can
be reduced when a somewhat expanded subspace is used for the eigenvalue iteration. For
that reason the default value for NITE is the minimum between NEIG+2 and the number of
unknowns. The iteration is terminated when the maximum number of iterations (default max
(15, 2 · NTE)) is reached or when the highest eigenvalue has only changed by a factor less
than 0.00001 compared to the previous iteration.

The method of Lanczos is significantly quicker than the vector iteration, when a large number
of eigenvalues is sought. A good accuracy is achieved when the number of vectors NITE is at
least double the number of sought eigenvalues (default). In case of NITE=NEIG, by contrast to
the vector iteration, the higher eigenvalues are usually worthless.

If a primary load case is selected with CTRL PLC, the geometric initial stiffness is included in the
eigenvalue analysis. So you will get the frequency zero if you are approaching a buckling case.

In that case you may however evaluate the buckling eigenform directly via TYPE BUCK (or
BULL).

Vektoriteration Lanczos
Number of Eigenvalues moderate high
Range of Eigenvalues Ritz-Step problematic no problems
multiple Eigenvalues yes yes
missing Eigenvalues very rare rare
negative Eigenvalues yes does not work
Memory requirement moderate high
Speed moderate fast

Overview of the algorithms

SOFiSTiK 2022 3-23

DYNA | Input Description

3.11 MODD – Modal Damping

See also: ECHO, CTRL, GRP, EIGE

MODD

Item Description Unit Default

NO Number of the eigenvalue − all

D direct Lehr’s modal damping − 0
A Mass proportional damping 1/ sec 0
B Stiffness proportional damping sec 0
PERS Persistent value LT Y
Y/N = YE/NO

The damping may be specified within the GRP record with different values for each group.
When using direct integration, these values will become effective in just this way. For a modal
analysis however the modal damping will be calculated, following the computation of the eigen-
values, from the defined damping values by a diagonalisation process. Each Eigenform will
then have one distinct modal damping value.

However the modal damping (Lehr’s damping factor), can also be defined separately for each
mode by three independent parts (direct value of D, mass proportional A and stiffness propor-
tional B). The values are stored in the database. The definition of this value will overwrite any
damping definitions in the GRP record or from explicit damper elements!

As the values in the literature are mostly given as modal damping values or logarithmic decre-
ments δ we will give some important formulas:
δ 1 A 1
d = =D+ · + · B · ω (3.11)
2π 2 ω 2

In the next pictures you will see the influence of the factors A and B depending on the eigenfre-
quencies of a SDOF-oscillator. The damping is shown as logarithmic decrement δ,describing
the ratio of two consecutive amplitudes A1 and A2.
A1
 
logarithmic dekrement δ = og (3.12)
A2

3-24 SOFiSTiK 2022

 Input Description | DYNA

Verschiebun g SY [mm]

A1
1600.00
0

A2
1400.00
0

1200.00
0

1000.00
0
N

800.00
0

600.00
0

400.00
0

200.00
0

Zeit
0.00
0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 [sec]

Figure 3.1: logarithmic dekrement

The decrement δ is related to the modal damping with a factor of 2π . Usual values for the
modal damping D are (Müller [9]):

elastic conditions [%] plastic conditions [%]

Reinforced concrete 1-2 7
Prestressed concrete 0.8 5
bolted steel constructions 1 7
welded steel constructions 0.4 4

Hint
MODD have to be specified as absolute value or with an explicit unit [%] !

For a direct integration without eigenvalues, there is no modal damping, thus it is necessary
to convert a given damping value to the parameters A and B. The conversion of parameters A
and B can be seen from the next picture.

SOFiSTiK 2022 3-25

DYNA | Input Description

B = 0.0005
50
B = 0.001
45
B = 0.002
40
B = 0.005
35
A = 0.1
30
A = 0.2
25 A = 0.5
20 A = 1.0

15 ALF = 0,3 DEL = 0,52

Dekremente [%]

10 ALF = 0,25 DEL = 0,55

ALF = 0,4 DEL = 0,52

5
ALF = 0,4 DEL = 0,55
0
0.5 1 2 5 10 20 50 ALF = 0,17 DEL = 0,5 THE
Eigenfrequenzen [Hz] = 1,4

Figure 3.2: Parameter A and B

To achieve a 10 % logarithmic decrement for a frequency of 5 Hz you may either define A as

1.0 or B as 0.001. In general you want to define the damping between two frequencies f1 and
f2 with a relatively constant decrement. You have then to specify a combination of A and B
given by: (circular frequencies ω = ƒ · 2 · π)
ξ1 · ω2 − ξ2 · ω1
A = 2 · ω1 · ω2 · (3.13)
ω2 2 − ω1 2

ξ2 · ω2 − ξ1 · ω1
B=2· (3.14)
ω2 2 − ω1 2

If the damping at the start of the interval should be equal to the damping at the end of the
interval and by converting to the standard frequencies ω = ƒ · 2 · π we have:

ƒ1 · ƒ2
A = ξ · 4π · (3.15)
ƒ1 + ƒ2

1
B=ξ· (3.16)
π · (ƒ1 + ƒ2 )

Example: A structural steel with bolted connections should have a mean modal damping of
0.01 between 2 Hertz and 10 Hertz. We thus have a decrement of 2 · π · ξ = 2 · π · 0.01 = 0.063
i.e. the amplitude of a free oscillation should reduce by 6.3 % from peak to peak within the
range from 2 Hertz to 10 Hertz.

3-26 SOFiSTiK 2022

 Input Description | DYNA

(ƒ2 −ƒ1 ) Š (10−2)

Factor is given by = 0.083. Thus:
(102 −22 )
€
ƒ2 2 −ƒ1 2

2 · 10
A = 4 · π · 0.01 · = 0.21 (3.17)
2 + 10

1 0.01
B= · = 0.000266 (3.18)
π 2 + 10

To check this we have from the diagram or the formulas:

at 2.0 Hertz at 10.0 Hertz at 5.0 Hertz

from A 0.0525 0.0105 0.021
from B 0.0105 0.0525 0.026
total 0.0630 0.0630 0.047

i.e. at the bounds of the interval we have the desired damping, but between we have a little bit
less. For 5.0 Hertz we have only d = 0.047.

For a direct integration there is an additional numerical damping effect possible with the selec-
tion of the integration constant BET, DEL and THE. The default (BET = 0,25; DEL = 0,5; THE
= 1) will not have any damping effect. The same is valid for modal analysis there is also no
damping effect, because the equations are integrated exactly.

SOFiSTiK 2022 3-27

DYNA | Input Description

3.12 STEP – Parameter of the Step-wise Integration

See also: ECHO, CTRL, GRP, MASS, EIGE, MODD, LC, CONT,
HIST, EXTR
STEP

Item Description Unit Default

N Number of time steps or divisor − 10

DT Time step or total time (sec) − 0.1
INT Output interval every INT steps −/ LT 1

A Mass proportional damping 1/ sec 0.

B Stiffness proportional damping sec 0.

BET Parameter of the integration method − 1/4

DEL − 1/2
THE relative time value for equilibrium − 1.

EIGL Estimate of lateral bending frequency (WIND) Hz -

EIGB Estimate of main bending frequency (WIND) Hz -
EIGT Estimate of torsional frequency (WIND) Hz -
EIGS Estimate of soil frequency (SBFEM) Hz -
DTF Number of steps for SBFEM convolution − *
STHE Extrapolation factor for SBFEM − 1.4

STEP prescribes the type of time-dependent analysis. Four cases must be distinguished:

3.12.1 Static Analysis

Static analysis will be perfomed if no STEP and no EIGE record is defined. All load cases will
be analyzed separately.

3.12.2 Steady State Analysis

Steady-state or stochastic excitations will be analyzed in the frequency domain if no STEP or
STEP ≤ 0 is defined and an EIGE record is also present.

Only periodic or spectral load functions will be used. Within any load case the phase shift
between the eigenforms will be considered to obtain the maximum response. The extremes of
all load cases will then be combined according to the CTRL STYP definition.

If N is < 0 a window of the periodic response is visualized including the phase shift. N is the
number of data points, DT is the size of the window, defaulting to the load frequency. In that
case all load cases will be added including the phase shift.

3-28 SOFiSTiK 2022

 Input Description | DYNA

3.12.3 Frequency Dependent Steady State Analysis

If the literal TIME or FREQ is input for INT, a response analysis for the periods or frequencies
in the region from 0 to N · DT times the specified load function frequency 2π/ T0 is performed.
If the literal STIM or SFRE is given, the load is normalised to a constant displacement instead
of the acceleration. If the literal VTIM or VFRE is selected, the normalisation is based on a
constant velocity.

All load cases are treated as with the steady state analysis.

3.12.4 Transient Analysis

Transient analysis of a time segment with duration N·DT by direct numerical or analytical modal
(if EIGE is present) integration is obtained via:
STEP N > 0
When N <1 is input, DT is interpreted as total time and the individual time step becomes N · DT .

All load cases and load functions are applied simultaneously.

For the numerical analysis there are four subtypes

• THE = 0.0 for the explicit integration (extra license reuqired)

• THE = 1.0 for the Newmark method (Default constant average)
• THE ≥ 1.4 for the Wilson method (BET = 1/ 6, DEL = 1/ 2)
• THE < 1.0 for the Hughes-Alpha method (0.7 < THE < 1.0)

A suitable size of the time step depends on the frequency of the expected response. In case
of the direct method components with periods smaller than about ten times the time step are
damped out of the solution. A comparison analysis should be performed if in doubt with a step
approximately equal to one fourth of the initial time step.

It should be taken care of the fact, that the standard Newmark-Method has no numerical damp-
ing. Thus small errors may amplify easily. Those errors may be introduced by a time step
chosen to small together with consistent mass matrices. In that or other cases the integration
constant should be modified, eg.:

• Definition of DEL > 0.5 for a damped Newmark-Method

• Definition of THE > 1.4 for the Wilson Method
• Definition of THE < 1.0 for the Hughes-Alpha Method

For those dynamic effects requiring convolution (Wind-Derivativa, SBFEM) the performance
and the precision depend on the number of the latest time steps, which may be controlled by
chracteristic frequencies.

Wind derivativas use the values EIGL, EIGB and EIGT.

SBFEM uses EIGS, without a definition the full convolution integral is evaluated, which creates
an effort quadratic in time. One component to reduce the effort is the evaluation of those
matrices in larger intervals and to interpolate between. The value DTF defines a multiple on
the system time step to be adopted to the properties of the far field. The essential paraemter
is the limit of the total convolution by the total number of matrices M. This will be defined by the

SOFiSTiK 2022 3-29

DYNA | Input Description

frequency EIGS = 1/(DT*DTF*M).

3-30 SOFiSTiK 2022

 Input Description | DYNA

3.13 LC – Load Case

See also: ECHO, CTRL, GRP, MASS, EIGE, MODD, STEP, CONT,
HIST, EXTR
LC

Item Description Unit Default

NO Load case number − *

FACT Factor for all loads of the load case − 1.0
DLX Dead weight factor in X direction − 0.0
DLY Dead weight factor in Y direction − 0.0
DLZ Dead weight factor in Z direction − 0.0
BASE Number of a static reference load case − -
STAT load case is included as static
AUTO load case is factorized as missing mass
nn load case is used as influence line
MODB Modal base load case number − -
TITL Identifier of the load case Lt32 -

The loading in DYNA is subdivided in load cases identified by a number. Each load case can
be assigned in program SOFiLOAD loadings, time-functions or a response spectrum. The use
of the load cases differs according to the computational procedure:

1. Transient analysis
During a time variation analysis (STEP N > 0) all the selected load cases and their functions
define the time dependence of the loading and the starting time. All functions act with their
loads simultaneously upon the structure. DYNA allows the extra definition of a contact
condition CONT for a moving load.
2. Static or steady-state analysis
In case of steady-state analysis the periodical loads are converted to corresponding re-
sponses according to Section 2.4. In case of static analysis the load cases are analysed
separately.

If spectra are defined, DYNA computes by double interpolation of all the spectra a system
response, which is then superimposed by statistical methods according to the input for CTRL
STYP.

With BASE STAT an static load cases is added to the modal superposition completely. With
BASE AUTO the fraction of the load case according to the "‘missing mass method"’ is evalu-
ated. The load case with the equivalent forces of the accelerations has to be calculated be-
forehand. BASE with a number is used to introduce influence functions, needed for separated
ground accelerations.

For a modal analysis the general case is to apply the same load vector for all eigenforms.
However if every eigenform should obtain a separate loading as in a modal wind analysis, the
item MODB allows to specify the load case number for the loading to the first eigen form. All
following eigenforms will be associated to the consecutive load case numbers.

SOFiSTiK 2022 3-31

DYNA | Input Description

In this case there is a special feature for spectral loading:

For the load cases the user defined parameter CRIT will be evaluated. If it is zero, no evaluation
will be taken from the spectra. The response is 1.0. If defined otherwise the period of the
eigenforms will be scaled with that value. Thus for a Wind spectrum this value is to be defined
with Ltrb / men / or z/ men / .

3-32 SOFiSTiK 2022

 Input Description | DYNA

3.14 CONT – Contact and Moving Load Function

See also: ECHO, CTRL, GRP, MASS, EIGE, MODD, STEP, LC, HIST,
EXTR
CONT

Item Description Unit Default

TYPE Selected contact value (obsoleted) LT -

REF Ident of a reference axis for load trains − 0
NR Number of an edge element − -
V Travel speed m/ sec -
YEX Local Eccentricity m 0.0
TMIN Time at start of travel sec 0.0
LCUV Load case for vertical track irregularities − -
LCUT Case for transverse track irregularities − -
LCUR Case for rotational track irregularities − -

Dynamic contact is governed by a changing location of contact point within time, as it is given
in the case of a vehicle travelling along a bridge.

This record allows the definition of the development in time of the contact point and a mecha-
nism to create loads based on current deformations of the system.

For the time dependant location you have to select a sequence of nodes and specify the time
value for each node when the contact point is exactly at that point. In most cases the selection
of the number NO of a boundary/edge element will select all nodes in the given sequence.
However explicit definitions with FUNC (program SOFiLOAD) records and mixing and concate-
nation of several elements is possible. Defining a travelling speed V will generate all the needed
time values from the distance either directly or taken along the reference axis and the optional
start time TMIN.

If the load case has a load train created within SOFiLOAD, all loads of the train will follow each
other with the appropriate distance. If the load train has also structural- or visualisation objects
created via the TREX command, the nodes of those objects will receive the current coordinates
as displacements and the absolute velocities. Only the point loads are processed by the CONT
command.

The three load cases LCUV, LCUT and LCUR allow to introduce track irregularities or uneven
pavements as additional displacements or rotations for the contact-point. The load functions
of these load cases must have the absolute displacements as a function of the travelling time
of the load reference point.

The reference displacement at the contact point is obtained by a linear interpolation between
the adjacent nodes. The loading at the contact point is similarly distributed between the ad-
jacent nodes. External nodal loads are placed at the contact point only if the node number is
specified as 0.

SOFiSTiK 2022 3-33

DYNA | Input Description

For the definition of a load train moving along a bridge according to DIN FB 101 / EC1 the input
may be as follows:

+PROG SOFILOAD
HEAD DEFINE A DEFAULT LANE GEOMETRY
ECHO FULL
GAX 'AXIS' 0.0 X 0.0 0.0 R 150 NZ +1.0
'AXIS' 3.0 X 30.0 0.0 R 150
LC 191
TRAI RFAT 4 P4 0.0
TREX 191 901 900 11 1
END

+PROG DYNA
GRP 1,2,3
GRP 901 FACS 0.0
CTRL RLC 1001 7
LET#1 30.0 ! SPEED in m/sec
STEP 0.01 300.0/#1 ! TOTAL TIME FOR TRAVELING
LC 191 ! LOAD TRAIN
CONT REF AXIS NO 10 #1 2.0 ! AUTOMATIC TIMEVALUES IN NODES FROM EDGE
END

Further variants may be seen within the example dyna9_travelling_loads.dat and at

SOFiLOAD loadtrains.dat.

3-34 SOFiSTiK 2022

 Input Description | DYNA

3.15 HIST – Results within Time

See also: ECHO, CTRL, GRP, MASS, EIGE, MODD, STEP, LC,
CONT, EXTR
HIST

Item Description Unit Default

TYPE Result value (see table) LT S

FROM Smallest node or element number − 1
TO Largest node or element number − FROM
INC Increment or Identifier −/ Lt 1
RESU Output request LT SAVE
SAVE save values in database
PRIN save and print values
PHAS save amplitudes and phases
LCST Number of case to store in database − *
XREF Beam section or reference point for the m 0.
YREF sum of the spring force components m 0.
ZREF or dP/P width m 0.
DUMP Filename for a dump of the values Lt48 -

The record HIST requests the time history of particular values. These will be saved into the
database for the presentation with DYNR, but it is also possible to print the values directly or to
save them to an external dump file. Up to 32 values can be addressed per input record.

The computed maximum and minimum values of the curves will be printed in any case.

Table 3.21: Possible literals for TYPE

TYPE Meaning
U UX, UY, UZ
U-X U- U-Z Displacements
U-RX U-RY U-RZ Rotations
V VX, VY, VZ
V-X V-Y V-Z Velocities
V-RX V-RY V-RZ Angular velocity
A AX, AY, AZ
AX AY AZ Accelerations
ARX ARY ARZ Angular acceleration
P PT M Spring forces and moment
PX PY PZ Spring force total global components
PT/P DP/P Spring force ratios

SOFiSTiK 2022 3-35

DYNA | Input Description

Table 3.21: (continued)

TYPE Meaning
SP Sum of all spring force components
SPX SPY SPZ Sum of the spring force components
SPRX SPRY SPRZ Sum of the spring moment components
TRUS Truss-bar axial force
CABL Cable axial force
BEAM All Beam forces
N VY VZ Beam forces normal and shear
MT MY MZ Beam moments torsion and bending
SECT SIG TAU Stresses in sectional points
QUAD All Shell forces
MXX MYY MXY Shell moments
VXX VYY Shell shear forces
NXX NYY NXY Shell membrane forces
BRIC All continua stresses
TXX TYY TZZ Stresses of 3D continuum
TXY TXZ TYZ Shear stresses of 3D continuum
DSX DSY DSZ Impedance of translatoric degrees of freedom
DSRX DSRY DSRZ Impedance of rotational degrees of freedom

For the beam results XREF is used to define the section where the results are evaluated. A
negative definition is taken as the ratio of the section to the total beam length, thus a value of
-1.0 selects the end of the beam.

For the stresses INC is used to define the identifier of the stress point (SPT) within the section
where the stresses should be evaluated.

The spring force ratios may be useful for vehicle-structure-interaction. They are defined as
follows:

PT/P The ratio of the resulting transversal force PT to the main force P of a spring
DP/P The ratio of the difference of the main forces of two springs to the mean value of
the same spring forces:
ΔP P1 − P2
= (3.19)
P P1 + P2

INC=0 only between FROM and TO

INC>0 change relative to the prestressing force or:
For all springs a following rotational spring is searched having the
same nodes. Then the value ΔP will be established from the moment
and the plan view distance derived from values XREF to ZREF.

3-36 SOFiSTiK 2022

 Input Description | DYNA

The impedance for frequency analysis is defined by the two factors of the complex stiffness:
R(t)
K= =k+·ω·c (3.20)
(t)

SOFiSTiK 2022 3-37

DYNA | Input Description

3.16 EXTR – Evaluation of Max. Internal Forces and Moments

See also: ECHO, CTRL, GRP, MASS, EIGE, MODD, STEP, LC,
CONT, HIST
EXTR

Item Description Unit Default

TYPE Structural magnitude LT !

MAX Load case number for maximum − 0
MIN Load case number for minimum − 0
0 print only
STYP Superposition type statistical/steady-state LT *1
ADD sum of values
SUM sum of absolute values
SRSS square root of sum of squares
CQC Complete Quadratic Combination
ACT Action names of the results LT
See below

With these records the evaluation of maximum results is selected, which may be saved within
a result load case. It has to be distinguished. if every result record contains all the individual
extreme values or if the extreme key value is accompanied with corresponding forces.

The first option is applied for all bold literals, the second for all other. The first option cannot be
used for any superposition with other results in a meaningful way.

The following literals are possible for TYPE.

Table 3.23: For nodes

TYPE Designation
U Displacements and support forces
V Velocities
A Accelerations

Table 3.24: For beams (BEAM)

TYPE Designation
BEAM Maximum values for beam elements
(not usable for superposition)
N Normal force
VY Shear force Vy
VZ Shear force Vz
1 The default superposition type is determined by CTRL STYP V2 settings.

3-38 SOFiSTiK 2022

 Input Description | DYNA

Table 3.24: (continued)

TYPE Designation
MT Torsional moment
MY Bending moment My
MZ Bending moment Mz
MB Warping moment
MT2 Secondary torsional moment

Table 3.25: For truss elements (TRUS)

TYPE Designation
TRUS Forces in truss elements

Table 3.26: For cables (CABL)

TYPE Designation
CABL Forces in cable elements

Table 3.27: For springs (SPRI)

TYPE Designation
SPRI Maximum values for spring elements
(not usable for superposition)
P Spring force in main direction
PT Resultant spring force in transverse direction
p
PT = PTX 2 + PTY 2 + PTZ 2
PTX Spring component in global X direction
PTY Spring component in global Y direction
PTZ Spring component in global Z direction
M Spring moment
SP Sum of spring forces
SPX Sum of spring force components in X direction
SPY Sum of spring force components in Y direction
SPZ Sum of spring force components in Z direction
SPRX Sum of rotational spring forces about X direction
SPRY Sum of rotational spring forces about Y direction
SPRZ Sum of rotational spring forces about Z direction

SOFiSTiK 2022 3-39

DYNA | Input Description

Table 3.28: For plates/shells (QUAD)

TYPE Designation
QUAD Maximum values for QUAD elements
(not usable for superposition)
MXX Bending moment m-xx
MYY Bending moment m-yy
MXY Torisonal moment m-xy
VXX Shear force v-x
VYY Shear force v-y
NXX Membrane force n-xx
NYY Membrane force n-yy
NXY Membrane shear force n-xy
NZZ Membrane force n-zz

Table 3.29: For volume elements (BRIC)

TYPE Designation
BRIC Maximum values for BRIC elements
(not usable for superposition)
TXX Stress in global X direction
TYY Stress in global Y direction
TZZ Stress in global Z direction
TXY Shear stress in global XY plane
TXZ Shear stress in global XZ plane
TYZ Shear stress in global YZ plane

Table 3.30: For result sets (RSET)

TYPE Designation
RSET Maximum values for RSET elements
(not usable for superposition)
RS1 the first entry of each RSET
RS2 the second entry of each RSET
...
RS31 the 31th entry of each RSET

3-40 SOFiSTiK 2022

 Input Description | DYNA

Table 3.31: For storey sets (SLVL)

TYPE Designation
SLVL Maximum values for storeys
(not usable for superposition)
S_PX Shear in global X
S_PY Shear in global Y
S_PZ Force in global Z
S_MX Moment in global X
S_MY Moment in global Y
S_MZ Moment in global Z
S_UX Displacement in global X
S_UY Displacement in global Y
S_UZ Displacement in global Z
S_DX Drift in global X
S_DY Drift in global Y

Hint
Storey rotation and rotation drift in global Z are not available as maximum result selection
since the rigid diaphragm method is not used when calculating these results.

The maximum values are stored in the database, if a load case number is input for MAX and/or
MIN. For nodal values (U,V,A) only the maximum values are output as default. Use ECHO
DISP, VELO, ACCE or REAC to see all nodal results.

Special remarks on support forces

As the evaluation of support forces requires quite a demanding effort and the amount of data
is kept small, there is no direct option to evaluate them in DYNA. There are only the following
possibilities:

support forces within a result set

spring elements
modal analysis with Eigenvalues from ASE

To get extended results with correspondig values, the use of RSET is recommended. But
saving complete results with CTRL RLC might be also a solution.

Special remarks on the extreme values of response spectra

SUM offers an upper limit, SRSS may overestimate or underestimate the probable magnitude.
CQC usually gives the most reasonable results. The default value (CQC) may be changed with
CTRL STYP V2.

SOFiSTiK 2022 3-41

DYNA | Input Description

The maximum forces will become always positive for SUM, SRSS and CQC.

For the global maximum values(e.g. U, V, A, REAC, BEAM, QUAD, BRIC) the extrema will
be calculated for every force independently. The extrema will be collected and stored within
a single record. But the real response may be any positive or negative combination of these
individual values. This is inconvenient for design purpose, and it is definitely not suited for a
display with the program ANIMATOR.

At superposition of a single internal force on the other hand the corresponding internal forces
are formed in the same ratio with a linear combination, so that one can employ the complete
set of the internal forces and moments. The algorithm used for that has been invented by
SOFiSTiK and is therefore hardly to be found in other programs.

Thus the calculated forces have the correct corresponding signs, but the leading force ist still
positive. Thus the results have to be inserted in other tasks alternatively with a positive or a
negative sign multiplier. In some cases it is a simplification to apply a sign based on a reference
load case, assuring that the results are always unfavourable.

This can be achieved with a definition of a refernce load case with CTRL STYP V3.

The base accelerations are also included within the resulting nodal accelerations. Displace-
ments and velocities however are always relative to the free field movements of the soil.

The types SP and SPX, SPY, SPZ address the total sum of all components of support spring
forces in the global coordinate directions. The output is done group-wise. Types PTX to PTZ
have only the tangential components.

Action names of the results

At ACT the results can be assigned to a specific action for a later superposition. All actions
preset in program SOFiLOAD record ACT are possible here.

3-42 SOFiSTiK 2022

 Input Description | DYNA

3.17 ECHO – Extent of Output

See also: CTRL, GRP, MASS, EIGE, MODD, STEP, LC, CONT, HIST,
EXTR
ECHO

Item Description Unit Default

OPT A literal from the following list: LT FULL

NODE Nodal values
SECT Cross section values
ELEM Elements
MASS Masses in nodes
EIGE Natural frequencies
LOAD Loads
DISP Displacements
FORC Internal forces and moments
VELO Velocities
ACCE Accelerations
REAC Support reactions
STAT Warning for convergence check

FULL All the above options

VAL The extent of the output LT FULL

OFF Option complete deactivate
NO No output
YES Regular output
FULL Extensive output
EXTR Extreme output

The name ECHO must be repeated in each record to avoid confusion with similar record names
(e.g. CROS).

The default value is NO for NODE, CROS, and ELEM; for all others it is YES.

The warning no. 10918 (No convergence of the iterative equation solver in load vector) for
convergence checks can be switched off with ECHO STAT NO.

Some characteristic output tables are show in the following.

SOFiSTiK 2022 3-43

DYNA | Input Description

Table 3.33: Eigenvalues (ECHO EIGE)

Eigenvalues
No. Z LC Z λ Z error Z ω Z f Z T Z ξ Meff
[rad2/sec2] Z [-] Z[rad/sec] Z [Hz] Z [sec] Z [o/o] ZX[o/o] ZY[o/o] ZZ[o/o]
1 n 9001 n8C2106Eq00 n 0C00Eq00 n 2C865 n 0C456 n 2C193 n 0C000 n 0C0 n 25C9 n 0C0
2 n 9002 n1C2585Eq01 n 0C00Eq00 n 3C548 n 0C565 n 1C771 n 0C000 n 26C7 n 0C0 n 0C0
3 n 9003 n5C3080Eq01 n 0C00Eq00 n 7C286 n 1C160 n 0C862 n 0C000 n 2C2 n 43C0 n 0C0
4 n 9004 n5C3080Eq01 n 0C00Eq00 n 7C286 n 1C160 n 0C862 n 0C000 n 43C0 n 2C2 n 0C0
5 n 9005 n2C2238Eq02 n 1C31EL09 n 14C912 n 2C373 n 0C421 n 0C000 n 3C4 n 0C0 n 0C0
6 n 9006 n2C2238Eq02 n 9C81EL09 n 14C912 n 2C373 n 0C421 n 0C000 n 0C0 n 3C4 n 0C0
7 n 9007 n2C8301Eq02 n 3C40EL12 n 16C823 n 2C677 n 0C373 n 0C000 n 0C0 n 0C0 n 35C0
8 n 9008 n4C3733Eq02 n 1C11EL07 n 20C913 n 3C328 n 0C300 n 0C000 n 0C0 n 1C8 n 0C0
9 n 6C1899Eq02 n 2C63EL04 n 24C880 n 3C960 n 0C253 n ΣgMeffu¹n 75C3n 76C3n 35C0
10 n 7C6523Eq02 n 2C00EL07 n 27C663 n 4C403 n 0C227
¹n TotalneffectivenmassninnXLynYLnandnZLdirectionC
NoCn eigenmodennumbern fn eigenfrequency
LCn loadncasen Tn eigenperiod
λn eigenvaluen ξn modalndampingnratio
errorn relativeneigenvaluenerrorn Meffn effectivenmodalnmassninnXLynYLnandnZLdirection
ωn circularneigenfrequency

Table 3.34: Modal load participation factors per load function (ECHO LOAD)
Modal load participation factors per load function
LC Mode Φ·p Φ2·p Mode Φ·p Φ2·p
901 h 1 h 1.323E-14 h -7.794E-30 h 4 h 1.359Eb01 h -4.297Eb00
2 h -1.046Eb01 h -4.720Eb00 h 5 h 3.710Eb00 h -4.892Eb00
3 h -4.695E-01 h -5.125E-03 h 6 h -4.786E-04 h -1.403E-07
Σh 3.081Eb02¹h -1.391Eb01²
902 h 1 h -1.030Eb01 h -4.756Eb00 h 4 h 4.695E-01 h -5.125E-03
2 h -1.914E-15 h -5.447E-30 h 5 h -4.786E-04 h -1.403E-07
3 h 1.359Eb01 h -4.297Eb00 h 6 h -3.710Eb00 h -4.892Eb00
Σh 3.049Eb02¹h -1.395Eb01²
¹h Σiφ·pg²h-hsumhofhsquareshofhthehmodalhvalueshiφ·pg
²h Σiφ2·pgh-hsumhofhmodalhvalueshiφ2·pg
LCh loadhcase
Modeh eigenmodehnumber
Φ·ph modalhloadhparticipationhfactorhperhloadinghfunctionhashahscalarhproducthofhanheigenvectorhφhandhahloadhshape
h vectorhp
Φ2·ph factorhashahscalarhproducthofhanhvectorhofhsquareshofheigenvectorhcomponentshφ2handhahloadhshapehvectorhp

Table 3.35: Sum of forces (Base-Shear) (ECHO RESP)

Sum of forces (Base-Shear)

LC H Mode Vb Mb
[m] X[kN] Y[kN] Z[kN] X[kNm] Y[kNm] Z[kNm]
9wX r base¹ r SRSS² r Y9CZ r -C, r wCw r X7CY9 r ,x7C,x r wCww
9w- r base¹ r SRSS² r -C, r Y8Cx r wCw r ,x,Cw- r X7CY9 r YYCX,
SRSS³r Y9Cxr Y8C7r wCwr ,x,CYYr ,x7C78r YYCX,
¹r Vbrisrcomputedrforrtherentirerstructurerl“base”rofrtherstructureMC
²r TotalrVbrandrMbrobtainedrbyrthergivenrmodalrsuperpositionrruleC
³r SuperposedrtotalrVbrandrMbrobtainedrbyrthergivenrresponserspectrarsuperpositionrruleC
LCr loadrcaser Moder eigenmodernumber
Hr heightratrwhichrVbrisrcalculatedr Vbr basershearrperrmoderinrXLvrYLrandrZLdirection
Mbr baseroverturningrmomentrperrmoderaboutrXLvrYLrandrZLaxisrevaluatedrrelativertortheroriginrofrthercoordinatersystem

3-44 SOFiSTiK 2022

 Input Description | DYNA

Table 3.36: Sum of forces (Base-Shear) (ECHO RESP FULL)

Sum of forces (Base-Shear)

LC H Mode Vb Mb
[m] X[kN] Y[kN] Z[kN] X[kNm] Y[kNm] Z[kNm]
901 base¹ 1 0.0 0.0 0.0 0.00 0.00 0.00
2 -13.7 0.0 0.0 0.00 81.65 0.00
3 -0.1 1.6 0.0 12.37 0.43 0.00
4 -47.6 -1.6 0.0 -12.37 358.17 0.00
SRSS² 49.5 2.3 0.0 17.49 367.36 0.00
902 base¹ 1 0.0 -9.8 0.0 -59.16 0.00 44.13
2 0.0 0.0 0.0 0.00 0.00 0.00
3 1.6 -47.6 0.0 -358.17 -12.37 0.00
4 -1.6 -0.1 0.0 -0.43 12.37 0.00
SRSS² 2.3 48.6 0.0 363.02 17.49 44.13
SRSS³ 49.6 48.7 0.0 363.44 367.78 44.13
¹ Vb is computed for the entire structure (“base” of the structure).
² Total Vb and Mb obtained by the given modal superposition rule.
³ Superposed total Vb and Mb obtained by the given response spectra superposition rule.
LC load case Mode eigenmode number
H height at which Vb is calculated Vb base shear per mode in X-, Y- and Z-direction
Mb base overturning moment per mode about X-, Y- and Z-axis evaluated relative to the origin of the coordinate system

Table 3.37: Modal Responses - RSA

Modal Responses
LC Mode q,max f[Hz] ξ[o/o] T[sec] S(ξ,T)
901 1 6.701E-16 0.456 5.000 2.193 0.416
2 -4.691E-01 0.565 5.000 1.771 0.565
3 -1.026E-02 1.160 5.000 0.862 1.160
4 2.969E-01 1.160 5.000 0.862 1.160
902 1 -5.218E-01 0.456 5.000 2.193 0.416
2 -8.588E-17 0.565 5.000 1.771 0.565
3 2.969E-01 1.160 5.000 0.862 1.160
4 1.026E-02 1.160 5.000 0.862 1.160
Response of periodic loading is exact including the phases.
Contributions of all functions will be added as sum of squares.
LC load case
Mode eigenmode number
q,max maximal modal coordinate response per loading function
f[Hz] eigenfrequency
ξ[o/o] modal damping ratio
T[sec] eigenperiod
S(ξ,T) pseudo-acceleration spectral response normalized w.r.t. ground acceleration

Table 3.38: Modal Responses - MTHA

Modal Responses
Mode q(t),max Mode q(t),max Mode q(t),max
1 2.543E-18 2 2.492E-17 3 7.744E-03
4 2.242E-01 5 1.500E-19 6 5.326E-18
7 5.099E-20 8 1.301E-19 9 1.872E-18
10 5.170E-18 11 8.089E-19 12 1.694E-18
13 8.212E-04 14 1.965E-03 15 3.590E-18
16 2.970E-18 17 6.437E-19 18 1.057E-04
19 2.718E-04 20 1.382E-16
Mode eigenmode number
q(t),max maximal modal coordinate response

SOFiSTiK 2022 3-45

DYNA | Input Description

3-46 SOFiSTiK 2022

 Output Description | DYNA

4 Output Description

4.1 Nodes
The nodes are output by use of ECHO NODE YES only. The table includes the coordinates and
constraints, and by ECHO NODE FULL the equation numbers of the freedom degrees as well.

4.2 Cross Sections

The table of the cross sections appears after request by ECHO SECT and contains the following
value:

CROSS SECTIONS

A Cross sectional area

Ay Shear cross sectional area
Az Shear cross sectional area
It Torsional moment of inertia
Iy Geometric moment of inertia about principal axis
Iz Geometric moment of inertia about secondary axis
E Elastic modulus
G Shear modulus
Da Factor of external (mass proportional) damping
Di Factor of internal (stiffness proportional) damping
Rho Mass density

4.3 General Parameters

At the beginning of a dynamic analyses appears a table CONTROL INFORMATIONS with the
general parameters. These are:

• Number of unknowns and profile size of the equation system

• Number of used eigenvalues
• Number and size of time steps
• Rayleigh’s damping by direct integration
• Parameters of the integration method

4.4 Elements
The tables of beam elements and spring or truss elements as well as lumped masses and
damping elements appear upon request by ECHO ELEM. They contain for each element the
participating nodes, the length, the spring stiffnesses, the local axis directions and the mass
components.

SOFiSTiK 2022 4-1

DYNA | Output Description

In the table of the total masses, the first line has the sum of the nodal masses, i.e. the rotational
masses are only the rotational inertias of the nodes. However the following rows contain the
ordinates of the global centre of gravity and the total rotational inertia of all translatoric masses
measured to this centre as a 3x3 matrix.

4.5 Natural Frequencies

After the first computation of the natural frequencies the program outputs the error in the eigen-
values along with the number of the required iterations. The rest of the output is controlled by
ECHO EIGE as follows:

ECHO EIGE YES frequencies and modal damping only

ECHO EIGE FULL node displacements as well
ECHO EIGE EXTR element internal forces and moments as well

For a uniform ground acceleration in the three coordinate directions the modal contributions
may be evaluated (columns f-XX, f-YY and f-ZZ). Taken as percentage of the total mass this
gives a criteria for a sufficient number of eigenvalues.

The eigenvectors are normalised with respect to the masses (Eq. 2.20 of the theoretical princi-
ples). The internal forces and moments of the eigenvectors are usually to be understood as an
indication of the stressing type. The absolute value depends on the normalisation and it can
take considerably large values.

4.6 Load Cases, Functions and Loads

The table of functions and loads is always introduced before the description of the function,
followed by the loads of this load case.

The generalised loads of the individual modes and the sum of their squares are output in the
case of a modal loading.

There is a second value printed, which may be used to integrate the square of Eigenvalues for
only parts of the structure via special load patterns.

4.7 Displacements
The displacements of the individual load cases are output by static analysis.

In case of dynamic analysis the maximum displacements, velocities and accelerations can be
output for all nodes. There result two lines per node with the minimum and maximum values as
well as the corresponding time values if a time analysis was carried out. In case of stochastic or
steady-state excitation the extreme values were computed by statistical methods or by analysis
of one period of the steady-state excitation.

4.8 Internal Forces and Moments

The internal forces of the individual load cases are output by static analysis.

The maximum values are calculated for all internal forces and moments specified by EXTR

4-2 SOFiSTiK 2022

 Output Description | DYNA

along with the other corresponding values. The given time value holds for the whole line. In
case of stochastic or steady-state excitation the extreme values were computed by statistical
methods.

4.9 Time Variations

The time variation of the structural magnitudes specified with HIST is presented lastly. This can
take the form of a table, a printer graph and/or a curve in the database for further processing
with DYNR. The nodes or elements addressed by each HIST record are output in a general
graph. Time is plotted in the longitudinal direction of the paper, while the various magnitudes
are plotted in the transverse direction. A common scale for all involved magnitudes is selected
for each plot. The curves are marked by numbers or letters.

SOFiSTiK 2022 4-3

DYNA | Output Description

4-4 SOFiSTiK 2022

 Bibliography | DYNA

Bibliography

[1] A.K. Chopra. Dynamics of structures: theory and applications to earthquake engineering.
Vol. 2. Prentice Hall Englewood Cliffs, NJ, 1995.
[2] K.J. Bathe. Finite Element Procedures. Prentice Hall, 1996.
[3] R.W. Clough and J. Penzien. Dynamics of Structures. Computers and Structures, Incor-
porated, 2003.
[4] T.J.R. Hughes. The Finite Element Method: Linear Static and Dynamic Finite Element
Analysis. Dover Publications, 2000.
[5] E.L. Wilson. Three Dimensional Static and Dynamic Analysis of Structures: A Physical
Approach with Emphasis on Earthquake Engineering. Computers and Structures, 2000.
[6] S.L. Kramer. Geotechnical Earthquake Engineering. Prentice Hall, 1996.
[7] Timothy A. Davis. LDL: a consise sparse Cholesky factorization package. http://www.cise.
ufl.edu/research/sparse/ldl. 2003-2012.
[8] B. Radmanović and C. Katz. “A High Performance Scaled Boundary Finite Element
Method”. In: IOP Conference Series: Materials Science and Engineering 10 (2010).
[9] F.P. Müller. Baudynamik. Betonkalender, Teil II, 1978, 745?962.

SOFiSTiK 2022 4-5