Hydra: Seepage and Thermal Analysis
Hydra: Seepage and Thermal Analysis
Hydra: Seepage and Thermal Analysis
SOFiSTiK | 2020
HYDRA
Seepage and Thermal Analysis
SOFiSTiK AG
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
Project: Queensferry Crossing | Photo: Bastian Kratzke
Contents | HYDRA
Contents
Contents i
SOFiSTiK 2020 i
HYDRA | Contents
ii SOFiSTiK 2020
General Description | HYDRA
1 General Description
A great number of physical problems may be described as potential problems governed by the
Laplace (ΔH = 0) or the Poisson differential equation (ΔH = q). Among these are magnetic
flow, heatflow and groundwater seepage. HYDRA is specialised on the the last and last but
one type of problem, but may be used for other problems with a similar background as well.
A finite element method discretizes the real continua in a finite number of small elements with
simplified physical properties. The elements are connected at nodes. An interpolation function
within the elements represents a numerical estimate to the real solution.
Thermal analysis allows the evaluation of the temperature distribution within a solid due to a
steady or transient heating or cooling or temperature gradient.
Groundwater models are used to compute the movement of water (seepage) within a porous
solid needed for geology, hydrology, soil mechanics or groundwater supply.
• Plane or 3D problems
The real situation is always 3D. For the analysis you may select a system which ne-
glects one of the dimensions. For a structure extending along a line like a dam, we may
select a plane cut with constant thickness. For aquifers with a vertical thickness much
less than the horizontal dimensions, we may use the hypothesis of Dupuit to switch to a
simplified, but not necessarily plane system.
• Linear or nonlinear flow models
Nonlinear flow is needed for turbulent flow, this is important for pipes and culverts
but not so for seepage. However one important case is the free surface problem. All
nonlinear problems may be solved with an iterative procedure.
• Free, confined or partial saturated flow
The free surface problem is governed by a surface with pressure zero, which is the
boundary of the flow domain and is not known a priori. A partial saturated flow may be
accounted for with a pressure dependant conductivity to some extend.
• Steady or transient
All estates which do not change in time are steady. For a transient analysis with
changing boundary conditions in time, we need to solve a linear or nonlinear problem for
every time step.
2 Theoretical Background
A value u which is defined in every point of space as a vector (e.g. displacements, flow veloci-
ties) is called a vector field.
In general the potential is only defined by the selection of a reference value in any arbitrary
point. As the fundamental solution (Greens function) of a 2D problem is the logarithm resp.
arcustangens, while in 3D it is the 1/r function, there is an important difference. While in 3D we
may easily select a zero potential at infinity, for 2D we have to select this point within a finite
distance. This makes 2D solutions in many cases less unique than 3D solutions.
Differentiating a vector field we have the divergence and the curl or rotation given by:
∂ ∂y ∂z
d = ∇ · = + + (2.2)
∂ ∂y ∂z
∂z ∂y
− ∂z
∂y
rot =
∂ ∂z
− ∂ (2.3)
∂z
∂y ∂
∂ − ∂y
A conservative field has its largest and smallest potential value at the boundary (Maximum
principle).
H = Ho on 1 (2.5)
T · n = q on 2 (2.6)
The principle of the finite elements is given by the approach to replace the exact solution by a
piece wise local (elemental) interpolation and to solve the minimum problem for this projected
system. The solution will fulfil the differential equation only in a weak sense, while the boundary
conditions are met exactly in most cases. The estimated solution has to fulfil the dirichlet
boundary conditions for the potential or temperature
H = ƒ (s) (2.8)
on the boundaries in advance. The minimisation will yield the natural boundary condition on
all other boundaries:
T · n = 0 (2.9)
The four resp eight noded isoparametric elements have a slightly more enhanced precision
especially for twisted saddle points. Moreover they have advantages for the mesh definition
which pay off the little extra effort quite well.
With some mathematical treatment (cmp. Zienkiewicz, 1975) the expression for the functional
may be written as matrix product:
Z
e = Ht · Bt · K · B · dV · H (2.10)
= 1/ 2 HT · A · H + H · Q (2.11)
where A is the total conductivity matrix and Q the vector of external quantities. The minimisa-
tion is equivalent to set the derivatives to zero:
A · H + Q = 0 (2.12)
This last equation is after the incorporation of the prescribed potential values with a penalty
function the equation system to be solved via Gauss elimination or an iterative conjugate gradi-
ent solver. Without any Potential boundary condition the equation system is still multi valued.
The differential equation respective the functional will be established at time t+ θ · Δt which
yields the following matrix equation:
1 1
(θ · A + · C) · H = (θ − 1) · A + · C H−1 − Q (2.14)
Δt Δt
Where A is the conductivity matrix and C the storage matrix, given for any element by:
Z
Cj = N S Nj dV (2.16)
with the storage coefficient S. For free surface problems the Matrix C has to be enhanced by
a part described by the effective porosity on the free surface.
For transient problems it is especially for free surface problems to choose the time step de-
pending on the mesh size neither to small than to large. Large time steps will make the Crank-
Nicholson scheme less precise, but the stability is kept unless the value of θ is chosen to small.
To small time steps might introduce an oscillating behaviour due to the violation of the discrete
maximum principle, which is physically unreasonable. This effect my be suppressed by using
diagonalized matrices, but the general rule of thumb is not to chose time steps less than the
speed of the variations across the element mesh will distribute. HYDRA has an automatic time
stepping facility therefore.
2.6 Iterations
The solution is obtained at the first step if neither a free surface nor any nonlinear boundary
conditions other nonlinear material effects occur. In those cases we need iterations.
This will be achieved by a piece wise linearisation from the first linear guess towards more and
more corrected potential distributions to meet the material laws or boundary conditions.
We can use the tangential stiffness or the initial stiffness of a quasi-Newton approach. If the
solution of Hi differs only slightly from H+1 we can rewrite the above equation:
This allows to keep the Matrix A several times for the solution. The iterations run faster but we
might need more. Thus it is recommended for large systems with only slight nonlinearities.
For transient problems there is also a derivative in time from H and H−1 to be accounted for
in an appropriate way.
The law of energy conservation requires that for any selected region the sum of entering and
leaving heat quantities has to be equal to the sum of sources or sinks and the heat stored
internally.
δT
d = q − ρ · cp · (2.19)
δt
Remarks:
The Fourier equation combines the temperature gradient grad T with the fluxes induced by it.
According to Biot/Fourier this is
= − K · grd T (2.20)
The conductivity is thus dependant on the flow direction. There are however three orthogonal
directions which let vanish the off diagonal terms. For an isotropic medium there are further
simplifications to yield:
δT
d(−K · grd T) = q − S · + (2.22)
δt
T = T0 (2.23)
= 0 (2.24)
This boundary condition also contains the so called natural boundary condition. All boundaries
not described otherwise will have the boundary condition, that the flux across the boundary is
zero (adiabatic surface):
t · n = 0 (2.25)
• Convective boundary condition The flux on the surface is proportional to the difference of
the surface temperature to the surrounding temperature.
= α · (T − T) (2.26)
• Radiation condition of black body The flux on the surface is proportional to the difference
of the fourth powers of the absolute surface temperature and the absolute surrounding
temperature (Law of Boltzmann)
q = ε · K · (T 4 − T4 ) (2.27)
Although there are some definitions around using the real time, it is far more sound to use the
effective age of the concrete τ . There are some definitions available for that purpose, but the
definition according to Freiesleben matches experimental data best and is used therefore as
default:
R = 8.3143 [ J/(mol K) ]
A = A0 + A1 · Max (0, 20-T)
A0 = 33500 [ J/mol ]
A1 = 1470 [ J/mol ]
For the function defining the hydration grad depending on the effective concrete age, there are
numerous suggestions available in literature. The most useful seems to be that of Jonasson,
an extension to the Byfors definition, which has been implemented therefore:
τ
α = ep b · n 1 + (2.32)
τk
Examples of those constants a,b and τk can be found in Appendix A of Heft 512 of the German
DAfStB, printed with the input record HMAT but the general rule is that you need tests!
Figure 2.1
· (τ − τk )
α = τ > τk (2.33)
1 + · (τ − τk )
or the Danish model according to Wesche including also a function for the development of
concrete strength within time:
τk b
α = ep − (2.34)
τ
ƒcc (t) 0
β = = · ep(c · t −0.55 ) c = c1 (2.35)
ƒcc (28d) z
For groundwater seepage the velocities are small enough to allow to neglect the last two terms.
For the coordinate z you may use any other coordinate axis with a positive or negative sign.
A movement of the groundwater will always occur from the higher value of the potential to the
lower through a connecting material with hydraulic conductivity.
If the soil does not only contain water but also air, one has a so-called multiphase flow which
cannot be used in HYDRA in the form of coupled differential equations. However, there are
possibilities for the description of partial saturated flows. One can introduce itself very well
the negative suctions as the height of capillary rise. A sufficient fine-grained soil may move
upward with its capillary water. Thus it has in the dry state a suction which corresponds to this
maximum height of capillary rise.
In the area of the saturated pore volume the pressure increases linearly with the depth, so
that the piezometric height and the position height are added to a constant value. It is very
important, that a constant pressure has to be available in all points above the capillary line. It
results consequently from top to bottom a gradient of the potential, which corresponds to the
geodetic height. The gradient is therefore 1.0.
The flow law describes the streaming velocities v depending on the potential gradient i = grad
H. The mostly used linear flow rule from Darcy is:
= − K · grd H (2.37)
The conductivity is thus dependant on the flow direction. There are however three orthogonal
directions which let vanish the off diagonal terms. For an isotropic medium there are further
simplifications to yield:
The permeabilities may vary in large magnitudes (up to 10 powers of 10). The numerical
treatment of such large differences may become difficult.
grdh = · + b · 2 (2.39)
= ∂h
∂
Figure 2.2
S is the specific storage coefficient, defining how much volume of the fluid may be stored inter-
nally if the pressure is rised by 1 m. For a fully saturated soil this is given by the compressibility
of the water and the soil particles by
where
n = porosity
K = Compression moduli
e.g. K = 2 000 000 kN/m2
For unconfined flow however S will be governed by the free surface and the effective porosity
for storage nsp . This value is much larger than the value according to equation (). However it
is recommended to specify S anyhow to avoid numerical difficulties if the saturation becomes
exactly 0.0 or 1.0.
As an extension HYDRA may account for partial saturated soils by defining the conductivity
and the effective storage volume as a function of the pore water pressure resp. suction. This
is not as powerful as a real three-phase flow and requires some experience therefore
2.14 Dupuit-Hypothesis
It is a common case, that we have to analyze a structure where the horizontal dimensions
are much larger than the vertical one. In this case we may use the benefits of the Dupuit-
Hypothesis. It assumes that the potential is constant across the thickness of the strata resp.
the part of which has saturated flow. This will reduce the 3D-Problem to a 2D-Problem, where
the geometry however is rather general in 3D and not necessarily plane. The conductivity
real
Piezometric head assumed
Piezometric Head
is replaced by the transmissivity, the integral of the first across the thickness of the layer.
confined aquifer:
Zd Zd
T = k dz , S = S0 dz (2.42)
0 0
L · 2
Δh = λ · (2.44)
R · 8 · g
where
RE = · 4 · R · kz (2.45)
where kz is the inverse 1/ν of the kinematic viscosity of the fluid. For the rare case of laminar
flow (RE < 2320) we have the law form Hagen-Poisseuille
λ = 64/ RE (2.46)
For the turbulent flow (RE>2320) we have the formula given from Prandtl-Colebrook with the
absolute roughness k:
1 2.51 k
p = − 2g p + (2.47)
λ λ · RE 14.84 · R
For channels the total section will be subdivided in parts with a similar flow characteristic. The
following assumptions will be made:
There are two subdivisions common. For compact sections there are vertical divisions perpen-
dicular to the isotaches. These lines do not have any roughness. The total resistance is then:
X
· λges = , · λ (2.48)
The value of l is the wetted part of the section. The value of λ depends not only on the
roughness but also on the hydraulic radius, which in turn is dependant on the water level.
For non compact sections this method is not admissible. For that case we the division lines will
be vertical and will have roughness assigned with a value of 1 to 3 times the roughness of the
bottom. The flow is then calculated for every partial section in separate.
H = H0 (2.49)
Hint: A surface with a wetting liquid has in all points the same value of the potential, be-
cause the geodetic height and pressure height will add up to the same constant value.
= 0 Q = Q0 (2.50)
This boundary condition is also the natural boundary condition. All edges which do not
have a boundary condition will have the impermeable boundary condition:
t · n = 0 (2.51)
Moreover this type of boundary condition also covers the effect of ground water genera-
tion, even if this condition is normally referred on to an area or the volume and not an edge
or point.
• Free Surface
The line bounding the flow region is not know a priori. The impermeable stream line
is also a line with given pressure (p=0). The location of this boundary has to be found
within an iterative process.
• Seepage areas (Groundwater leaving the domain)
At the surface of seepage we have atmospheric pressure. The point where the free surface
hits the seepage area has to be found iteratively.
Although the formulation of the free surface with the natural boundary condition is mathemat-
ically elegant and easy to implement, one should keep in mind, that very steep gradients will
create skewed intersections and an inherent instability of the iterations. In some cases a switch
to an unsaturated flow model may become necessary.
• Wells
Wells with a given positive outlet or a negative inlet may be handled like a given potential
or seepage area, but the potential will be varied iteratively until the desired quantity of
These are the general case of a Robin-Problem and are needed for the leakage of a mud-
ded soil of a river. The quantity which is entering or leaving the system is dependant on
the potential difference of two points. This is the hydraulic analogue to the spring element
in static analysis.
2.18 Elements
The total thermal or hydraulic system can be described with an assemblage of the following
element types. Each element has a specific flow property. All elements are connected to each
other at nodes. The sum of all flux quantities at any node has to be equal to the external take
in or take out, which establishes the equation system.
to describe boundary conditions on areas of a volume, but have now own physical properties.
Other quantity conditions will be accounted for by surface integrals. The matching to the ele-
ments will be obtained by the same firm functions of the elements:
Z
Q = q H · N dF (2.52)
boundary condition. By repeating this process the desired location will be found by iteration.
An error encountered quite often is to define faulty boundary conditions which let the total
domain become dry. The solutions are of course not correct, but a beginner may have problems
in finding that fact.
• The total system is subdivided in node and elements. Material constants will be defined for
nodes or elements. Parts of the structure my be organised in groups and missing material
values may be calculated with linear regression.
• With the given boundary conditions there will be now an iterative steady or transient analy-
sis. Each intermediate step follows the scheme:
- Evaluate results of step i-1
- Create new conductivity and capacity matrixes or
corrective parts on the right hand side
- Check the boundary conditions
- Evaluate potentials at step i
• At the end of any time step or the total analysis, the flow in each element and the total
quantities for all nodes will be evaluated:
Q = − A · H − C · dH/ dt (2.53)
The quantities will be summed up for any partial system (group). They will be integral
results for any node and will fulfil the total balance exactly within the computers precision.
For transient analysis however the differential equation si fulfilled at a time between the
time stepping values, thus there might be small deviations.
The velocities or fluxes within the elements are obtained by differentiation of the potentials.
They are therefore less precise and will jump from one element to its neighbour. That’s why
they are only evaluated at the centre of the element.
3 Input Description
The program calculates problems of groundwater flow and thermal analysis for an existing
structure in the database. The different variants are selected by the unit of the quantities with
record SYST (e.g. Watt or l/min)
general
Highest node number 9999999
Number of nodes 9999999
Number of materials 999
Highest element number 9999999
Number of elements 9999999
Number of nodes per well 240
Record Items
SYST TYPE SNO GDIR DIMT DIMQ
CTRL OPT VAL
GRP NO LEV LOAD T HC TF HP
STEP N T TFAK THET HOPT TMIN
PLC NO T HP DHPX DHPY DHPZ DHTF
PU LCS LCU LCU0
NVAL NO TYPE VAL VAL2 VAL3 VAL4 VAL5
VAL6 COPY
HMAT NO TYPE TEMP KXX KYY KZZ KXY
KXZ KYZ S NSP A B C
QMAX TK TITL
LINK NO A B S EPS AREF
PIPE NO L
LC NO TITL
BOUN TYPE VAL FROM TO INC F VP
EPS NB DVX DVY DVZ
SURF TYPE VAL FROM TO INC F VP
Table continued on next page.
Record Items
EPS NB
FUNC T F NO TITL
HIST TYPE FROM TO INC
ECHO OPT VAL
The records HEAD, END and PAGE are described in the general manual SOFiSTiK: ’FEA /
STRUCTURAL Installation and Basics’.
HYDRA 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-Systems of the sections are
saved in a subdirectory and has to be created for general sections in AQUA via option CTRL
STYP FEMX. HYDRA may create those meshes from scratch for simple rectangular T-Beam
and circular or annular sections if not yet available. The mesh density is controlled by CTRL
HDIV defined before the SYST record.
The specification of unit W, kW or MW respective AW, AKW or AMW will select a thermal
problem. The unit KG is used for a vapour diffusion analysis.
The specification of a unit for the time will allow the usage of handsome units in input and
output for the large range of possible physical problems. Dependant are time values of the
functions and time steps as well as all quantities given for boundary conditions or obtained as
a result.
The unit is only valid within the current run. It is not saved in the database and has to be
specified for every single task.
CTRL
The method of tangent stiffness is recommended in general for free surface flow as the change
of the boundary conditions require a new system matrix quite often. For larger systems which
do not change so much, the quasi Newton-Methods will be favourable instead.
To enhance the stability and speed of the iteration methods there are multiple facilities to damp
oscillating iterations or to extrapolate for monotone convergence. The extrapolation is limited
by a maximum factor of FMAX and a maximum increase of the factor by FMUL. Values of
FMAX < 1.0 will introduce a damping. FMUL 0.0 will deactivate the increase.
A transient analysis will be initiated by the specification of N. There are several possibilities to
subdivide the analysis time.
On the other side one can define the total time of the analysis explicit or via the time steps.
• With an input of N > 1 , N is the number of time steps. The total time is the sum of all time
steps.
• As an alternate definition T may be the total analysis time if N is defined less than 1.0. N
is than the factor for the first time step. (N 0.001 yields a first time step of one/thousand of
the total time T.)
Start of any transient analysis or the deviation to a given result case will be a primary estate
with number NR at time value T. Default for T is of course the last time value stored at NO. The
T-value is needed for correct interpolation within the time functions and identification of results
especially for graphics required.
Additional linear varying potentials are also possible. For each node we will have an extra
component of the primary estate by:
Eventually it is also possible to add a primary value for every group via GRP. Nodes blonging
to multiple groups will get the mean value of all defined values of all groups connecting to that
node.
For transient analysis the time derivative of the last estate may facilitate the startup of the
current analysis. If desired a specification of DTHF 1.0 will be required. The time derivative
itself will be calculated from the difference of the last two values and the involved time step.
As it is not always that this condition is valid, the default for DTHF has to be 0.0. PU may be
needed if for the transition from the primary estate to the initial step, significant changes for the
boundary conditions are encountered.
If for item T is given the literal CONT, all potential values of the primary estate - even those
from earlier time steps - will be copied to the current load case. This creates more data in the
data base but is convenient for the graphical postprocessing. Results of the elements however
can not be processed this way.
With LCS it is possible to specify a load case the stresses of which are used to modify the
transmissivities according the given definitions (HMAT FVOL/FSIG).
A structural load case can be input with LCU/LCU0. The volumetric stress of this load
case is then used as initial conditions for the pore water pressure of a consolidation (H=h-
1/3(Δsigx+Δsigy+Δsigz)/gam).
For very special cases it is possible to modify results in the database. The record COPY, allows
to combine several transient results or to shift and scale pressure values.
COPY
The total structure may be organized within groups. The group number will be multiplied with
the group divisor (record SYST GDIV) and added to the element number during creation after
the GRP-record. Later analysis steps may then deactivate parts of the system by selecting
LEV OFF or selecting only other groups.
HYDRA may save as a result loading in the database to be used by TALPA or ASE. A pure
temperature filed does not need to be saved as loading as those programs have better options
to use the temperature field directly.
BUOY All elements within the active flow region (= not above the free surface) will receive
a buoyancy loading.
P = + γ · grd Z
STRE All elements within the active flow region (= not above the free surface) will receive
a streaming pressure loading..
P = − γ · grd H
PRES All elements within the active flow region (= not above the free surface) will receive
P = − γ · grd (H − Z)
For free surface problems the non active flow region will keep a small conductivity for numerical
reasons, obtained by multiplying the permeability with a factor TF. The definition of HC as
the height of the suction region will be a simple tool for partial saturated soils with a smooth
transition between those two regions.
Linear Material values may be specified either in a separate material table or at any node via
this record NVAL to be used by the connecting elements.
With COPY any result case within the database may be used to set the material values. Items
VAL to VAL6 are then factors. This option is intended for those cases where an solution of the
identification problem has been run and the results should be converted to material values.
HMAT
NO Material number − !
TYPE Type of material law LT *
DARC Seepage (Darcy, linear)
FORC Seepage (Forchheimer, nonlin.)
MISS Seepage (Missbach, nonlin.)
FLOW General diffusion (e.g. vapor)
FOUR Heatflow (Fourier, linear)
Hydration of concrete:
JONA Jonasson model
HSCM ”Shrinkage-Core” model
WESC Danish model acc. to Wesche
HMAT allows to specify three individual sets of constants for three tasks diffusion, seepage
and heat flow for a material. Material properties have to be selected according to literature or
experimental data. But some rough estimates (without warranty) are given here.
For any material there might be up to 15 different sets of materials for different temperatures
[ ◦ Celsius] or pore water pressures [ kPa = kN/m2 ] . With a nonlinear analysis the material
values will then be interpolated between those values.
Positive pressures represent saturated flow regions while negative values describe unsaturated
soils. Conductivities and Capacities will be interpolated. Free surface problems also use a
variation of the porosity to account for the effective capacity of the free surface.
Material values may be defined isotropic or anisotropic depending on a stress state from the
database. This is performed by additional data given with the same material number and the
types FVOL, FSIN or FSIT. The difference between FSIN and FSIT is given by the fact that
FSIN modifies the values across the crack, while FSIT does this perpendicular to the crack. It
is recommended to use a linear stress field for that purpose and not a plasticity field.
Within HYDRA the user has the possibility to define material properties for the elements either
via a given material number (engineering constructions) or via the nodes (element material
number = 0), especially for ground water models. Both methods may be used together within
the same system.
For a general diffusion according to potential theory, the unit for conductivity (KXX, KYY, KZZ)
is [ sec/m] , for storage coefficient (S) [ kg/m3/Pa] =[ sec2/m] .
e.g.:
If the materials have been defined with CONC, STEE, TIMB, BRIC or TYPE at MATE, all
constants will be preset for FOUR according EN 12524. For concrete and timber a definition of
NSP will select the moisture ratio.
With explicit definition of a TEMP-value it is possible to create and modify tabulated values.
The default is to use the formulas defined in the Eurocode. The thermal conductivity of the
4500
1.800
4000
1.600
3500
1.400
3000
1.200
1.000 2500 S [kJ/m3/K]
0.800 2000
K [W/m/k]
0.600 1500
0.400 1000
0.200 500
0.00 [◦ C] 0.00 [◦ C]
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
100.00
100.00
1000.00
1100.00
1200.00
1000.00
1100.00
1200.00
0.00
0.00
concrete is given by an upper (A=1.0) and a lower (A=0.0) limit. The special effect for S is the
evaporation of pore water.
.0 50.00 18000
18000
.0 45.00
16000 16000
.0 40.00
14000 14000
.0 35.00
12000 12000
.0 30.00 K [W/m/k]
k [W/m/K] 10000 10000
.0 25.00
8000 8000
.0 20.00
6000 6000 S [kJ/m3/K]
.0 15.00 S [kJ
.0 10.00 4000 4000
0.00
100.00
300.00
400.00
500.00
700.00
1000.00
1200.00
1100.00
0.00
100.00
200.00
300.00
500.00
600.00
700.00
800.00
900.00
1000.00
1200.00
1100.00
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
1000.00
1100.00
1200.00
1000.00
1100.00
1200.00
0.00
0.00
Figure 3.2: Conductivity and Capacity of Structural Steel
1.50 7.00
1.40 6.50
1.30 6.00
1.20 K [W/m/k] 5.50
1.10 5.00
1.00
4.50
0.90
4.00
0.80
0.70 3.50
0.60 3.00
0.50 2.50
0.40 2.00
0.30 1.50
0.20 1.00
0.10 0.50 S [kJ/m3/K]
[◦ C] [◦ C]
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
100.00
200.00
300.00
400.00
500.00
600.00
700.00
800.00
900.00
1000.00
1100.00
1200.00
1000.00
1100.00
1200.00
0.00
0.00
TYPE JONA
Function of Jonasson, an extension to the Byfors definition:
τ
α = ep b n 1 + (3.1)
τk
Examples of those constants a,b and τk can be found in Appendix A of Heft 512 of the Ger-
man DAfStB, printed with the input record HMAT but the general rule is that you need tests!
Unfortunately there are publications with exchanged parameters a and b.
The values have to be determined from experiments, values for the total heat in the literature
are often defined in [ kJ/kg] . However the following values might give a rough idea:
Q-max [ kJ/m3] a b τ k [ h]
B25 (Heft 512 DAfStB) 167700 -0.925 -0.42 41.82
B35WU (Heft 512 DAfStB) 135750 -1.185 -1.00 17.53
B35 (Heft 512 DAfStB) 105940 -1.605 -1.00 37.46
TYPE HSCM
”Shrinkage-Core” model:
· (τ − τk )
α = τ > τk (3.2)
1 + · (τ − τk )
τk is a period with very low chemical reaction, named d in the original formula. Values for a
mass concrete are given by Dussinger:
TYPE WESC
Danish model according to Wesche:
τk b
α = ep − (3.4)
τ
ƒcc (t) 0
β = = · ep c · t −0.55 c = c1 · (3.5)
ƒcc (28d) z
The water cement ratio w0 /z has to be incorporated in the input value c. Please note that
parameter a is used in the original with two different meanings. The following parameters for
the Wesche model can be found in the literature.
Q-max [ kJ/kg] a c1
w0 /z 0.35 0.55 0.75
Z 55 / Z 45 F 380 1.10 1.15 1.20 -1.50
Z 45 L / Z 35 F 295 1.20 1.30 1.40 -2.80
Z 35 L 216 1.30 1.50 1.70 -4.40
Z 25 1.50 1.90 2.30 -7.10
τk = 24.87[h]
b = 0.84[−]
0 / z = 0.68[−] (3.6)
= 1.63[−]
c1 = −4.4[−] ⇒ c = −4.4 · 0.68 = −2.99
Hint: If the maturity function according to Saul is used, the exponent s there might be specified
with item S.
With this record spring elements from the data base will become effective for the analysis.
These elements can be redefined at any time which is especially useful for systems where
temperature and seepage analysis has to be performed. The link element allows a rather
general coupling within hydraulic or thermal analysis problems. The difference of the potential
or temperature is given by the following equation:
Q Q 2
dh = A · + B ·
Areƒ Areƒ
For example:
L dh
dh = · Q ⇐⇒ Q = Areƒ · K ·
K · Areƒ L
for a prismatic body with area Areƒ , the conductivity K and the length L. The term B can be
used only with a nonlinear analysis.
Moreover a radiation condition according to the law of Boltzmann may be applied. EPS is than
the emission ratio of the black body. For the correct definition of the boundary conditions it must
then be distinguished between absolute or relative temperatures in degree Kelvin or Celsius
and the correct literal selected at record SYST . Of course this effect is also nonlinear.
Hydraulic Conduction
A m per m/sec K ·m2 /W
B m per m/sec not reasonable
S 1 = m3 /m/m2 J/m2 /K
EPS not reasonable 1
Hydraulic conductive elements like pipes or culverts are introduced by truss elements. As a
section of a pipe the simplest case is defined with a TUBE where the nominal width and the
roughness defined be MEXT is used. The more general case of an open or closed section may
be described with a polygon section.
The pipe element allows the analysis of full flow governed by the rule of Prandtl-Colebrook and
of open channels with the Darcy-Weissbach formulas. The storage coefficient S will be derived
from the compressibility of the fluid and the elastic properties of the pipe material or from the
cuurent width of the free surface (flow width).
An analysis requires the setting of TURB=1 and KITE=1 (Record CTRL) in general.
A section with VERT-elements is composed from any number of polygon segments collected
within partial sections. Each edge of the polygons may have a different roughness assigned
with EXP and MEXT.
All results of an analysis are saved in the data base with a number LC (loading case). For
every input block only one LC-record is allowed.
It is also possible to save differences to the initial estate as a separate loading case. This
will be done be defining with record PLC the initial estate and with CTRL LDIF an positive or
negative offset of the load case number.
There are variants possible which design to any node the same value or which integrate the
values between the adjacent nodes.
The flow will always be from the higher to the lower potential value. The sign of the potential
is for free surface problems always to the top positive. For those problems the potential value
will be only applied if the resulting pressure is positive. If this is not true, the potential defined
by pressure zero will be set and maintained as long as the resulting quantities are negative
(outflow) . If the gravity direction is in a positive direction (GDIR POSZ), we have for a pressure
free seepage boundary therefore H = -z.
In special cases e.g. for the specification of water levels within a BRIC element edge this
behaviour may be not appropriate. In those cases the user may act as follows:
• If NB is negative or zero, the check will be completely disabled, i.e. the potential is always
set without checking the quantities.
• VP describes the possible suction (negative pore pressure) at the node. The potential will
not be corrected until this values has been reached. For the type HP this value is preset to
-9999 which will always apply the value of the primary estate. .
The resulting quantities at the nodes applied with the H, HP or SPEC boundary condition will
be evaluated separately for all values of NB and may be saved with the HIST record to the
database for later display.
It is also possible to define global gradients for the boundary value. For any selected node, the
value at that point is calculated via:
2):
Conductivity α
[ W/m2 K]
Inner wall 7.8
Outer side of ceilings and walls 25.0
Protected outer side of ceilings and walls 12.5
Floors, Heat transfer downwards 5.8
Floors, Heat transfer upwards 10.0
Fire case, hot side 25.0
Fire case, cool side 5.0
The emissivity EPS = εƒ · εm contains the factors of the source εƒ (e.g. 0.8) and of the ma-
terial surface εm (e.g. 0.7). (cmp. EC1-2, Abs. 4.2.1) A total value of 0.5 to 0.7 is therefore
quite common. However Kordina has stated, that for a numerical fire analysis of a standard
fire curve (ETK of DIN 4102) a resulting factor of 0.4 yields much more reasonable results.
(Betonkalender, 2003, page 180). Further one may find in the literature:
The record SURF acts as an area boundary condition on all quadrilateral elements. You may
address standard QUAD-elements wit a positive number or surface elements also with a neg-
ative number. The conditions QAV and QIV however act on volume elements.
The parameters of the SPEC-Condition are explained in detail at the BOUN record descrip-
tion.
T Time value ∗ !
F Function value or predefined type −/ LT !
ISO Standard fire curve
EXT External fire curve
CARB Hydro carbon-fire curve
NO Number of the time function − 1
TITL Title of this function Lt32 -
The record FUNC describes a boundary condition which changes in time during a transient
analysis. The time values are expected in the unit specified at record SYST. Without specifica-
tion of FUNC the boundary condition will be treated as constant during time. With the definition
the function values of BOUN or SURF will be multiplied with the function value at any given
time value. For type SPEC only the VP-value will be multiplied. In that case it is advisable to
set VP to 1.0 and to describe the potential values with FUNC.
A function is composed by at least two pairs of values with the same function number. The
functions are expected to be sorted by time. The function value before the first and after the
last time value is assumed to be zero. Instead of discrete functions a literal may be used to
select predefined curves as specified in EC1-2 chapter 4.2. for design fires:
1100 HYDRO−CARBON
1050 ISO/ETK
1000
950
900
850
800
750
700
EXTERN
650
600
550
500
450
400
350
300
250
200
150
100
100
105
110
115
120
5.00
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
55.0
60.0
65.0
70.0
75.0
80.0
85.0
90.0
95.0
50.0
0.0
[min]
HIST
For transient analysis there are either the method to save the complete results for any time step
at a different load case number or to select some important results like potentials, temperature
or flux quantities of elements or groups (max 1024) and to save their development in time in a
compact form in the database. DYNR will then be used to show these diagrams
If a selection set has already been established for the primary load case, a definition of HIST
REST will take this and continue with the time history automatically.
COPY
Correction of pressures
PCOR by a value to add ∗ 0.0
PREF by a reference to divide ∗ 1.0
COPY is a command for special cases where separate result cases should be combined to a
single one. For the evaluation of wind pressures which have been calculated with PHYSICA,
the conversion to pressure coefficients is a useful aid.
IMPO
With IMPO it is possible to import results from other programs into HYDRA. The import file
is a text file and may contain any comments starting with a $ or #, the data values have to
be specified in lines each containing the node number and the data value at this node in free
format.
A line *END terminates the import, a line *TIME=timevalue may be used to structure a transient
result file.
ECHO
4 Output Description
4.2 Nodes
The first table contains the maximum coordinates within the system followed by the explicitly
defined nodes. Option ECHO NODE FULL will switch to the total table. The same mechanism
will be used for the printout of nodal results.
The nodal table will be printed when creating or modifying a system or with an explicit option
ECHO FULL or NODE.
It contains:
4.3 Materials
If materials have been defined a material table follows:
NO Material number
P pore pressure level
S[ 1/m] Storage coefficient
NSP[ -] effective porosity
Kxx[ m/sec] Permeability
Kyy[ m/sec]
Kzz[ m/sec]
NO Material number
TEMP Temperature level
4.4 Sections
The table of pipe and culvert sections is hierarchically organized:
C R O S S S E C T I O N S
No PNO VNO area CIRCUM EFFECT. K LEAKAGE Y Z
[m2] [m] [m] [mm] (m2/sec) [m] [m]
2 57.0000 59.301 31.301 4.9 1.55E+02 S = 1.250E+01
2 1 6.5000 15.414 8.414 12.6 2.54E+01 15.000 -2.000
1 2 20.0 1.00E+00 14.000 -1.000
1 3 12.0 4.00E+00 8.000 -1.000
1 4 -8.0 0.00E+00 8.000 -2.000
2 2 46.0000 36.472 20.472 5.6 1.12E+02 8.000 -2.000
2 12 -4.0 0.00E+00 8.000 -1.000
2 13 3.0 0.00E+00 7.000 1.000
2 14 7.0 8.00E+00 -7.000 1.000
2 15 3.0 0.00E+00 -8.000 -1.000
2 16 -4.0 0.00E+00 -8.000 -2.000
...
The first number gives the total sectional values with the mean roughness and leakage factors.
Then (with ECHO SECT FULL) one will get the sectional values of every partial section, and
with (ECHO SECT EXTR) the individual vertices of the section.
4.5 Elements
The second part of the system description are the elements. These tables will be printed if a
system is defined or the option ECHO ELEM FULL has been selected.
Link Elements
Pipe Elements
Plane Elements
Volume Elements
Boundary Conditions
Surface Boundary Conditions
dt is the size of the time step, the variation of the potential within the time, the following iteration
lines reflect the changes to the last iteration. "f" is the correction factor or the parameter of the
line search.
For every well there will be a line with the current water head and the resulting take out. For
transient analysis and option ECHO FUNC one will get functional values of the function for
every time step.
Plane Elements
For open culverts a "flooded" estate will be marked. Elements with velocities above a given
limit will be flagged with asterix.
For LINK-elements only ECHO VELO FULL will print potential differences and flux quantities.
With ECHO VELO NO those tables may be suppressed. Option ECHO VELO FULL will also
print the flow rates. If the porosity is not specified, the real velocity is undefined, then the flow
rates will be printed instead in the only printed line.
All nodes which yield a quantity entering or leaving the system will be printed within a separate
table. With nonlinear problems there might be small quantities in other nodes. Option ECHO
QUAN EXTR will print all nodes even those with zero or very small quantities.
Capacity effects of derivatives in time of a transient analysis will be also calculated and in-
cluded. For larger time steps the total balance of flow will therefore not match. ECHO ERG 6
or 7 instead of FULL or EXTR will suppress this feature. All nodes will then show the quantities
stored or supplied at the node.
For coupled nodes, one will get two quantities with opposite signs in the master and slave
node.
4.10 Balancing
For every group there will be the balance of the quantities going in and out and internally
absorbed or generated. A total balance of all groups will be printed whenever more than two
groups are present in the analysis.
For graphical results there are the modules WING/WinGRAF and DYNR available. As only a
few commands from DYNR are required for HYDRA we repeat them here for the ease of use.
SIZE
The horizontal scale of the time will be automatically selected based on the given time val-
ues. The unit will be switched to minutes or days whenever the size of the values makes this
reasonable. At HIST you may also specify an interval at your own discretisation.
For further remarks see the general SOFiSTiK-Manual "Installation and Basics"
SCHH
The default of H1 will be corrected for smaller formats to enable the display of 72 characters.
A value of 0.0 will suppress the labelling.
FUNC
One can define any function within DYNR to be used as comparison within the HIST displays.
NO and TITL are evaluated only for the first function value. If no time value is given, the last
time increment will be added to the last time value. For regular time steps the time values may
be omitted therefore for all higher values.
HIST
C6 − C5
The command HIST can request up to six time variation curves per graph. The displayed range
of values is determined automatically based on the found values and the scale data given in
SIZE. YMIN and YMAX or TMIN and TMAX can be used to modify this range.
H = Potentials
QH QBH QHPI = Quantities nodes, regions, pipes
VQUA VBRI = velocities QUAD, BRIC-Elements
T = Temperature
QT QBT QTPI = Heat flux nodes, regions, links
QQUA QBRI = Heat flux densities QUAD, BRIC-Elements
Beside the variations that are stored in the database, one can also define and plot comparison
curves with FUNC. The negative value of the function number must be given to one of the
NOi-entries for this purpose. In order to plot the functions only, one must enter the value of 0
for LCS.