Professional Documents
Culture Documents
Caesar Ii Ug
Caesar Ii Ug
Caesar Ii Ug
=
summed over i = 1 to n, by 6
(X-direction degrees of freedom)
%Active MassY
[ ]
[ ]
100( )
e
M i
M i
=
summed over 1 = 2 to n, by 6
(Y-direction degrees of freedom)
%Active Massz
[ ]
[ ]
100( )
e
M i
M i
=
summed over 1 = 3 to n, by 6
(Z-direction degrees of freedom)
Where:
Me = vector (by degree-of-freedom) of sum (over all extracted modes) of effective modal
masses
M = vector corresponding to main diagonal of system mass matrix
The maximum possible percent of active mass which is theoretically possible is of course 100%, with 90-
95% usually indicating that a sufficient number of modes have been extracted to provide a good dynamic
model.
The percent of active force is calculated by the following factors:
separately summing the components of the effective force acting along each of the three directional
degrees-of-freedom
combining them algebraically
doing the same for the applied load
taking the ratio of the effective load divided by the applied load
For example:
Fe
x
= EFe
[i]
Fx = EF[i]
66 CAESAR II Technical Reference Manual
summed over i = 1 to n, by 6
(X - direction degrees of freedom)
Fe
y
= EFe
[i]
Fy = EF[i]
summed over i = 2 to n, by 6
(Y - direction degrees of freedom)
Fe
z
= EFe
[i]
Fz = EF[i]
summed over i = 3 to n, by 6
(Z - direction degrees of freedom)
% Active Force
2 2 2
2 2 2
[ ]
100*
[ ]
e e e
F x F y F z
Fx Fy Fz
+ +
=
+ +
Where:
FeX,FeY,FeZ = effective force (allocated to extracted modes) acting along the global X-, Y-,
and Z-axes, respectively
Fr = vector of effective forces (allocated to extracted modes)
FX,FY,FZ = total system forces acting along the global X-, Y-, and Z-axes, respectively
F = vector of total system forces
The maximum possible percent which is theoretically possible for this value is also 100%; however, in
practice it may be higher, indicating an uneven distribution of the load and mass in the system model.
There is nothing inherently wrong with an analysis where the included force exceeds 100%if the
missing mass correction is included, the modal loadings will be adjusted to conform to the applied loading
automatically. Often the percent of included force can be brought back under 100% by extracting a few
more modes. At other times, the situation can be remedied by improving the dynamic model through a
finer element mesh, or, more importantly, equalizing the mass point spacing in the vicinity of the load.
Chapter 5 Controlling the Dynamic Solution 67
Frequency Cutoff (HZ)
(Active for: Spectrum, Modal, and Time History)
As noted above, CAESAR II permits the user to specify either a number of modes or a frequency cutoff for
extracting modes to be considered in the dynamic analysis. Modal extraction ceases when the Eigensolver
extracts either the number of modes requested, or extracts a mode with a frequency above that of the
Frequency Cutoff, whichever comes first.
One recommendation for selection of a frequency cutoff point is that the user extract modes up to, but not
far beyond, a recognized rigid frequency, and then include the missing mass correction (discussed in the
section Include Missing Mass Components). Choosing a cutoff frequency to the left of the response
spectrums resonant peak will provide a non-conservative result, since resonant responses may be missed.
During spectrum analysis, using a cutoff frequency to the right of the peak, but still in the resonant range,
will yield either overly- or underly-conservative results, depending upon the method used to extract the
ZPA from the response spectrum. (In the case of time history analysis, selecting a cutoff frequency to the
right of the peak, but still in the resonant range, will probably yield non-conservative results, since the
missing mass force is applied with a dynamic load factor of 1.0). Extracting a large number of rigid modes
for calculation of the dynamic response may be conservative in the case of Spectrum analysis, since all
spectral modal combination methods (SRSS, GROUP, ABS, etc.) give conservative results versus the
algebraic combination method (always used during time history analysis), which gives a more realistic
representation of the net response of the rigid modes. Based upon the response spectrum shown in the
following figure, an appropriate cutoff point for the modal extraction would be about 33 Hz.
Non-conservative cutoff (Misses amplification of any modes in resonant range)
Conservative cutoff (Multiplies missing mass contribution by excessive DLF1.6)
Optimal cutoff (Includes all modes in resonant range, uses low DLF1.05for missing mass
contribution, minimizes combination of rigid modes)
Conservative Cutoff (Too many rigid modes combined using non-conservative summation methods)
68 CAESAR II Technical Reference Manual
When the analysis type is SPECTRUM, MODES, or TIMEHIST, either this parameter or the previous one
must be entered.
Closely Spaced Mode Criteria/Time History Time Step (ms)
(Active for: Spectrum/GROUP and Time History)
This parameter does double duty, depending upon the analysis type. For a Spectrum analysis type with
GROUP modal Combination Method (as defined by USNRC Regulatory Guide 1.92), this parameter
specifies the frequency spacing defining each modal groupi.e., the percent (of the base frequency)
between the lowest and highest frequency of the group. Regulatory Guide 1.92 specifies the group spacing
criteria as 10% (entered here as 0.1), so it is unlikely that the user would ever wish to change the Closely
Spaced Mode Criteria from the CAESAR II default value of 0.1. The GROUP modal combination method is
described in detail in the section Modal Combination Method found later in this chapter.
For a Time History analysis type, this parameter is used to enter the length of the time slice, in
milliseconds, to be used by the program during its step-by-step integration of the equations of motion for
each of the extracted modes (CAESAR II uses the unconditionally stable Wilson q integration method, so
any size time step will provide a solution, with a smaller step providing greater accuracyand more strain
on computational resources). The time step should be sufficiently small that it can accurately map the
force vs. time load profile (i.e., the time step should be smaller than typical force ramp times).
Additionally, the time step must be small enough that the contribution of the higher order modes is not
filtered from the response. For this reason, it is recommended that the time step should be selected such
that Time Step (in seconds) times Maximum Modal Frequency (in Hz) be less than 0.1. For example, if the
modal frequency cutoff is set to 50 Hz, the time step should be set to a maximum of 2 milliseconds:
0.002 sec x 50 Hz = 0.1
Chapter 5 Controlling the Dynamic Solution 69
Load Duration (Time History or DSRSS Method) (Sec.)
(Active for: Spectrum/DSRSS and Time History)
This parameter is used to specify the duration of the applied dynamic load. For a Time History analysis,
this parameter is used to specify the total length of time (in seconds) over which the dynamic response is
to be simulated. The load duration, divided by the time step size (see the previous section) gives the total
number of integration steps making up the solution (currently CAESAR II limits the number of time steps to
5000, or as permitted by available memory and system size). It is recommended that, if possible, the
duration be at least equal to the maximum duration of the applied load, plus the period (in seconds) of the
first extracted mode. This allows simulation of the system response throughout the imposition of the
external load, plus one full cycle of the resulting free vibration. After this point, the response will die out,
according to the damping value used. For example, if the applied load is expected to last 150 milliseconds,
and the lowest extracted frequency is 3 hz, the load duration should be set to a minimum of 0.150 plus 1/3,
or 0.483 seconds.
For a Spectrum analysis using the Double Sum (DSRSS) modal Combination Method (as defined by
USNRC Regulatory Guide 1.92), this parameter is used to specify the duration of the earthquake, in
seconds. This duration is used to compute the modal correlation coefficients based on empirical data. The
DSRSS modal combination method is described in detail in the section Modal Combination Method later in
this chapter.
Damping (Time History or DSRSS) (Ratio of Critical)
(Active for: Spectrum/DSRSS, Harmonics, and Time History)
This parameter is used to specify the system damping value, as a ratio of critical damping. Typical values
for piping systems, as recommended in USNRC Regulatory Guide 1.61 and ASME Code Case N-411,
range from 0.01 to 0.05, based upon pipe size, earthquake severity, and the systems natural frequencies.
Generally, damping cannot be considered in the mathematical solutions required for spectrum or harmonic
analysis. It is therefore ignored (or solved as specialized cases) in most analyses, and must be instead
considered through adjustment of the applied loads (generation of the response spectrum) and/or system
stiffness.
For a Time History analysis, damping is used explicitly, since this method uses a numeric solution to
integrate the dynamic equations of motion.
For a Spectrum analysis using the Double Sum (DSRSS) modal Combination Method (as defined by
USNRC Regulatory Guide 1.92), the damping value is used in the computation of the modal correlation
coefficients. (Note that CAESAR II does not permit the specification of damping values for individual
modes.) The DSRSS modal combination method is described in detail in the section Modal Combination
Method later in this chapter.
For a Harmonic analysis, this ratio is converted to Rayleigh Damping, where the damping matrix can be
expressed as multiples of the mass and stiffness matrices:
[C] = a [M] + b [K]
70 CAESAR II Technical Reference Manual
On a modal basis, the relationship between the ratio of critical damping Cc and the constants a and b is
given as:
Where:
e= undamped natural frequency of mode (radians/sec)
For many practical problems, a is extremely small, and so may be ignored, reducing the relationships to:
o = 0 | = 2 Cc / e
CAESAR II uses this implementation of damping for its harmonic analysis, with the exception that a single
b is calculated for the multi-degree-of-freedom system, and the w used is that of the load forcing
frequency. When the forcing frequency is in the vicinity of a modal frequency, this gives an accurate
estimate of the true damping value.
Chapter 5 Controlling the Dynamic Solution 71
ZPA (Reg. Guide 1.60/UBC- G's)/# Time History Output Cases
(Active for: Spectrum/1.60/UBC and Time History)
This parameter does double duty, depending upon the analysis type. When used with certain pre-defined
normalized response spectra, it is used as the acceleration factor (in g's) by which the spectrum is scaled.
For example, when a spectrum analysis uses one of the pre-defined spectra names beginning with "1.60"
(i.e., 1.60H.5 or 1.60V7), CAESAR II constructs an earthquake spectrum according to the instructions
given in USAEC (now USNRC) Regulatory Guide 1.60. That guide requires that the shape of the response
spectrum be chosen from the curves shown in the following figures, based upon the system damping value
(for example, the .5 or 7 in the spectrum names 1.60H.5 or 1.60V7). If the analysis uses one of the pre-
defined spectra names beginning with "UBC" (i.e., UBCSOIL1), CAESAR II uses the normalized seismic
response spectra for the corresponding soil type from Table 23-3 from the Uniform Building Code (1991
Edition). Both the Reg Guide 1.60 and the UBC curves are normalized to represent a ground acceleration
(ZPA) of lg; the true value is actually site dependent. Therefore, entering ZPA value here appropriately
scales any Regulatory Guide 1.60 or the Uniform Building Code response spectra.
72 CAESAR II Technical Reference Manual
When performing Time History analysis, this parameter is used to specify the number of distinct times at
which the results of the load cases (the dynamic load as well as all static/dynamic combinations) should be
generated. In addition, CAESAR II generates one set of results (for each load case) containing the maximum
of each output value (displacement, force, stress, etc.) along with the time at which it occurred. The times
for which results are generated are determined by dividing as evenly as possible the load duration by the
number of output timesfor example, if the load duration is 1 second, and 5 output cases are requested,
results will be available at 200, 400, 600, 800, and 1000 milliseconds (in addition to the maximum case).
The total number of results cases generated for an analysis is the product of the number of load cases (one
dynamic case plus the number of static/dynamic combination cases) times the number of results cases per
load (one maxima case plus the requested number of output cases). Currently the total number of results
cases is limited to 99:
(1 + # Static/Dynamic Combinations) x (1 + # Output Cases) <= 99
At least one output case (in addition to the automatically generated maxima case) must be requested; more
than one is not really necessary, since the worst case results are reflected in the Maxima case and
individual results at every time step are available through the ELEMENT command when animating the
Time History results.
Chapter 5 Controlling the Dynamic Solution 73
Re-use Last Eigensolution
(Active for: Spectrum and Time History)
When repeating a dynamic analysis, this parameter may be set to Yes, causing CAESAR II to skip the
eigensolution (reusing the results of the earlier analysis), and only perform the computations for
displacements, reactions, forces, and stresses. Activating this option is only valid after an initial
eigensolution has been performed and is still available. Additionally, the mass and stiffness parameters of
the model must be unchanged or the previous eigensolution is invalid.
Spatial or Modal Combination First
(Active for: Spectrum)
This directive tells CAESAR II whether to combine the Spatial components or the Modal components of the
load case first.
When performing a spectrum analysis, each of the modal responses must be summed. In addition, if
multiple shocks have been applied to the structure in more than one direction, the results from different
directions must be combinedfor example, spatially combining the X-direction, Y-direction, and Z-
direction results. The question arises as to whether the spatial summations should precede or follow the
modal summations. A difference in the final results (of Spatial first vs. Modal first) arises whenever
different methods are used for the spatial and modal combinations.
The combination of Spatial components first implies that the shock loads are dependent, while the
combination of Modal components first implies that the shock loads are independent.
Dependent and Independent refer to the time relationship between the X, Y, and Z components of the
earthquake. With a dependent shock case, the X, Y, and Z components of the earthquake have a direct
relationshipa change in the shock along one direction produces a corresponding change in the other
directions. For example, this would be the case when the earthquake acts along a specific direction having
components in more than one axissuch as when a fault runs at a 30 angle between the X- and Z-axes.
In this case, the Z-direction load would be a scaled (by a factor of tan 30), but otherwise identical version
of the X-direction load. In this case, spatial combinations should be made first.
An Independent shock is one where the X, Y, and Z time histories produce related frequency spectra but
have completely unrelated time histories. It is the Independent type of earthquake that is far more
common, and thus in most cases the modal components should be combined first.
For example, IEEE 344-1975 (IEEE Recommended Practices for Seismic Qualification of Class 1E
Equipment for Nuclear Power Generating Stations) states:
Earthquakes produce random ground motions which are characterized by simultaneous but statistically
INDEPENDENT horizontal and vertical components.
74 CAESAR II Technical Reference Manual
This is usually less of an issue for force spectrum combinations, since normally there are no separate
spatial components to combinei.e., there are not X-, Y-, and Z-shocks acting simultaneously. However,
in the event that there is more than one potential force load (such as when there is a bank of relief valves
that can fire individually or in combination), the spatial combination method may be used to indicate the
independence of the loadings. For example, if two relief valves may or may not fire simultaneously (i.e.,
they are independent), the two shocks should be defined as being in different directions (for example, X-
and Y-), and the combination method selected should be Modal before Spatial. If under certain
circumstances, the two valves will definitely open simultaneously (i.e., the loadings are dependent), the
combination method should be Spatial before Modal. (Otherwise, the direction defined for a force
spectrum loading has no particular meaning.)
Nuclear Regulatory Guide 1.92 (published in February, 1976) describes the requirements for combining
spatial components when performing seismic response spectra analysis for nuclear power plants.
Note: Since all Time History combinations are done algebraically (in-phase) this parameter has no
effect on Time History results.
Spatial Combination Method (SRSS/ABS)
(Active for: Spectrum)
This parameter is used to define the method for combining the spatial contributions of the shocks in a
single spectrum load case. This option is only used for spectrum runs with more than a single excitation
direction. Since directional forces are usually combined vectorially, this points to a Square Root of the
Sum of the Squares (SRSS) combination method as being most appropriate. An Absolute method is
provided for additional conservatism.
Note: Since all Time History combinations are done algebraically (in-phase) this parameter has no
effect on Time History results.
Modal Combination Method (GROUP/10%/DSRSS/ABS/SRSS)
(Active for: Spectrum)
During a spectrum analysis, responses are calculated for each of the individual modes; these individual
responses are then combined to get the total system response. Considering that the response spectrum
yields the maximum response at any time during the course of the applied load, and considering that each
of the modes of vibration will probably have different frequencies, it is probable that the peak responses of
all modes will not occur simultaneously. Therefore an appropriate means of summing the modal responses
must be considered.
Nuclear Regulatory Guide 1.92 (published in February, 1976) defines the requirements for combining
modal responses when performing seismic response spectra analysis for nuclear power plants. The four
options presented there are also available, along with one other, for modal combinations under non-
nuclear seismic and force spectrum analyses.
There are five available modal combination methods:
Grouping Method
Ten Percent Method
Double Sum Method
Absolute
Chapter 5 Controlling the Dynamic Solution 75
Square Root of the Sum of the Squares
Grouping Method
This method is defined in USNRC Regulatory Guide 1.92. The Grouping Method attempts to eliminate
the drawbacks of the Absolute and SRSS methods (see below) by assuming that modes are completely
correlated with any modes with similar (closely spaced) frequencies, and are completely uncorrelated with
those modes with widely different frequencies. Therefore, the total system response is calculated as
1/ 2
2
1
1
(where 1 )
N
j P j
q m i lq mq l i
k
R Rk R R m
= = =
=
(
= + + + =
(
Where:
R = total system response of the element
N = number of significant modes considered in the modal response combination
Rk = the peak value of the response of the element due to the kth mode
P = number of groups of closely-spaced modes (where modes are considered to be closely-spaced if
their frequencies are within 10% of that of the base mode in the group), excluding individual
separated modes. No mode can be in more than one group.
i = number of first mode in group q
j = number of last mode in group q
Rlq = response of mode l in group q
Rmq = response of mode m in group q
Effectively, this method dictates that the responses of any modes which have frequencies within 10% of
each other first be added together absolutely, with the results of each of these groups then combined with
the remaining individual modal results using the SRSS method.
Note: The 10% figure controlling the definition of a group may be changed by using the Closely Spaced
Mode Criteria/Time History Time Step (ms) parameter. For more information see the corresponding section
earlier in this chapter.
Ten Percent Method
This method is defined in the USNRC Regulatory Guide 1.92. The Ten Percent Method is similar to the
Grouping method in that it assumes that modes are completely correlated with any modes with similar
(closely spaced) frequencies, and are completely uncorrelated with those modes with widely different
frequencies. The differences between this one and the preceding method is that the Grouping Method
assumes that modes are only correlated with those that fall within the group -i.e., are within a 10% band,
while this method assumes that modes are correlated with those that fall within 10% of the subject mode-
effectively creating a 20% band - 105 up and approximately 10% down. The total system response is
calculated as
1/ 2
2
1
2 (Where i j)
N
i j
k
R Rk R R
=
(
= + =
(
This method is based upon the statistical assumption that all modal responses are completely independent,
with the maxima following a relatively uniform distribution throughout the duration of the applied load.
This is usually non-conservative, especially if there are any modes with very close frequencies, since those
modes will probably experience their maximum DLF at approximately the same time during the load
profile.
Note: Since all Time History combinations are done algebraically (in-phase) this parameter has no
effect on Time History results.
Include Pseudostatic (Anchor Movement) Components (Y/N)
(Active for: Spectrum/ISM)
This option is only used when Independent Support Motion (anchor movement) components are part of a
shock load case. The excitation of a group of supports produces both a dynamic response and a static
response. The static response is due to the movement of one group of supports or anchors relative to
another group of supports/anchors. These static components of the dynamic shock loads are called
pseudostatic components. USNRC recommendations, as of August 1985, suggest that the following
procedure be followed for pseudostatic components:
1 For each support group, the maximum absolute response should be calculated for each input direction.
2 Same direction responses should then be combined using the absolute sum method.
3 Combination of the directional responses should be done using the SRSS method.
4 he total response should be formed by combining the dynamic and pseudostatic responses, using the
SRSS method.
Therefore pseudostatic components should be included whenever Independent Support spectral loadings
are used.
78 CAESAR II Technical Reference Manual
Include Missing Mass Components (Y/N)
(Active for: Spectrum and Time History)
During spectrum (either seismic or force spectrum) or time history analyses, the response of a system
under a dynamic load is determined by superposition of modal results. One of the advantages of this type
of modal analysis is that usually only a limited number of modes are excited and need to be included in the
analysis. The drawback to this method is that although displacements may be obtained with good accuracy
using only a few of the lowest frequency modes, the force, reaction, and stress results may require
extraction of far more modes (possibly far into the rigid range) before acceptable accuracy is attained.
CAESAR II provides a feature, called the Missing Mass Correction, which helps solve these problems.
This feature offers the ability to include a correction which represents the contribution of the higher order
modes not explicitly extracted for the modal/dynamic response, thus providing greater accuracy without
additional calculation time. When this option is activated (by entering Yes for this parameter), the
program automatically calculates the net (in-phase) contribution of all non-extracted modes and combines
it with the modal contributionsavoiding the long calculation time associated with the extraction of the
high order modes and the possible excessive conservatives of the summation methods. This feature is
described in Chapter 6 of this manual.
Pseudostatic (Anchor Movement) Comb. Method (SRSS/ABS)
(Active for: Spectrum)
This directive specifies the method by which the pseudostatic responses (see description in the earlier
section Include Pseudostatic (Anchor Movement) Components (see "Include Pseudostatic (Anchor
Movement) Components (Y/N)" on page 77)) are to be combined with the dynamic (inertial) responses;
therefore it is applicable only when there is at least one Independent Support Motion excitation component
in a shock load case. Pseudostatic combinations are done after all directional, spatial, and modal
combinations. Absolute combination gives conservative results, but, as noted in the section Include
Pseudostatic (Anchor Movement) Components, the USNRC recommends using the SRSS method for
pseudostatic combinations.
Missing Mass Combination Method (SRSS/ABS)
(Active for: Spectrum)
This directive defines the method used to combine the missing mass/force correction components (see
description in an earlier section, Include Missing Mass Components (see "Include Missing Mass
Components (Y/N)" on page 78)) with the modal (dynamic) results. Research suggests that the modal and
rigid portions of the response are statistically independent, so the SRSS combination method (CAESAR IIs
default) is usually most accurate. The Absolute combination method provides a more conservative result,
based upon the assumption that the modal maxima occur simultaneously with the maximum ground
acceleration. Missing mass components are combined following the modal combination.
Note: Even though missing mass components may be included during Time History analyses, all Time
History combinations are done algebraically (in-phase), so this parameter has no effect on Time History
results.
Chapter 5 Controlling the Dynamic Solution 79
Directional Combination Method (SRSS/ABS)
(Active for: Spectrum)
This directive specifies the method used for combining shock components acting in the same direction.
This directive is used most typically with Independent Support Motion load cases, where it defines the
way in which responses from different support groups caused by excitation in the same direction are
combined. Additionally, if there are multiple uniform shock spectra acting in the same direction (although
this is unusual), this directive would govern their combination. In general, directional combinations should
be made using the absolute method. (As noted in the earlier section, Include Pseudostatic (Anchor
Movement) Components (see "Include Pseudostatic (Anchor Movement) Components (Y/N)" on page 77),
this is the USNRC recommendation for directional combination of pseudostatic responses.) However, in
the case of force spectrum loads, if several loads (for example, several relief valve loads) are all defined
with the same shock direction, using an SRSS combination method would be a way of modeling these
as independent loads, while using the Absolute method would model them as dependent loads.
Note: Since all Time History combinations are done algebraically (in-phase) this parameter has no
effect on Time History results.
Sturm Sequence Check on Computed Eigenvalues (Y/N)
(Active for: Spectrum, Modal, and Time History)
In almost all cases, the eigensolver will detect modal frequencies from the lowest frequency to the highest.
Sometimes, when there is some strong directional dependency in the system, the modes may converge in
the wrong order. This could cause a problem if the eigensolver reaches the cutoff number of modes (i.e.,
20), but has not yet found the 20 modes with the lowest frequency (it may have found modes 1 through 18,
20, and 21, and would have found number 19 next). CAESAR II checks for this anomaly using the Sturm
Sequence calculation. This procedure determines the number of modes that should have been found
between the highest and lowest frequencies found, and compares that against the actual number of modes
extracted. If those numbers are different, the user is given a warning. For example, if 22 natural
frequencies are extracted for a particular system, and if the highest natural frequency is 33.5 Hz, the Sturm
Sequence check makes sure that there are exactly 22 natural frequencies in the model between zero and
33.5+p Hz, where p is a numerical tolerance found from:
10
Log [(Highest Eigenvalue)-(Number of Significant Figures+1.5)]
10
p=
2n
80 CAESAR II Technical Reference Manual
The Sturm Sequence check would fail in the case where there are two identical frequencies at the last
frequency extracted. The significance of this failure can only be estimated by the user. For example,
consider a system with the following natural frequencies:
0.6637 1.2355 1.5988 4.5667 4.5667
If the user asks for only the first four natural frequencies, a Sturm Sequence failure would occur because
there are five frequencies, rather than four, which exist in the range between 0.0 and 4.5667 + p (where p
calculates to 0.0041). To correct this problem, the user can do either of the following:
Increase the frequency cutoff by the number of frequencies not found. (This number is reported by the
Sturm Sequence Check.)
Increase the cutoff frequency some small amount, if the frequency cutoff terminated the
eigensolution. This will usually allow the lost modes to fall into the solution frequency range.
Fix the subface size at 10 and rerun the job. Increasing the number of approximation vectors improves the
possibility that at least one of them will contain some component of the missing modes, allowing the
vector to properly converge.
The default here is Yes, and should be left alone unless the user has some specific reason for
deactivating the check.
Chapter 5 Controlling the Dynamic Solution 81
Advanced Parameters
Estimated Number of Significant Figures in Eigenvalues
(Active for: Spectrum, Modal, and Time History)
This is the approximate number of significant figures in the computed eigenvalues (e
2
, where eis the
angular frequency in rad/sec). For example, using the default value of 6, if a computed eigenvalue was
44032.32383, then the first digit to the right of the decimal is probably the last accurately computed figure.
The eigenvectors, or mode shapes, are computed to half as many significant figures as are the eigenvalues.
If the eigenvalues have 6 significant figures of accuracy, then the eigenvectors have 3.
This number should typically never be decreased. Increases to 8 or 10 are not unusual but result in slower
solutions with typically little change in response results.
82 CAESAR II Technical Reference Manual
Jacobi Sweep Tolerance
(Active for: Spectrum, Modal, and Time History)
Eigen analyses are done using an NxN subspace for calculating the natural frequencies and mode shapes
for a reduced problem. The first step is to perform a Jacobi denationalization of the subspace. Iterations
are performed until the off-diagonal terms of the matrix are approximately zero. The off-diagonal terms
are considered to be close enough to zero when their ratio to the on-diagonal term in the row is smaller the
Jacobi Sweep Tolerance.
The default is 1.0E-12. Users wishing to change this value should be aware of the computers precision
(the IEEE-488 double precision word on the IBM PC has approximately 14 significant figures) and the
approximate size of the on-diagonal coefficients in the stiffness matrix for the problem to be solved
(which may be estimated from simple beam expressions).
Decomposition Singularity Tolerance
(Active for: Spectrum, Modal, and Time History)
During the eigensolvers decomposition of what may be a shifted stiffness matrix, a singularity check is
performed to make sure that the shift is not too close to an eigenvalue that is to be calculated. If a singular
condition is detected, a new shift, not quite as aggressive as the last one, is computed and a new
decomposition is attempted. If the new composition fails, a fatal error is reported from the eigensolver. In
certain cases, increasing the singularity tolerance is warranted and eliminates this fatal error. Values
should not be entered greater than 1.0 E13. Singularity problems may also exist when very light, small
diameter piping is attached to very heavy, large diameter, or when very, very short lengths of pipe are
adjacent to very, very long lengths of pipe.
Subspace Size (0-Not Used)
(Active for: Spectrum, Modal, and Time History)
During an eigensolution, the NDOFxNDOF problem constructed by the user is reduced to an NxN
problem during each subspace iteration, where N is the subspace size. If a zero is entered in this field,
CAESAR II selects what is expected to be an optimal subspace size (so this value usually need not be
changed); if a non-zero value is entered here, it will override CAESAR IIs calculation and will be used as
the subspace size.
CAESAR IIs default is to use the square root of the bandwidth (with a minimum of 4) as the subspace size,
resulting in sizes of 4 to 8 for typical piping configurations. Increasing the subspace size slows the
eigensolution, but increases the numerical stability. Values in the range between 12 and 15 should
probably be used when unusual geometries or dynamic properties are encountered, or when a job is large
(has 100 elements or more, and/or requires that 25 or more frequencies be extracted).
Chapter 5 Controlling the Dynamic Solution 83
No. to Converge Before Shift Allowed (0 - Not Used)
(Active for: Spectrum, Modal, and Time History)
A zero in this field lets CAESAR II select what it thinks will be the most optimal shifting strategy for the
eigen problem to be solved. One way to speed the eigensolution is to improve the convergence
characteristics. The convergence rate for the lowest eigenpair in the subspace is inversely proportional to
e1 / e2, where e1 is the lowest eigenvalue in the current subspace and e2 is the next lowest eigenvalue in
the current subspace. A slow convergence rate is represented by an eigenvalue ratio of approximately one,
and a fast convergence rate is represented by an eigenvalue ratio of zero. The shift is employed to get the
convergence rate as close to zero as possible. The cost of each shift is one decomposition of the system set
of equations. The typical shift value is equal to the last computed eigenvalue plus 90 percent of the
difference between this value and the lowest estimated eigenvalue still nonconverged in the subspace.
As e1 is shifted closer to zero, the ratio e1/e2 will become increasingly smaller thus increasing the
convergence rate. In certain instances where eigenvalues are very closely spaced, shifting can result in
eigenvalues being lost (the Sturm Sequence Check will detect this condition). A large value entered for
this parameter will effectively disable shifting, so no eigenvalues will be missed; however, the solution
will take longer to run. When the system to be analyzed is very large, shifting the set of equations can be
very time consumingin these cases, the user is advised to set this parameter to somewhere between 4
and 8.
No. of Iterations Per Shift (0 - Pgm computed)
(Active for: Spectrum, Modal, and Time History)
A zero in this field lets CAESAR II compute what it thinks is an optimal number of subspace iterations per
shift. This parameter, along with the next one (% of iterations per shift before orthogonalization) can work
together to control solution shifting. These two parameters are used to limit the number of Gram-Schmidt
orthogonalizations that are performed. Trying to limit this number is very dangerous for small subspace
problems, but less dangerous when the subspace size is large (around 10-20 percent of the total number of
eigenpairs required).
The Gram-Schmidt orthogonalization is by default performed once during each subspace iteration. This
orthogonalization makes sure that the eigenvector subspace does not converge to an already found
eigenpair. When a large number of eigenpairs are to be computed this repeated computation can
appreciably slow down the extraction of the highest eigenpairs. Proper setting of these parameters can
cause the eigensolution to perform the orthogonalization every second, third, fourth, etc. iteration, thus
speeding the solution. Unfortunately, once orthogonalized, the subspace may still converge to earlier
eigenpairs during subsequent non-orthogonalized subspace iteration passes. Users setting these
parameters are urged to use caution. The Force Orthogonalization After Convergence (see "Force
Orthogonalization After Convergence (Y/N)" on page 84) parameter (see corresponding section later in
this chapter) should probably also be set if the frequency of orthogonalization is slowed.
84 CAESAR II Technical Reference Manual
Percent of Iterations Per Shift Before Orthogonalization
(Active for: Spectrum, Modal, and Time History)
CAESAR II computes a number of iterations per shift that are to be performed, which the user can alter if
desired. A maximum of N eigenpairs can conceivably converge per subspace pass, where N is the
subspace size (although this is highly unlikely). By default a Gram-Schmidt orthogonalization is
performed for each subspace pass. This directive allows the user to alter this default. For example, if there
are 12 iterations per shift, and the percentage of iterations per shift is 50 percent (an entry of 0.50), the
Gram-Schmidt orthogonalization would be performed every 6 iterations. Users employing this option
should also set the Force Orthogonalization After Convergence (see "Force Orthogonalization After
Convergence (Y/N)" on page 84) directive to Yes. The Percent of Iterations per Shift Before
Orthogonalization parameter is most often used in conjunction with the No. of Iterations per Shift (see
"No. of Iterations Per Shift (0 - Pgm computed)" on page 83) parameter because then the user knows
exactly how many iterations will go by without an orthogonalization.
Force Orthogonalization After Convergence (Y/N)
(Active for: Spectrum, Modal, and Time History)
This parameter is only needed for eigensolutions for which the Percent of Iterations per Shift Before
Orthogonalization (on page 84) (the previous section) has been set to a non-zero value. When set to
Yes in this case, whenever a subspace pass that sees at least one eigenpair convergence completes, a
Gram-Schmidt orthogonalization is performed whether the specified percentage of iterations has been
completed or not.
Use Out-Of-Core Eigensolver (Y/N)
(Active for: Spectrum, Modal, and Time History)
This parameter is used primarily as a benchmarking and debugging aid. When entered as Yes, the out-
of-core eigensolver is automatically invoked regardless of the problem size. Using this solver can take
considerably more time than the in-core solver, but should in all cases produce exactly the same results.
Note that if the problem is too big to fit into the in-core solver (the capacity of which is based upon the
amount of available extended memory), the out-of-core solver will be invoked automaticallythis
parameter does not need to be changed to have this automatic switch occur when necessary.
Frequency Array Spaces
(Active for: Spectrum, Modal, and Time History)
This is the maximum number of eigenpairs that can be extracted for the problem. The default value of 100
is arbitrary. If the user needs to extract more than 100 eigenpairs, then some number greater than the
number to be extracted must be entered.
Chapter 5 Controlling the Dynamic Solution 85
Pulsation Loads
Unexpectedly, and sometimes after support changes or process modifications, an operating line will begin
experiencing large amplitude, low frequency vibration. The first step in the solution is the construction of
the dynamic model. Particular attention should be paid when modeling the piping system in the area of the
field vibration. This might include accurately representing valve operators, in-line flange pairs, orifice
plates and measuring equipment. It is also a good idea to add extra nodes in the area where vibration is
experienced. The extra nodes would be put at bend near nodes and at span midpoints.
The next step is the eigenvalue/eigenvector extraction. If the system is large, then degrees of freedom far
removed from the area of local vibration should be eliminated. (6-10) natural frequencies should be
extracted. Natural frequencies and mode shapes define the systems tendency to vibrate. The mode
shapes extracted should show how the system in the area of the local vibration problem is tending to
displace. In most cases acoustic resonances are coupled with mechanical resonances to produce the large
amplitude vibrations experienced in the field.
Very typically one of the first mode shapes will show exactly the shape displayed by the pipe vibrating in
the field. If the mode shapes extracted do not show movement in the area of the local vibration, then not
enough degrees of freedom were removed from other areas. If the lowest mode shape in the area of the
local vibration problem is above (15) Hz. then there is a good possibility that either the vibration is
mechanically induced or the fluid pulsation peak pressures are very high. Either of these cases may
represent critical situations which should be evaluated by an expert.
When the mode shape is identified which corresponds to the observed field vibration, the pulsation load
model can be developed. Pulsation loads will exist at closed ends, at bends, and at changes in diameter.
Harmonically varying forces are put at these points in an attempt to get the mathematical model to vibrate
like the real piping system.
The driving frequency for the applied harmonic load should be equal to the frequency that pressure pulses
are introduced into the line.
The magnitude of the harmonic load can be estimated within a range of tolerances. The actual design
value is selected from this range such that resulting displacements of the model are close to those observed
in the field.
Output from the harmonic analysis can be processed in the static output processor and maximum restraint
loads due to the dynamic forces calculated.
It is critical when redesigning supports for dynamic loads that static thermal criteria are not violated by
any new support configuration designed.
Important: Static thermal criteria and dynamic displacement criteria must be satisfied
simultaneously.
The ultimate objective of the harmonic analysis will be to find the elbow pair whose unbalanced load
results in the observed field vibration.
Unbalanced loads exist between adjacent elbows because the pressure peak in the traveling wave hits each
elbow at a slightly different time.
86 CAESAR II Technical Reference Manual
If the pressure at elbow a is denoted by Pa(t) and the pressure at elbow b is denoted by Pb(t), then the
unbalanced force which acts along the pipe connecting the two elbows is:
F = A * Pa(t) - A * Pb(t) EQ. (1)
Where A is the inside area of the pipe. The expression for Pa(t) can be found assuming the pressure peak
hits the elbow a at time t = 0:
Pa(t) = Pavg + 0.5 (dP) cos et EQ. (2)
Where:
(Pavg) - average pressure in the line,
(dP) - alternating component of the pressure, (Pmax-Pmin)
(e) - driving frequency.
If the straight pipe between the elbows a and b is (L) inches long, then the pressure peak that has just
passed elbow a will get to elbow b (ts) seconds later, where
(ts) = (L) / c,
(c) being the speed of sound in the fluid. (Remember, pressure pulses travel at the speed of sound, not the
speed of the fluid ! ! !) The expression for the pressure at b can now be written:
Pb(t) = Pavg + 0.5(dP) cos (et + Q) EQ.(3)
Q is the phase shift between the pressure peaks at a and b,
Q = e* (ts). (Where Q is in radians, and eis in radians/second)
Combining equations 1, 2, and 3 the unbalanced pressure force can be written:
F(t) = 0.5(dP)A * [ cos et - cos (et-Q) ]
This function has a maximum:
Fmax = 0.5(dP)A sin Q/cos (Q/2)
and a period of 1/w, and will be approximated with:
f(t) = 0.5(dP)A (sin Q/cos (Q/2)) cos et
The formulation of the harmonic loads can be summarized as follows:
1 Decide which elbow-elbow pair is most likely to have an unbalanced force which could cause the
displacements observed in the field.
2 Find upper and lower estimates for the following variables:
dP Alternating pressure in the line (Pmax - Pmin)
e Driving frequency.
c Speed of sound in the fluid.
L Length between the two elbows.
A Area of the pipe.
Chapter 5 Controlling the Dynamic Solution 87
3. Find the time it takes the pressure wave to get from one elbow to the other.
ts = L / (c-)
(c-) is the lower estimate for the speed of sound in the fluid.
4. Find the largest estimated magnitude of the unbalanced pressure force:
Fmax = (0.5) (dP+)A * sin [(e+) (ts)] / cos [(et) (ts)/2]
(dP+) is the upper estimate for the alternating pressure.
(e+) is the upper estimate for the driving frequency.
5. Run a single harmonic analysis with a force of F = Fmax [cos (et)] acting along the axis of the pipe
between the two elbows. If the pattern of the displacement approximately that seen in the field, and if the
magnitude of the calculated displacement is greater than or equal to the magnitude of the displacement in
the field, then the harmonic load to be used for the design of the new restraints has been found.
88 CAESAR II Technical Reference Manual
Relief Valve Thrust Load Analysis
There can be two types of destructive dynamic forces associated with relief devices:
Thrust at the valve/atmosphere interface
Acoustic shock due to the sudden change in fluid momentum and the associated traveling pressure
wave(s).
The analyst must evaluate the effective contribution of both types of loads. Dynamic forces associated
with relieving devices can cause considerable mechanical damage to equipment and supports. The
discussion below concerns only the thrust at the valve/atmosphere interface. The acoustic traveling
pressure wave can be dealt with similar to the water hammer problem, addressed elsewhere.
The first step in performing a relief load analysis is to compute the magnitudes of the relieving thrust
forces. For open-type vent systems CAESAR II has a RELIEF LOAD SYNTHESIZER that will make these
computations automatically for the user. There are two procedures incorporated into the synthesizer, one is
for gases greater than 15 psig, and the other is for liquids. Both are discussed as follows.
Relief Load Synthesis for Gases Greater Than 15 psig
CAESAR II assumes that a successful vent stack/relief system design maintains the following gas
properties:
Chapter 5 Controlling the Dynamic Solution 89
The input for the gas relief load synthesis is shown as follows:
Line Temperature
Enter the stagnation condition temperature of the gas to be relieved. (Usually just the gas temperature
upstream of the relief valve.)
Line Pressure
Enter the stagnation pressure of the gas to be relieved. (Usually just the gas pressure upstream of the relief
valve.) Note that stagnation properties can vary considerably from line properties if the gas flow velocity
in the line is high.
ID of Relief Valve Orifice
Enter the flow passage inside diameter for the smallest diameter in the relief valve throat. (This
information is usually provided by the relief valve manufacturer).
ID of Relief Valve Piping
Enter the inside diameter of the piping attached directly to the exhaust of the relief valve.
ID of Vent Stack Piping
If CAESAR II is to size the vent stack then leave this ID blank. If the vent stack piping is the same size as
the relief valve piping, i.e. it is one-in-the-same, then this field may be left blank. Otherwise enter the
inside diameter of the vent stack piping.
90 CAESAR II Technical Reference Manual
Length of the Vent Stack
Enter the length of the vent stack. This is a required entry. Add double the lengths of fittings and elbows
(or compute the appropriate equivalent lengths for non-pipe fittings and add the lengths).
Some typical values for these constants are given below:
Ratio of Gas-Specific Heats (k) Gas Constant (R) (ft. lbf./lbm./deg. R
Superheated Steam 1.300 Nitrogen 55.16
Saturated Steam 1.100 Carbon Dioxide 35.11
Nitrogen 1.399 Acetylene 59.35
Carbon Dioxide 1.288 Ammonia 90.73
Acetylene 1.232 n-Butane 26.59
Ammonia 1.304 Ethane 51.39
n-Butane 1.093 Ethylene 55.09
Ethane 1.187 Methane 96.33
Ethylene 1.240 Propane 35.05
Methane 1.226
Propane 1.127
Chapter 5 Controlling the Dynamic Solution 91
Does the Vent Pipe Have an Umbrella Fitting (Y/N)
Enter a Y or a N. See the following figures to determine if the connection of the vent stack to the vent
piping is via an umbrella fitting.
92 CAESAR II Technical Reference Manual
Should CAESAR II Size the Vent Stack (Y/N)
Enter a Y if CAESAR II should size the vent stack. The sizing algorithm searches through a table of
available inside pipe diameters starting at the smallest diameter until a vent stack ID is found that satisfies
the thermodynamic criteria shown in the figure above. The computed ID is automatically inserted into the
input.
Example input and output from the relief load synthesizer is shown and discussed as follows:
Relief Load Synthesis Input (Gas)
Relief Load Synthesis Output (Gas)
Chapter 5 Controlling the Dynamic Solution 93
Computed Mass Flowrate (Vent Gas)
This is CAESAR IIs computed gas mass flow rate based on choked conditions at the relief orifice. If
greater mass flow rates are expected, then the error in either the approach used by CAESAR II or in the
expected mass flow rate should be investigated.
Thrust at Valve Pipe/Vent Pipe Interface
If there is an umbrella fitting between the vent stack and the relief valve piping then this is the thrust load
that acts back on the relief valve piping. (See the following figure.) If the vent stack is hard piped to the
relief valve piping then this intermediate thrust will be balanced by tensile loads in the pipe and can be
ignored.
Thrust load acts directly on valve opening. Only the valve pipe/vent stack interface thrust
acts in this configuration.
Thrust at the Vent Pipe Exit
When there is an elbow in the vent stack piping, this is the thrust load that acts on the elbow just before
the pipe opening into the atmosphere. (See the following figures for clarification.)
94 CAESAR II Technical Reference Manual
Transient Pressure Rise on Valve Opening
This is the estimated magnitude of the negative pressure wave that will be superimposed on the line
pressure when the relief valve fist opens. This negative pressure wave will move back through the relief
system piping similar to the pressure wave in the downstream piping of a water hammer type system. The
magnitude of this wave is estimated as (Po-Pa)*Ap, where Po is the stagnation pressure at the source, Pa is
atmospheric pressure, and Ap is the area of the header piping.
Transient Pressure Rise on Valve Closing
This is the estimated magnitude of the positive pressure wave that will be superimposed on the line
pressure when the relief valve slams shut. This positive pressure wave will move back through the relief
system piping similar to the pressure wave in the supply side piping of a water hammer type system. The
magnitude of this wave is estimated from: r*c*dv where r is the gas density, c is the speed of sound in the
gas and dv is the change in the velocity of the gas.
Thermodynamic Entropy Limit /Subsonic Vent Exit Limit
These values should always be greater than 1. If either of these computed limits fall below 1.0 then the
thermodynamic assumptions made regarding the gas properties are incorrect and the computed thrust
values should be disregarded.
Valve Orifice Gas Conditions /Vent Pipe Exit Gas Conditions/Subsonic Velocity Gas Conditions
These are the thermodynamic properties of the gas at three critical points in the relief system. These three
points are shown in the figure on the opposite page. The entire formulation for the thrust gas properties is
based on an ideal gas equation of state. If the pressures and temperatures displayed above for the gas being
vented are outside of the range where the ideal gas laws apply then some alternate source should be sought
for the computation of the systems thrust loads.
In addition, all three of these points should be sufficiently clear of the gas saturation line. When the exit
gas conditions become saturated, the magnitude of the thrust load can be reduced significantly. In this case
the manufacturer should be consulted. In several instances at COADE, saturated exhaust thrust loads were
50 to 75% less than the CAESAR II computed values.
Relief Load Synthesis for Liquids
CAESAR II assumes that the liquid vent system has one of the two following configurations:
Chapter 5 Controlling the Dynamic Solution 95
The input for the liquid relief load synthesis is shown as follows:
Relief Valve or Rupture Disk
Enter RV if the relieving device is a relief valve and RD if the relieving device is a rupture disk. If the
user has his own relief exit coefficient it can be entered here in place of the letters RV or RD. An entry of
zero represents No appreciable head loss due to the relief opening configuration. The exit coefficient for a
relief valve is 0.25 and for a rupture disk is 0.5.
Supply Overpressure
Enter the stagnation, or zero velocity pressure in the fluid upstream of the relief valve.
ID Relief Orifice or Rupture Disk Opening
Enter the manufacturers inside diameter of the contracted opening in the particular relieving device. (For
special purpose calculations this ID may be equal to the ID of the Relief exit piping.)
ID Relief Exit Piping
Enter the inside diameter of the piping connected to the downstream side of the relief valve.
ID Manifold Piping
If the relief exit piping runs into a manifold then enter the inside diameter of the manifold. Leave this field
blank or zero if there isn't a manifold.
ID Supply Header
Enter the inside diameter of the supply header.
96 CAESAR II Technical Reference Manual
Fluid Density
Enter the specific gravity of the fluid being relieved.
Length of Relief Exit Piping
Enter the equivalent length of the relief exit piping. Add twice the piping length for fittings and elbows, or
the calculated fitting equivalent length.
Length of Manifold Piping
Enter the equivalent length of the manifold piping, if any. If there isn't a manifold system then leave this
field blank or zero. Add twice the piping length for fitting and elbows. If the manifold is not filled by the
relieving fluid then leave the manifold length zero.
Fluid Bulk Modulus
Enter the bulk modulus of the fluid. If omitted a valve of 250,000 psi will be used as the default. Some
typical values for use are given as follows. These are the values for an isothermal compression as taken
from Marks Standard Handbook for Engineers, p. 3-35, 8th edition.
Supply Header Pipe Wall Thickness
Enter the wall thickness of the supply header.
Note: When running the relief load synthesis for liquids, the error message: NUMERICAL ERROR OR
NO FLOW CONDITION DETECTED, means a physically impossible configuration was described.
Flashing of volatile relief liquids is not considered. If the relieving liquid flashes in the exhaust piping as
its pressure drops to atmospheric then some other means should be used to compute the resulting gas
properties and thrust Loads.
Output From the Liquid Relief Load Synthesizer
Computed Mass Flow rate
The computed exhaust mass flow rate in U.S. Gallons per minute. CAESAR II makes the necessary pressure
drop calculations between the stagnation pressure upstream of the relief device and atmospheric
conditions at the exit of the manifold.
Thrust at the End of the Exit Piping
The computed thrust load at the last cross section in the exit piping. If there is no manifold then this is the
external thrust load that acts on the piping system. If there is a manifold then this thrust is opposed by
tension in the pipe wall at the junction of the exit piping and manifold. See the figures that follow for
clarification.
Thrust at the End of the Manifold Piping
The computed thrust load at the last cross section in the manifold piping. If there is no manifold system
then this thrust will be equal to the thrust at the end of the exit piping. See the figures that follow for
clarification.
Chapter 5 Controlling the Dynamic Solution 97
Transient Pressure Rise on Valve Opening
This is the estimated magnitude of the negative pressure wave that will be superimposed on the line
pressure when the relief valve fist opens. This negative pressure wave will move back through the relief
system piping similar to the pressure wave in the downstream piping of a water hammer type system. The
magnitude of this wave is estimated as (Po-Pa)*Ap, where Po is the stagnation pressure at the source, Pa is
atmospheric pressure, and Ap is the area of the header piping.
Transient Pressure Rise on Valve Closing
This is the estimated magnitude of the positive pressure wave that will be superimposed on the line
pressure when the relief valve slams shut. This positive pressure wave will move back through the relief
system piping similar to the pressure wave in the supply side piping of a water hammer type system. The
magnitude of this wave is estimated from: r*c*dv where r is the gas density, c is the speed of sound in the
gas and dv is the change in the velocity of the gas.
98 CAESAR II Technical Reference Manual
Orifice Flow Conditions/Exit Pipe End Flow Conditions/Manifold Pipe End Flow Conditions
These are the computed fluid properties at the three critical cross-sections in the relief piping. If pressures
or velocities here do not seem reasonable then some characteristic of the relief model is probably in error.
Note: If the L dimensions are significant in any of the previous figures (several feet) then unbalanced
thrust loads will act between the elbow-elbow pairs that is very similar to a water hammer load. Water
hammer pulses travel at the speed of sound in the fluid, while the fluid/atmosphere interface pulses
travel at the velocity of the flowing fluid. For this reason, these unbalanced loads can cause significant
piping displacements in much shorter pipe runs. The magnitude of these loads is equivalent to the
computed thrust and the duration may be found from the computed fluid velocity and distance between
each elbow-elbow pair.
1
In This Chapter
Rigid Element Application..........................................................2
Cold Spring .................................................................................4
Expansion Joints..........................................................................7
Hanger Sizing Algorithm.............................................................10
Class 1 Branch Flexibilities.........................................................14
Modeling Friction Effects ...........................................................17
Nonlinear Code Compliance .......................................................19
Sustained Stresses and Nonlinear Restraints ...............................20
Static Seismic Loads ...................................................................24
Wind Loads .................................................................................27
Hydrodynamic (Wave and Current) Loading ..............................30
Evaluating Vessel Stresses ..........................................................44
Inclusion of Missing Mass Correction.........................................49
Fatigue Analysis Using CAESAR II............................................54
Pipe Stress Analysis of FRP Piping.............................................70
Code Compliance Considerations ...............................................93
Local Coordinates .......................................................................127
C H A P T E R 6
Technical Discussions
2 CAESAR II Technical Reference Manual
Rigid Element Application
CAESAR II forms rigid elements by multiplying the wall thickness of the element by 10. The inside
diameter, and the weight of the element, remain unchanged.
The rigid element in CAESAR II is rigid relative to the pipe around it. If a 6-in. line ties into a 72-in. heat
exchanger, then the rigid elements modeling the heat exchanger should have a diameter closer to 72 than
6.
The user that is sensitive to the rigidness of the rigid element can increase or decrease the diameter or
wall thickness of the rigid to simulate any order of magnitude stiffness.
Rigid Material Weight
The weight of the rigid element is entered by the user. If no value is input then the weight of the rigid is
taken to be zero. The entered weight is the weight of the rigid excluding insulation or fluid. If the weight
of the rigid element is entered as zero or blank, then no additional weight due either to insulation or fluid
will be added.
Rigid Fluid Weight
CAESAR II automatically adds fluid loads for rigid elements if a non-zero fluid density is entered on the
pipe spreadsheet. The fluid weight in a rigid element is assumed to be equal to the fluid weight in an
equivalent straight pipe of similar length and inside diameter.
Rigid Insulation Weight
CAESAR II also automatically adds insulation loads if the line containing the rigid element is insulated. The
insulation weight for the rigid is assumed to be equal to 1.75 times the insulation for an equivalent length
of straight pipe of equal outside diameter.
The cumulative rigid element weight calculation is as follows:
Weight = 0.0 Wu = 0.0
Weight = Wu + Wf + 1.75Wi Wu > 0.0
Where:
Wu = User entered rigid weight
Wf = Calculated fluid weight for equivalent straight pipe
Wi = Calculated insulation weight for equivalent straight pipe
Chapter 6 Technical Discussions 3
The users entered weight for the rigid element is not reflected in the Thermal Expansion/Pipe Weight
Report optionally printed during error checking.
Stresses are not calculated on Rigid elements since they are often used to simulate components that have
variable cross-sections along the length of the element, i.e. a valve, and is normally not of concern for this
type of analysis anyway. Forces and Moments are not normally printed on nodes between two rigid
elements, but can be by selecting the appropriate check box found in Kaux-Special Execution Parameters
from the Piping Input Spreadsheet.
Zero-weight rigids ("dummy" rigids) are often used to model components whose weight is not important
to the analysis, but where thermal growth may be a consideration. Dummy rigids are often used to model
restraints. Tie rods in an expansion joint, rod hangers, and trunnions are examples of restraints modeled as
dummy rigids. Dummy rigids may also be used to provide connectivity between the center line of an
element and its outside edge. The most common example of this is the addition of a dummy rigid that runs
from the node at the center line of the vessel to the edge where a nozzle is to be connected. Sometimes
equipment is modeled through a series of rigid elements. This is particularly true when multiple nozzles
are attached and the equipment is restrained such that the interactions between the various nozzles must
be taken into account due to the thermal growth of the attached piping system. The use of dummy rigids is
explained in the CAESAR II Applications Guide in various sections as appropriate to a particular modeling
technique.
4 CAESAR II Technical Reference Manual
Cold Spring
Cold spring is the process of offsetting (or pre-loading) the piping system with displacement loads
(usually accomplished by cutting short or long the pipe runs between two anchors) for the purpose of
reducing the absolute expansion load on the system. Cold spring is used to do the following:
hasten the thermal shakedown of the system in fewer operating cycles
reduce the magnitude of loads on equipment and restraints, since often, only a single application of a
large load is sufficient to damage these elements
Several things should be considered when using cold spring:
Cold reactions on equipment nozzles due to cold spring should not exceed nozzle allowables.
The expansion stress range should not include the effect of the cold spring.
The cold spring should be much greater than fabrication tolerances.
Note: No credit can be taken for cold spring in the stress calculations, since the expansion stress
provisions of the piping codes require the evaluation of the stress range, which is unaffected by cold
spring (except perhaps in the presence of non-linear boundary conditions, as discussed below). The cold
spring merely adjusts the stress mean, but not the range.
Many engineers avoid cold spring due to the difficulty of maintaining accurate records throughout the
operating life of the unit. Future analysts attempting to make field repairs or modifications may not
necessarily know about (and therefore include in the analysis) the cold spring specification.
Due to the difficulty of properly installing a cold sprung system, most piping codes recommend that only
2/3 of the specified cold spring be used for the equipment load calculations.
The cold spring amount is calculated as:
Ci = 1/2Li o dT
Chapter 6 Technical Discussions 5
Where:
Ci = length of cold spring in direction i (where i is X, Y, or Z), (inches)
Li = total length of pipe subject to expansion in direction i, (inches)
o = mean thermal expansion coefficient of material between ambient and operating temperature,
(in/in/F)
dT = change in temperature, (F)
6 CAESAR II Technical Reference Manual
Note that the 1/2 in the equation for the cold spring amount is used such that the mean stress is zero. In
some cases it is desirable to have the operating load on the equipment as close to zero as possible. In this
latter case the 1/2 should be omitted. The maximum stress magnitude will not change from a system
without cold spring, but will now exist in the cold case rather than the hot.
To model a cold spring in CAESAR II specify the elements as being made of cut short or cut long
materials. Cut short describes a cold sprung section of pipe fabricated short by the amount of the cold
spring, requiring an initial tensile load to close the final joint. Cut long describes a cold sprung section of
pipe fabricated long by the amount of cold spring, requiring an initial compressive load to close the final
joint. The software models cut shorts and cut longs by applying end forces to the elements sufficient to
reduce their length to zero (from the defined length) or increase their length to the defined length (from
zero) respectively. (It should be remembered to make the lengths of these cold spring elements only 2/3 of
their actual lengths to implement the code recommendations.) This is effectively what occurs during
application of cold spring. The end forces applied to the elements are then included in the basic loading
case F (for force), whereby they can be included in various load combinations.
Special material numbers 18 and 19 are used to signal CAESAR II that the element currently in the
spreadsheet actually represents a length of pipe that is to be cut short or long during fabrication.
Material # 18 - Cut Short
Material # 19 - Cut Long
The user should be sure to reset the material property on the element following the cold spring element.
The following load cases are recommended when analyzing a cold spring system:
Load Case 1 (OPE) W+T1+P1+CS includes all of the design cold spring
Load Case 2 (OPE) W+P1+CS includes all of the design cold spring but not the temperature.
Load Case 3 (SUS) W+P1 standard sustained case for Code Stress check
RUN # 1
Load Case4 (EXP) L1-L2 expansion case for code stress check.
Cold spring is allowed to reduce the magnitude of equipment loads because, often, only a single
application of a large load is sufficient to cause damage to rotating machinery.
Cold spring does not change the range of stresses that the piping system is subject to, and so, no
allowance is given for stress reduction. (The maximum value of the stress is lowered, but the range is
unchanged.)
Both the sustained loads and the operating loads should be within the manufacturers allowables for the
particular piece of equipment. If the designer isn't careful, the installation of the cold spring in the ambient
state can overload a piece of rotating equipment as the unit starts up.
Chapter 6 Technical Discussions 7
Expansion Joints
To define an expansion joint, activate the Expansion Joint check box (see "Expansion Joints" on page
19) on the pipe element spreadsheet.
Expansion joint elements may have a zero or nonzero length. The expansion joint will have a zero length
if the Delta fields in the spreadsheet are left blank or zero. The expansion joint will have a nonzero length
if at least one of the elements spreadsheet Delta fields is non-blank and non-zero. When an expansion
joint has a finite length CAESAR II evenly distributes the expansion joint stiffnesses over the entire length
of the element. This will usually result in a more accurate stiffness model in what is typically a very
sensitive area of the piping system.
Four stiffnesses define the expansion joint
Axial Stiffness
Transverse Stiffness
Bending Stiffness
Torsional Stiffness
These stiffnesses are defined as shown in the following figure:
8 CAESAR II Technical Reference Manual
The transverse and the bending stiffnesses are directly related when a finite length joint is defined. In this
case the bending stiffness should be left blank and the transverse stiffness entered. CAESAR II will
compute the proper bending stiffness from the relationship between the bending and transverse stiffnesses.
Bending stiffnesses from manufacturers catalogs should generally only be entered for zero length
expansion joints modeling hinges or gimbals. Before a manufacturers bending stiffness is used for a finite
length bellows it should be multiplied by 4.0 (note that in this case the transverse stiffness would be left
blank).
Torsional stiffnesses are often not given by expansion joint manufacturers. In this case the user is
recommended to insert a large torsional stiffness value and ensure that the resulting load on the bellows is
not excessive. When the piping system is tight, and the diameter large, the magnitude of this large
torsional stiffness can significantly effect the magnitude of the torsion carried by the bellows, i.e.
stiffnesses of 100,000 in.lb./deg. and 1E12 in.lb./deg. can produce considerably different torsional load
results. The tendency would be to go with the larger stiffness, i.e. being conservative, except that the
torsional stiffness value is probably closer to the 100,000 in.lb./deg. In the instance where the largeness
of the torsional stiffness value is important, the manufacturer should be pressed for his best-guess at the
stiffness, or the following equation should be used to get an estimate, which the user can then
conservatively increase to get reasonable torsional loads on the bellows and surrounding equipment.
The equation for estimating bellows torsional stiffness is
3
( ) ( )( )
(1 )
Re t E
L
t
+ v
Where
t = 3.14159
Re = Expansion joint effective radius
t = Bellows thickness
E = Elastic Modulus
v = Poissons Ratio
L = Flexible bellows length
Chapter 6 Technical Discussions 9
When the expansion joint has a zero length, none of the expansion joint stiffnesses are related. The user
must be sure that a value is entered into all four of the Stiffness fields.
CAESAR II will calculate pressure thrust on the expansion joint if the bellows effective id is given in the
expansion joint auxiliary screen. The mathematical model for pressure thrust applies a force equal to the
pressure times the effective area of the bellows at either end of the expansion joint. The force will tend to
open the bellows if the pressure is positive, and close the bellows if the pressure is negative. Users should
note that this model does not exactly distribute the pressure loads correctly in the vicinity of the expansion
joint. In most cases the misapplied load does not effect the solution. There are two components of the
pressure thrust to be applied in practice, rather than the one component applied in the model. The first
component is equal to the pressure times the inside area of the pipe and acts at the first change in direction
of the pipe on either side of the expansion joint. This load will tend to put the pipe wall between the
change in direction and the expansion joint in tension. The second component is equal to the pressure
times the difference between the bellows effective area and inside pipe area. This load acts at the end of
the expansion joint and tends to open the bellows up, putting the pipe between the expansion joint and the
change in direction in compression. In the mathematical model the full component of the pressure thrust
force is placed on the ends of the bellows instead of having a portion shifted out on either side of the
expansion joint. A large number of expansion joint examples can be found in Chapter 5 of the Applications
Guide.
10 CAESAR II Technical Reference Manual
Hanger Sizing Algorithm
The basic function of the hanger design algorithm is to calculate the hot load and travel for user-specified
hanger locations.
Once the hot load and travel are known, spring tables are entered and the theoretical cold load is
calculated for each spring in the table.
Spring Design Requirements
The smallest single spring that satisfies all design requirements is selected as the designed spring.
The spring design requirements are
1 Both hot and the cold loads must be within the spring allowed working range.
2 If the user specified an allowed load variation then the absolute value of the product of the travel and
the spring rate divided by the hot load must be less than the specified variation.
3 If the user specified some minimum available clearance then the spring selected must fit in this space.
If a single spring cannot be found that satisfies the design requirements, CAESAR II will try to find two
identical springs that do satisfy the requirements.
If satisfactory springs cannot be found, CAESAR II recommends a constant effort support for the location.
There are several variations of this approach that arise due to the different design options available in
CAESAR II, but for the most part the general algorithm remains unchanged.
Restrained Weight Case
Hanger hot loads are calculated in the restrained weight case.
In any job, if a hanger is to be designed, the first analysis case that must be run is the restrained weight
case. This case usually includes weight, pressure and concentrated loads.
For the restrained weight run, rigid Y restraints are placed at each hanger location, and any anchors to
be freed are properly released.
Loads on the Y restraints at hangers, calculated from the restrained weight case, are the hanger hot
design loads.
Chapter 6 Technical Discussions 11
Operating Case
Immediately after the restrained weight case, an operating analysis is performed. The Y restraints are
removed from the hanger locations and the hot loads just calculated are inserted. Any anchors that were
freed for the restrained weight analysis are fixed.
The operating case vertical displacement at each hanger location defines that hangers travel. If there
were single directional restraints or gaps in the system that changed status in the operating case then the
possibility exists that loads on hangers will be redistributed. When a nonlinear status change is detected
CAESAR II reruns the restrained weight case with the restraints left as they were at the end of the
operating case. New restraint loads are calculated and another operating case is run to get the updated
travel.
The operating case must always be the second load case in the set of defined analysis cases.
The user has the ability to define the restrained weight or operating load cases for hanger design any
way he sees fit. For simplicity, CAESAR II recommends the load cases it thinks should be run whenever it
detects the first attempt to analyze a particular system. The user can accept or reject CAESAR IIs
recommendations. The user that sets up his own hanger design load cases should be sure he understands
exactly what is done in the restrained weight and operating passes of the hanger design algorithm.
Installed Load Case
If the user requested the calculation of the actual hanger installed loads, the third analysis level
combination case must define the weight configuration that will exist in the field when the spring is
installed. Typically this case includes weight without fluid contents and concentrated loads.
The theoretical cold, or installed, load is the load on the spring when the pipe has exactly zero
displacement. The actual installed load may differ from the theoretical installed load by (K)(d), where (K)
is the spring stiffness and (d) is the displacement of the pipe in the installed condition. In essence, the
actual installed load is calculated by taking the piping system and freezing all displacements at zero.
With the pipe in this condition, the hangers are installed and the theoretical cold load is applied. The pipe
is then defrosted and allowed to adjust its weight position due to the hanger, restraint, and anchor
stiffnesses and the installed hanger loads. Once the system settles out, the total load on each of the hangers
is read and recorded as the actual hanger installed load.
12 CAESAR II Technical Reference Manual
Setting Up the Spring Load Cases
The load cases that must exist for hanger design, as described above, are
Restrained Weight
Operating
Installed Weight ...if the user requested actual hanger installed loads.
After the hanger algorithm has run the load cases it needs to size the hangers. The newly selected springs
are inserted into the piping system and included in the analysis of all remaining load cases.
The spring rate becomes part of the global stiffness matrix, and is therefore added into all subsequent load
cases. Hanger installed loads are concentrated forces and are only included in subsequent load cases that
contain the first concentrated force set, (i.e., +H).
The user may specify any number of his own load cases after the required spring load cases are set up.
Spring hanger design does not affect CAESAR IIs ability to check code compliance. In fact, in CAESAR IIs
recommended load cases, the normal code compliance cases always follow the set of load cases required
for hanger design.
Multiple operating case spring hanger design implies that hanger loads and travels from more than one
operating case are included in the spring hanger selection algorithm.
Each spring in a multiple operating case hanger design has a multiple load case design option. This design
option tells CAESAR II how the multiple loads and travels for a single hanger are to be combined to get a
single design load and travel. The set-up of the analysis cases is slightly different for multiple operating
case hanger design, and as might be expected, the difference is that now there is more than one operating
case. The actual number of operating cases is specified by the user on the Hanger Design Control dialog
and can be up to 9.
Load cases that must be set up for a multiple load case hanger design that considers two hanger design
operating cases are:
Restrained Weight (this doesn't change)
Operating case #1
Operating case #9
Installed Weight ...if the user requested that actual installed loads are to be calculated.
Constant Effort Support
The specification of the support load for a constant effort hanger completely defines the hanger location.
If the user enters this value it will be included in all hanger design runs and all analysis cases following the
hanger cases that include concentrated loads in their formulation. This value is the load on each support at
this location.
Chapter 6 Technical Discussions 13
Including the Spring Hanger Stiffness in the Design Algorithm
The operating cases for hanger travel are normally analyzed with no stiffness included at the hanger
locations (hence these cases are traditionally referred to as "free thermal" cases). However, when the
piping system is very flexible, or the selected springs are very stiff, the actual resulting spring loads in the
hot condition can vary significantly from the theoretically calculated results. In that case, CAESAR II offers
the option to include (via an iterative process) the stiffness of the selected springs in the operating cases
for hanger travel. This can be activated by setting the Hanger Stiffness Load Case option to "As
Designed" for that operating case. (Activating the Configure/Setup option "Include Travel cases to
default to "As Designed".) The user is warned that selection of this option may lead to convergence
problems. If this option is used, it is very important that the hanger load in the cold case (In the physical
system) be adjusted to match the reported hanger Cold Load.
Spring Hanger Hot Loads for as designed springs are always included in all Operating Hanger Travel
cases. Cold loads can be included in subsequent load cases through the use of the H load component. (Note
that applying thermal and displacement effects to the system should make the Cold Load move to the Hot
Load in the operating case.)
Other Notes on Hanger Sizing
Users should note that whenever a hanger location is found to hold the pipe down, a beep and a warning
message is flashed to the user. These locations in output are flagged as zero load constant effort supports.
These supports are usually found to be at poor hanger design locations.
Hanger design load cases, unless specifically designed with a "KEEP" status by the user, show up in the
output report as being NOT ACTIVE. Results from these analysis are reflected in the spring hanger
table only.
14 CAESAR II Technical Reference Manual
Class 1 Branch Flexibilities
This analytical option was added to CAESAR II for the following reasons:
Automatic local flexibilities at intersections help the user bound the true solution. Because the
computer time to do an analysis is getting less expensive, more frequently an analyst is running
several solutions of the same model using slightly different input techniques to determine the effect of
the modeling difference on the results. (This gives the analyst a degree of confidence in the numbers
he is getting.) For example, structural steel supporting structures may be modeled to see the effect of
their stiffnesses, nozzle flexibilities may be added at vessel connections to see how these features
redistribute load throughout the model, friction is added to watch its effect on displacements and
equipment loads, and with CAESAR II users may include Class 1 intersection flexibilities. The
characteristic that makes this option convenient to use is that the use can turn the Class 1 flexibilities
on and off via a single parameter in the setup file. There is no other modification to the input
required.
In WRC 329, there are a number of suggestions made to improve the stress calculations at
intersections. These suggestions are fairly substantial, and are given in order of importance. The most
important item, as felt by Rodabaugh in improving the stress calculations at intersections is given, in
part, as follows:
In piping system analyses, it may be assumed that the flexibility is represented by a rigid joint at the
branch-to-run centerlines juncture. However, the Code user should be aware that this assumption can be
inaccurate and should consider the use of a more appropriate flexibility representation.
User of the Class 1 branch flexibility feature may be summarized as follows:The user adds the option:
CLASS_1_BRANCH_FLEX to the setup file. This option is a flag, and merely has to appear in the
setup file to activate the option.
Where reduced branch geometry requirements are satisfied, CAESAR II constructs a rigid offset from
the centerline of the header pipe to its surface, and then adds the local flexibility of the header pipe,
between the end of the offset, at the header, and the start of the branch. Stresses computed for the
branch, are for the point at its connection with the header.
Where reduced branch geometry requirements are not satisfied, CAESAR II constructs a rigid offset
from the centerline of the header pipe to its surface. The branch piping starts at the end of this rigid
offset. There is NO local flexibility due to the header added. (It is deemed to be insignificant.)
Stresses computed for the branch, are for the point at its connection with the header.
The reduced branch geometry requirements checked by CAESAR II are
d/D s 0.5 and D/T s 100.0
Where:
d = Diameter of branch
D = Diameter of header
T = Wall thickness of header
Chapter 6 Technical Discussions 15
When the Class 1 branch flexibilities are used, intersection models in the analysis will become stiffer
when the reduced geometry requirements do not apply, and will become more flexible when the reduced
geometry requirements do apply. Stiffer intersections typically carry more load, and thus have higher
stresses (lowering the stress in other parts of the system that have been unloaded). More flexible
intersections typically carry less load, and thus have lower stresses, (causing higher stresses in other parts
of the system that have picked up the extra load).
The branch flexibility rules used in CAESAR II are taken from ASME III, Subsection NB, (Class 1), 1992
Edition, Issued December 31, 1992, from Code Sections NB-3686.4 and NB-3686.5.
When the reduced branch rules apply, the following equations are used for the local stiffnesses:
TRANSLATIONAL:
AXIAL = RIGID
CIRCUMFERENTIAL = RIGID
LONGITUDINAL = RIGID
ROTATIONAL:
AXIAL = RIGID
CIRCUMFERENTIAL = (kx)d/EI
LONGITUDINAL = (kz)d/EI
Where:
RIGID = 1.0E12 lb./in. or 1.0E12 in.lb./deg.
d = Branch diameter
E = Youngs Modulus
I = Cross Section Moment of Inertia
D = Header diameter
T = Header thickness
Tb = Branch fitting thickness
kx = 0.1(D/T)
1.5
[(T/t)(d/D)]
0.5
(Tb/T)
kz = 0.2(D/T)[(T/t)(d/D)]
0.5
(Tb/T)
16 CAESAR II Technical Reference Manual
Users are referred to WRC 329 Section 4.9 Flexibility Factors. A brief quote from this section follows:
The significance of k depends upon the specifics of the piping system. Qualitatively, if k is small
compared to the length of the piping system, including the effect of elbows and their k-factors, then the
inclusion of k for branch connections will have only minor effects on the calculated moments.
Conversely, if k is large compared to the piping system length, then the inclusion of k for branch
connections will have major effects. The largest effect will be to greatly reduce the magnitude of the
calculated moments acting on the branch connection. To illustrate the potential significance of ks for
branch connections, we use the equation [above] to calculate k for a branch connection with D=30 in.,
d=12.75 in. T=t=0.375 in.:
k = 0.1(80)
1.5
(0.425)
0.5
* (1.0) = 46.6
This compares to the more typical rigid-joint interpretation that k=1, rather than k=46.6 !
Further discussion in section 4.9 illustrates additional problems that can arise by overestimating the
stiffness at branch connections. Problems arise by believing mistakenly that the stress at the intersection
is too high. Further reference should be made to this section in WRC 329.
The branch automatic flexibility generation can be used where the user has only defined the branch
element in the model, i.e. has left the header piping out of the analysis. In this case there will be no
offset equal to one-half of the header diameter applied to the branch end. A partial intersection is one
where either the header pipe is not modelled, is modelled with a single element, or is part of a geometric
intersection where the header pipes are not colinear. In the case where there is no header pipe going to the
intersection there will be no modification to the model for the class 1 branch flexibilities. When at least a
single header pipe is recognized, the local flexibility directions are defined by the branch alone and in
accordance with the CAESAR II defaults for circumferential and longitudinal directions for the branch and
header. Users are recommended to build full intersection models at all times (not only when employing
the class 1 branch flexibility.) In most cases building full intersection models will eliminate problems
caused by the assumptions necessary when a partial intersection is described.
In the equations in NB-3686.5 for tn, the thickness of the branch pipe is used in all cases.
When branches are skewed with respect to the header pipe, and where the two header pipes are colinear,
the local Class 1 flexibilities are still taken to be the longitudinal and circumferential directions that are
tangent to the header surface at its intersection with the branch.
Class 1 branch flexibilities can be formed at both ends of a single pipe element.
Note: The offsets necessary to form the class 1 intersections are automatically generated by CAESAR II.
There is no extra input required by the user to have CAESAR II build these intersections.
(If there are already user-defined offsets at an intersection end, the computed offset to get from the header
centerline to its surface along the centerline of the branch will be added to the already entered user offset.)
Automatic offsets will be generated providing that the distance from the header centerline to the header
surface along the branch centerline is less than or equal to 98% of the total pipe straight length.
When a bend curved element is part of an intersection model, the offset and flexibility calculations will
not be performed.
Chapter 6 Technical Discussions 17
Modeling Friction Effects
There are two approaches to solving the friction problem; insert a force at the node which must be over
come for motion to occur, or insert a stiffness which applies an increasing force up to the value of Mu *
Normal force. CAESAR II uses the restraint stiffness method. (An excellent paper on this subject is
Inclusion of a Support Friction Into a Computerized Solution of a Self-Compensating Pipeline by J.
Sobieszczanski, published in the Transactions of the ASME, Journal of Engineering for Industry, August
1972. A summary of the major points of this paper can be found below.)
Ideally, if there is motion at the node in question, the friction force is equal to (Mu * Normal force).
However, since we have a non-rigid stiffness at that location to resist the initial motion, the node can
experience displacements. The force at the node will be the product of the displacement and the stiffness.
If this resultant force is less than the maximum friction force (Mu * Normal force), the node is assumed to be
not sliding, even though we see displacements in the output report.
The maximum value of the force at the node is the friction force, Mu * Normal force. Once this value is
reached, the reaction at the node stops increasing. This constant force value is then applied to the global
load vector during the next iteration to determine the nodal displacements.
Basically here is what happens in a friction problem.
1 The default friction stiffness is 1,000,000 lb./in. This value should be decreased to improve
convergence.
2 Until the horizontal force at the node equals Mu * Normal force, the restraint load is the displacement
times the friction stiffness.
3 Once the maximum value of the friction force is reached, the friction force will stop increasing, since
a constant effort force is inserted.
18 CAESAR II Technical Reference Manual
By increasing the friction stiffness in the setup file, the displacements at the node will decrease to some
degree. This may cause a re-distribution of the loads throughout the system. However, this could have
adverse affects on the solution convergence.
If problems arise during the solution of a job with friction at supports, reducing the friction stiffness will
usually improve convergence. Several runs should be made with varying values of the friction stiffness to
insure the system behavior is consistent.
Summary of J. Sobieszczanskis ASME Paper
For dry friction, the friction force magnitude is a step function of displacement. This discontinuity
determines the problem as intrinsically nonlinear and eliminates the possibility of using the
superposition principle.
The friction loading on the pipe can be represented by an ordinary differential equation of the fourth
order with a variable coefficient that is a nonlinear function of both dependent and independent
variables. No solution in closed form is known for an equation of this type. Solution has to be sought
by means of numerical integration to be carried out specifically for a particular pipeline configuration.
Dry friction can be idealized by a fictitious elastic foundation, discretized to a set of elastic (spring)
supports.
A well-known property of an elastic system with dry friction constraints is that it may attain several
static equilibrium positions within limits determined by the friction forces.
THE WHOLE PROBLEM THEN HAS CLEARLY NOT A DETERMINISTIC, BUT A
STOCHASTIC CHARACTER.
Chapter 6 Technical Discussions 19
Nonlinear Code Compliance
Nonlinear piping code compliance can be directly satisfied by
1 Performing an operating and sustained analysis of the system including in each case the effect of
nonlinear restraints.
2 Subtracting the sustained case displacements from the operating case displacements to find the
displacement range.
3 Calculating the expansion stresses from the displacement range solved for in #2 above.
Approximate approaches usually involve some combination of the above. The approximate combination
used depends typically on the inherent limitations of the base program. In several commonly used
programs, the approach taken is
1 Formulate and solve for operating case displacements including an iteration to deal with the effect of
nonlinear restraints in the system.
2 Run the thermal-only analysis of the system to calculate expansion stresses with restraints in the same
condition as they were at the end of #1.
3 Run the weight+pressure only analysis of the system to calculate sustained stresses, again with
restraints in the same condition as they were at the end of #1.
This alternate approach is identical to the first method only when the sustained analysis final stiffness
matrix is the same as the operating analysis final stiffness matrix. The resulting error in the displacement
range can be found from
{[Fo] - [Fs]}fs.
Where:
[Fo] is the operating analysis final flexibility matrix. (i.e. the inverse of the stiffness matrix.)
[Fs] is the sustained analysis final flexibility matrix.
fs is the sustained analysis load vector.
CAESAR II uses the exact method described above for calculating the expansion stress range. In addition
CAESAR II scans the users input and recommends loading cases and combinations for performing the
operating, sustained and expansion stress calculations. This recommendation can prove very useful when
performing spring hanger analysis of a multiple operating case system.
20 CAESAR II Technical Reference Manual
Sustained Stresses and Nonlinear Restraints
The proper computation of sustained stresses has been an issue since the late 1970s, when computerized
pipe stress analysis programs first attempted to address the problem of non-linear restraints. The existing
piping codes offered little guidance on the subject, since their criteria were developed during the era when
all analyses were considered to behave in a strictly linear fashion. The problem arises because the codes
require that a piping system be analyzed separately for sustained loadings the engineer must determine
which stresses are caused by which loadings. Sustained loads are force loadings which are assumed not to
change, while expansion loadings are displacement loadings which vary with the system operating
conditions. Determination of the sustained loads is the simple part most everybody agrees that those
forces consist of weight, pressure, and spring preloads these forces remain relatively constant as the
piping system goes through its thermal growth. However, confusion occurs when the status of nonlinear
restraints change (pipes lift off of supports, gaps close, etc.) as the pipe goes from its hot to cold state in
this case, which boundary conditions should be used when evaluating the applied forces? Or in other
words, what portion of the stress in the operating case is caused by weight loads, and what portion is
caused by expansion effects? (Note that there is no corresponding confusion on the question of calculating
expansion stresses, since the codes are explicit in their instructions that the expansion stress range is the
difference between the operating and cold stress distributions, both of which are known.)
The obvious answer to this question, to the developers of some pipe stress programs, was that the
sustained stress calculation should be done using the operating, or hot boundary condition. This
compounded the problem, in that the laws of superposition no longer held in other words, the results of
sustained (W+P) and thermal (T) cases, when added together, did not equal the results of the operating
(W+P+T) case! One pioneering program, DYNAFLEX, attempted to resolve this by introducing the
concept of the thermal component of weight an oxymoron, in our opinion. Other programs, notably
those which came from the mainframe/linear analysis world, had to approximate the behavior of these
non-linear restraints. Their approach to the problem is to run an operating case, obtain the restraint status,
and modify the model according to these results. All subsequent load cases analyzed use this restraint
configuration. The fact that the laws of static superposition didn't hold was hopefully not noticed by the
user. CAESAR II, on the other hand, represents new technology, developed expressly for operation on the
PC, and therefore incorporates directly the effects of non-linear restraints. This is done by considering
each load case independently the restraint configuration is determined for each load case by the
program as it runs, based upon the actual loads which are considered to be present.
Some users have asserted that there are actually two sustained load cases. In fact, there has been a B31.3
code interpretation that indicates that the sustained stress may also be checked with the operating restraint
configuration. Calculating the sustained stresses using the operating restraint status raises several other
issues; what modulus of elasticity should be used, and which sustained stresses should be used for
occasional cases.
Chapter 6 Technical Discussions 21
It is COADEs assertion that there is only one sustained case (otherwise it is not sustained) there can
be, however, multiple sustained stress distributions. The two most apparent are those associated with the
cold (installed) and hot (operating) configurations, however, there are also numerous in-between, as the
piping system load steps from cold to hot. Whether the true sustained load case occurs during the
installed or operating case is a matter of the frame of reference. If an engineer first sees a system in its
cold condition, and watches it expand to its operating condition, it appears that the first case (since weight
and pressure primary loads are present) is the sustained case, and the changes he viewed are thermal
effects (due to heat up) secondary loads due to displacements. If a second engineer first sees the same
system in the operating case and watches it cool down to the cold case, he may believe that the first case
he saw (the operating case) is the sustained case, and changes experienced from hot to cold are the thermal
expansion effects (the thermal stress ranges are the same in both cases). Consider the further implications
of cryogenic systems where changes from installed to operating are the same as those experienced by
hot systems when going from operating to installed. Once elastic shakedown has occurred, the question
becomes clouded even further, due to the presence of thermally induced pre-stresses in the pipe during
both the cold and hot conditions. We feel either the operating or installed case (or some other one in-
between) could justifiably be selected for analysis as the sustained case, as long as the program is
consistent.
We have selected the installed case (less the effect of cold spring) as our reference sustained case, since
thermal effects can be completely omitted from the solution (as intended by the code), and this best
represents the support configuration when the sustained loads are initially applied. If the pipe lifts off of a
support when going from installed to operating, we view this as a thermal effect consistent with the
piping codes view of thermal effects as the variation of stress distribution as the piping system goes from
cold to hot (this view is explicitly corroborated by one code the French petrochemical code, which
states that weight stress distributions due to thermal growth of the pipe should be considered as expansion
stresses). For example, we feel that a change in a rigid support load from 2,000 lbs to zero should be
treated no differently than would be a variable spring load changing from 6,000 lbs to 4,000 lbs (or
another rigid support load going 2,000 lbs to 1 lb). In the former case, if the pipe became overstressed, it
would yield, and sag back to the support, relieving the stress. This process is identical to the way that all
other expansion stresses are relieved in a piping system.
We are confident that our interpretation is correct. However, we understand that our users may not always
agree with us that is why CAESAR II provides the greatest ability to custom tailor the analysis to ones
individual specifications. If desired, a hot sustained case can be analyzed by adding two load cases to
those normally recommended by CAESAR II. This would be done by assuming that the pipe expands first,
and then the sustained loads are applied (this is of course an idealized concept, but the stresses can only be
segregated by segregating the applied loads, so the sustained loads can only be applied either before, or
after, the expansion loads). Following are the default load cases, as well as those required for a hot
sustained.
Default New
W+P1+T1(OPE) W+P1+T1(OPE)
W+P1(SUS) W+P1(SUS)
L1-L2(EXP) T1 (EXP)
L1-L2(EXP)
L1-L3(SUS)
22 CAESAR II Technical Reference Manual
In the new load case list, the second case still represents the cold sustained, while the fourth case
represents the expansion case (note that L1-L2, or W+P1+T1-W-P1, equals T1, with non-linear effects
taken into account). The third case represents the thermal growth of the weightless, non-pressurized
pipe, against the non-linear restraints.
The fifth case (L1-L3, or W+P1+T1-T1, equals W+P1) represents the application of weight and pressure
to that expanded case, or the hot sustained case. Note that when the piping system is analyzed as above,
the actual effects of the non-linear restraints are considered (they are not arbitrarily removed from the
model), and the laws of superposition still hold.
An alternative school of thought believes that a "hot sustained" is only valid if (1) the sustained, primary
loads are applied, (2) all springs are showing their Hot Load settings, and (3) any supports that lift off (or
otherwise become non-active) have been removed from the model. An analysis such as this is achievable
by setting the "Keep/Discard" status of the Restrained Weight case (the first hanger design load case) to
"Keep", thus permitting the results of that case to be viewable as for any other load case. The Restrained
Weight case automatically removes restraints that become non-active during the designated operating
case, and apply the Hot Load at each of the hanger locations.
Chapter 6 Technical Discussions 23
Notes on Occasional Load Cases
Several piping codes require that the stresses from occasional loads (such as wind or earthquake) be added
to the sustained stresses (due to weight, pressure, and other constant loads) before comparing them to their
allowables. This combination is easily created in CAESAR II:
CASE # 1 W+P+F1 (SUS) :Sustained stresses
2 WIND (OCC) :Wind load set
3 U1 (OCC) :Uniform (g) load set for earthquake
4 L1+L2 (OCC) :Code stresses for wind *
5 L1+L3 (OCC) :Code stresses for earthquake*
* Scalar Summation Method required
If nonlinear effects are modeled in the system these combinations may not be so straight forward. Friction,
one-direction restraints and double-acting restraints with gaps are the nonlinear items which present this
complication. Wind loading on a long vertical run of pipe with a guide will serve as an example. Assume
there is a one inch gap between the pipe and guide. Under normal operation, the pipe moves 3/4 inch
towards the stop leaving a gap of 1-3/4 inch on either side of the pipe and a 1/4 inch gap on the other side.
If wind loads are analyzed alone, the pipe is allowed to move 1 inch from its center point in the guide to
the guide stop. Since occasional loads are usually analyzed with the system in operation, the pipe may be
limited to a 1/4 inch motion as the gap is closed in one direction, and 1-3/4 inch if the gap is closed in the
opposite direction. With nonlinear effects modeled in the system, the occasional deflections (and stresses)
are influenced by the operating position of the piping.
The following list of CAESAR II load cases take this point into consideration. Note that the load cases
shown below are only for wind acting in one direction, i.e., +X. Depending on the system, the most critical
loads could occur in any direction, i.e., +/-X, +/-Z or skewed in an XZ direction.
The intention of the following load case construction is to find the occasional loads effect on the piping
system in the operating condition. The stress due to the moment change from the operating to the
operating plus wind case is added to the stress from the sustained case.
The isolated wind effect on the piping system in the operating condition in is computed in Case 5. Case 6
adds the stresses from Case 5 to the sustained stresses from Case 2.
CASE # 1 W+T+P (OPE) :Operation analysis
2 W+P (SUS) :Sustained stresses
3 W+T+P+WIND (OPE) :Operating analysis with wind
4 L1-L2 (EXP) :Expansion stresses (Algebraic summation)
5 L3-L1 (OCC) :Winds net deflection (Algebraic summation)
6 L2+L5 (OCC) :Code stresses for wind (Scalar summation)
24 CAESAR II Technical Reference Manual
Static Seismic Loads
Static earthquake loads are applied in a manner very similar to static wind loads. The static loading
magnitude is considered to be in direct proportion to the elements weight. Earthquake load magnitudes
are given in terms of the gravitational acceleration constant, i.e. gs. If an earthquake is modeled as having
a 0.5-g load in the X direction, then half of the systems weight is turned into a uniform load and applied in
the X direction.
Earthquake static load cases are set up exactly as they are for wind occasional loads, i.e. the same load
case, nonlinearity, and directional sensitivity logic. In some cases the client specifies the magnitude of the
earthquake loading in gs and the direction(s). In others, the analysis is left to the sole discretion of the
analyst. It is not unusual to see only X or X-Y components of an earthquake. It is not uncommon to see Y
only components, or X, Y, and Z simultaneous components.
Dynamic earthquakes are discussed later in this chapter, in the dynamic analysis and output chapters, and
in the screen reference chapter.
The ASCE #7 method for determining earthquake coefficients is described below. Once calculated, the g-
factors should be entered as uniform loads on the piping spreadsheet.
Note: The Uniform Load in G's (on page 112) check box must also be enabled in the spreadsheet
special execution parameters.
The total lateral force at the base of a structure is to be computed from:
V = ZIKCSW
Where:
V - total lateral force or shear at the base
Znumerical coefficient from table 22
Knumerical coefficient from table 23
Cnumerical coefficient from Sect. 9.4
Ssoil factor from table 25
Wtotal dead load
The g-factor can be found by dividing Eq. 6 through by W.
gs = V/W = ZIKCS
Chapter 6 Technical Discussions 25
The product CS does not need to exceed the value 0.14. Use this value as a conservative maximum.
The following table provides the seismic zone coefficient (Z)
.
Seismic Zone
Coefficient, Z
4 1
3 3/4
2 3/8
1 3/16
0 1/8
From the following table, the importance factor can be found: (However use a value for I = 1.0. The
categories in this table are identical for those used in the wind load calculation.)
The following table shows K varying from 0.67 to 2.0. Use K=2.0 for Structures other than buildings.
So the equation for the g load:
g = ZIKCS
reduces to:
g = Z (1.0) (2.0) (0.14)
and for the various value of Z:
26 CAESAR II Technical Reference Manual
Seismic Zone Product "g" Load
4 (1)(1)(2)(0.14) 0.28
3
(3/4)(1)(2)(0.14)
0.21
2
(3/8)(1)(2)(0.14)
0.105
1
(3/16)(1)(2)(0.14)
0.0525
0
(1/8)(1)(2)(0.14)
0.035
Seismic Zones from A58.1 - 1982 fig. 13, p.50
ASCE #7 - 1990 is the 1990 revision to ANSI A58.1 1982. There are no revisions to this code which
affect CAESAR II.
ASCE #7 - 1993 has completely changed the approach for "static" seismic analysis. These changes
are not addressed by this discussion.
Chapter 6 Technical Discussions 27
Wind Loads
Wind loads are generated by multiplying the pipe exposed area, including insulation, and considering
angle to the wind, by the equivalent wind pressure and the pipe shape factor. There are typically three
different ways to get at the equivalent wind pressure:
ASCE #7 (1995)
Pressure vs. elevation table entry
Velocity vs. elevation table entry
The total wind force on the element is calculated from
F = PeqSA
Where:
F = the total wind force on the element
Peq = the equivalent wind pressure (dynamic pressure)
S = the pipe element wind shape factor
A = the pipe element exposed area as shown in the figure as follows:
Peq is calculated for each end of the element and the average taken. The average applies uniformly over the
whole length of the element. Note, the wind force is applied in the three global directions as a function of
the element direction cosines.
If the user enters a velocity vs. elevation table then the velocity is converted to a dynamic pressure using
the following equation:
P = 1/2 V
2
where V is the wind velocity and is the air density.
The WIND SHAPE FACTOR is entered on the pipe spreadsheet and, for cylindrical elements, the value
from Table 12 is between 0.5 and 0.7. A value of 0.65 is typical. The wind shape factor as entered is
distributive. This means that the shape factor applies for all following elements until zeroed or changed.
Important The user does not have to enter the shape factor on each pipe spreadsheet. Zero (or turn
"Off") the wind shape factor if the piping system runs inside of building or similarly protective structure.
Wind load data is entered on the Wind Loads (on page 59) tab of the Static Load Case Builder. Up to four
different wind loads can be entered per analysis. These typically might be set up to model wind loads in
the +X, -Y, and -Z directions.
The ASCE #7 ( 1995) Method for computing equivalent pressure requires several computerized table look
ups and interpolation. The user enters the following parameters:
1 Basic wind speed (mph) - The minimum allowed basic wind speed is 85 mph. This does not include
averages for abnormally high wind loading events such as hurricanes or tornadoes.
ASCE #7 refers to fig. 6-1 for basic wind speeds in the continental United States. The following
description is a crude representation of Figure 1:
28 CAESAR II Technical Reference Manual
California 85 mph (124.6 ft/sec)
Other West Coast Areas 85 mph (124.6 ft/sec)
Great Plains 90 mph (132.0 ft/sec)
Non-Coastal Eastern United States 90 mph (132.0 ft/sec)
Gulf Coast 130 mph (190.6 ft/sec)
Florida Carolinas 130 mph (190.6 ft/sec)
Miami 145 mph (212.6 ft/sec)
New England Coastal Areas 120 mph (176.0 ft/sec)
2 Wind Exposure Options
Large oily center
Urban, suburban, and wooded areas
Open terrain
Flat coastal areas
3 Structural Classification Options
Everything except the following options (used most often)
Primary occupancy more than 100 people
Essential facilities, i.e. hospitals
Failure represents low hazard
4 Topographic Factor Parameters (sec. 6.5.5)
Height of hill or escarpment
Crest distance
Height above ground level
Distance from crest to site
Hill type
The following procedure from the appendix is used to calculate the effective wind pressure:
1 Get the Importance Factor from Table 6-2 (p.17)
2 Get (Alpha), Zg, from Table C 6-2.
3 Calculate Kz from Eq. C2 (p.152)
4 Calculate Kzt from Eq. 6-2 (p.34)
5 Calculate qz from Eq. 6-1, (p.17)
6 Calculate Gz from sec 6.6
Chapter 6 Technical Discussions 29
7 Calculate the effective wind pressure from
PRESSURE = Gz * qz * Shape Factor
Note: Winds of 20 to 40 mph can cause vortex shedding and excitation in the 30 Hz and higher range
that can cause fatigue failure in smaller line sizes particularly susceptible to fatigue type failures. To
analyze vortex shedding, use harmonic analysis methods.
Elevation
The accurate elevation of each individual piping element may, or may not be important depending on the
total height, diameter and rigidity of the piping system and attachments. By default, CAESAR II starts the
first node on the first element at an elevation of 0.0. If this is not close enough to the true elevation then
the user should set the true coordinates of the piping system through the command EDIT - GLOBAL. This
presents a dialog requesting coordinates for the first node of any disconnected section. The coordinates for
up to 100 node points can be specified and saved as part of the input data from the model.
30 CAESAR II Technical Reference Manual
Hydrodynamic (Wave and Current) Loading
Ocean waves are generated by wind and propagate out of the generating area. The generation of ocean
waves is dependent on the wind speed, the duration of the wind, the water depth, and the distance over
which the wind blows is referred to as the fetch length. There a variety of two dimensional wave theories
proposed by various researchers, but the three most widely used are the Airy (linear) wave theory, Stokes
5
th
Order wave theory, and Dean's Stream Function wave theory. The later two theories are non-linear
wave theories and provide a better description of the near surface effects of the wave.
(The term two dimensional refers to the uni-directional wave. One dimension is the direction the wave
travels, and the other dimension is vertical through the water column. Two dimensional waves are not
found in the marine environment, but are somewhat easy to define and determine properties for, in a
deterministic sense. In actuality, waves undergo spreading, in the third dimension. This can be easily
understood by visualizing a stone dropped in a pond. As the wave spread, the diameter of the circle
increases. In addition to wave spreading, a real sea state includes waves of various periods, heights, and
lengths. In order to address these actual conditions, a deterministic approach can not be used. Instead, a
sea spectrum is utilized, which may also include a spreading function. As there are various wave theories,
there are various sea spectra definitions. The definition and implementation of sea spectra are usually
employed in dynamic analysis. Sea Spectra and dynamic analysis will not be discussed in this article.)
The linear or Airy wave theory assumes the free surface is symmetric about the mean water level.
Furthermore, the water particle motion is a closed circular orbit, the diameter of which decays with depth.
(The term circular should be taken loosely here, the orbit varies from circular to elliptical based on
whether the wave is in shallow or deep water.) Additionally, for shallow water waves, the wave height to
depth ratio (H/D) is limited to 0.78 to avoid breaking. (None of the wave theories address breaking
waves!) The figure below shows a typical wave and associated hydrodynamic parameters.
SWL - The still water level.
L - The wave length, the horizontal distance between successive crests or troughs.
H- The wave height, the vertical distance between the crest and trough.
D - The water depth, the vertical distance form the bottom to the still water level.
q - The surface elevation measured from the still water level.
Chapter 6 Technical Discussions 31
Ocean Wave Particulars
The Airy wave theory provides a good first approximation to the water particle behavior. The nonlinear
theories provide a better description of particle motion, over a wider range depths and wave heights. The
Stokes 5
th
wave theory is based on a power series. This wave theory does not apply the symmetric free
surface restriction. Additionally, the particle paths are no longer closed orbits, which means there is a
gradual drift of the fluid particles, i.e. a mass transport.
Stokes 5
th
order wave theory however, does not adequately address steeper waves over a complete range
of depths. Deans Stream Function wave theory attempts to address this deficiency. This wave theory
employs an iterative numerical technique to solve the stream function equation. The stream function
describes not only the geometry of a two dimensional flow, but also the components of the velocity vector
at any point, and the flow rate between any two streamlines.
The most suitable wave theory is dependent on the wave height, the wave period, and the water depth.
Based on these parameters, the applicable wave theory can be determined from the figure below (from
API-RP2A, American Petroleum Institute - Recommended Practice 2A).
32 CAESAR II Technical Reference Manual
Applicable Wave Theory Determination
The limiting wave steepness for most deep water waves is usually determined by the Miche Limit:
H / L = 0.142 tanh( kd )
Where:
H is the wave height
L is the wave length
k is the wave number (2t)/L
d is the water depth
Pseudo-Static Hydrodynamic Loading
CAESAR II allows individual pipe elements to experience loading due to hydrodynamic effects. These fluid
effects can impose a substantial load on the piping elements in a manner similar to, but more complex than
wind loading.
The various wave theories incorporated into CAESAR II as well as the various types of current profiles are
discussed below. The wave theories and the current profile are used to compute the water particle
velocities and accelerations at the node points. Once these parameters are available, the force on the
element can be computed using Morrisons equation:
F = 1/2 * * Cd * D * U * |U| + t/4 * * Cm * D
2
* A
Where
- is the fluid density
Cd- is the drag coefficient
D - is the pipe diameter
U - is the particle velocity
Cm - is the inertial coefficient
A - is the particle acceleration
The particle velocities and accelerations are vector quantities which include the effects of any applied
waves or currents. In addition to the force imposed by Morrisons equation, piping elements are also
subjected to a lift force and a buoyancy force. The lift force is defined as the force acting normal to the
plane formed by the velocity vector and the elements axis. The lift force is defined as:
Fl = 1/2 * * Cl * D * U
2
Where
- is the fluid density
Cl - is the lift coefficient
D - is the pipe diameter
U - is the particle velocity
Chapter 6 Technical Discussions 33
The buoyancy force acts upward, and is equal to the weight of the fluid volume displaced by the element.
Once the force on a particular element is available, it is placed in the system load vector just as any other
load is. A standard solution is performed on the system of equations which describe the piping system.
(The piping system can be described by the standard finite element equation:
[K] {x} = {f}
Where
[K] - is the global stiffness matrix for the entire system
{x} - is the displacement / rotation vector to solve for
{f} - is global load vector
The element loads generated by the hydrodynamic effects are placed in their proper locations in {f},
similar to weight, pressure, and temperature. Once [K] and {f} are finalized, a standard finite element
solution is performed on this system of equations. The resulting displacement vector {x} is then used to
compute element forces, and these forces are then used to compute the element stresses.)
Except for the buoyancy force, all other hydrodynamic forces acting on the element are a function of the
particle velocities and accelerations.
AIRY Wave Theory Implementation
Airy wave theory is also known as linear wave theory, due to the assumption that the wave profile is
symmetric about the mean water level. Standard Airy wave theory allows for the computation of the
water particle velocities and accelerations between the mean surface elevation and the bottom. The
Modified Airy wave theory allows for the consideration of the actual free surface elevation in the
computation of the particle data. CAESAR II includes both the standard and modified forms of the Airy
wave theory.
To apply the Airy wave theory, several descriptive parameters about the wave must be given. These values
are then used to solve for the wave length, which is a characteristic parameter of each unique wave.
CAESAR II uses Newton-Raphston iteration to determine the wave length by solving the dispersion
relation, shown below:
L = (gT
2
/ 2t) * tanh(2tD / L)
Where
g - is the acceleration of gravity
T - is the wave period
D - is the mean water depth
L - is the wave length to be solved for
Once the wave length (L) is known, the other wave particulars of interest may be easily determined. The
parameters determined and used by CAESAR II are: the horizontal and vertical particle velocities ( UX and
UY ), the horizontal and vertical particle acceleration ( AX and AY ), and the surface elevation above (or
below) the mean water level ( ETA ). The equations for these parameters can be found in any standard text
(such as those listed at the end of this section) which discusses ocean wave theories, and therefore will not
be repeated here.
34 CAESAR II Technical Reference Manual
STOKES Wave Theory Implementation
The Stokes wave is a 5th order gravity wave, and hence non-linear in nature. The solution technique
employed by CAESAR II is described in a paper published by Skjelbreia and Hendrickson of the National
Engineering Science Company of Pasadena California in 1960. The standard formulation as well as a
modified formulation (to the free surface) are available in CAESAR II Stokes 5th Order Wave Theory.
The solution follows a procedure very similar to that used in the Airy wave, characteristic parameters of
the wave are determined by using Newton-Raphston iteration, followed by the determination of the water
particle values of interest.
The Newton-Raphston iteration procedure solves two non-linear equations for the constants beta and
lambda. Once these values are available, the other twenty constants can be computed. After all of the
constants are known, CAESAR II can compute: the horizontal and vertical particle velocities ( UX and UY
), the horizontal and vertical particle acceleration ( AX and AY ), and the surface elevation above the
mean water level (ETA).
Stream Function Wave Theory Implementation
In addition to the forces imposed by ocean waves, piping elements may also be subjected to forces
imposed by ocean currents. There are three different ocean current models in CAESAR II; linear, piece-
wise, and a power law profile.
The linear current profile assumes that the current velocity through the water column varies linearly from
the specified surface velocity (at the surface) to zero (at the bottom). The piece-wise linear profile
employs linear interpolation between specific depth/velocity points specified by the user. The power law
profile decays the surface velocity to the 1/7 power.
While waves produce unsteady flow, where the particle velocities and accelerations at a point constantly
change, current produces a steady, non-varying flow.
Ocean Currents
In addition to forces imposed by ocean waves, piping elements may also be subjected to forces imposed
by ocean currents. There are three different ocean current models in CAESAR II; linear piece-wise linear
profile, and a power law profile. The linear current profile assumes that the current velocity though the
water column varies linearly from the specified surface velocity (at the surface to zero (at the bottom). The
piece-wise linear profile employs linear interpolation between specific "depth /velocity" points specified
by the user. The power law profile decays the surface velocity to the 1/7 power.
While waves produce unsteady flow, where the particle velocities and accelerations at a point constantly
change, current produces a steady, non-varying flow.
Chapter 6 Technical Discussions 35
Technical Notes on CAESAR II Hydrodynamic Loading
The input parameters necessary to define the fluid loading are described in detail in the next section. The
basic parameters describe the wave height and period, and the current velocity. The most difficult to
obtain, and also the most important parameters, are the drag, inertia, and lift coefficients, Cd, Cm, and Cl.
Based on the recommendations of API RP2A and DNV (Det Norske Veritas), values for Cd range from 0.6
to 1.2, values for Cm range from 1.5 to 2.0. Values for Cl show a wide range of scatter, but the
approximate mean value is 0.7.
The inertia coefficient Cm is equal to one plus the added mass coefficient Ca. This added mass value
accounts for the mass of the fluid assumed to be entrained with the piping element.
In actuality, these coefficients are a function of the fluid particle velocity, which varies over the water
column. In general practice, two dimensionless parameters are computed which are used to obtain the Cd,
Cm, and Cl values from published charts. The first dimensionless parameter is the Keulegan-Carpenter
Number, K. K is defined as:
K = Um * T / D
Where:
Um - is the maximum fluid particle velocity
T - is the wave period
D - is the characteristic diameter of the element.
The second dimensionless parameter is the Reynolds number, Re. Re is defined as
Re = Um * D / v
Where:
Um - is the maximum fluid particle velocity
D - is the characteristic diameter of the element.
v - is the kinematic viscosity of the fluid (1.26e-5 ft
2
/sec for sea water).
36 CAESAR II Technical Reference Manual
Once K and Re are available, charts are used to obtain Cd, Cm, and Cl. (See Mechanics of Wave Forces on
Offshore Structures by T. Sarpkaya, Figures 3.21, 3.22, and 3.25 for example charts, which are shown in
the figures below.)
Chapter 6 Technical Discussions 37
In order to determine these coefficients, the fluid particle velocity (at the location of interest) must be
determined. The appropriate wave theory is solved, and these particle velocities are readily obtained.
Of the wave theories discussed, the modified Airy and Stokes 5
th
theories include a modification of the
depth-decay function. The standard theories use a depth-decay function equal to cosh(kz) / sinh(kd),
Where:
k - is the wave number, 2t /L
L - is the wave length
d - is the water depth
z - is the elevation in the water column where the data is to be determined
The modified theories include an additional term in the numerator of this depth-decay function. The
modified depth-decay function is equal to cosh(od) / sinh(kd),
Where:
o - is equal to z / (d + q)
38 CAESAR II Technical Reference Manual
The term od represents the effective height of the point at which the particle velocity and acceleration are
to be computed. The use of this term keeps the effective height below the still water level. This means that
the velocity and acceleration computed are convergent for actual heights above the still water level.
As previously stated, the drag, inertia, and lift coefficients are a function of the fluid velocity and the
diameter of the element in question. Note that the fluid particle velocities vary with both depth and
position in the wave train (as determined by the applied wave theory). Therefore, these coefficients are in
fact not constants. However, from a practical engineering point of view, varying these coefficients as a
function of location in the Fluid field is usually not implemented. This practice can be justified when one
considers the inaccuracies involved in specifying the instantaneous wave height and period. According to
Sarpkaya, these values are insufficient to accurately predict wave forces, a consideration of the previous
fluid particle history is necessary. In light of these uncertainties, constant values for Cd, Cm, and Cl are
recommended by API and many other references.
The effects of marine growth must also be considered. Marine growth has the following effects on the
system loading: the increased pipe diameters increase the hydrodynamic loading; the increased roughness
causes an increase in Cd, and therefore the hydrodynamic loading; the increase in mass and added mass
cause reduced natural frequencies and increase the dynamic amplification factor; it causes an increase in
the structural weight; and possibly causes hydrodynamic instabilities, such as vortex shedding.
Finally, Morrisons force equation is based the small body assumption. The term small refers to the
diameter to wave length ratio. If this ratio exceeds 0.2, the inertial force is no longer in phase with the
acceleration of the fluid particles and diffraction effects must be considered. In such cases, the fluid
loading as typically implemented by CAESAR II is no longer applicable.
Additional discussions on hydrodynamic loads and wave theories can be found in the references at the end
of this article.
Chapter 6 Technical Discussions 39
Input: Specifying Hydrodynamic Parameters in CAESAR II
The hydrodynamic load analysis requires the specification of several measurable parameters which
quantify the physical aspects of the environmental phenomenon in question.
Note: Users can enter four different wave loads here. Use the Editing Load Case buttons to move up or
down between the Wave Load Input Spreadsheets.
The necessary hydrodynamic parameters are discussed in the following paragraphs and a CAESAR II
hydrodynamic loading dialog is shown in the figure below.
Wave Loading Editing in the Load Case Editor
Current Data
Profile TypeThis entry defines the interpolation method used by CAESAR II to determine the current
velocity as a function of depth. Available options for this entry are: a power law profile, a piece-wise
linear profile, and a linear profile.
The power law profile determines the current velocity at depth D according to the equation:
Vd = Vs * [di / D]
p
Where
Vd - is the velocity at depth di
Vs - is the specified velocity at the surface
D - is the water depth
p - is the power, set to 1/7
40 CAESAR II Technical Reference Manual
The piece-wise linear profile performs a linear interpolation of a velocity verse depth table (provided by
the user) to obtain the current velocity at depth di. When this type profile is specified, a table of depths and
velocities must be provided. The table should start at the surface (a depth of zero) and progress in the
direction of increasing depth, to the sea bed.
The linear profile also performs a linear interpolation to obtain the current velocity at depth di. However,
this method assumes the current velocity varies linearly from the specified surface velocity to zero at the
sea bed.
Current Speed This entry defines the current speed at the surface. The units for this entry are
(length/time) as defined by the active units file at the time of input. This value should always be a positive
entry.
Current Direction Cosines These entries define the direction of fluid transport due to the current. These
fields are unitless, and follow the standard software global axis convention.
Chapter 6 Technical Discussions 41
Wave Data
Wave Theory Indicator This entry specifies which wave theory is to be used to compute the water
particle velocities and accelerations. The wave theories presently available are:
Standard Airy Wave This is also known as linear wave theory. Discussion of this theory can be found in
the previously mentioned references.
Modified Airy Wave This is a modification of the standard Airy theory which includes the free surface
effects due to the wave. The modification consists of determining a depth scaling factor equal to the depth
divided by the depth plus the surface elevation. Note that this scale factor varies as a function of the
location in the wave train.
Standard Stokes 5th Wave This is a 5th order wave theory, also discussed in the previously mentioned
references.
Modified Stokes 5th Wave This is a modification of the standard Stokes 5th theory. The modification is
the same as applied to the Airy theory.
Stream Function Wave This is Deans Stream Function theory, also discussed in the previously
mentioned references.
Modified Stream Function Wave This is Deans Stream Function theory, modified to directly consider
current in the wave solution.
Stream Function Order When the Stream Function theory is activated, the solution order must be
defined. Typical values for the stream function order range from 3 to 13 (see API-RP2A figure).
Water Depth This entry defines the vertical distance (in units of length) from the still water level (the
surface) to the sea bed.
Wave Height This entry defines the height of the incident wave. The height is the vertical distance (in
units of length) from the wave crest to the wave trough.
Wave Period This entry defines the time span (in seconds) for two successive wave crests to pass a
fixed point.
Wave Kinematic Factor Because the two dimensional wave theories do not account for spreading, a
reduction factor is often used for the horizontal particle velocity and acceleration. Wave kinematic
measurements support values in the range of 0.85 to 0.95. Refer to the applicable offshore codes before
using this item.
Wave Direction Cosines These entries define the direction of wave travel. These fields are unitless, and
follow the standard software global axis convention.
Wave Phase Angle This entry defines the position of the wave relative to the starting node of the piping
system. The phase angle is a measure (in degrees) of position in the wave train, where 0 is the wave crest,
180 is the wave trough, and 360 is the following crest. Since the wave propagates over the piping
structure, each point in the structure experiences all possible wave phase angles. One analysis technique
specifies the wave phase at the system origin, and then the phase at each node point in the model is deter-
mined. From these exact phase locations, the water particle data is computed from the wave theory.
42 CAESAR II Technical Reference Manual
Alternatively, a conservative engineering approach is to use the same phase angle (usually zero) for all
points in the model. This technique produces higher loads, however, the extra conservatism is warranted
when given the unknowns in specifying environmental data.
Seawater Data
Free Surface Elevation This entry defines the height of the free surface, from the global system origin. If
the system origin is at the free surface, this entry should be specified as zero. If the system origin is at the
sea bottom, this entry is equal to the water depth. By default, the first node in a CAESAR II model is at an
elevation of zero. This elevation can be changed using the [Alt-G] key sequence.
Kinematic Viscosity This entry is used to define the kinematic viscosity of water. This value is used to
determine the Reynolds number, which is subsequently used to determine they hydrodynamic coefficients
Cd, Cm, and Cl. Typical values of kinematic viscosity for sea water are listed in the table below.
Temp Deg (F) v(ft
2
/sec) Temp (C) v(m
2
/sec)
60 1.26
e-5
15.556 1.17058
e-6
50 1.46
e-5
10.000 1.35639
e-6
40 1.55
e-5
4.444 1.44000
e-6
30 2.00
e-5
-1.111 1.85807
e-6
Fluid Weight Density - This entry defines the weight density of the fluid. For sea water, this value is
approximately .037037 pounds per cubic inch (.001025 kg/cm3, 1.0256SG).
Piping Element Data
Element Exposure In implementing hydrodynamic loading in a software program, one must be able to
indicate that elements are either exposed to the fluid or not exposed to the fluid. In CAESAR II, this is
accomplished by a set of radio buttons, which indicate that the particular element is exposed to
hydrodynamic loads, wind loads, or not exposed. This specification carries forward for all subsequent
elements, until changed.
Hydrodynamic Coefficients Piping elements which are to be subjected to hydrodynamic loading must
have a drag (Cd), an inertia (Cm), and a lift (Cl) coefficient defined. The specification of these items is
optional. A user may specify these values as constants to be applied to all subsequent exposed elements,
regardless of depth or phase position in the wave. Alternatively, these values may be left blank, which will
cause CAESAR II to interpolate their values from the charts previously discussed.
Marine Growth This entry defines the amount of marine growth on the piping elements. The value of
this entry is used to increase the diameter of the piping elements. The units for this field are the current
diameter units. The diameter used in the computation of the hydrodynamic forces is equal to the pipe
diameter plus twice the marine growth entry.
References
1 Mechanics of Wave Forces On Offshore Structures, Turgut Sarpkaya and Michael Isaacson, Van
Nostrand Reinhold Co., 1982, ISBN 0-442-25402-4.
2 Handbook of Ocean and Underwater Engineering, Myers, Holm, and McAllister, McGraw-Hill Book
Co., 1969, ISBN 07-044245 -2.
Chapter 6 Technical Discussions 43
3 Fifth Order Gravity Wave Theory, Lars Skjelbreia and James Hendrickson, National Engineering
Science Co., Pasadena, California, 1960.
4 Planning and Design of Fixed Offshore Platforms, McClelland and Reifel, Van Nostrand Reinhold
Co., 1986, ISBN 0-442-25223-4.
5 Intercomparison of Near-Bottom Kinematics by Several Wave Theories and Field and Laboratory
Data, R. G. Dean and M. Perlin, Coastal Engineering, #9 (1986), p399-437.
6 A Finite Amplitude Wave on a Linear Shear Current, R. A. Dalrymple, Journal of Geophysical
Research, Vol 79, No 30, 1974.
7 Application of Stream Function Wave Theory to Offshore Design Problems, R. G. Dean, OTC #1613,
1972.
8 Stream Function Representation of Nonlinear Ocean Waves, R. G. Dean, Journal of Geophysical
Research, Vol 70, No 18, 1965.
9 American Petroleum Institute - Recommended Practice 2A (API-RP2A), American Petroleum
Institute, July 1993.
10 Improved Algorithm for Stream Function Wave Theory, Min-Chih Huang, Journal of Waterway, Port,
Coastal, and Ocean Engineering, January 1989.
11 Stream Function Wave Theory with Profile Constraints, Min-Chih Huang, Journal of Waterway, Port,
Coastal, and Ocean Engineering, January/February 1993.
44 CAESAR II Technical Reference Manual
Evaluating Vessel Stresses
The ASME Section VIII, Division 2 code provides for a fairly elaborate procedure to analyze the local
stresses in vessels and nozzles. Only the elastic analysis approach will be discussed in this manual. The
user should always refer to the applicable code if any of the limits described in this section are
approached, or if any unusual material, weld, or stress situation exists, or there are non-linear concerns
such as the material's operation in the creep range.
The first step in the procedure is to determine if the elastic approach is satisfactory. Section AD-160
contains the exact method and basically states that if all of the following conditions are met, then fatigue
analysis need not be done:
1 The expected design number of full-range pressure cycles does not exceed the number of allowed
cycles corresponding to an Sa value of 3Sm (4Sm for non-integral attachments) on the material fatigue
curve. The Sm is the allowable stress intensity for the material at the operating temperature.
2 The expected design range of pressure cycles other than startup or shutdown must be less than 1/3 (1/4
for non-integral attachments) the design pressure times (Sa/Sm), where Sa is the value obtained on the
material fatigue curve for the specified number of significant pressure fluctuations.
3 The vessel does not experience localized high stress due to heating.
4 The full range of stress intensities due to mechanical loads (including piping reactions) does not
exceed Sa from the fatigue curve for the expected number of load fluctuations.
Once the user has decided that an elastic analysis will be satisfactory, either a simplified or a
comprehensive approach may be taken to the vessel stress evaluation. Both methods will be described in
detail below, after a discussion of the Section VIII Div. 2 Requirements.
ASME Section VIII Division 2 - Elastic Analysis of Nozzle
Ideally, in order to address the local allowable stress problem, the user should have the endurance curve
for the material of construction and complete design pressure / temperature loading information. If any of
the elastic limits are approached, or if there is anything out of the ordinary about the nozzle/vessel
connection design, the code should be carefully consulted before performing the local stress analysis. The
material Sm table and the endurance curve for carbon steels are given in this section for illustration. Only
values taken directly from the code should be used in design.
There are essentially three criteria that must be satisfied before the stresses in the vessel wall due to nozzle
loads can be considered within the allowables. These three criteria can be summarized as:
Pm < kSmh
Pm + Pl + Pb< 1.5kSmh
Pm + Pl + Pb + Q < 3Smavg
Chapter 6 Technical Discussions 45
Where Pm, Pl, Pb, and Q are the general primary membrane stress, the local primary membrane stress, the
local primary bending stress, and the total secondary stresses (membrane plus bending), respectively; and
K, Smh, and Smavg are the occasional stress factor, the hot material allowable stress intensity, and the average
material stress intensity (Smh + Smc) / 2.
Due to the stress classification defined by Section VIII, Division 2 in the vicinity of nozzles, as given in
the Table 4-120.1, the bending stress terms caused by any external load moments or internal pressure in
the vessel wall near a nozzle or other opening, should be classified as Q, or the secondary stresses,
regardless of whether they were caused by sustained or expansion loads. This causes Pb to disappear, and
leads to a much more detailed classification:
PmGeneral primary membrane stress (primarily due to internal pressure)
PlLocal primary membrane stress, which may include
--Membrane stress due to internal pressure
--Local membrane stress due to applied sustained forces and moments
QSecondary stresses, which may include
--Bending stress due to internal pressure
--Bending stress due to applied sustained forces and moments
--Membrane stress due to applied expansion forces
--Bending stress due to applied expansion forces and moments
--Membrane stress due to applied expansion moments
Each of the stress terms defined in the above classifications contain three parts: two stress components in
normal directions and one shear stress component. To combine these stresses, the following rules apply:
Compute the normal and shear components for each of the three stress types, i.e. Pm, Pl, and Q;
Compute the stress intensity due to the Pm and compare it against kSmh;
Add the individual normal and shear stress components due to Pmand Pl; compute the resultant stress
intensity and compare its value against 1.5kSmh;
Add the individual normal and shear stress components due to Pm, Pl, and Q, compute the resultant
stress intensity, and compare its value to against 3Smavg.
If there is an occasional load as well as a sustained load, these types may be repeated using a k value
of 1.2.
These criteria can be readily found from Figure 4-130.1 of Appendix 4 of ASME Section VIII, Division 2
and the surrounding text. Note that the primary bending stress term, Pb, is not applicable to the shell stress
evaluation, and therefore disappears from the Section VIII, Division 2 requirements. Under the same
analogy, the peak stress limit may also be written as:
Pl + Pb + Q + F < Sa
The preceding equation need not be satisfied, provided the elastic limit criteria of AD-160 is met based on
the statement explicitly given in Section 5-100, which is cited below:
If the specified operation of the vessel meets all of the conditions of AD-160, no analysis for cyclic
operation is required and it may be assumed that the peak stress limit discussed in 4-135 has been
satisfied by compliance with the applicable requirements for materials, design, fabrication, testing and
inspection of this division.
46 CAESAR II Technical Reference Manual
Procedure to Perform Elastic Analyses of Nozzles
The procedure for checking stresses in vessel shells using WRC 107 can be summarized as follows:
1 Check geometric limitation to see whether WRC 107 is applicable;
2 If yes, check to see whether or not the elastic approach as outlined in Section VIII, Division 2, AD-
160 is satisfactory;
3 Compute the sustained, expansion and occasional loads in the vessel shell due to the applied nozzle
loads. Consider the local restraint configuration in order to determine whether or not the axial pressure
thrust load (P * Ain) should be added to the sustained (and occasional loads). If desired by the user,
this thrust load will be automatically calculated and added to the applied loads.
4 Calculate pressure stresses, Pm, on the vessel shell wall in both longitudinal and circumferential (hoop)
directions for both sustained and occasional cases. Notice that two different pressure terms are
required in carrying out the pressure stress calculations. P is the design pressure of the system
(sustained), while Pvar is the DIFFERENCE between the peak pressure and the design pressure of the
system, which will be used to qualify the vessel membrane stress under the occasional load case.
Note: The Pm stresses will be calculated automatically if a pressure value is enter by the user.
1 Run WRC 107 to calculate the Pl, and Q stresses as defined earlier. Note that the local stresses due to
sustained, expansion and occasional loads can now be compute simultaneously.
2 Various stress components can be obtained from combining the stress intensities computed from
applying the sustained, expansion and occasional loads, if applicable. These stress intensities can then
be used to carry out the stress summations and the results are used to determine acceptability of the
local stresses in the vessel shell. Notice now CAESAR II can provide the WRC 107 stress summation
module in line with the stress calculation routines
Under the above procedure, the equations used in CAESAR II to qualify the various stress components can
be summarized as follows:
Pm(SUS) < Smh
Pm(SUS + OCC) < 1.2Smh
Pm(SUS) + Pl(SUS) < 1.5Smh
Pm(SUS + OCC) + Pl(SUS + OCC) < 1.5(1.2)Smh
Pm(SUS + OCC) + Pl(SUS + OCC) + Q(SUS + EXP + OCC) < 1.5(Smc + Smh)
Chapter 6 Technical Discussions 47
Description of Alternate Simplified ASME Sect. VIII Div. 2 Nozzle Analysis
The most difficult problem associated with the comprehensive ASME Sect. VIII, Div. 2 nozzle/vessel
analysis involves the pressure calculation. Hoop and longitudinal hand pressure calculations can not be
considered dependable, and axial pressure loading on the junction is often calculated incorrectly or
omitted. A smaller, yet significant problem with the comprehensive calculation is the time it takes to
organize and manipulate the stress data.
For these reasons, an alternate simplified approach was developed. To eliminate the concern for pressure,
both the pressure term in the loading on the left side of the inequality and the pressure term in the
allowable on the right side of the inequality are cancelled.
The first check is Pm (due to pressure) must be less than or equal to 1.0 Smh. Assuming that the area
reinforcement around the nozzle will satisfy the pressure requirements, let this first check equal the
maximum value.
The second check is Pm + Pl + Pb must be less than or equal to 1.5 Smh. Subtracting the stresses due to
pressure (assumed equal to Smh) reduces this check to: Pl + Pb (due to external sustained forces without
pressure) < 0.5 Smh.
Unfortunately, the third check on the Pm + Pl + Q terms are at the root of an application controversy. There
are primarily three schools of thought:
Pm+Pl+Q is an operating loading condition, and as such, includes the loads due to pressure and weight.
Pm+Pl+Q is the range of loads, i.e. the expansion loading condition, and as such, excludes the effects
of sustained, or primary loads. Primary sustained loads, such as weight and pressure, should be
excluded.
Pm+Pl+Q is the range of loads and should exclude the primary load weight, but should include the
varying pressure load at least in those thermal load cases where the system goes from a startup
(ambient temperature and pressure condition to operating condition).
For the simplification, it is assumed that the Pm component due to pressure should be included in both the
left and right side of the Pm+Pl+Pb+Q < 3Sm inequality, thus assuming that the area reinforcement
requirements are exactly satisfied, i.e.
Again, letting Pm = Sm and subtracting this pressure term from the expansion allowable (Pm + Pl + Q <
3Sm) provides a simplified allowable limit.
The expansion (or operating, or both) loads from the CAESAR II restraint report should satisfy the
computed stress requirement:
Pl + Pb + Q (operating or expansion excluding pressure) < 2Sm.
In summary
Ensure proper nozzle reinforcement for pressure and assume pressure stresses are at their maximum.
Compare primary stresses (without pressure) to 1/2 Smh.
Compare stresses due to the sum of primary and secondary loads to 2Sm(avg); where Sm(avg) is the average
of the hot and cold allowable stress intensities (Smh & Smc).
48 CAESAR II Technical Reference Manual
Simplified ASME Sect. VIII Div. 2 Elastic Nozzle Analysis
1 Perform a CAESAR II analysis of the piping loads on the vessel/nozzle junction. Use WRC 297
flexibilities to compute loads more accurately, but less conservatively (or do two analysis, one with
flexibilities and one without). From this analysis the user should have sustained, operating, and
expansion loads on the vessel/nozzle junction.
2 Find Smh and Smc from the Sect. VIII allowable stress tables. Smh is the vessel material hot allowable,
and Smc is the vessel material cold allowable.
3 Run WRC 107 with the sustained loads on the vessel/nozzle junction from CAESAR II, and make sure
that the computed stress intensities are less than 0.5 Smh. This conservatively considers bending
stresses from internal pressure and sustained moments to have a primary classification; if it fails, the
stresses must be reviewed in more detail.
4 Run WRC 107 with the operating loads on the vessel/nozzle junction from CAESAR II, and make sure
that the computed stress intensities are less than Smh + Smc.
5 Run WRC 107 with the expansion loads on the vessel/nozzle junction from CAESAR II, and make sure
that the computed stress intensities are less than Smh + Smc.
Should any of the checks described fail, then the more comprehensive analysis (described earlier) of the
junction should be performed.
Chapter 6 Technical Discussions 49
Inclusion of Missing Mass Correction
The response of a system under a dynamic load is often determined by superposition of modal results,
with CAESAR II specifically providing the spectral analysis method for use. One of the advantages of
modal analysis is that usually only a limited number of modes are excited and need be included in the
analysis. The drawback to this method is that although displacements may be obtained with good accuracy
using only a few of the lowest frequency modes, the force, reaction, and stress results may require
extraction of far more modes (possibly far into the rigid range) before acceptable accuracy is attained.
CAESAR IIs Missing Mass option offers the ability to include a correction which represents the quasi-
static contribution of the higher order modes not explicitly extracted for the modal/dynamic response, thus
providing greater accuracy with reduced calculation time.
The dynamic response of a linear multi-degree-of-freedom system is described by the following equation:
Ma(t) + Cv(t) + Kx(t) = F(t)
Where:
M= n x n mass matrix of system
C = n x n damping matrix of system
K = n x n stiffness matrix of system
a(t) = n x 1, time-dependent acceleration vector
v(t) = n x 1, time-dependent velocity vector
x(t) = n x 1, time-dependent displacement vector
F(t) = n x 1, time-dependent applied force vector
Assuming harmonic motion and neglecting damping, the free vibration eigenvalue problem for this system
is
Ku - Mu e
2
= 0
Where:
u = n x n mode shape matrix
e
2
= n x n matrix where each diagonal entry is the frequency squared of the corresponding mode
The modal matrix u may be normalized such that u
T
M u = I (where I is the n x n identity matrix) and u
T
K u = e
2
.
The modal matrix u may be partitioned into two submatrices:
u = [ ue ur ]
Where:
ue = mode shapes extracted for dynamic analysis (i.e., lowest frequency modes)
ur = residual (non-extracted) mode shapes (corresponding to rigid response, or the missing mass
contribution)
50 CAESAR II Technical Reference Manual
The extracted mode shapes are orthogonal to the residual mode shapes, or:
ue
T
x ur = 0
The displacement components can be expressed as linear combinations of the mode shapes:
x = uY = ue Ye + ur Yr = xe + xr
Where:
x = total system displacements
xe = system displacements due to extracted modes
xr = system displacements due to residual modes
Y = generalized modal coordinates
Ye = partition of Y matrix corresponding to extracted modes
Yr = partition of Y matrix corresponding to residual modes
The dynamic load vector can be expressed in similar terms:
F = K u Y = K ue Ye + K ur Yr = Fe + Fr
Where:
F = total system load vector
Fe = load vector due to extracted modes
Fr = load vector due to residual modes
Y = generalized modal coordinates
Ye = partition of Y matrix corresponding to extracted modes
Yr = partition of Y matrix corresponding to residual modes
Normally, modal superposition analyses completely neglect the rigid response the displacements X r
caused by the load Fr. This response, of the non-extracted modes, can be obtained from the system
displacement under a static loading Fr. Based upon the relationships stated above, Fr can be estimated as
follows:
F = K ue Ye + K ur Yr
Multiplying both sides by ue
T
(and considering that ue
T
ur = 0):
ue
T
F = ue
T
K ue Ye + ue
T
K ur Yr = ue
T
K ue Ye
Substituting ee
2
for ue
T
K ue and solving for Ye:
ue
T
F = ee
2
Ye
Ye = ue
T
ee
-2
F
The residual force can now be stated as
Fr = F - K ue Ye = F - ue
T
K ue ee
-2
F
Chapter 6 Technical Discussions 51
As seen earlier
u
T
M u e
2
= I e
2
= u
T
K u
Substituting ue
T
Mue ee
2
for ue
T
K ue:
Fr = F - ue
T
M ue ee
2
ee
-2
F = F - ue
T
M ue F
Therefore, CAESAR II calculates the residual response (and includes it as the missing mass contribution)
according to the following procedure:
1 The missing mass load is calculated for each individual shock load as
Fr = F - ue
T
M ue F
Note: The load vector F represents the product of the force set vector and the rigid DLF for force
spectrum loading; the product of the mass matrix, ZPA, and directional vector for non-ISM seismic loads;
and the product of the mass matrix, ZPA, and displacement matrix (under unit ISM support displacement)
for seismic anchor movement loads. Note that the missing mass load will vary, depending upon the
number of modes extracted by the user and the cutoff frequency selected (or more specifically, the DLF or
acceleration corresponding to the cutoff frequency). "Rigid, for the purposes of determining the rigid
DLF, or the ZPA, may be designated by the user, through a setup parameter, to be either the
DLF/acceleration associated with the frequency of the last extracted mode, or the true spectral DLF/
ZPAthat corresponding to the largest entered frequency of the input spectrum.
2 The missing mass load is applied to the structure as a static load. The static structural response is then
combined (according to the user-specified combination method) with the dynamically amplified
modal responses as if it were a modal response. Actually this static response is the algebraic sum of
the responses of all non-extracted modes representing in-phase response, as would be expected
from rigid modes.
3 The Missing Mass Data report is compiled for all shock cases, whether missing mass is to be included
or not. The percent of mass active is calculated according to:
% Active Mass = 1 - ( Fr[i] / F [i])
summed over i = 1 to n
52 CAESAR II Technical Reference Manual
The maximum possible percent that is theoretically possible for this value is of course 100%, however
numerical inaccuracies may occasionally cause the value to be slightly higher. If the missing mass
correction factor is included, the percent of mass included in the correction is shown in the report as well.
Since CAESAR IIs procedure assumes that the missing mass correction represents the contribution of rigid
modes, and that the ZPA is based upon the spectral ordinate value at the frequency of the last extracted
mode, it is recommended that the user extract modes up to, but not far beyond, a recognized rigid
frequency. Choosing a cutoff frequency to the left of the spectrums resonant peak will provide a non-
conservative result, since resonant responses may be missed. Using a cutoff frequency to the right of the
peak, but still in the resonant range, will yield conservative results, since the ZPA/rigid DLF will be
overestimated. Extracting a large number of rigid modes for calculation of the dynamic response may be
conservative, since all available modal combination methods (SRSS, GROUP, ABS, etc.) give
conservative results versus the algebraic combination method which gives a more realistic representation
of the net response of the rigid modes. Based upon the response spectrum shown below, an appropriate
cutoff point for the modal extraction would be about 33 Hz.
CAESAR II provides two options for combining the missing mass correction with the modal (dynamic)
resultsSRSS and Absolute. The Absolute combination method of course provides the more conservative
result, and is based upon the assumption that the dynamic amplification is going to occur simultaneously
with the maximum ground acceleration or force load. Literature (References 1, 2) states that the modal and
the rigid portions of the response to typical dynamic loads are actually statistically independent, so that an
SRSS combination method is a more accurate representation of reality. For this reason, CAESAR IIs
default missing mass combination method is SRSS.
References
1 A. K. Gupta, Response Spectrum Method in Seismic Analysis and Design of Structures, CRC Press,
1990
2 K. M. Vashi, Computation of Seismic Response from Higher Frequency Modes, ASME 80-
C2/PVP-50, 1980
Chapter 6 Technical Discussions 53
3 O. E. Hansteen and K. Bell, On the Accuracy of Mode Superposition Analysis in Structural
Dynamics, Earthquake Engineering and Structural Dynamics, Volume 7, John Wiley & Sons, Ltd.,
1979
54 CAESAR II Technical Reference Manual
Fatigue Analysis Using CAESAR II
For most piping codes supported by CAESAR II, performance of fatigue analysis is an extension to, rather
than an explicit part of, the code requirements (however, it is an explicit part of the IGE/TD/12 Pipework
Stress Analysis for Gas Industry Plant code).
Fatigue Basics
Piping and vessels have been known to suffer from sudden failure following years of successful service.
Research done during the 1940s and 1950s (primarily advanced by A. R. C. Markls Piping Flexibility
Analysis, published in 1955) provided an explanation for this phenomenon, as well as design criteria
aimed at avoiding failures of this type. The explanation was that materials were failing due to fatigue, a
process leading to the propagation of cracks, and subsequent fracture, following repeated cyclic loading.
Steels and other metals are made up of organized patterns of molecules, known as crystal structures.
However, these patterns are not maintained throughout the steel producing an ideal homogeneous
material, but are found in microscopic isolated island-like areas called grains. Inside each grain the pattern
of molecules is preserved. From one grain boundary to the next the molecular pattern is the same, but the
orientation differs. As a result, grain boundaries are high energy borders. Plastic deformation begins
within a grain that is both subject to a high stress and oriented such that the stress causes a slippage
between adjacent layers in the same pattern. The incremental slippages (called dislocations) cause local
cold-working. On the first application of the stress, dislocations will move through many of the grains that
are in the local area of high stress. As the stress is repeated, more dislocations will move through their
respective grains. Dislocation movement is impeded by the grain boundaries, so after multiple stress
applications, the dislocations tend to accumulate at grain boundaries, eventually becoming so dense that
the grains lock up, causing a loss of ductility and thus preventing further dislocation movement.
Subsequent applications of the stress cause the grain to tear, forming cracks. Repeated stress applications
cause the cracks to grow. Unless abated, the cracks propagate with additional stress applications until
sufficient cross sectional strength is lost to cause catastrophic failure of the material.
The fatigue capacity of a material can be estimated through the application of cyclic tensile/compressive
displacement loads with a uniaxial test machine. A plot of the cyclic stress capacity of a material is called
a fatigue (or endurance) curve. These curves are generated through multiple cyclic tests at different stress
levels. The number of cycles to failure usually increases as the applied cyclic stress decreases, often until a
threshold stress (known as the endurance limit) is reached below which no fatigue failure occurs,
regardless of the number of applied cycles. An endurance curve for carbon and low alloy steels, taken
from the ASME Section VIII Division 2 Pressure Vessel Code is shown in the following figure.
Chapter 6 Technical Discussions 55
Fatigue Analysis of Piping Systems
The IGE/TD/12 code does, on the other hand, present specific requirements for true fatigue evaluation of
systems subject to a cyclic loading threshold. Furthermore, ASME Section III, Subsection NB and ASME
Section VIII Division 2 provide guidelines by which fatigue evaluation rules may be applied to piping
(and other pressure retaining equipment). These procedures have been adapted, where possible, to
CAESAR IIs methodology.
To perform fatigue analyses:
1 Assigning fatigue curve data to the piping material: This is done on the Allowable auxiliary screen.
Fatigue data may be entered directly, or read in from a text file (a number of commonly used curves
have been provided). Users may define their own fatigue curves as defined later in this section.
2 Defining the fatigue load cases: This may be done in either the static or dynamic load case builders.
For this purpose, a new stress type, FAT, has been defined. For every fatigue case, the number of
anticipated cycles must also be defined.
3 Calculation of the fatigue stresses: This is done automatically by CAESAR II the fatigue stresses,
unless explicitly defined by the applicable code are calculated the same as CAESAR II calculates stress
intensity, in order to conform to the requirements of ASME Section VIII, Division 2 Appendix 5. (The
IGE/TD/12 is currently the only piping code supported by CAESAR II which does have explicit
instructions for calculating fatigue stresses.) The equations used in the calculation of fatigue stresses
are documented at the end of this section.
4 Determination of the allowable fatigue stresses: Allowables are interpolated logarithmically from the
fatigue curve based upon the number of cycles designated for the load case. For static load cases, the
calculated stress is assumed to be a peak-to-peak cyclic value (i.e., thermal expansion, settlement,
pressure, etc.), so the allowable stress is extracted directly from the fatigue curve. For harmonic and
dynamic load cases, the calculated stress is assumed to be a zero-to-peak cyclic value (i.e., vibration,
earthquake, etc.), so the extracted allowable is divided by 2 prior to use in the comparison.
56 CAESAR II Technical Reference Manual
5 Determination of the allowable number of cycles: The flip side of calculating the allowable fatigue
stress for the designated number of cycles is the calculation of the allowable number of cycles for the
calculated stress level. This is done by logarithmically interpolating the Cycles axis of the fatigue
curve based upon the calculated stress value. Since static stresses are assumed to be peak-to-peak
cyclic values, the allowable number of cycles is interpolated directly from the fatigue curve. Since
harmonic and dynamic stresses are assumed to be zero-to-peak cyclic values, the allowable number of
cycles is interpolated using twice the calculated stress value.
6 Reporting the results: CAESAR II provides two reports for viewing the results of load cases of stress
type FAT. The first of these is the standard stress report, which displays the calculated fatigue stress
and fatigue allowable at each node. Stress reports may be generated individually for each load case,
and show whether any of the individual load cases in isolation would fail the system.
However, in those circumstances where there is more than one cyclic load case potentially contributing to
fatigue failure, the Cumulative Usage report is appropriate. In order to generate this report, the user selects
all of the FAT load cases which contribute to the overall system degradation. The Cumulative Usage
report lists for each node point the usage ratio (actual cycles divided by allowable cycles), and then sums
these up for total Cumulative Usage. A total greater than 1.0 indicates a potential fatigue failure.
Static Analysis Fatigue Example
Consider a sample job that potentially has several different cyclic load variations:
1 Operating cycle from ambient (70F) to 500F (12,000 cycles anticipated)
2 Shut down external temperature variation from ambient (70F) to -20F (200 cycles anticipated)
3 Pressurization to 1800 psig (12,000 cycles anticipated)
4 Pressure fluctuations of plus/minus 30 psi from the 1800 psig (200,000 cycles anticipated)
Chapter 6 Technical Discussions 57
In order to do a proper fatigue analysis, these should be grouped in sets of load pairs which represent the
worst-case combination of stress ranges between extreme states. These load variations can be laid out in
graphical form. The figure below shows a sketch of the various operating ranges this system experiences.
Each horizontal line represents an operating range. At the each end of each horizontal line, the
temperatures and pressures defining the range are noted. At the center of each horizontal line, the number
of cycles for each range is defined.
58 CAESAR II Technical Reference Manual
Using this sketch of the operating ranges, the four fatigue load cases can be determined. The procedure is
as follows.
Case 1: Cover the absolute extreme, from 20F and 0 psi to 500F and 1830 psi. This occurs 200 times.
As a result of this case, the cycles for the ranges defined must be reduced by 200. The first range (-20,0 to
70,0) is reduced to zero, and has no contribution to additional load cases. The second range (70,0 to
500,1800) is reduced to 11,800 cycles. The third and fourth ranges are similarly reduced to 199,800
cycles.
These same steps can be used to arrive at cases 2 through 4, reducing the number of considered cycles at
each step. This procedure is summarized in the table below.
Segment
-20, 0 to 70, 0
70, 0 to 500, 1800 500, 1700 to 500, 1800 500, 1800 to 500, 1830
Case
Initial 200 12,000 200, 000 200,000
After 1 0 11,800 200, 000 199,800
After 2 0 0 200, 000 188,000
After 3 0 0 12,000 0
After 4 0 0 0 0
This table is then used to set the load cases as cycles between the following load values:
Between -20F, 0 psig and 500F, 1830 psig (200 cycles)
Between 70F, 0 psig and 500F, 1830 psig (11,800 cycles)
Between 500F, 1770 psig and 500F, 1830 psig (188,000 cycles)
Between 500F, 1770 psig and 500F, 1800 psig (12,000 cycles)
These temperatures and pressures are entered as operating conditions accordingly:
Chapter 6 Technical Discussions 59
Static Analysis Fatigue Example
It is next necessary to enter the fatigue curve data for the material. This is done by clicking the Fatigue
Curves button, revealing the Material Fatigue Curve dialog box. This can be used to enter the fatigue
curve for the material (note: for the IGE/ TD/12 code it is necessary to enter five sets of fatigue curves, for
fatigue classes D, E, F, G, and W). Up to eight Cycle vs. Stress data points may be entered to define the
curve; interpolations are made logarithmically. Cycle/Stress pairs should be entered in ascending order
(ascending by cycles). Stress values should be entered as allowable Stress Range, rather than allowable
Stress Amplitude.
60 CAESAR II Technical Reference Manual
Material Fatigue Curves Dialog
Fatigue curves may be alternatively acquired from a text file, by clicking on the Read from file button.
This displays a list of all \CAESAR\SYSTEM\*.FAT files.
Read from File Dialog
Shipped with the program are the following fatigue curve files (the user may easily construct additional
fatigue curve files, as described in Appendix A below):
5-110-1A.FAT ASME Section VIII Division 2 Figure 5-110.1, UTS < 80 ksi
5-110-1B.FAT ASME Section VIII Division 2 Figure 5-110.1, UTS = 115-130 ksi
5-110-2A.FAT ASME Section VIII Division 2 Figure 5-110.2, Curve A
5-110-2B.FAT ASME Section VIII Division 2 Figure 5-110.2, Curve B
5-110-2C.FAT ASME Section VIII Division 2 Figure 5-110.2, Curve C
Chapter 6 Technical Discussions 61
In this case, for A106B low carbon steel, operating at 500F, 5-110-1A.FAT is the appropriate selection.
This fills in the fatigue curve data:
A106B Low Carbon Steel Example Fatigue Curve Data
At this point, the job can be error checked, and the load cases can be set up.
The static load case builder offers a new stress type, FAT (fatigue). Selecting this stress type does the
following:
1 invites the user to define the number of cycles for the load case (dragging the FAT stress type into the
load case or pressing the Load Cycles button opens the Load Cycles field),
2 causes the stress range to be calculated as per the fatigue stress method of the governing code
(currently this is stress intensity for all codes except IGE/TD/12),
3 causes the calculated stress range to be compared to the full value extracted from the fatigue curve,
and
4 indicates that the load case may be included in the Cumulative Usage report.
62 CAESAR II Technical Reference Manual
The last four load cases represent the load set pairs defined earlier.
Example with Fatigue Load Cases Defined in the Load Case Editor
Once the job has been run, note that the presence of a FAT stress type adds the Cumulative Usage report to
the list of available reports.
Static Output Processor
Chapter 6 Technical Discussions 63
The fatigue stress range may be checked against the fatigue curve allowable for each load case by simply
selecting it along with the Stresses report. Review of each load case shows that all stress levels pass.
Example of a Fatigue Stress Report
However, this is not a true evaluation of the situation, because it is not a case of either-or. The piping
system is subjected to all of these load cases throughout its expected design life, not just one of them.
Therefore, we must review the Cumulative Usage report, which shows the total effect of all fatigue load
cases (or any combination selected by the user) on the design life of the system. This report lists for each
load case the expected number of cycles, the allowable number of cycles (based upon the calculated
stress), and the Usage Ratio (actual cycles divided by allowable cycles). The Usage Ratios are then
summed for all selected load cases; if this sum exceeds 1.0, the system has exceeded its fatigue
capabilities. In this case, it is apparent that the sum of all of the cyclic loadings at node 115 can be
expected to fail this system:
64 CAESAR II Technical Reference Manual
Cumulative Usage Report
Chapter 6 Technical Discussions 65
Fatigue Capabilities in Dynamic Analysis
Fatigue analysis capability is also available for harmonic and dynamic analyses as well. Harmonic load
cases are entered as they always have been; they may be designated as being stress type FAT simply by
entering the number of expected load cycles on the harmonic input screen:
Harmonic Input Screen
This produces the same types of reports as are available for the static analysis; they can be processed as
discussed earlier.
66 CAESAR II Technical Reference Manual
Static Output Processor
The only difference between the harmonic and static fatigue analyses is that for harmonic jobs, the
calculated stresses are assumed to be zero-to-peak calculations, so they are compared to only half of the
stress value extracted from the fatigue curve. Likewise, when creating the Cumulative Usage report, the
number of allowable cycles is based upon twice the calculated stress.
For other dynamic applications (response spectrum and time history), the stress type may be identified as
fatigue by selecting the stress type from the drop list for the Load Case or Static/Dynamic Combination,
and by entering the number of expected cycles in the provided field.
Note that as with the harmonic analyses, the calculated stresses are assumed to be zero-to-peak
calculations, so they are compared to only half of the stress value extracted from the fatigue curve.
Likewise, when creating the Cumulative Usage report, the number of allowable cycles is based upon twice
the calculated stress.
Chapter 6 Technical Discussions 67
Creating the .FAT Files
The .FAT file is a simple text file, containing the data points necessary to describe the fatigue curve for
the material, for both butt welded and fillet welded fittings. A sample FAT file is shown below.
* ASME SECTION VIII DIVISION 2 FATIGUE CURVE
* FIGURE 5-110.1
* DESIGN FATIGUE CURVES FOR CARBON, LOW ALLOY, SERIES 4XX,
* HIGH ALLOY AND HIGH TENSILE STEELS FOR TEMPERATURES NOT
* EXCEEDING 700 F
* FOR UTS <= 80 KSI
*
0.5000000 - STRESS MULTIPLIER (PSI); ALSO CONVERTS AMPLITUDE TO FULL
RANGE
*
10 580000.0
100 205000.0
1000 83000.0
10000 38000.0
100000 20000.0
500000 13500.0
1000000 12500.0
0 0.0
*
This text file can be created using any available text editor. Any line beginning with an asterisk is treated
as a comment line. It is highly recommended that comment lines be used so that the data can be related
back to a specific material curve.
The first actual data line in the file is a stress multiplier. This value is used to adjust the data values from
zero to peak to peak to peak and/or to convert the stress levels to psi (the entered values will be
divided by this number -- i.e., if the stress values in the file represent a stress amplitude, in psi, rather than
a range, this "stress multiplier should be 0.5). Following this line is the fatigue curve data table. This table
consists of eight lines, of two columns. The first column is the Cycle column, the second column is the
Stress column. For each value in the cycle column, the corresponding stress value from the material
fatigue curve should be listed in the stress column.
68 CAESAR II Technical Reference Manual
Fatigue curves intended for use in the IGE/TD/12 code are built slightly different. The first data line
contains not one, but three values: the stress multiplier described above, a modulus of elasticity
correction, and a modulus of elasticity multiplier (the correction factor is divided by this to convert to
psi) upon file read, the modulus of elasticity correction is inserted into the appropriate field on the
fatigue curve screen. Furthermore, the IGE/TD/12 fatigue files include five fatigue curves (sequentially
Fatigue Class D, E, F, G, and W), rather than one. Optional comment lines may be used to separate the
tables these comments aid in the readability of the data file. The format of the IGE/TD/12 fatigue files
can best be determined by reviewing the contents of the file TD12ST.FAT.
In all tables, the number of cycles increases as you work down the table. If there is not enough data to
utilize all eight lines, unused lines should be populated with zeroes.
Calculation of Fatigue Stresses
For the IGE/TD/12 piping code, the computation of fatigue stresses are detailed in Section 5.4.4 of that
code. This section of the code states: "The principal stress in any plane can be calculated for any set of
conditions from the following formula:"
2 2
1
2
( ) ( ) 4
h a h a q
S S S S S
(
+ +
(
Where,
S
h
= Hoop stress
S
a
= Axial stress
S
q
= Shear stress
"This should be used for establishing the range of stress, due regard being paid to the direction and sign."
For all other piping codes in CAESAR II, the fatigue stress is computed as the stress intensity, as follows:
3D Maximum Shear Stress Intensity (Default)
SI = Maximum of:
S1OT - S3OT
S1OB - S3OB
Max(S1IT,RPS) - Min(S3IT,RPS)
Max(S1IB,RPS) - Min(S3IB,RPS)
Where:
S1OT=Maximum Principal Stress, Outside Top
= (SLOT+HPSO)/2.0+(((SLOT-HPSO)/2.0)
2
+TSO
2
)1/2
Chapter 6 Technical Discussions 69
S3OT=Minimum Principal Stress, Outside Top
=(SLOT+HPSO)/2.0-(((SLOT-HPSO)/2.0)2+TSO2) 1/2
S1IT=Maximum Principal Stress, Inside Top
=(SLIT+HPSI)/2.0+(((SLIT-HPSI)/2.0)
2
+TSI
2
) 1/2
S3IT=Minimum Principal Stress, Inside Top
=(SLIT+HPSI)/2.0-(((SLIT-HPSI)/2.0)
2
+TSI
2
) 1/2
S1OB=Maximum Principal Stress, Outside Top
=(SLOB+HPSO)/2.0+ (((SLOB-HPSO)/2.0)
2
+TSO
2
) 1/2
S3OB=Minimum Principal Stress, Outside Bottom
=(SLOB+HPSO)/2.0- (((SLOB-HPSO)/2.0)
2
+TSO
2
) 1/2
S1IB=Maximum Principal Stress, Inside Bottom
=(SLIB+HPSI)/2.0+ (((SLIB-HPSI)/2.0)
2
+TSI
2
) 1/2
S3IB=Minimum Principal Stress, Inside Bottom
=(SLIB+HPSI)/2.0- (((SLIB-HPSI)/2.0)
2
+TSI
2
) 1/2
RPS=Radial Pressure Stress, Inside
HPSI=Hoop Pressure Stress (Inside, from Lame's Equation)
HPSO=Hoop Pressure Stress (Outside, from Lame's Equation)
SLOT=Longitudinal Stress, Outside Top
SLIT=Longitudinal Stress, Inside Top
SLOB=Longitudinal Stress, Outside Bottom
SLIB=Longitudinal Stress, Inside Bottom
TSI=Torsional Stress, Inside
TSO=Torsional Stress, Outside
70 CAESAR II Technical Reference Manual
Pipe Stress Analysis of FRP Piping
Underlying Theory
The behavior of steel and other homogeneous materials has been long understood, permitting their
widespread use as construction materials. The development of the piping and pressure vessel codes
(Reference 1) in the early part of this century led to the confidence in their use in piping applications; the
work of Markl et. al. in the 1940s and 1950s was responsible for the formalization of todays pipe stress
methods, leading to an ensuing diversification of piping codes on an industry by industry basis. The
advent of the digital computer, and with it the appearance of the first pipe stress analysis software
(Reference 2), further increased the confidence with which steel pipe could be used in critical applications.
The 1980s saw the wide spread proliferation of the micro computer, with associated pipe stress analysis
software, which in conjunction with training, technical support, and available literature, has brought stress
analysis capability to almost all engineers. In short, an accumulated experience of close to 100 years, in
conjunction with ever improving technology has led to the utmost confidence on the part of todays
engineers when specifying, designing, and analyzing steel, or other metallic, pipe.
For fiberglass reinforced plastic (FRP) and other composite piping materials, the situation is not the same.
Fiberglass reinforced plastic was developed only as recently as the 1950s, and did not come into wide
spread use until a decade later (Reference 3). There is not a large base of stress analysis experience,
although not from a lack of commitment on the part of FRP vendors. Most vendors conduct extensive
stress testing on their components, including hydrostatic and cyclic pressure, uniaxial tensile and
compressive, bending, and combined loading tests. The problem is due to the traditional difficulty
associated with, and lack of understanding of, stress analysis of heterogeneous materials. First, the
behavior and failure modes of these materials are highly complex and not fully understood, leading to
inexact analytical methods, and a general lack of agreement on the best course of action to follow. This
lack of agreement has slowed the simplification and standardization of the analytical methods into
universally recognized codes (BS 7159 Code (Design and Construction of Glass Reinforced Plastics Piping
Systems for Individual Plants or Sites) and UKOOA Specification and Recommended Practice for the Use of
GRP Piping Offshore being notable exceptions). Secondly, the heterogeneous, orthotropic behavior of FRP
and other composite materials has hindered the use of the pipe stress analysis algorithms developed for
homogeneous, isotropic materials associated with crystalline structures. A lack of generally accepted
analytical procedures has contributed to a general reluctance to use FRP piping for critical applications.
Stress analysis of FRP components must be viewed on many levels. These levels, or scales, have been
called Micro-Mini-Macro levels, with analysis proceeding along the levels according to the MMM
principle (Reference 4).
Chapter 6 Technical Discussions 71
Micro-Level Analysis
Stress analysis on the Micro level refers to the detailed evaluation of the individual materials and
boundary mechanisms comprising the composite material. In general, FRP pipe is manufactured from
laminates, which are constructed from elongated fibers of a commercial grade of glass (called E-glass),
which are coated with a coupling agent or sizing prior to being embedded in a thermosetting plastic
material, typically epoxy or polyester resin.
This means, on the micro scale, that an analytical model must be created which simulates the interface
between these elements. Since the number and orientation of fibers is unknown at any given location in
the FRP sample, the simplest representation of the micro-model is that of a single fiber, extending the
length of the sample, embedded in a square profile of matrix.
Evaluation of this model requires use of the material parameters of
1 the glass fiber
2 the coupling agent or sizing layer (normally of such microscopic proportion that it may be ignored)
3 the plastic matrix
It must be considered that these material parameters may vary for an individual material based upon
tensile, compressive, or shear applications of the imposed stresses, and typical values vary significantly
between the fiber and matrix (Reference 5):
Young's Modulus Ultimate Strength Coefficient of Thermal Expansion
Material tensile (MPa) tensile (MPa) m/m/C
Glass Fiber 7.25 x 10
3
1.5 x 10
3
5.0 x 10
-6
Plastic Matrix 2.75 x 10
3
7.0 x 10
1
7.0 x 10
-3
The following failure modes of the composite must be similarly evaluated:
failure of the fiber
failure of the coupling agent layer
failure of the matrix
failure of the fiber-coupling agent bond
failure of the coupling agent-matrix bond
Because of uncertainties about the degree to which the fiber has been coated with the coupling agent and
about the nature of some of these failure modes, this evaluation is typically reduced to
failure of the fiber
failure of the matrix
failure of the fiber-matrix interface
72 CAESAR II Technical Reference Manual
Micro-Level GRP Sample-- Single Fiber Embedded in Square Profile of Matrix
Stresses in the individual components can be evaluated through finite element analysis of the strain
continuity and equilibrium equations, based upon the assumption that there is a good bond between the
fiber and matrix, resulting in compatible strains between the two. For normal stresses applied parallel to
the glass fiber:
cf = cm = oaf / Ef = oam / Em
oaf = oam Ef / Em
Where:
cf = strain in the fiber
c = strain in the matrix
oaf = normal stress parallel to fiber, in the fiber
Ef = modulus of elasticity of the fiber
oam = axial normal stress parallel to fiber, in the matrix
E = modulus of elasticity of the matrix
Due to the large ratio of the modulus of elasticity of the fiber to that of the matrix, it is apparent that nearly
all of the axial normal stress in the fiber-matrix composite is carried by the fiber. Exact values are
(Reference 6):
oaf = oL / [| + (1-|)Em/Ef]
oam = oL / [|Em/Ef + (1-|)]
Where:
oL = nominal longitudinal stress across composite
| = glass content by volume
Chapter 6 Technical Discussions 73
The continuity equations for the glass-matrix composite seem less complex for normal stresses
perpendicular to the fibers, since the weak point of the material seems to be limited by the glass-free
cross-section shown in the following figure.
For this reason, it would appear that the strength of the composite would be equal to that of the matrix for
stresses in this direction; in fact, its strength is less than that of the matrix due to stress intensification in
the matrix caused by the irregular stress distribution in the vicinity of the stiffer glass. (Since the
elongation over distance D1 must be equal to that over the longer distance D2, the strain, and thus the stress
at location D1 must exceed that at D2 by the ratio D2/D1.) Maximum intensified transverse normal stresses
in the composite are:
( )
1.25
2
2 2
1 ( ) /(1 )
(1 0.85 )[1 (2 3 )1 ( )(1 )]
m f m
m f m
E E V
E E V
|
(
o | + |
(
o =
+ | | t
Where:
ob = intensified normal stress transfer to the fiber, in the composite
o^ = nominal transverse normal stress across composite
V= Poissons ratio of the matrix
Note: Because of the Poisson effect, this stress produces an additional s'