Mechanical Engineering News For the Powe: ‘The COADE Mechanical Engincering News Bulletin x published periodically from the COADE offices in Houston, Texas. The ulletin is intended to provide information about software applications and development for Mechanical Engineers serving the power, petachemical, and related industries. Addliionally,the Bulletin will serve as theofficial notification vehicle for software errors discoveredin those Mechanical Engineering programs offered by COADE. (Please note, this bulletin is published only two to Lee times peryear As ofthis ine, the neue number will appear in the banner block atthe top ofthe frst page.) CAESAR I Version 3.19 has recently shipped to users in out effort to maintsin COADE as the leader in the Feld of piping engineering software, we have made over 60 modifications to the rogram, Several new festures. suchas the missing mass correction (p.8),the B31G code (p. 27), and virtual memory support (p.2) are siscussed in det inthis bulletin. Several useful feature, added to Version 3.19 too late 19 document in the release notes, are Aescribed on page 3. {In adion to our usual technical anticles, we also are using this bulletin to discuss recent developments at COADE, as well as 0 Provide a proview of the enhancements we have “inthe works” for ‘our piping and vessel software, We continue to weleome any articles hy users of our software for inclusion in future issues of| Mechanical Engineering News. New Features of CAESAR II Version 3.19 * Dynamics modules provide Missing Mass and Seis mie Anchor movements. B3IG criteria for corroded pipelines added. RCC-M C piping code added. High Resolution Animated Graphics. All dynamics modules are now in 32 bit “protected” ‘mode, yielding substantial speed increases. ‘Thermal bowing loads can be analyzed. Multiple analysis via Batch Stream Processing. Additional expansion joint and spring hanger manu- facturers data bases added, * French and Spanish language support. * Additional output reports, Petrochemical and Related Industries COADE, Inc Venn evar TABLE OF CONTENTS PC Hardware forthe Engineering User (Part 16) CAFSAR II Version 3.19 Undocumented Features ‘What's new at COADE. COADE Bulletin Board os CAESAR I Quality Assurance Manual ‘AO? Pressure Vessel Code Updates CAESAR II Development Schedule Fe‘Pipe Version 27 Announcement. COADE Seminars Technology You Can Use CAESAR IL, Use ofthe Missing Mass Correction Option in Spectral ANALYSIS... innsnnne 8 ‘The Role of Stress Analysis in the Life Extension ‘of Piping Systems - Evaluation of Creep Stresses. 18 ‘The ASME B3I-G Criteria ame ‘The CAESAR I Expansion Joint Mode 00... 29 Expansion Case for Temperatutes Below Ambient... 32 CAESAR U Specifications... u PC Hardware for the Engineering User (Part 16) ‘As many users are aware, CAESAR II is being converted {rom 16bitoperationto 32 bitoperation. Therearetwomain advantages to this conversion: frst, the 32 bit modules can ize the extended memory found on the computer, and second, there ia significant speed increase due othe 32 bit addressing. This conversion of the CAESAR IL source code involves, switching from the Microsoft FORTRAN 5.1 compiler to the WATCOM FORTRAN 9.0 compiler. The WATCOM compiler utilizes the 32 bit DOS extender from Rational Systems. Am item of importance to CAESAR I users isthe act that this DOS extender has the potential to support “virtual memory”. By activating this Viral memory op- tion, the computer system can use the hard disk as main ‘memory, just like RAM, only slower. This gives all CAESAR TH users the ability o analyze a system requiringCOADE Mechanical Engineering News May, 1993 CAESAR II, Use of the Missing Mass Correction Option in Spectral Analysis During spectral (either seismic or force spectrum) analysis, the response of a system under a dynamic loadis 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, reac- tion, and stress results may require extraction of far more modes (possibly far into the rigid range) before acceptable accuracy is attained. CAESAR Il permits the user to specify — cither through a ‘mode number cutoff ora frequency cutoff — the number of ‘modal responses to be included inthe system results. Ifthe analysts more interested in providing an accurate represen- tation of the system displacements, it may only be necessary to request the extraction of a few modes, allowing a rapid calculation time. However, if an accurate estimate of the forces, stresses, etc. in the system is the objective, calcula tion time grows as it becomes necessary to extract far more ‘modes. This is particularly true in the case when solving a fluid hammer problem in the presence of axial restraints; ‘often modes with natural frequencies of up to 300 Hz must be extracted. How does the analyst know how many modes are sufficient? ‘The usual procedure is o extract a certain number of modes and review the results; then to repeat the analysis while extracting 5 to 10 additional modes, comparing the new results othe old. Ifthere isa significant change between the results, anew analysis smade, again extracting 5 to 10 more modes above those that were extracted for the second analysis. This iterative process continues until the results taper off, becoming asymptotic, as shown in Figure 1. a ane attanet Figure 1 ~ Maximum Stress vs, Extracted Modes ‘This procedure has two drawbacks, the firstone obvious — the time involved in making the multiple analyses, as well as the time involved in extracting the potentially large number of modes. The second drawback is less obvious — ‘a degree of conservatism s introduced when combining the contributions of the higher order modes. Possible modal summation methods include SRSS, ABSOLUTE, and GROUP—all methods that combine modal results as same- sign (positive) values. In reality, theory states thatthe rigid ‘modes actually act in phase with each other, and should therefore be combined algebraically, thus permitting the response of some rigid modes to cancel the effect of other rigid modes (this is exactly what occurs in a time history analysis). Because of this conservatism, it is actually possible to get results which exceed twice the applied load, despite the fact that the Dynamic Load Factor (DLF) of an impulse load cannot be greater than 2.0. With Version 3.19, CAESAR II has introduced a feature, called the "Missing Mass Correction”, which helps solve these problems. Tis feature offers the ability to include @ ‘correction which represents the contribution of the higher ‘order modes not explicitly extracted for the modal/dynamic response, thereby providing greater accuracy with reduced calculation time. When the option is activated (by entering YES for the appropriate dynamic control parameter) the program automaticaly calculates the net (in-phase) contri- bution of all non-extracted modes and combines it with the ‘modal contributions — avoiding the long ealeulation time associated with the extraction of the high order modes and the excessive conservatism of the summation methods. This article explains this feature in detail, describing the mathematics behind it, providing instructions on its use, demonstrating its effect through an example problem, and finally discussing the Included Mass Repor, a new report added to the Dynamic Output processor. Missing Mass Theory ‘The dynamic response of a linear multi-degree-of-freedom system, such asa piping system, is described by the follow sng equation: My()+CZ0+Kx()=FO Where: M_ =nxnmass matrix of system © =nxn damping matrix of system K =n xnstiffness matrix of systemCOADE Mechanical Engineering News May, 1993 (0 = nx L, time-dependent acceleration vector (0 =n 1, time-dependent velocity vector x(Q) =m x I, time-dependent displacement vector FQ) =n. |, time-dependent applied force vector ‘Assuming harmonic motion and neglecting damping, the free vibration eigenvalue problem for this system is: Ko -Mow Where: @ =nxnmode shape matrix 2 = nx mmatrix where each diagonal entry isthe frequency squared of the corresponding mode ‘The modal matrix @ may be normalized such that TM =1 (where lis then x nidentity matrix)and®" K@= 2, Each modal response represents the response of a single degree-of-freedomoscillato, withits own natural frequency (oj) and share ofthe total system mass, and acted upon by a share of the total external laad. Fach mode's share of the total system mass and external load (exclusive of DLF) are calculated as: mjy=MO, 07M §=MO, OTF Where: ‘matrix of modal mass (systemmass active in mode) for mode i ‘@; = mass normalized mode shape for mode i vector of modal force (force exciting the mode) for mode i This can be verified by summing the modal masses and forces overall modes, which shows that they equal the total system mass and extemal force, respectively Em =M 0 M=@™M@M=IM=M Efj=MOoTF= OTM @F= Rarely are all modes extracted during a spectral analysis; therefore the mass and applied force actually considered “during the spectral analysis can rarely be identical tothe true system mass and applied force. Furthermore, as differing ‘numbers of modes are extracted forthe analysis, the amount of mass and force included in the solution will vary. The accuracy of the solution will likewise vary as modes with large modal masses and forces are included or omitted, ‘The matrix of all system modes ® may be partitioned into ‘two submatrices — those which are extracted for the analy- sis and those which are not: P= [Ge %] Where: @, = mode shapes extracted for dynamic analysis (i... lowest frequency modes) , = residual (non-extracted) mode shapes (correspond ing to the “missing mass” contribution) ‘The system mass is equal to the sum of the modal masses allocated to the extracted modes plus the sum of the modal ‘masses allocated to the residual modes. Likewise, the applied force is equal to the sum of the applied forces allocated to the extracted modes plus the sum ofthe applied forces allocated to the residual modes: M=Mg+M,=M®, OTM +M0,0,7M FeFe+Fr=M, 0,0 F+MO, 0,0 F [deally, the effective mass included in the extracted modes should be as close to 100% as possible, indicating that as ‘much of the modal response as possible hasbeen considered. Effective mass is calculated as: cffective mass = Me /M Normally, modal analysis completely neglects the rigid response — the response of the non-extracted modes, due to the load F,. However, since M, e, and F are all known, Fr ‘can be easily calculated, without actually extracting the high-order modes, as: Fp=F-FesF-M@_@_1F ‘This residual force can then be applied statically t0 the structure, with the resulting displacements, forces, stresses, and reactions representing the net response of all rigid modes:COADE Mechanical Engineering News May, 1993 Kx= (LFF, Where: x = maximum displacements of residual (non-ex- tracted) modes. DLF= dynamic load factor fornon-extracted modes (maxi- mum DLF for any frequency above the highest ‘extracted frequency) ‘This responses then added o he responses the extracted, low-order modes in order to getthe total system response — effectively including 100% of the system mass and 100% of the external force while obviating the need to extract a large number of modes. Note that although this procedure is called a “missing mass” ‘correction, it might possibly be better called a “mis: force” correction, since it is the force acting on the rigid ‘modes which is added back into the solution. The term “missing mass” isused since this correction has traditionally ‘beenused for improving seismic analyses, where the applied load isan inertial load —i.e., MA. In this case, the applied missing force is the missing mass (i., the difference be- tween the total system mass and the sum of the mass associated with all ofthe extracted modes) times the ground acceleration, ot ZPA: Where: €, = transient creep strain, infin a, b= constants, dependent upon material properties, stress, lemperature, etc. t= Toad duration, her ‘Therefore, the transient creep strain rate (de/at), as a fune- tion of time, is: defdt = alt + (6/3) Iti clear that this approaches zero as time increases. There- fore, this component does not affect the life extension of a system, ‘The third portion of the curve, viscous creep, is linear with respect to time, and is expressed simply as: , = viscous creep strain, infin ‘constant, dependent upon material properties, stress, temperature et. Toad duration, hr At high temperatures, transient creep is often negligible ‘compared to the viscous creep; therefore the later is often called “hot creep". This component of the creep curve ‘determines the deformation ofthe loaded system; failure can bbe designated as occurring when the viscous creep strain ‘exceeds that point where system deformations prevent proper functionality ‘The fourth region ofthe curve shows an accelerating rate of, deformation, followed by rupture of the material. The accelerating rate of deformation is fictitious, since this is actually the result of a constant load being applied to a reducing cross-section. Constant stress is shown to produce no accelerated creep rate, but to still lead to rupture — therefore the path to rupture is simply an extension of the viscous creep curve. Obviously rupture represents system failure as well — this point might be reached before the system deformations grow large enough to cause loss of function, ‘Therefore, there are two potential failure modes due to creep excessive deformations, or catastrophic rupture. Ineither case, fora given material, temperature, stress level, etc, the “hot creep” damage rate is linearly dependent on tt In order for this information to be very useful, it must be possible to express the total creep strain in terms of the other parameters (material, temperature, stress level) as well. Viscous eeep isa thermal reaction, soitsrate is governed by Boltzmann's expression forthe rate of thermal reactions: deat = ce" Where: de/at = mean viscous creep strain rate, inin/hr © = constant, proportional to applied stress = fo" f= constant, dependent upon material, temperature, structural features, ete. = applied tri-axial stress, psi ‘n= constant, dependent upon material, temperature, structural features, etc ‘A. = activation energy for creep process, electron- volts = approximately 1 eV = 3.826E-20 calories k = Boltzmann’s constant = 1.832E-24 calorie"R T = temperature, ° Rankine 9COADE Mechanical Engineering News ‘This demonstrates thatthe viscous creep rate is constant as Tongas the stress and temperature are constant, sototal creep strain can be defined as the product of the rate and the time: fore Where: = time, he Due to the fact that the constants f and n are actually functions of the temperature, stress, ec. this formula is too complex to be of much use in calculating an exact erecp strain value. Instead, it can be fit to existing test data by ‘considering the relationship ofthe parameters, oat atime, while the third is held constant. First, considering the case where the creep strain is normalized and the stress level is held constant: 10= get Where: 2 = constant, based upon material, stress level, etc. Rearranging this equation, taking the log (tothe base 10) of each side, and multiplying each side by T: 0.434 Ark = Tog g + 10g) A simplified version ofthis expression, which clearly dem- onstrates the relationship between time and temperature, was first postulated by F. R. Larson and James Miller in 1982: P=TB+logt) Where: P_ =constant (known as the Larson-Miller param- een), dependent upon total creep strain (€,), stress level (6), structural features, et. B= constant, dependent upon material total creep strain, ses level, structural features, et. (usu- ally taken as 20 for steel) [Next, when the creep strains normalized and the time (load duration) is held constant: 1.0 =hote** May, 1993 Where: hh =-constant, based upon material, load duration, et. Rearranging this equation as before: 0.434 A/k = Tog h + 010g 0) Q=TC+l0g0) Where Q = constant, dependent upon total ereep strain.) structural features, et. c constant, dependent upon material, total creep strain, structural features, etc. Finally, considering when the temperature is held constant: iort Where: J = constant, based upon material, temperature, etc. ‘The constants B, C, P,Q, j, and n can be determined from these relationships if experimental creep data i available, yielding @ generalized equation linking any combination of viscous creep.) stress(o,), temperature (T,)andtime (i For example, four viscous creeplstresstemperaturetime combinations defining two temperature lines are available, as shown for 2-1/4% chromium, 1% molybdenum stel in Figure 2: Point A—e,.,,T,,t, PointB—€,,0,,T,.t, Point C—e,.6,.T,.t, PointD —e,.6,.T;.t, Figure 2 20COADE Mechanical Engineering News May, 1993 Using the relationship for constant time at both time, and, two points (Point —e, 0, T, t, PointF—e, 6, 'T, t.) defining the equation for temperature T, can be found: Q=T,C,+lo0,)=7, (C, +1080) (C,+1og 0.) 21, C,+1og0,)=7,(C, + log0,)=7,(C,+ 10g). 50: og = tl MTT 086, -(WE,IM,L0 9 HIT IT) y= 1! 47176 090) Since the curve for the desired temperature line is now defined, vscouscreep straincan be calculated forthe desired stress and time using the relationship for constant tempera- ture: (enn) oer” 30259 (17, 17,)log 6, + (1,1) og a) (UT, 1.) UT ese 4e5) FWloas logs, HVT, 1 KlogT0g@,) ‘This formula shows the compl of the creep phenom- enon. Its accuracy can be verified by calculating the ex- pected creep (for the steel shown in Figure 2) for a tempera- ture of 1 100°F (15600R), a stress level of 8300 psi, and aload S,-6,and 0,2 S, -$,, oF: 0,=0,6,/S, foro, $2429h0 demonstrat thatthe aystcu cayricines uly 32,429 hours at an equivalent of 10SO°F per every 11.4 years of operation. Since the use of the code stress allowables implies a creep life of 100,000 hours, this would correspond toa design life of 11.4 x 100,000 / 52,429, of 21.2 years Review of the actual vs. allowable stress level: Itmay be thatthe system has not only operated uninterrupted for its emtre life, butit may have operated atthe anticipated temperature as well. In this case, the last remaining possi- bility is to determine whether the design level can be extended based upon stress levels falling below the allow- able. Itmay be that a re-evaluation ofthe operating circumstances ‘may show that the original stress estimates were not realistic — design pressure may far exceed actual operating pres- ‘sures, spring hangers may have been installed with different pre-loads than intended, or as-built deviations may have led to a significantly different sustained stress distribution, However, the most likely source of extended life due to reduced stress is found inthe difference between the actual sustained stress and the code allowable — itis rare that 100% of the allowable is used in a piping system. For example, the Maximum 3-D Stress Intensity for the system shown in Figure 3, is $419 psi, found at nodes 10 and 15: 25