Cx The American Society of ‘Mechanical Engineers Raprintad From PVP — Vol. 169 Design and Analysis of Piping and Components Editors: Q. N. Truong, EC, Goodling, JR. J.J. Balaschak, and G. E. 0. Widera Book No, H00484 — 1989 THE ART OF CHECKING PIPE STRESS COMPUTER PROGRAMS: Liang-Chuan Peng Peng Engineering Houston, Texas ABSTRACT With the computer getting more and more soph- isticated, the chance of getting a bug in a program er a misapplication fn an analysts also becomes more and more likely. Analyste necd aome rules of thumb to quickly spot problem areas and to make a auick check if necessary, This paper outlines some of the generai rules used in checking boundary con- ditions, anbalanced forces, and irregularities. it aiso uses specific examples to demonstrate tae checking of some elementary functions. Special dis- cussions are given on advanced features such as Support friction, thermal bowing, and expansion bellow elements. INTRODUCTION With the uew requirements given on the design of a modern plant piping, the only practical toot for the design analysis Is the computer. ‘The computer program: designed for pipe stress analysis gets more aad more sophisticated every day. Some programs have gone through several generations of develop~ ment employing completely differem background of personne), ‘The ney genpéation normally will aot touch the good work done by their predecessors. instead, they make layers of shelis around the exis ting work. The completed program becomes very Glsorganized. Therefore, it is safe ta say that a modern pipe stress computer program is bound to have Somp inconsistoncies. Pipe stress analysts are normally too timid sn challenging a well established compnter program However, if we recognize that to err i6 computer program, we mey be able to more objecuvely ensure the quality of our analysis, It is important to real- ize that everylhing has its se called norm, Tp other words, if something Jocks unrealistic then it probably n i8 unreal. Therefore, it is important to be able to ook at the oatpet and point out the irregularities that might exist, That is the art. From time to time we have sean some experienced engineers who are able to judge whethor 4 system is satisfactory just by Joking st the model, The computer analysis is just a comfirming check. However, they are the exceptional rather than the normal. The inconsistent resis in an analysis comes either from the bug in the program or from the mis- applicslion of the program, Nowadays, people Like ito boast that you don't even need to read tue manual to use their compuier program, The so called asor friendly is probably wbat they intended to say, but eomebow ths impression they give is avi. You type in some data, then you get some results. it sounds easy, bot is scary. ‘To ensure a good analysis the analyst has to have gl leusi a clear picture of what the program functions are, He or she should also be able to epot the inconsistencies when they occur. PROGRAM VERIFICATION A program in systematically verified before beiag released for production. The verifiestion invo~ wes alsost every slep of the program's operation and function. ‘The resaiis of the verification are do- cumentes in the verification reports. This is the function of the program developer and should not be a burden to the asers Verification by the user is occasionally required by the inhouse QA procedure, or to simply satisfy the curiosity of the user or the boss. To an analyst, to be able to personally verify a couple of analyses will definitely increase his or her cowlidence is the program, ‘The mosi common approach of the veriZi- cation is lo encek against known results. ‘The book by Kellogg Company [1] contains quite a ew band calculation results which can be checked against theexpansion stress calculation. A more formal. calou~ lution intended to be a benchmark was published by ASME (2] in 1072, Unfortunately, this benchmerk problem contains some sisprints, which bave never deen corrected, and also the anusual non-cireular cross section elemente. Recause of these difficulties the problem bes created © huge frustration in the piping industry. Everywhere, engineers are trying to make ® comparison in vain, Later in 1960 U.S, NRC pablished » get of representative piping bench~ mark problems [3]. This set of problems was taken from real systems laidout in nuclear power plants. Tt ie mainly used to check the earthquake annlysis using the response spectra method, ‘The benchmark problems check only the general dehaviors of the program. The general behavior of a given program differs very little from the original black box oa whic most of the programm are bused, ‘Therefore, very little deviation stall be expected from these tesis, The most important items to be concerned with are the ones particular to individual programs. These items need to be checked very discretely. DEVIATION In comparing the test results with publisued or benchmars results, ile relative deviation is used. ‘The term error is not used because ine difference might be caused hy the error of the published or so called known results. Bven the so called exact solu~ tion might have some seomingly insignificant terme ignored. However, if the deviation in small then there i a good chance that both the testing progrem and the benchmark are correct, This is more so when then testing program uses au entirely different solution technique thun that used by the bencumark, Im oveluating the deviation, come common sense has to be applied to avoid unnecessary argumente, Yor given quantity, R, wiose exact solution ie shwon in Figure 1 (a). Tks corresponding result, R', from the test program may be auified to as shown in Figure 1 0b), ‘Then by some methods of evaluation, y y Ry R fa} () Figure 1, Standard Deviation it _may be concluded thut there is no comparison &t ali, Because ihe deviation is essentially infinite on component Ry. But we all kaow that the real differ- fence between the two solutions is very small. This ean be easily proved because if we rotate the axes by 45 degrees, the deviation will almost disappear a completely, The point is that a number is meaning- lose if ite quantity is entizely dependent of the sel- evtion of the coordinate axes, Therefore, it is xm- portant to have the devistion properly defined a2 follows : dev (Ry) > (RyRy )/ Ry (Meaningless) dev (Ry) * ( Ry'-Ry )/R (Local) dey (Ry) = ¢ Ry!-Ry }/ Ro (Global) Where H is the resultant quantity at the poiat of interest, and No is the maximum regultant quantity in the entire system analyzed, The global deviation is introduced, because at a given point the resultant quantity itself may be Insignificant. Whether it is Significant or not, the too} to measure is the global comparison, ‘The evaluation of the local deviation requires some personal judgement, but the globsl deviation should be limited to about 19 pereent, BOUNDARY CONDITIONS: ‘The first step In quick checking sn analysia is ta make sure that the resalts, mateh the boundary Conditions of the aystema. Tais can be dane easily with the hely of « good oulpul arrangement. Most compuicr programs lave a operate regort for the anchor end support ferces and momeats as shown in Table 1, Yor this particular one (4) the friction and the pipe displacoments are also given. This makes the checking of the boundary condition very eaxy. By using reports such an Table 1, the boundary conditions can be checked divectly by looking ai the pipe displacements, At an anchor point the pipe Aispiacement should be the same as tho inpst displ~ acement, and at ihe Limit stop location the pipe Gisplacerent all be equal @ or smaller than the gop specified, However, i should be noted that the Support displacement specified in the input is for the support stracture, ‘The actual pipe displacement at that point may or may not be the same as the cup= port copending on the cigisity of the support. If the Support is rigid then the pipe and the structure will have the same displacement, But if the support is flexible then the pipe displacement and the support movement are different sa shown in Figure 2. Spring Constant, K Supyort Force = K (Dp - De) Pipe Displacement, Dp (output) Support Displacement, Ds (input? Figure 2, Sapport and Pipe DinplacementaJe 1, Anchor and support Loads Rrra, FENG EINEERIME - SMPLEX- RELA.09 GNEI-L.SD FASE SPLE PICSLEM, 198 ASHETONE rH SOSRENCE SHIEION FACTOR 4,4 0h ALL YEAICAL SUPIETE case We wt peas tomes i HS, FR, FC, PRES ‘4 AOMOR 3ND SUPBRE FREES = MIRUDMG FAETION CACTIAG OM BUFPDED ose ' ‘PFN FORCE an REA aT wT ane FREES CE > we or oo OR tener con RROD = TO sores oa ee WS 3 HL "ess. ‘oie AERIEN TANT, TORSION AEUNE_LOSTOL a a wt 8 kw cE eos mba wy 8 ime ew stow mm be wk myo 0 ot it ee wom 8 ew hr TI my 0m ew i m8 Tk a ves ea atts meow Ww eh ea Sars (moinL PERN TANT. TRS CCIE oA, sm bat St vi wk mom bem a a pean UOEEEEGHUERELASUSRELORECALLSLESEEASLRAEOACSOTAOIFAAECURUOAOEL IDSC SAAEEADEAOL AREACODE ser routes 7) tures te 6 say