8/12/2019 C StructuralForum Powell Nov08

1/3STRUCTURE magazine November 2008

discuss

ionon

top

icso

fcurrent

importance

tostruc

tura

leng

ineers

STRUCTU

RALFORU

M

50

Structural Forum is intended to stimulate thoughtful dialogue and debate among structural engineers and other participants in thedesign and construction process. Any opinions expressed in Structural Forum are those of the author(s) and do not necessarily reflect theviews of NCSEA, CASE, SEI, C3Ink, or the STRUCTURE

magazine Editorial Board.

Structural Analysis: Are We Relying Too Much on Computers?Part 1: The ProblemBy Graham H. Powell, Ph.D., P.E.

I have noted with interest some recent editorials in STRUCTURE magazine on the subjectAre We Relying Too Much onComputers(February 2007 and August 2007), and also the article Basic Education for a Structural Engineer(April 2007). Idisagree mildly with the editorials. I disagree strongly with the Course Content in the Basic Educationarticle.

I also consider a number of second-

ary points, in particular (a) the role ofstructural analysis and (b) the coordinationof courses in structural analysis withcourses in structural design.

Developing a Feel forStructural Behavior

Once upon time it was necessary toanalyze structures by hand (I graduatedin 1958, so I was trained in that era). Anumber of useful, and occasionallyelegant, analysis methods were developed,

the most important of which was MomentDistribution. It is often argued that whenengineers used Moment Distribution theydeveloped a feel for structural behavior,but that when they use a computer thisfeel is lost. I strongly disagree with thisargument. I firmly believe that if you givea computer program to a student, and ifyou teach him or her to use that programproperly (a big if, which I address later),then that student can develop a betterfeel for structural behavior in a couple ofsemesters than he or she could develop in

a lifetime of using Moment Distribution.The following is a list of some struc-

tural analysis exercises. I believe that byexamining and explaining the analysisresults for exercises of this type, a stu-dent can quickly develop a feel forstructural behavior.

1) Analyze a truss structure as asimple truss (assuming pinnedconnections) and as a frame(accounting for continuousmembers and estimatingconnection stiffnesses). Do this for

trusses with slender and not-so-slender members, and comparethe results. Discuss whether thebending moments can be ignoredfor design.

2) For a simple cantilever shear wall,calculate the moments, shears anddeflections both including andignoring shear deformation. Vary

the wall aspect ratio. Compare the

deflected shapes for a slender wall(mainly bending deformation)

with that for a squat wall (mainlyshear deformation). Do the samefor a coupled shear wall. Vary thecoupling beam span to see howit affects the results. Both includeand ignore shear deformations inthe walls and coupling beams. Fora cantilever wall and a coupled

wall, estimate the rotationalstiffness of the foundation, and see

whether foundation deformation

has much effect on the behavior.3) Apply support settlements on a

statically determinate structureand a similar, but staticallyindeterminate, structure. Examinethe moments, shears, reactions anddeflections. Explain the results. Dothe same with thermal expansionloads. Use this example to considethe advantages and disadvantagesof redundancy, not in terms of thenumber of unknowns for analysisby the force method, but as a

physical concept.4) For a reinforced concrete frame,consider different methods toestimate the bending stiffnessesof the beams and columns. Seehow different stiffness estimatesaffect the calculated momentsand deflections. Use this exampleto explain how cracking affectsstiffness. Emphasize that there is alot of uncertainty in modeling foranalysis, and that analysis resultsare at best approximate.

5) Analyze a multi-story frame anda shear wall separately for lateralload, and note that they havedifferent deflected shapes. Thencouple them together and explain

what is happening (one way is toconnect them with stiff bars, andlook at the forces in the bars).

I do agree that problems can arise when

computers are used for structural analysis.However, the computer is not at fault.The computer is merely a tool, and likeall tools it must be used in a craftsman-like fashion. In my opinion, the problemlies not with the computer but with thecraftsmanship. In this article I explain

why I believe this is the case. In a follow-up article I propose a solution.In this article, I argue two main points,

as follows:1) Older engineers often complain

that young engineers lack a

feel for structural behavior.For example, a young engineermay get an analysis result thatis obviously wrong, yet maynot realize that there is an error,and may even believe that theresult is accurate to 6 or moresignificant figures. A viewpointthat is often expressed is thathand calculation methods, sucha Moment Distribution, help anengineer to develop a better feel,and that computer analysis does

not. I emphatically disagree withthis viewpoint. In my opinion theopposite is the case computeranalysis can give a young engineera much better feel for structuralbehavior than hand calculation.

2) In structural analysis there are threephases, which I will identify astheModeling, Number Crunchingand Interpretationphases. Inmy opinion, the modeling andinterpretation phases are by far themost important, and should be

given by far the most attention. Asfar as I can discern, the structuralanalysis courses at most universitiesdevote almost all of their attentionto the number crunching phase,and very little to modeling andinterpretation. This does notprovide students with the skillsthat they really need.

8/12/2019 C StructuralForum Powell Nov08

2/3STRUCTURE magazine November 200851

6) Analyze a steel frame with semi-rigidconnections. Vary the rotationalstiffness of the connections and see theeffects. Estimate the stiffness for anactual bolted connection.

7) Analyze a frame with stiff end zones

(or rigid offsets) in the connectionregions. Examine the effect on theframe deflections. Compare the beambending moments and shear forcesat the column faces (which would beused to design the beams) and at the

joint center-lines.8) Analyze a steel frame with and

without deformable connectionpanel zones at the beam-to-columnconnections. Calculate how much ofthe frame deflection originates in thepanel zones. Explain how loads are

transmitted through the connectionand why the shear stress in a panelzone is much larger than in the websof the adjacent beams and columns.

For these exercises, the analyses are all easyto do with a computer but would requirean impossible amount of time by hand (notto mention a high level of expertise in handcalculation methods). Using a computer, astudent can quickly build the model and runthe analyses, see the structure in 3D, varythe dimensions, loads and other properties,and display the results graphically. Most

importantly, a student can focus on themodeling and interpretation, and not getbogged down in number crunching. I believethat a student can learn much more aboutstructural behavior by completing exercisesof this type than by learning how to analyzestructures by hand, and can also have morefun doing it.It is important to note that to gain benefit

from such exercises it is not enough just to runthe analyses it is essential to understand andexplain the behavior. The computational de-tails of the analysis are relatively unimportant.

The important things arewhat procedures and as-sumptions are used to setup the analysis model, and

what the analysis resultsmean in terms of structuralbehavior and performance.

What isModeling?

It may seem obvious,but it is worth keepingin mind that we analyze amodel of a structure, notthe structure itself, andthat the behavior of themodel may or may not be

close to the behavior of the actual structure.The challenge is to create a model that isaccurate enough for practical purposes.To create a meaningful model an engineer

must (1) understand the behavior of the com-ponents that make up the structure, and (2)

know how to capture the important aspectsof this behavior in an analysis model. This isnot a simple task. It requires an understandingof such things as axial and shear forces, bend-ing and torsional moments, beam and columnbehavior, load transfer through connections,connection deformations, composite action,cracking, yield, bolt and bond slip, buckling,and many other aspects, including the differ-ence between actual and design loads. It thenrequires decisions on how (or whether) tomodel different aspects of behavior, given thecapabilities of the available analysis methods

(or computer programs). Modeling is especiallydifficult for nonlinear analysis, because of themany types and causes of nonlinear behavior.It is also more difficult for dynamic loads.An important skill for an analyst is the abilityto create useful models of real structures (not

just simple two-dimensional frames of the typethat are often considered in analysis courses). Auseful model must capture the behavior of thestructure with sufficient accuracy for designpurposes, it must produce results that areaccurate enough for making design decisions,and it must not be so large or complex that it

takes too long to analyze. A model does nothave to be, and never can be, exact.

What is Interpretation?There are two parts to interpretation, namely

(1) checking the analysis results to make surethat they are reasonable, and (2) organizingthe results in a form that supports decisionmaking for design.Checking analysis results requires a number

of skills. There are several tools that can beused, including free bodies and equilibriumequations, and methods for checking dis-

placement compatibility. Common sense a feel for structural behavior are indispeable. Experience is always valuable.Organizing the results requires an und

standing of how analysis results are used assessing performance and making desdecisions. Almost always, design decisionsmade using strength and/or deformation mands, corresponding capacities, and heDemand/Capacity ratios. In the analysisa frame structure, the deflected shape a

the bending moment diagram are useful checking the analysis results. They are particularly useful for assessing performaor making design decisions.

What is Number CrunchingThe number crunching phase is everyth

that is not included in the modeling ainterpretation phases. It includes the numercomputations and the underlying theory.

Do We Have a Problem?

I believe that we do have a problem, and tit is a serious one.For an engineer who uses structural analy

the most important phases are modeland analysis. These are tasks that requengineering skills and judgment, and not (yet) be done by computer. The numcrunching phase is not unimportant. Howe

with modern computer programs, this phcan be handled entirely by the computer. engineer needs a basic understanding of theory (i.e., the Direct Stiffness Method) the computational procedure, but for the mpart the analysis details are not importantThe problem, I believe, is that young engin

are not being taught the skills required for meling and analysis. In most structural analcourses, the emphasis is overwhelmingly the number crunching phase, even to extent that students are being taught M

ment Distribution (which I find horrifying)a consequence, young engineers have to lethe really important skills on the job. At the vleast, this is an inefficient use of resources.In my opinion, we spend far too m

time teaching unimportant (and relativuninteresting) skills, and far too little tteaching the skills that are important (amore interesting).This is the main problem. There are a

some secondary ones.

The computer is merely a tool, andlike all tools it must be used in a

craftsman-like fashion.

continued on next p

8/12/2019 C StructuralForum Powell Nov08

3/3STRUCTURE magazine November 200852

The Purpose ofStructural Analysis

In structural analysis courses, students caneasily get the impression that analysis is an endin itself. This is rarely, if ever, the case. The endresult for the analysis of a building frame is notthe deflected shape and the bending momentdiagram. Usually, the required result consistsof the strength and deformation demands on

the structure and its components, so that adesigner can assess the performance of thestructure by calculating Demand/Capacityratios. When students present the results ofa structural analysis, they should considerhow to organize those results in a way thatsupports decision making for design. This israrely done in structural analysis courses.

Coordination withDesign Courses

Design courses frequently make use ofstructural analysis, but often the analysis

methods are different from those that aretaught in analysis classes. Students can getthe impression that analysis and design areunrelated disciplines. It is important, if onlyin the interests of efficiency, that the coursematerial in analysis be coordinated with thatin design.

AccuracyStudents can easily get the impression that

structural analysis results are somehow exact.This relates to the common complaint is thatinexperienced engineers often present analysisresults to six-figure accuracy or more. Even

worse, students may be led to believe thatstructural analysis can predict the behavior ofa structure exactly. Researchers who shouldknow better often suggest that advanced

structural analysis can accurately simulateand predict structural behavior, accountingfor all types of nonlinearity and all types ofload. The truth is that structural analysis isalmost always very approximate, and for realstructures accurate simulation is impossible.Even if this were not the case, over-relianceon analysis can lead to design by analysis,

which is a very bad idea.Students should be taught that for real

structures there is a great deal of uncertainty,in the properties of the structural components,the support conditions, the loads, and the

contributions of nonstructural components.Students should not be under the illusionthat analysis results are exact.

ConclusionI have seen it argued that back in the days

when engineers did structural analysis by

Graham Powell is Professor Emeritus ofStructural Engineering at the Universityof California at Berkeley. Over his career,Professor Powell has developed, or supervisedthe development of, several computer programsthat have been used widely in research andindustry. He currently works as a consultantto Computers and Structures Inc., developingand supporting the CSI PERFORM computer

program. Professor Powell may be reached atgraham.powell@sbcglobal.net.

hand, they developed a better feel fostructural behavior than engineers who uscomputers. In my opinion this is, or shoulbe, incorrect. With proper instruction, aengineer can get a much better feeling fostructural behavior by running computeanalyses and examining the results rathethan by doing repetitive and boring hancalculations. Using a computer, an engineecan analyze complex 3D structures, quicklexplore the effects of different modelin

assumptions, and consider dynamic loadand nonlinear effects. None of this is possibl

when hand calculation is used.The key point is that engineers are no

being educated to use computers effectively (deliberately use educated here, not trainedbecause this is an academically demandinprocess). Engineers are also not being educateto think critically about computer results. If aengineer believes that a computer analysis modeis exact, or that it is the only possible modeor that the results are accurate to 6 significanfigures, that engineer has received a poo

education. An engineer has also been poorleducated if he or she thinks that structuraanalysis is an end in itself, rather than just a wato get useful information for design.The problem, in my opinion, is that structura

analysis, as it is presently taught in most of ouuniversities, incorrectly emphasizes the number crunching aspect of analysis, and greatlunderestimates the importance of modelinand interpretation. Modeling and interpretation are not simple tasks. They can be compleeven for structures that have static loads anlinear behavior, and they can be very comple

when dynamic loads and nonlinear behavioare considered. Modeling and interpretatioare academically demanding processes tharequire an understanding of structural behavior, the modeling capabilities of computeprograms, and the needs of designers. Typicauniversity courses in structural analysis providvery little instruction in these important areasThe problem is not with the computer, it

with the professor.In the second part of this article, in a

upcoming issue, I propose a solution.