Need a book? Engineering books recommendations...

Return to index: [Subject] [Thread] [Date] [Author]

SAP2000, RISA3D, and STA

[Subject Prev][Subject Next][Thread Prev][Thread Next]
Greg Luth writes:

. > My advice is to treat all results from analysis programs as wrong until 
. > proven right.


Well said, Greg!

I am amazed at how engineers accept the output of commercial structural 
analysis/design programs when they don't even know how the computer handles 
the problem, whether it is a legitimate FEM analysis or whether the 
programmer decided to just generate random numbers.

My first introduction to computers was when the IBM 650 was *the* computer.  
If I recall correctly, its memory was a drum with storage for about *1850* 
"words."  To speed up access, IBM increased the read heads from one to four 
so that the drum only needed to make a quarter turn for memory access.  Later 
magnetic core memory was installed which really *speeded* up access.  My 
first programming was in FORTRAN when there was Basic FORTRAN and Full 
FORTRAN.

As a graduate student taking a Plates and Shells course in which the 
classical analysis was studied (infinite series), I recognized that the 
computer could perform this analysis much easier and faster than I, using a 
slide rule, could and proceeded to write a FORTRAN program.  In each term of 
the infinite series there was -1 raised to the nth power, a very simple 
mental calculation in that the term alternated between negative and positive 
depending on whether the exponent was odd or even.  As such, I programmed 
that as (-1)**N.  When I got the program debugged and the output back and 
compared it to Timoshenko's solution, I noticed a slight difference.  After 
much consternation, I remembered that we were told that the compiler, when it 
saw **, took the logarithm of the number, multiplied it by the exponent and 
then took the antilogarithm of the result.  However, the logarithm of a 
negative number does not exist, and I don't know what the computer did with 
it.  However, when I changed the program to alternate between negative and 
positive, the results compared favorably with Timoshenko.

>From that time on, I have been very skeptical of computer programs that do 
not provide the source code in readable form, where the programmer and 
his/her qualifications are not given, and, for structural analysis and design 
programs, where the structural engineer responsible for the program does not 
seal the listing and the users' manual.  After all, it is the person writing 
the program that is doing the analysis and/or design, not the person 
inputting data.  (Any clerk can input data from a coding sheet.)

A. Roger Turk, P.E.(Structural)
Tucson, Arizona