stiffness matrix solution of unkown eqns

huangyhg

stiffness matrix solution of unkown eqns.
well the situation is this: i am working on making a computer programme of stiffness matrix. i have actually made a complete programme of continous beam, 2d trusses and am starting to work on frames. i followed the method described in "matrix analysis of framed structure" by weaver and gere. made the full structural matrix (irrespective of restrains), did the shuffling in order to cater the restrains, and then exrtacted the matrix ,from this reshuffled one ,reqired to calc deformation and so on. the inverse was calculated by gausians elimination method.
   i know there are other mothods like cholskey modified sq. root method and skyline method to compute the unkowns. lets keep skyline aside for a moment.
    i went through cholesky`s method. in the end i came across the chart comparing the number of processes invlove in modified cholesky vs compact gaussian elimination and to my surpirse the total number of processes invloved were same. this means that both the processes are same, if they are measured in terms of no. of processes. is it true? actually ive been advised by a peer that i should revise my programme and base it on atleast cholesky if not skyline. according to him the process that ive followed is going to take a lot of time during processing large number of frames. is there huge difference? can you explain?
hi, rednyx
i am very interested in the work you describe.
a basic interpreter called mathserv (which runs in any word processor) has these commands: mattranspose(), matmult() and matinverse(), which are useful in solving continuous beams, 2d trusses, and frames.
can we pursue this further by email?
warmest regards, helmut
hi rednyx,
i am very much interested in the work you are doing. actually even i have the ambition of doing such programs.i was thinking of writing a program on moment distribution for any sructure.please send me some details about how must i start.
bye varun
dear varun n engcomp
   its nice to hear ur interest in particular field.
actually i have coded in vb for the inverse , transpose and stuff. and i did made excel sheet for moment dist (with limited facilities). any help that i can be of ,in these fields, i am willing to do so.
    i was hoping that some one will tell me about the pros of using choleskey or sky line over normal conventional method.
hi rednyx,
read ur email.thanks for replying.right now i am in the final year of civil engineering and i do have matrix method of structural analysis as a subject for which i refer the book by weaver and gere.i know c and java languages. i wanted to know which language would be suitable for moment distribution method.also i am interested in the work u are doing.please send me the code of your programs that u have done so that i can learn the way i can approach my task.
bye,
varun
albeit i have done limited fem work in mathcad i am not competent enough  to help you. i suggest you posting this same question as well in the fem forum that exists in this same site to gain further concurrence.
this question is already floating in fem forum ..yet unanswered uptill now.
varun: my sugesstion for moment distribtion is that u make a sheet in excel and generalize it to ur desire. you can make macros to cater spans,cycles, sway etc. it is very tidy the way excel present the calculated sheet. i wont suggest java as programming tool for eng calculation as its main aim is to improve the web base solutions. c is a very powerful language, but y make ur life complicated when you can do the same in excel with much ease.
dear varun
the matrix method of structural analysis is fine.
the moment distribution method is from the old days when solving more than three simultaneous equations was a pain and a stepped approach was more convenient. hardy cross has our admiration but the method is as dead as the sliderule.
the days are gone now that matrix analysis is part of most computer programs.
please have a look at mathserv and email your questions or solutions.
regards, helmut
i agree with engcomp, whats all this trouble using excel to do moment distribution when you have matrix methods right at hand?!  moment distribution is good to know, don't get me wrong, but it is not what programs of today's genre are made of.
incidentally, check out bathe's finite element method for more on the numerical methods used to facilitate the fea solution.