Would it not be prudent to get the facts of the criticism and present these to Algor? Theres plenty of rumour and gossip in the computer industry to destabilise a product.
I have used Algor's Pipepak since the early 80's and would be interested in any criticism. They do provide ebnchmark and verifying examples with their software.
I would question someone's criticism of a well-known product like Algor. I have used it since the DOS days and always found it to be reliable. The DOS linear version was based on the well-known SAP IV program introduced years ago by Ed Wilson at UCB. I heard rumors that this code sometimes produced erroneous results, but I never found that to be true. It was an easy code in which you could incorrectly define a model mesh due to lack of graphics and the load case multipliers were somewhat difficult to apply. In my opinion, Algor brought over some of those complications, but the graphics helped considerably to spot problem areas. I think the Windows version of Algor is much easier to use and the user has considerably less chance to make an error. FEA is a lot like flying...there is a better chance of pilot error than any other cause.
That said, general purpose FEA software is so powerful and complex that there is always the potential for a bug in a seldom used option or problem type. I would ask for specifics from the naysayers.
One persons unreliability is another’s challenge. There are several aspects to unreliability.
Program crashes or errors
==========================
Generally FEA packages are very complex and have a relatively small user base. This means that not all combinations of options can be tested either by the vendor or by beta testing. Hence in my view FEA programs tend to be relatively unreliable for certain users, those that push the capabilities of the system while others have no problem. This is true even for expensive high-end systems; look up the bug reports for any major system and it will probably run to tens or even hundreds of potential problems.
Numerical errors
================
Push any system too far and it will fall over: elements will distort, non-convergence etc. In my experience some systems are better than others in that the defined out of tolerance warning and error messages are set wider than others. That one system will carry on where another will fail is not necessarily good.
Modelling Errors
=================
Basically user induced - read the manual.
It is best to ask 'some people' what they mean by un relaability and to give examples. I think it would be fool hardy to expect the same level of reliability from and FEA package as from say a word processor (even a Microsoft one). What is probably more important is the companies attitude to correctng any problems and user support.
As afar as I am concerned I have had tremendous support from Algor for the Pipepak module. In fact being based in Australia and they in the USA I need quick responses because of the time difference. They have web broadcasts of their product applications that are useful for distance learning.
The product is benchmarked and verified against nuclear regulatory authority models.
This morning I viewed an FEA linear model verification aginst a NAFEMS model.CHeck out the website if you are in doubt. Send some questions in about verified models.
My understanding of FEA is that the key is in the boundary conditions that you set up not the maths engine. So ease of use is a big factor in choice of product. There is so much investment in time in an FEA product that users become fiercely loyal for fear of having to change. So anyone using one product and criticising another may just be trying to justify to themselves why they made the decision in the first place.
Its like buying cars in some ways. Who's going to own up to buying a lemon? But how many people criticise the other car they didnt buy or couldnt afford?
If you are concerned about Algor's error, simply build a model using the elements you intend to use in your work and solve a closed form solution problem. Compare the results and you will know firsthand if Algor is a reliable package.
This is how all the "major League" packages prove their software and you can too.
![[pc2]](/data/assets/smilies/pc2.gif "[pc2] [pc2]")