Are High End FEA Software Packages Worth The Investment? 6

[chris9]

I've been struggling with this question ever since I started using FEA over 4 years ago.

I appreciate that in some cases it might be worth spending £40,000 on an FEA package for specialist types of analysis such as the forging process or crash simulation etc.

But why spend this amount of money on a high end FEA package when a mid-range FEA package for about £10,000 or less will do the same job in the vast majority of circumstances.

I don't think there is much of a difference in accuracy if you compare the different software types. The judgement of the person driving the software is far more important.

I can't think of a reason to buy a high end FEA software package. Can anyone else?

 

[corus]

Why bother even buying a FEA package at all? Engineers used to just manage with a slide rule and their bridges are still standing.

corus

 

[ThomasH]

I don't understand the question. If you don't se any reason to but expensive software, then don't do it.

I have to agree with corus on this one.

Thomas

 

[johnhors]

In many cases, many FEA users will draw the following conclusions:-

1). You should purchase software that matches your requirements.

2). Why pay for exotic analysis options if you're never going to use them?

However, I know of instances where sub-contractors have their choice of software enforced upon them as a requirement for getting the contract in the first place.

Also, unfortunately, certain industries have de facto "standard sofware" solutions by which FEA users are compelled to use specific codes (and very often not the best!)

 

[nagi]

I believe that if you don't need any of the advanced features of high end FEA software, and you customer does not dictate the choice of FEA software then you might be better off with a cheaper FE software.

But I would be careful before concluding that I don't need the advanced features. Let's say I need shells and plasticity in small deformation problems. Just because a cheaper FE program claims that they have these features doesn't mean that they are implemented there in a general and reliable way.

 

[GregLocock]

If there were two companies competing for a job, and one had a high end FE program, and the other had a cheap FE program, but had spent the difference on establishing a correlation facility, then I'd always give the job to the second one.

The problem is that in the overall scheme of things the price of the FE package is not all that significant - any improvement in robustness, portability, speed or flexibility will save enough time to justify the difference in purchase cost.

Let's see, take a single seat $50000 package. Interest is about 5k, maintenance is about 7k, analyst's time is about 80k, overheads for employing analyst is say 40k (ha). Note that implies a breakeven charge out rate of around $60 ph - rather low in my experience.

Oh, so I can save roughly 10% by going to a non-industry standard solution? No thanks

That being said, if it is all in-house and you are sure that in the short term the cheaper package is sufficient then you'd be silly to go for the expensive one.



Cheers

Greg Locock

 

[chris9]

I've been a user of ANSYS structural, IDEAS Simulation, Abaqus, Patran and Pro/Mechanica.

They all do the job and they all have their quirks. I think johnhors has hit the nail on the head. I think it is mainly driven by companies enforcing a particular software packages onto suppliers and subcontractors.

I don't think there is much difference in functionality between the major software packages.

 

[sean01]

Chris9,

I would suggest that the decision on what software to use should in part be based on what your customers are using and also what service you are providing....

Some companies I work with flatly refuse to use you as a supplier unless you are capable of taking 'native code'. This attitude, while being a pain in the @*&" forces you to adopt the license they are using. The trick more often than not is in getting either enough work from them - or getting them to pay a fair hourly rate for it !

However, if your customer doesn't specify the software you should use, you are free to offer the best tool (in your opinion) for the job. Such an approach however, shouldn't be based purely on commercial issues (such as license cost recovery) but whether it can do the job - whether all your customers will accept using this solver output etc.

For me personally, the actual solver used in the analysis is incidental. I have taken a number of contracts this year where an F.E.A. study has been initially done - the results reacted to only for there to be a problem later. In some instances this has been caused by bad design - but also by bad modelling and analysis. The point of digital testing (ie F.E.A.) is accurate identification of a products performance without incurring the cost and leadtimes of manufacturing prototype parts. If you accept this, you understand that it's the quality and detail of the model build (pre-processing) and its susequent analysis - not only in terms of understanding of principles behind performance but also the direction of effort to ensure the product meets the required performance. The solver is only part of the equation that gets you this result. To make sure you get it right, you need good people who not only understand the tools used - but can also prove and correlate their conclusions as Greg has already indicated.

Sean

 

[ThomasH]

Chris9

I'm a bit curious as to the reason for your question.

What kind of software do you use today, high-end (expensive) or mid/low - end (less expensive)?

Are you satisfied with this software?
Is it worth the money and does it do the job?
Is it reliable (in your opinion)?

I think that if you want to suggest that buying high-end (expensive) software is "stupid" (and I don't think you want to do that) you have missed something.

So, I'm curious why you ask.

Regards

Thomas

 

[chris9]

Thomas,

The reason I ask this question is that I will be buying some software this year. I wanted ANSY structural but can't afford it.

Now I'm looking at Cosmos DesignStar, which is a fraction of the cost, and I'll bet it can do the majority of what ANSYS structural can.

Cosmos DesignStar costs about a third of the price of ANSYS Structural. I don't know why there is such a huge price difference and thought maybe someone could enlighten me.

 

[magicme]

hello

i have to agree with sean01....it depends on what your customer wants. because my big customer uses ansys, i need to license ansys (at ridiculous annual fees!) though i would love to outright purchase nastran rather than license ansys every year.

of course if it brings in a few 1000 hour jobs and keeps the customer happy.... buy what you need, add $10 to your bid rate and generate that cash flow.

daveleo

 

[ThomasH]

Chris9

I'm by no means an expert in neither Cosmos DesignStar nor Ansys. But a quick glance on their sites gave me perhaps(?) some insights.

In my humble opinion they do not compare.
Ansys is a more or less (there are always limitations) complete FEA-tool with extensive analysis capabilities and a more or less complete element library.
On Cosmos I can only see that the analysis capabilities seem "OK" but the element library seem very limited. The information that I found is very limited but I don't know if this is due to limatations in the software or due to something else.

If I were to buy new software I would not (based om this information) further evaluate Cosmos.

So, It's very much depending on what YOU do on a daily basis and if Cosmos can solve YOUR needs.

But based on what I found on the web I'm not surprised there is a difference in price.

Good Luck

Thomas

 

[chris9]

I don't think I'll ever need a software package with a huge element library.

I've probably only ever used about twenty element types in four years.

I've decided I'm going to evaluate ANSYS Designspace, NE/Nastran, Cosmos DesignStar and maybe Strand.

If the customer wants it in ANSYS then I'll just lease it.

 

[izax1]

We use both low-end and high-end SW. (DesignSpace, I-DEAS, MSC/NASTRAN, ABAQUS, FEDEM (which I guess none of you have heard about)) If you are only to perform linear, static analysis on structural parts, almost any SW will do. But if you need to perform non-linear analysis, dynamics (or even non-linear dynamics) analysis of cyclic structures, composites etc. I think you need more than COSMOS or MECHANICA. What about the non-linearities in bolted connections? Finally, engineers have realised that the world is non-linear and dynamic (not linear and static), and today we have the computer power to simulate it. Yes, I agree. You can solve a lot of problems assuming linear static behaviour. But the technology is there to squeeze the rest out of it. And it will cost, yes. But that is what it takes to be up front.

And, anyway, what is the cost of FE analysis on a total development project? I think it is worth it.

 

[chris9]

I have worked in the automotive industry and white goods. In my experience everyone wants the results yesterday. They don't want to wait a minute longer than necessary or pay a penny more than they can get away with.

Every time I suggest more advanced types of analysis I find there is not enough time or money in the budget. I guess the aerospace and nuclear industry will be more willing to invest and wait a little longer to get more accurate results.

 

[imagitec]

Chris9,

Based on your requirements, I think COSMOS/M is the package you should look at; perhaps it's what you meant, rather than DesignStar. COSMOS/M is a full blown FEA package, roughly equivalent to ANSYS. Models can be created from a variety of 1D, 2D, and 3D elements, and it has the greatest range of analysis capabilities of any product in the COSMOS family. My understanding (from following the COSMOS list at

http://www.topica.com/lists/cosmosfea)

is that it is significantly less expensive than ANSYS. Of course, the cost/benefit analysis will be different for every potential user.

CosmosWorks (CW) is COSMOS' package that is tightly integrated with SolidWorks. It can only create solid and shell meshes, and the two may not be mixed in the same analysis. Also, meshes may not be manually created. If the automatic meshing fails, the analysis can not be done. CW's analysis capabilities are also the most limited; for example, it doesn't support temperature dependent convection coefficients.

DS is a standalone application. Like CW, it generates meshes from solid models and surfaces. I think it has some manual mesh editing tools and the ability to mix solids and shells. Unlike CW, it is CAD-agnostic: it can work and maintain associativity with geometry that's created natively or imported from various CAD packages. It also has more analysis capabilities; for example, it can use temperature dependent convection coefficients.

I hope this has helped.

Rob Campbell

http://www.livejournal.com/users/robcampbell