RISA 3D - Limits on Deflection Analysis 1

[dcmiles]

Hi, I am a fairly new Structural EIT (<4 months on the job) at a small design firm in Ontario, Canada. I am the only engineer in the office (my superior works remotely) and I am using a different FEA software as RISA has a connection to Revit -- which is what the designers here use.

I like RISA 3D for the most part but I am frustrated by the deflection analysis that it is capable of. As far as I can tell, it will only give you an absolute deflection diagram and NO relative deflection diagram considering the support conditions. The latter seems like the most relevant for code design while absolute deflection is useful contextually.

I know you can adjust the support conditions (cantilever and supported ends) in which the program will divide the beam into 'spans' and calculate the deflection ratios of each span, displayed in the Member Deflections spreadsheet. This only seems to work, though, when the spans are broken up by vertical support members (columns).

In our projects -- namely unique high-end boathouses and cottages -- we often have conditions where long spans are hung onto the ends of cantilevered beams, and column point loads will strike from the top. RISA 3D ends up dividing up these beams at the 'vertical supports', which are actually point loads, and gives inappropriate deflections ratios.

I'm wondering if other users have come across problems like this and how they managed to get around it. RISA support literally told me to use a different software until they 'improved the software's behaviour', but if someone had found any kind of workaround which would prevent me from having to learn yet another software I would be VERY grateful for the wisdom. Deflection usually governs our designs and I have been floundering trying to make working designs with this (what I would consider) a HUGE design flaw of the software. Moment capacity, shear capacity, and deflection criteria are the holy trinity of structural design and I never in a million years would have thought that a major software would be lacking in any one of these :S

Signed,

Halp!

 

[jayrod12]

Could you post a model that is exhibiting the behaviour you speak of? Perhaps with some description of what you're expecting versus what the program is giving you?

I've never had much of an issue gleaning the information important to me from Risa. I've actually found it generally more user friendly than the other softwares.

I do understand the desire to have the risa/revit link. However I would caution using a program that you're the only designer in the office doing it. Without your mentor having access to the same program and having some experience in the program, it makes it tough for them to review your work or understand the issues you're having. Therefore they may not be able to help you when you need it.

 

[dcmiles]

Hi @jayrod12

I've posted a file that I have been working on which illustrates the problems I have been having.

Take FB104 first. Both of the ends are cantilevered and the beam itself is supported from below on two columns -- support conditions set to 'cantilevered both ends'. The program correctly divides this beam into three spans and hand calculations verify that the deflection ratios in the Member Deflections spreadsheet are correct.

Now take FB102 -- an eave beam which is supported by the I-ends of FB104 and FB105. Support conditions set to "cantilever J-end", though because there are no vertical supports the program still shows just one span and incorrect deflection ratios.

Lastly take FB103. It has a roughly 12' cantilever on one end and is hung on the cantilevered ends of both FB108 and FB114. It also has two column point loads striking it from above, neither is in line with the supporting cantilever beams. FB103 is set to "cantilever I-end". It is broken into three spans, however these spans are defined by the points at which the point loads from above strike it.

As far as I can tell there is no way to manually adjust the program's limited intelligence.

This seems fundamental, and I can't believe that there isn't a way to work around this in such a popular software.

 

[dold]

Is your problem simply that it's not calculating the deflection ratio 'correctly'? If it's giving you such heartburn you can just look at the node displacement by hovering your mouse over the node of interest while looking at the deflected model. Then just calculate the ratio yourself. Software is a tool, not an engineer. You need to be able to interpret the results instead of scrolling through results spreadsheets and making sure all the numbers aren't red. I'm guessing that's why you're so hung up on the deflection ratios?

 

[jayrod12]

I'm inclined to agree with Dold. Although in that same breath, I understand that you're not wanting the overall displacement taking into account the deflection of the supporting member, however often the cumulative deflections are actually the thing that kills you the most even if the beam itself passes the individual member deflection ratios.

When I switched member FB102 back to default for end support conditions, the deflection ratio makes much more sense. L/280 at 4.7m which appears to match the deflected shape diagram.

 

[JLNJ]

Don't forget to turn off frame softening when looking at these types of deflections.

All but the most simple of projects will have these sorts of complications. Knowing how to look for them and resolving them is how you show your engineering chops.

 

[dcmiles]

@dold

What you have been suggesting is how I have been going through it until now. It isn't 100% accurate since it requires some 'eyeballing' and only shows those values where there is a node.

I'm not quite sure why you put the word 'correctly' in quotations -- isn't there a correct and incorrect way to determine a deflection ratio? Especially when I am working with large cantilevers with point loads at the end it is helpful to have a quantified value to point to rather than a ballpark figure.

Software is a tool, and I have seen softwares (S-Frame for example) which make the analysis that I am looking for very easy, clear, and quantified.

 

[dcmiles]

@JLNJ

I can see that is going to be the case! As I'm developing those chops there is a period where I will probably overthink everything until I gain better intuition at the outset of a project/problem. I am doing my best to be thorough and cover all my bases until I'm confident that I'm not overlooking anything.

As for frame softening - are you speaking in terms of concrete frames?

 

[dold]

Well, I can tell you pretty confidently that I've never made or used a 3d model in any software that is 100% accurate. There are definitely features that could improve a lot of software I use but you just have to learn how to use the tools as they are. I've been using risa for maybe 7 years or so and I'm still learning new stuff every now and then. 4 months is a pretty short time to expect to be able to fully utilize any piece of software.

I'm not sure how accurate you're trying to be, but I suspect that you're trying to be too accurate, or precise. There's a difference between "ballpark numbers" and "wrong". If I was working with that model I would happily "ballpark" deflections of the offending members to within 1/4" or so and call it a day. If I were designing this sort of complicated multi-cantilevered system I'd probably just ignore the program calculated deflection ratios for everything except simple span beams anyway since it's not entirely clear to me what it's doing, as you have discovered. It's pretty darn easy to just look at the difference in deflection between 2 nodes and divide by a member's length, or whatever portion of it. To say that this issue is a HUGE design flaw seems to be a bit of an over reaction. Plus I'm a risa fanboy so I'm a bit biased.

You're asking for advice on how to work around this and the method I described is how I do it. You can always add a node at any location along a member if you want to investigate deflection at a specific point.

FYI your snow loads are defined as live loads in basic load case spreadsheet.

 

[Aesur]

I have noticed in my use of RISA that global deflections are shown including the deflection of the supporting member. I typically look at the detail report of the member to get a better idea of local member deflection, however as jayrod12 said, the global deflection is still a concern that should be thought about. Another location that I believe you can see local deflections is in the "Member Deflections" report.

Just a quick comment on your model, I would take another look at your unbraced lengths "Lb y-y, z-z and Lcomp top and bottom". You have for FB124 Lby-y set to 1 m, however the distance between intersecting beams is 6.833m, I believe the correct settings for this member is Lbyy = distance between intersecting beams, 6.833m and Lcomp top = 1m (assuming it's braced by the decking). I suggest going to the help menu and searching for unbraced lengths, this gives some very handy diagrams for setting the unbraced lengths in RISA.

Another note, what is your lateral system? You have various global pinned locations in the model for stability. Every member appears to be pin-pin with no moment frames or fixed bases, did you add these global pinned nodes for stability? If so you should discuss this with your project manager and this doesn't appear to be correct, however please note that my knowledge of this project and it's lateral system is fairly limited.

 

[JoshPlumSE]

Okay, it's been about 5 years since I really used RISA. But, I did work there for 16 years and was manager of the tech support group during most of that time. And, this was a reasonably common tech support question.

Presuming that they haven't changed this aspect of the program much in the last few years, my response was typically along the lines of the following:

Basic concept of member deflections:
RISA reports relative deflection for a member, not absolute. Take a joist that is supported by a girder on each end. If one girder deflects 0.25" on one end and the other girder deflects 0.35" on the other end, then the reported member deflections will be based on a the deflection from a straight line between the two deflected supports.

Where it starts to break down in RISA:
For the most part, RISA is NOT smart enough to fully understand member framing. The program looks at the member deflection as a whole and tries to interpret whether it is supported on both ends or whether it's only supported on one end. If it's supported on both ends, then it's member deflections are calculated as I already suggested above.

If, however, it's deflected shape looks more like a cantilever beam, then RISA will report it's relative deflection considered the deflection from the supported end. So, it will treat it as a cantilever.

The problem is that there are many cases where the RISA algorithm makes a bad choice and the program thinks that it's a cantilever when it's not. This is especially true for physical members that have multiple spans. I suspect that this is what's happening for your model.

Other considerations:
When you use RISAFloor, the program has an inherent understanding of where a beams support points are. So, that program can do a much better job with member deflections. I believe that may even be the case for a RISAFloor model that was brought into RISA-3D. Though I'm not 100% sure about that.

Normally, I would tell you to model your beams as single members spanning between supports. So, you might divide what was a single physical member into three separate members. Left Cantilever to left support. Left support to right support. Right support to right cantilever tip.

That being said, the Revit interaction may not like that type of division and may prefer the single cantilever member. So, you may be kind of stuck and have to do the member relative deflection calculations by hand.


Additionally, you probably should check in with their tech support group. the company has had nearly a complete turnover in staff over the last 5 years, but I still know two people over there (Andrew and Rachelle). They're both very capable and should be able to address your questions more completely than I have.... especially concerning what happens with Revit integration.

 

[dcmiles]

@dold

That all sounds reasonable -- I can back off somewhat with my level of precision. I have a problem of overthinking things. Up here we tend to use the snow load as the 'live' load rather than applying them both concurrently to the exterior which is considered overkill. This is how I've been guided by several local engineers.

 

[dcmiles]

@JoshPlumSE

Thanks for the insight. I'll tinker with what you suggested and keep using the workarounds like @dold mentioned.

Y'all are showing me I was a bit too frustrated earlier and I can relax a bit and make this work.

Thanks everyone!

 

[dold]

Up here we tend to use the snow load as the 'live' load rather than applying them both concurrently to the exterior which is considered overkill. This is how I've been guided by several local engineers.

Hmm. I don't really know anything about Canadian codes, but that's not really how load combinations work. There can certainly be live loads and snow loads acting concurrently, and the intent of load combos is to essentially assign probablities to each event (building will most likely not be loaded to 100% design live load at the same time you have 100% of the design snow load). For example the combo below (NBCC 2 (b)) assigns a 1.5 factor to LL, but a 1.0 factor to SL. So this is actually overly conservative, and in combo (NBCC 3 (b)) it's the opposite, so you might end up with unconservative results. It all depends on the magnitude of the loads.

This would also apply to deflection/service conditions, but it looks like Canadian combos are slightly different than ASCE combos when considering LL along with SL. That said, I would always always always define them separately.

But I can see how you might arrive at the same result if you just lump everything in as live load, since the LL and SL factors are the same (but reversed) in a couple of combos. But, I'd be careful doing this when you start considering wind and seismic loads as the factors are likely different, and that will affect results such as uplift for example.

I can back off somewhat with my level of precision. I have a problem of overthinking things.

Once you get a little more experience under your belt you'll start to get a feel for which values/results you can "ballpark" and which values/results might be of greater importance and warrant extra precision.

 

[MotorCity]

I assume the OP meant to say that snow load and ROOF live load are not applied simultaneously which is correct for US codes. Snow load and FLOOR live load could occur simultaneously.

 

[dcmiles]

@dold

That said, I would always always always define them separately.

Yes, you're right on this one. Good to start proper habits from the beginning.

 

[jayrod12]

When I worked for a smaller firm a province over, we would lump snow and live together. When the old code was enforced, it was more conservative to account for it all acting concurrently as it was 1.5S and 0.5L (and vice versa). New codes have 1.5S and 1.0L (and vice versa) and so the difference between the load combos is less. Although still more conservative to lump them together so you essentially have 1.5S and 1.5L.