Structural Analysis Format

J by Jesusbill on 14 Oct 2020, edited 25 Aug 2022 # +6 votes

A new open format for structural analysis has come to my attention, thanks to Jan @brewsky, initiated from the Nemetschek Group - Structural Analysis Format - SAF.

It is based on the Excel format and it is supported by various vendors at the moment (SCIA, Graphisoft, Allplan, RISA, FRILO, Strusoft, Axis VM, Dlubal, Sofistik, SCAD and LIRA land).

Also, it looks like buildingSmart is approving its development; see the post by Léon van Berlo in LinkedIn and the comments below where he states:

I think everyone will agree that the structural analysis part of IFC is not properly worked out. Improving it would be a collaborative effort between all the Structural Analysis vendors. Logically the result will be the same as SAF, since that is the result of consensus and collaboration between the important stakeholders. The most pragmatic standpoint is to use and adopt SAF and figure out later how it can be used by vendors that prefer to exchange IFC.

Looking at the site, I can definitely see that the documentation is more clear and concise, without room for interpretation, the schema is based on the ifc structural schema with modifications and enriched or new entities and it does look from a quick view to handle better various issues related to structural analysis. For example I saw a dedicated "rigid" element that does not explicitly exist in IFC (only implicitly for line elements with eccentricity); a few fundamental and specific types of geometry for each type of structural element, division between connections and supports.

Tomorrow there is the monthly IFC development meeting on the topic of structural analysis and a presentation on IFC and SAF workflows from Herman Oogink from SCIA who is coordinating the development of SAF. I will be participating so as to get more info on the potential of this new format. Meanwhile, I continue working on porting IFC structural to Code_Aster.

J by Jesusbill on 14 Oct 2020 #

Here is a figure comparing a specific element between SAF and IFC, posted by Akos Rechtorisz

L by Luis on 14 Oct 2020 # +1 votes

Thanks for the info @Jesusbill.

The documentation looks good.

R by ReD_CoDE on 14 Oct 2020 #

@Jesusbill is the schema available on Excell? If yes, could you please share the Excel-based version of SAF and/or SAV?

J by Jesusbill on 14 Oct 2020 # +1 votes

@ReD_CoDE the site I indicated above contains documentation and for each documented type and page there is an online version of excel with an example that you can download. See for example here

I do not have at the moment any specific example in Excel to share

M by Moult on 14 Oct 2020 #

Fantastic @Jesusbill ! Please let us know your thoughts on what this means for the future of structural model translation.

R by ReD_CoDE on 14 Oct 2020 #

Also, Thomas @aothms is familiar with this, it's Excel-based STEP which Thomas has worked on it at https://github.com/stepcode

R by ReD_CoDE on 14 Oct 2020 #

The idea is simple, step file is nothing more than some sets or arrays (or nested arrays) that can read and write its values even manually, or though Excel (tables, tuples) or through a machine automatically :)

A by aothms on 15 Oct 2020 # +1 votes

Interesting, good to have seen this in advance. It looks like the schema is profoundly differently (or extended at least) comparing e.g structural curve member:

https://saf.guide/Default.htm#A_Objects/7_StructuralCurveMember.htm%3FTocPath%3DStructural%2520analysis%2520elements%7C_____6

https://standards.buildingsmart.org/IFC/RELEASE/IFC4_1/FINAL/HTML/schema/ifcstructuralanalysisdomain/lexical/ifcstructuralcurvemember.htm

@ReD_CoDE step file is nothing more than some sets or arrays (or nested arrays)

Nested arrays is a key point here though, isn't this where excel will begin to struggle? I'm not actually very familiar with STEP in Excel.

In the case of SAF it also appears the references are more symbolic, like CS3; N9; N10; for cross section and node respectively, as they need to point to a sheet and and a row. Could have been done with generic express references, but harder to read I guess.

R by ReD_CoDE on 15 Oct 2020 #

@aothms said: Nested arrays is a key point here though, isn't this where excel will begin to struggle? I'm not actually very familiar with STEP in Excel.

It would be hard to define the structure on Excel that everything works well, but it's possible, the existing example is: NIST STEP File Analyzer and Viewer (https://www.nist.gov/services-resources/software/step-file-analyzer-and-viewer)

But there's a trick, you just need to put data inside that nasty array, something like Solving Sudoku rows:

#N= IFCPERSON('Ehsan Azari','Azari','Ehsan',$,$,$,#N,#N);

| :-----------------: | :-----------------: | :----------------: | :------------------: | :---------------: | :----------------: | :--------: | :---------------: |

| Ehsan Azari | Azari | Ehsan | $ | $ | $ | #N | #N |

R by ReD_CoDE on 15 Oct 2020 #

Also, there's BNF (TBNF, EBNF) and WSN

For instance, the whole IFC TYPEs in EXPRESS are just these patterns (it's just preliminary):


TYPE , END_TYPE, FIXED, WHERE, AND , OR, ;, [], (), {}

Name,  > , >= , = , <= , < , Number

Type, Type(Number), Value

[Number:?] OF , [Number:Number] OF

Value, (Value1, ...)

WR+Number : , Name : // WR21:

SELF, SELF[Number], ABS(), SIZEOF() 

IN [Array] 

{Number <= SELF <= Number}

J by Jesusbill on 18 Oct 2020 # +1 votes

A few words about the monthly IFC development meeting held last Thursday on the Structural Analysis Format.

Excel was selected as the feedback of a lot of engineers was that they want to be able to view and modify the model in Excel, as it is a tool that they are well acquainted with and use it daily in their work;
SCIA started and is leading this initiative but the intention is to pass the management to a neutral organization, hence the (possible?) involvement of buildingSMART;
There is a good number of structural software vendors that have implemented, are currently implementing, or are planning to implement support for this format;
Current version is 1.0.5 released on 14/04/2020 while the next release will be version 2.0 towards the end of the year (if I understood well);
No current intention to include results from finite element analyses but there is intention to include key design results for elements based on analysis results.

I am glad to see that there is a good adoption from various software vendors, possibly attributed to the fact that the format is more clear and easier to consider with respect to IFC, which at the moment has a low level of adoption. I have the feeling that it will be relatively straightforward to include this format in the current implementation of ifc2ca. I think that when version 2.0 will be released we will be close to a good level of implementation of the IFC schema, and that will be a good time for the integration of SAF.

J by JanF on 18 Oct 2020 #

So if I understand this correctly, it is only a standard shared between structural analysis software, so it has nothing to do with interdisciplinary collaboration.

Do you guys have experience with using ifc models for structural analysis? Our structural engineers are still unhappy with the results and prefer to remodel from 2d dwgs. Could you give me some tips on how to make it work better? (We're architects)

R by ReD_CoDE on 18 Oct 2020 #

@JanF IFC has a lot of limitations in many domains, so these kinds of movement are good to improve parts of the IFC, but building such a hilarious format would be just good for structural engineers

J by Jesusbill on 18 Oct 2020 # +1 votes

@JanF said:

So if I understand this correctly, it is only a standard shared between structural analysis software, so it has nothing to do with interdisciplinary collaboration.

Yes, absolutely. But the issue is not, in my view, the lack of an interdisciplinary collaboration standard but the lack of tools that implement interdisciplinary collaboration workflows. Since the geometry of a structural engineer's analysis model (analytical geometry - lines for beams/columns, faces for slabs/walls) is different than the real geometry (3D), the engineer has to redefine/reconstruct the geometry of his model. Surely, a 3D geometry can be helpful but it does not alleviate the need to define everything, and often 2D drawings can be useful for this task, especially plan views. On this matter, the autoconverter tool that was also presented during the call and uses SAF as the final export file schema is certainly an important step forward as it converts automatically (also through a number of user-defined options) a structural 3D openBIM model of the structure into a valid structural analysis analytical model. I can't deny that this would be the way to go for the future development of ifc2ca as well, but requires quite a lot of development.

Do you guys have experience with using ifc models for structural analysis? Our structural engineers are still unhappy with the results and prefer to remodel from 2d dwgs.

The ifc2ca project, currently under development, is about using the structural analysis parts of the IFC schema and defining the analytical structural model, to be then analyzed with the open-source code Code_Aster. In theory, structural software should be able to import structural ifc models but the adoption is very low at the moment.

Could you give me some tips on how to make it work better? (We're architects)

It's difficult to give specific advice as it depends on the way the engineer works and the tools he uses. I would advise you to discuss with them how you can help them in the specific workflow they have. Anything that can help reducing the time needed to construct the analytical model is a big plus, but it is difficult to achieve this, especially if there is only a 3D architectural model (so no indication effectively of the structural elements) and even for the case of a 3D structural model, as obtaining a valid structural analysis model is not straightforward.

B by bernd on 19 Oct 2020 # +1 votes

How do we parse this file format? Is there anything available? As FreeCAD is able to read geometry of IFC architectureal and analysis view, i would like to implement this format as well.

How does it work? Does it only save the analysis geometry, means a face with a thickness and a line with a crosssection? How about the mesh? Is it possible to save the mesh as well? How about solid analysis and solid mesh? Is it possible to save such geometry? Structural IFC is not capable of saving solid analysis geometry.

cheers bernd

B by brunopostle on 19 Oct 2020 #

@bernd said:

How do we parse this file format? Is there anything available?

An XLSX file is just a ZIP archive containing XML data, so there are python modules for accessing the files.

J by Jesusbill on 19 Oct 2020 #

Yes @bernd openpyxl referenced by @brunopostle should do the work.

The format is similar to structural ifc regarding the elements and geometries supported, so no solid geometries, no meshes, only line and faces with section properties and thickness. In my first post there is the link to the documentation, which is pretty clear, you can take a look for the features.

J by jtm2020hyo on 27 Oct 2020 #

@Jesusbill said:

Here is a figure comparing a specific element between SAF and IFC, posted by Akos Rechtorisz

any format similar to CSV is too easy to understand, I wish all BIM formats were too simple to understand like this.

J by Jesusbill on 27 Oct 2020 #

@jtm2020hyo can definitely see your point but not all data can be nicely represented in a table and I question the use of a single cell of a string with ";"s to represent lists in this case. For my taste the json format is the best

B by brunopostle on 27 Oct 2020 # +2 votes

@Jesusbill said:

I question the use of a single cell of a string with ";"s to represent lists in this case. For my taste the json format is the best

Not using proper references loses much of the advantage of a spreadsheet. This would make it very hard (for example) to calculate member lengths needed to generate a bill of quantities.

J by JanF on 5 Nov 2020 #

So to sum up what I gathered from trying to make the digital cooperation between architecture (A) and strucutral engineerig (ES) happen, there are some options and the issues:

A manages the whole model, model shared as ifc - ineffective, ES builds its model from scratch, everything is done twice
A manages the whole model, model shared as saf - maybe better, ES only corrects the model, but most software doesn't know saf and the whole communication clarity from ifc/bcf exchange is gone (unless saf becomes a part of ifc)
A manages the whole model except the loadbearing constructions, model shared as ifc - ES software does not so often export ifc?

What are your opinions on 3.? Would it be an ok way for ES? It certainly takes a portion of freedom from A, but also brings more clarity (would be interesting to know which concrete walls are actually load bearing) also moves theoretically a portion of work from A to E (coordination of openings between MEP and ES)

Can Code_Aster export ifc?

T by theoryshaw on 5 Nov 2020 # +3 votes

I think the ultimate goal is to emulate software development as much as possible.

That is the 'content' (the code) is not managed or controlled by any party, but through standards and distributed version tracking, all parties have access and can modify the content of the other, at all times. Through pull requests, each party vets the other. At the end of the day, in the software world, the code, is managed by all parties-- there's no unique code base for the frontend, backend, data engineer, and QA developer (like we have in our AEC industry) where each party has to recode the other's content or act as their gate keeper.

H by Herman on 6 Nov 2020 #

HI All, Happy to see that you like the SAF initiative. We have been working on it about 2.5 years now and we cover already a lot objects needed for structural analysis. We keep ongoing and hope to get next year a lot special stuff into it especially for non lin analysis and dynamics. Another initiative we have done is the creation of the SCIA AutoConverter. This is a whole new approach for collaboration between modeler and engineer also based on a lot of feedback coming from the market that are not happy with the current available workflows. Now analysis models can be created in a CAD system but that is normally not the tool for the engineer. And the analysis model is dictating the results. So we made a neutral web app that can take any structural model (physical load bearing model) and can transfer that to an analysis model steered by the engineer. The upload of the structural model is done via IFC and the analysis model is transferred to CAE software via SAF. Changes are managed and if structural model is changed we can update the analysis model not losing all loads and boundary conditions.

L by Luis on 6 Nov 2020 #

@Herman said:

HI All, Happy to see that you like the SAF initiative. We have been working on it about 2.5 years now and we cover already a lot objects needed for structural analysis. We keep ongoing and hope to get next year a lot special stuff into it especially for non lin analysis and dynamics. Another initiative we have done is the creation of the SCIA AutoConverter. This is a whole new approach for collaboration between modeler and engineer also based on a lot of feedback coming from the market that are not happy with the current available workflows. Now analysis models can be created in a CAD system but that is normally not the tool for the engineer. And the analysis model is dictating the results. So we made a neutral web app that can take any structural model (physical load bearing model) and can transfer that to an analysis model steered by the engineer. The upload of the structural model is done via IFC and the analysis model is transferred to CAE software via SAF. Changes are managed and if structural model is changed we can update the analysis model not losing all loads and boundary conditions.

Just a question: SAF is "open source" (anyone can use the specification without a license) but SCIA AutoConverter is not, right?