NetworkRail Parametric Overbridge tool?

NetworkRail is looking at a number of 'prefab' overbridge designs. This one by Danish architects Gottlieb Paludan is written in Shapediver.

Pretty good effort. But, what is really needed is something that you can use within the dgn environment.

Questions:

1. Could Bentley cook up something to match the Shapediver tool above using OBD/GC? Should be pretty easy.

2. Can the results be hosted on a website like Shapediver does? I guess that this would involve publishing to itwins and using react etc to connect the sliders on the web site to a copy of GC running in the background or silent mode? Or would this be something the Optioneering Center on the Connect Center already supports?

3. Analytics - I suppose you could write a GC node to replicate the old Structural Modeler tools to assign analytical attributes to the structural members?

  • this would involve publishing to itwins and using react etc to connect the sliders on the web site to a copy of GC running in the background or silent mode? Or would this be something the Optioneering Center on the Connect Center already supports?

    Hi Dominic.

    Linking in with iTwin and other cloud based technologies is definitely features on the roadmap. The detail of what and when are still in the mixing pot right now but feel free to keep adding ideas and its always valuable to hear more detail on the requirements. 

    Stuart


    This is a test

  • Linking in with iTwin and other cloud based technologies is definitely features on the roadmap.

    Good to know! Shapediver is pretty interesting here as they are heavily based on GH on one hand and web/cloud based on the other. Their 'configurator' beginings are also quite relavant as we increasing find that the old long chain 'algorithmic' script based form generation starts to transition to a greater emphasis on using modular and Commercial Off the Shelf products.

    Modularity and COTS is definitely the trend in rail sector and is also being driven by the way the specification / product data portals are getting more cloud-based. I think that we are seeing a lot less 'if I can model it and performance spec it, someone will find the real world products to build it for reals' and more of 'let's model it and incrementally verify we can source the products for reals as we go along, and not at the last minute'.

    This has big implications for BIM working... and the apps that are intended to support it.

    NBS Chorus suffers a lot as it does a poor job of filtering for product properties. It happily points the user to some products but the user will need to manually verify the performance and other requirements. BIMproduct.cloud on the other hand takes a much more designer-friendly approach.

    1. User selects the type of object he wants to place in the model. The categories are much more BIM tool aligned than Uniclass.. which is good.

    2. User defines some basic parameters and properties.

    3. User switches to 'product catalog' mode and is presented with a selection of real world products that match the criteria inputed previously.

    4. User selects one (or the best fit)

    5. User selects type of parametric object to generate... and places the object in OBD / Mstn etc. If he does not find one, he can return to 'custom object' mode and place a low LOD 'generic' placeholder object.

    I think this kind of cloud functionality can be easily provided by ComponentCenter... with a bit of help or collaboration with BIMproduct.cloud (I note that CC is also collaborating with BIMstore here in the UK). Or maybe the CC team could roll their own using something like Figma?

  • CC at the moment is also a bit too passive and takes a 'library' approach. There is some search functionality but its not as fine tuned to help the user 'configure' content to be downloaded.

    Shapediver also starts off very similar to BIMproduct.cloud.

    1. User selects the type of object he wants to place in the model.

    2. User defines some basic parameters (no properties, they are not very BIM-savvy but hey, its just another parameter)

    3. User places the object and Shapediver provides some very nice snapping behaviour to help speed things up (important when you are trying out new options).

    Some ability to change and experiment with different materials and finishes... in the brower environment, not in the BIM model. 

    Some nice animation functionality. I guess there are a lot of freely available code libraries for this kind of stuff for browsers these days. Useful for explodes and explaining stuff, I suppose.

    I suppose GH is used to provide the parametric control and run on the cloud. What would be the Bentley equivalent? GC for itwins? GC.js?

  • But of course this is all very nice but what if we put the Bentley 'federated approach' spin on it?

    1. As the configuration is done in a browser. You could save the browser 'configs' as a tab or separate web page for each (or group of) object(s) placed. No impact on model size. The webpage(s) would also link back to the product libraries hosted by ComponentCenter. CC would also store all the ancilary 2d drawing etc info all without bloating the BIM model.

    2. But the kitchen would also part of a larger imodel hosted by itwins. The user would select the room and its contents and 'check it out' briefcase or OpenPlant ModelServer MCS-style. iTwins are already browser-based so everything is just a tab away. When the config is done, the user just synchs the configured kitchen back into the main imodel... and through to the dgn-based model on the desktop?

    Or the user may decide to keep the config and mix and match later. I can see this being used a lot for developing facades for example.

    This much more dynamic link between an object and its alternatives paves the way for the type of smart 'data driven' optioneering that will be required to meet today's financial and carbon budgets.

    A model should not be a static document, which is more often than not the case today. It should be a working dynamic model that can react and respond and have more 'depth' to it.

  • This much more dynamic link between an object and its alternatives paves the way for the type of smart 'data driven' optioneering that will be required to meet today's financial and carbon budgets.

    A lot of chatter about Chat GPT and how Artificial Intelligence will change design and designers' jobs.

    Interesting parallels between how LLMs use neural nets to generate 'embedding' vectors to help guess the next word in sentences. Vector databases are used to search for the best guess. But it is still a guess and LLM's 'hallucination' problems will probably never really go away.

    Maybe something like causaLens will help make these models a bit more engineering-rules friendly and controllable. Something AIWorx, OpenSite, Darwin and even DesignPower might recognise?

    Currently, there is a big disconnect between the generic BIM design and the real-world products that will be needed to realise the model. This is being done manually. Most of the spec portals like NBS Chorus are pretty unhelpful here as they do not really allow real searching by criteria. You generally pick a manufacturer and are then channeled through the manufactuer's offering from there.

    Yes, 'geometry-first' BIM modeling is probably here to stay. Sketch, CAD model, add or use BIM objects with Building object behaviours, add spatial and other classification, business data- all of which then gets handed to the contractor who has to map the objects to real-life products, compare and select, fill in any gaps, before procuring and installing.

    This process will probably be 'still valid' for awhile... but this linear sequence is also like a verbal sentence... that can be also be expressed in script, code.

    BricsCAD already has its BIMify, Propagate tools. Others like Swapp.net are looking to automate annotation. Still others like Architechtures, Higharc, Finch, Swallow etc etc are using AI or fast rules processing to automate schematic building design.

    As you design, you could call on the OBD Chatbot to 'complete your sentences' by proposing plausible cladding, connections, accessories etc based on your predefined rules and/or you favourite past models. The Chatbot would also record your edits and will over time anticipate your moves.

    Training a large 'building product' model to auto-complete building elements or fit out rooms based on some rules should be something that would ideally be done 'centrally'... not really something that can be left to the users to tinker up on their own... although their telemetry, models would probably be used to train a collective model.

    And those bsmart guys might find that AI models will probably be able to do much better IFC models than they can ! Stuck out tongue winking eye