#25 – Capturing reality…
Image by Anonymous.
Here we are then, several days post-Brexit and no countrywide implosion to report – unless you’re a tragically unfortunate victim in Streatham or a disgruntled England Rugby fan, the last few days have been relatively uneventful. Time for us all to get back to work.
Today’s topic is point clouds. Reality capture technologies, as they are often referred, are an emerging trend in the AEC industry. It’s become one of those subjects that everyone is wading into; every event sees new companies, new technology, new software and hardware all designed to record our built environment. I touched on this in Post #4 after Digital Construction Week 2019. My current issue is this: How is all of this data is being used? Are we really seeing the benefits?
Being honest, it’s become the latest band-wagon for everybody to jump on. Contractors and consultants are lapping up the opportunity to get to the ‘front of the crowd’ in this increasingly popular scene, but is it a worthwhile investment? The scanners are expensive, the software is expensive, capturing and processing is a huge time investment and everyone needs some level of training before they get started (whether it be for the hardware, software or both). Is anybody currently realising an actual return from all of this expense? I’m struggling to see it.
It may well be my limited experience on this particular topic, but at the moment I can’t seem to obtain the information that I really need. I’ve worked on projects, recently, that have captured thousands of scans during construction. From what I’ve seen the data is relatively useful for the building structure and fabric (slabs, columns, walls etc.), but it’s seriously lacking when you get down to anything more complex – I’m of course talking about MEP.
The building structure/fabric are typically large surface objects; the scanner is able to place thousands, if not millions, of points across the surface of each object and as a result the scans are detailed, accurate and able to become ‘surfaces’ in certain software packages (surfaces are much easier to view than millions of tiny points). You can then start to use the scans for something useful, with various pieces of software to choose from that enable you to not only view the scan data but to overlay it with BIM model elements; Virtual Vs. Reality. It’s a good tool, but is this enough? Does this warrant the time and cost investment? And, crucially, do we really need it for this purpose?
For me, projects are won and lost in the MEP package (see Post #19 for more on this topic). It’s therefore no surprise that I want the reality-capture technologies to record the services installation; this is something that I need. Once a building is finished, the vast majority of building services are all hidden away – they’re under the floors, above the ceiling and within the walls. When it comes to facilities management, what do you need access to? Filters, valves, dampers, access panels… the list goes on. I need good, accurate MEP as-built data more than I need an accurate representation of the partition wall location (that I can easily see). Yes, accurate data for the building fabric may be useful elsewhere but I cannot see that it’s more important than the services, especially when we’re considering as-built data for FM. This is where I feel that the point clouds are letting me down.
Services above a ceiling, for example, are often so intertwined with one another that even to the naked eye it takes time to assess what’s going where. The pipes, containment and ductwork appear to twist and set around one another in a seemingly impossible way – you’ll find yourself looking up at the installation and walking around in circles, trying to understand what’s happening from different angles. Depending on the depth of the void, you may also have several ‘layers’ of equipment; this is a crucial issue for the scanner as the lowest level services obstruct the elements above. A point cloud scanner, in my experience, simply cannot capture enough data to be able to clearly distinguish the individual services that are layered in that void. There is too much going on, especially when you’re capturing wire baskets or perforated cable trays – the reduced flat surface area (compared to something like a duct) makes it even more difficult to see. Overlay this scan data with a BIM model and even the slightest misalignment between model and reality ensures that any on-screen comparison is barely usable. The problem is exacerbated as the difference between the model and the install gets worse, to the point where you can’t recognise anything.
Further issues we’ve encountered include poor temporary lighting on site reducing the quality of images and impacting the capture of coloured point clouds. Temporary sheeting, pipework, lights and cables as well as bundled wiring hanging down from the soffit have also added confusion to the scan data. And finally, as any surveyor will tell you, surveying a live construction site is a near impossible task. Most scans capture blurred high-vis jackets moving around as well as people up on lifts and scaffold towers working at ceiling height, all of which is further complicating the scans. Even scanning out of hours can’t exclude the access equipment like lifts and scaffolding; it’s a very difficult task.
This, for me, puts point clouds in the corner. Whether its the quality of the scan or the abilities of the different software packages we’ve used to view them, nothing so far has been up to the challenge of complex MEP installations. I don’t think that this is any revelation, I actually think that most vendors in this field know that it’s an issue. Images and videos show people scanning perfectly clean, empty rooms that look nothing like a real construction site. Many will also show examples of scans overlaid with models; it looks great for a slab, but when it comes to MEP you’ll be shown a single pipe on a wall or a pump in a plant room neatly placed over a model object. I’m yet to see any convincing examples of multiple services that are easily visualised, as either a standalone point cloud or overlaid with a model. From what I can see, it’s simply not there yet.
Point cloud has it’s uses, I’m not trying to discredit an entire sector, but you need to define your needs and manage your expectation before you join everyone else on the band wagon – a mistake we may have made. If you’re capturing structure, architecture, FF&E, existing buildings or facades then you’ve got a good chance of obtaining the data you need. If you’re looking for something more intricate, particularly MEP, you’ll need to think long and hard about what can actually be achieved. You need to test different hardware and software packages to really understand what is currently possible and you’ll certainly need to manage any expectations you have. You’ll be in a much better position further down the line if you invest money and time in your investigation, before you invest more money and time in the implementation of new technology.
Have I got this completely wrong? Are there particular scanners, processes or software packages that can visualise the complex MEP that I need? Let me know in the comments, I’d be gladly proven wrong.
Mr. Smith
smith@dbe.careers