This presentation in part of the 2017 3D Digital Documentation Summit.
Reality Capture, 3D Modeling, and Historic Preservation: A Structural Engineer’s Perspective
Kate: Good afternoon. Good to see everybody back. Everybody’s still awake after lunch, right? Okay, good. All right. I’d like to take just a minute to tell you a little bit about Silman before I dive into the presentation. We are a structural engineering firm, headquartered in New York. We also have offices in D.C. and Boston. We’re about 150 people right now, and we just celebrated our 50th anniversary last year.
We are a generalist firm. We do everything from tiny restaurant renovations, up through big new museum projects. We do a lot of new construction. We do a lot of adaptive reuse. We do a lot of historic preservation. Those last two categories are why we’re here today.
Why are we talking about digital documentation? Why is it important for structural engineers? Well, as we’ve already heard today, contemporary documentation is one of the best sources that we can have for an existing building. If we’re really lucky, that existing documentation looks something like this. It’s nice. It’s clear. It’s complete. It’s legible. But far too often, it looks like that, in which case we have to come up with alternate methods.
Silman has actually been involved in these alternate methods for quite a long time. This is St. Thomas Church in New York City. Our work on this actually started back in 1998, a project to analyze the vaults, and we laser scanned it. You might notice, it’s a little tall. They didn’t really think they could get a tape measure up all the way across there.
After the laser scan, we were able to generate a SAP model, that helped us analyze the work on the vaults and come up with the numbers that we needed for this project. I will not say that this was easy. 20 years ago the scanning technology was no where near what it is now. The digital modeling technology was no where near what it is now. It involved a lot of back-and-forth between our team, and the scanning team, and I think we even had a third consultant who was actually helping to build the SAP model. Lots of round and round, but we did eventually get the results that we need.
You might recognize this project. This is another one of our early forays into laser scanning. This is the Guggenheim Museum, also in New York. Back in 2005, they started a major renovation project, mainly focusing on the façade, and the mechanical systems. Silman’s work centered around the façade. The big problem, as you can see from the diagram on the bottom was that there were a lot of cracks in the sprayed concrete walls, all the way around the building. It’s a really unusual building, really unusual geometry, difficult to document with conventional methods. Again, we scanned it.
The cracks, we were able to determine … I wrote a list of all the reasons that there were cracks in these walls. Thin cross sections, poor detailing and implementation of rebar, including discontinuities in the rebar and other structural systems, weathering, from being in New York, and inadequacies in the original structural design. Sorry, Frank, but there was a lot of work to do.
We combined laser scanning with nondestructive evaluation rate GPR to identify some of the locations of those discontinuities in the rebar. We were eventually able to combine everything into another SAP model, the deformed shape of which you can see in the upper left hand corner. That’s representing the building under … I forget whether that condition is wind or weather, or temperature. We were able to evaluate the building for its response to wind conditions and the temperature conditions.
Again, I will say this was not an easy process. 15 years ago, lots of back and forth, lots of reiteration. But eventually we were able to get the data we need. That was 10 years ago though. We’ve made a lot of progress since then.
I’m going to turn it over to Nathan, who can tell you about some of the things we’re doing today.
Nathan: Thanks Kate. In the past, if we had a project where laser scanning was being performed, often, we as structural engineers were just getting the in-deliverable, the CAD drawings, or the [inaudible 00:04:17] model that resulted from the laser scanning.
In the present, We’re pushing more and more to actually get the raw data, the Point Cloud. As we’ve been doing that, really the story of the present is, the utilization from a structural engineer’s perspective of some of that raw data.
This is one of our recent projects. This is the Richmond Old City Hall in Virginia. It’s an 1894 structure. It’s actually across the street from the Virginia State Capital. We’re going to look at this one real quick as one of the case studies we present to talk about the present uses of laser scanning for the field of structural engineering.
This is a photo from Richmond old City Hall. There’s actually a four story atrium space in there. Early on in the project, the design team commissioned a laser scanner to go out there and develop a Point Cloud from the laser scan. Then actually, The architect, who was Quinn Evans, put together a Revit model. Rather than just getting the Revit model, we did push to get the Point Cloud data.
You can see right here, this is actually the laser synchrony and Point Cloud data from that same atrium space. One of the technologies that’s commonly used today is, you actually have the camera with the part of the laser scanner. You can associate an RGB value with each of those individual points to get you that colored Point Cloud.
This is from the space that’s above the atrium. There’s actually a laylight and a skylight assembly. Early on in the project, it became evident that we weren’t going to be able to get up into this space to take measurements of the trusses. This presented a bit of an issue though, because we were replacing the entire skylight and laylight assembly. We didn’t know if the existing trusses were adequate for the increased loads that we were putting on them.
In conversations with the owner, and the rest of the design team, we decided it wasn’t adequate just to carry a large contingency about potential reinforcement of these trusses. We needed to figure out an alternate means of evaluating the trusses and their adequacy for these increased loads.
As you can see, this is the Point Cloud. Luckily, the laser scanning company did go up into the attic space and was able to capture these trusses. We were able to take that laser scan data and cut sections, essentially, to give us a good understanding of the overall geometry of the trusses, and the individual numbers.
Now I won’t lie. I think we got pretty good measurements for the overall geometry. But, when you’re talking about the individual numbers, it was hit or miss. For instance, the top part of this truss, we can pretty confidently say it’s a 10 inch deep channel section that’s two inches wide. But this laser scan, in particular, wasn’t accurate enough that we could tell you the thickness of the flange, or the thickness of the web.
There was still a little bit of guesswork that went into this. But we were able to develop an accurate enough structural model that we could run the numbers, and really refine the amount of reinforcement that we called for, without ever having laid hands on any of these trusses.
This is a view of the overall Point Cloud. A laser scan, for us in the structural engineering profession, is not just for looking at these inaccessible spaces though. Our office is a couple hours from Richmond. It really supplemented our site time. Having an accurate and reliable 3-D representation of the building through this Point Cloud, it’s very valuable for us in the engineering profession.
We can go through this and determine, do the walls line-up? Do the columns line-up? If they don’t, do we need to do some additional probes there to understand what’s happening? Do we need to figure out if there is a transfer girder that we hadn’t been accounting for in the past? We can go in and take fairly accurate measurements of how far that beam is spanning.
You know when you get back from a site visit there’s always that one room where you forgot to take a photo. By having the Point Cloud data readily accessible, you can easily go into that model and get a good visual of what that room looks like.
The second case study I want to talk about, is a project that we’re actually presently working on. It’s the Corcoran Gallery of Art. This is in Washington D.C. The Corcoran dates back … Portions of it date back to the late 1890s. This is actually right across the street from the White House. It was recently purchased by GW or George Washington University. They’re adaptively reusing it to support some classrooms, and also exhibits for the National Gallery of Art.
At the Corcoran, the laser scan was performed early in design. This figure shows the site. Each of those dots represents an individual scan. The different colors on those dots correspond to different elevations at which the scan was taken, the different floor levels.
At the Corcoran, rather than working with the entire laser scan merged together, that can often be a lot of data. With some of our computers, they can work pretty slow. We actually would go to the locations at the individual scans, and we were able to take pretty accurate measurements from those.
This is one of the auditoriums there. You’re able to see exactly what the laser scanner saw at that location. The Point Cloud data on the right, we can go into that and figure out pretty accurately, without being on site, where can we hang these new loads from the trusses?
This is from a boiler room on the left here. We can, from the comfort of our office, really … This is only five minutes from our office, but still, there’s not a need to go out to the site if we can measure how far those ducts are, and if we can fit in this new stair, just with the Point Cloud data.
The Corcoran Gallery really challenged us to balance traditional methods of data collection with new methods. The traditional methods would be things like, probes, hand measurements, photography. Some of the more modern technologies would be, NDE, non-destructive evaluation, and laser scanning.
I really think that this is best illustrated by our investigations into the floor systems at the Corcoran Gallery. They used an archaic system, that’s called a metropolitan slab. That’s actually what this is a photo of during construction.
What is a metropolitan slab? A metropolitan slab is a floor system where steel beams are spaced at every few feet on center. Then there’s actually a cinder concrete slab spanning between, that has draped wires in it.
When we first got brought on to this project, we did some NDE, non-destructive evaluation, using ground-penetrating radar. That’s actually what that image is on the left. You can almost think of it as an x-ray of the floor, to identify where the steel beams were, the spacings, then also, where the draped wires are.
We wanted to be able to correlate what we were finding with the GPR with some selective probes. That’s actually what that image is at the bottom right, is a view to basically confirm what we’re finding with the GPR.
One of the benefits that we’ve been finding as structural engineers for laser scanning, is illustrating the flatness of different building assemblies, such as walls and floors. This is a masonry wall at a historic building that we were working on in Virginia.
We could tell that the wall was bowing, but it was difficult to determine a true extent. I think in the past we would have taken a plumb bob out there, and taken a lot of measurements to try to develop essentially a contour map of how far this wall was bowing. What we did in this case is, we actually laser scanned that wall. We were able to cut sections through the Point Cloud data to get a far more accurate representation of how much that wall was bowing at the different elevations, and different locations.
It really helped us to identify the true deformation of the wall. This is good for us, as structural engineers, one, because it provides us with a better analysis. Rather than having to make conservative assumptions that this wall is two inches out of plane over the whole height, we can really refine that analysis.
It also helps us to focus our repair efforts. If through the Point Cloud, we can identify a few hotspots where the wall is really bowing. We can focus our investigations and future repairs.
That was in the past. Actually, more recently what we’ve been doing is, actually developing contour maps, similar idea though. We basically drape a surface over the top of the Point Cloud, and then we determine the distance away from a reference plane. For a wall, this would be a vertical plane, for a floor, a horizontal plane.
You can see how far out that wall is, out of plane from that reference. At floors, similar approach, you can determine how much the floor is sloping, or how much that deflection is at mid-span. It also provides a good visual for the client.
In terms of the structural analysis, this is Low Library at Columbia University. We did a laser scan of the dome there. A laser scan is critical for us, as structural engineers, especially when it forms a basis for a finite element model. You can see finite element model on the right there.
As we were translating between a Point Cloud, and a numeric or structural model, it requires a lot of filtering of the data. You’re going from potentially millions of data points to, hopefully, a few hundred, maybe a few thousand. That does take a little bit of an effort.
We basically have to create a regularly spaced grid that still communicates the geometry, but won’t bog down our structural analytical programs. There’s no way that we would be able to incorporate millions of data points. It would just crash the structural programs.
This is actually from a different project, Iglesia San Jose, in Puerto Rico. The reason I bring this up is, a laser scan is going to capture the exposed surfaces. You might catch the underside and the topside of a dome. But really, when we doing a structural modeling, we’re interested in the centroid of that section. It’s not this direct translation. We actually need to figure out where the centroid is.
Finally, we’ve talked about the present, and what we’re doing in the structural engineering world with laser scanning and Point Clouds. But what does the future hold?
We’re starting to see shape extraction already. The idea is that you would have a Point Cloud. The Point Cloud through running it through software, would actually automatically be able to identify why flange sections, rather than having to go in and make those measurements yourself.
Drones, again, this is something that’s already happening. But mounting laser scanners to drones, I think we’ll see that continue to develop in the next few years.
Then, just the merging of non-destructive evaluation techniques, such as ground-penetrating radar, and thermography, and others, with laser scanning. So often we, as structural engineers, are interested in what’s hidden. That’s the end goal, is a way to basically communicate the hidden in three dimensions.
Kate: I want to wrap up today by talking about the other side of digital documentation. We’ve talked a lot so far about Point Clouds and NDE. But, for most of us, the deliverable is still 2-D plans, or sections, or details, or elevations. Even when something like the HABS and the HAER drawings, they have to translated back into 2-D.
For us, that typically means BIM, typically means something like this. It’s a beautiful looking model. Is it useful? You can’t tell by looking at it. You also can’t tell without having the conversation whether what you’re going to get is going to be useful.
Just to make a quick point, I’m probably preaching to the choir. But a Point Cloud and a model are not the same thing. The skills that it takes to have a good, useful Point Cloud are very different from the skills that it takes to have a useful Revit model, or useful 3-D AutoCAD model.
Sometimes those can be done by the same firm. Sometimes maybe not a good idea. I’m sure we could trade war stories about having outsourced the development of these sorts of models. But, make sure that you’re … If you’re going to work with somebody who is doing these for you, have a conversation. Make sure that you’ve got the right team for the things that you want to accomplish.
How can you determine those things? This is a phrase that we have found very useful recently as we start to develop RFPs for delegating this kind of model creation. We are asking our consultants to respect the orthogonal design and intent of the building. For us, that means walls are straight, floors are flat, columns are vertical, corners are 90 degrees.
Sounds great, right? Except we know that in existing conditions, none of those things are true, which is why there’s an asterisk. Those are all the things are within reason. Even that gets an asterisk, because who decides what is reasonable?
That’s a conversation that you have to have with your design team. How accurate do you need to be? At what point are you willing to accept that a corner might really be 89 1/2 degrees?
I have to tell you, we had one building where the client said, “We’re going to absolutely obey the Point Cloud. We’re going to trace it exactly.” No two walls in that building were parallel. It was a nightmare.
However, if you hold everything to 90 degrees, you may find that you’re Revit model is off by six inches at one end of the building. That may not be acceptable either. I can’t tell you which way to go, but it’s a conversation that you need to have.
Some questions to ask … How many people work with consultants to do scanning, as opposed to doing it yourself? Not many … Everybody else doing your own scanning? Okay, good.
All right. If you are engaging a scanning consultant, ask them, what point density … Ask your team. Have the conversation with the scanner, what point density is required? The denser you need your grid, the more setups will be required, the closer access to your walls. Maybe for some things you can get away with being setup farther back, and scanning the whole wall.
Do you need a color scan? You saw some of our color scans earlier today. We’ve seen some nice black and white ones earlier. RGB may add some cost to your project. It may add some time. It may add cost because they need to turn the lights on for you. If you’re scanning an occupied space, and you can afford to do it after dark with the lights off, you might save some time and get in and out of the building quicker.
Are the spaces accessible? I’m not actually sure how they scanned that skylight at Richmond Old City Hall. It’s possible they had to reach through the wall, and held still. But, if you have a tiny space that needs to be scanned, maybe there’s a small scanner that can be used. But, if you don’t know ahead of time, you’re going to incur additional mobilization costs, and they have to go back and get the right scanner.
Are there elements that need to be moved prior to scanning? Is there something temporary in this space that you can get out of your way, so that you don’t have to deal with it in post processing. Is there furniture in the room that you’d rather not scan?
Are there cars parked next to the building that you could move? Are you going to scan the outside of your building in the winter, maybe when there aren’t any leaves on the trees?
Demolition, if you’re going to demo all the interior finishes, do you really want to do the scan before you’ve done that work? Maybe you can wait a couple months, let the demo happen, then go scan.
The last question for your scanner is, does the Cloud need to be geo-located? Does it need to be tied to a specific survey data? We heard from the Port Jefferson team how they’ve specifically set up their survey network, because you needed to link things in. Maybe you’re just going out to grab a façade, and it doesn’t really matter what its survey coordinates are, something to ask, a conversation to have.
Let’s say that somebody else is doing the modeling, based off of your Point Cloud. We mentioned existing documentation. If you have some, why not use it, unless you know it’s unreliable? See what you’ve got. The more you can get about the design intent, the better off you are again, assuming accuracy and completeness.
What details are needed? Do you need to have a solid model of every Coping stone, at every cornice, and every window molding, or can you get away with generic elements? No right answers, just cost implications.
If you can’t see something, how are you going to identify it? An example might be, a basement wall. We can scan the inside, can’t scan the outside. What are you going to do with that? Maybe you’ve got your GPR, and you can figure out how thick it is. But, are you going to tag it somehow, to say, this wall is an assumed size, or, we know this side, but not the other side.
There are things that you can do with the meta-data and object parameters to identify those kind of elements. You can’t avoid them, so you need to talk about how you’re going to handle them.
Mentioned demolition earlier, same thing goes for modeling. If it’s going to be demolish, do you need to model it? Sometimes the answer might be yes, because you don’t know if you’re going to demolish it at the time you do the model. You might know that an entire wing is going to go away, and you can get away without actually putting the effort into creating a solid model.
Last one, are there structural analysis requirements for the model elements? This one’s for us, everything else applies to all disciplines. The structural is for us. Do you need solid elements? Do you need shell elements? Do you need plates? Those are different … Each of those have their own requirements. Easy to do by a team, if they know what you want ahead of time. Do you come back to them later and say, oh by the way, I needed the midpoint of that wall? Probably asking for a change order.
I wanted to close with one image, which is what happens if you maybe don’t have that conversation about what you want out of the model, or the scan. I call this, my ghost in the machine. What is a little difficult to tell in this image is, he’s actually up there about eight times. Every time you move the scan, I guess he just, he couldn’t get out of the way.
Something like that is very difficult to take care of in post-processing. You’ve got to go and find all those little tiny points. I think in future, we would make a note, scanner operator to avoid being scanned.
Thank you very much.
3D modeling has been part of the preservation engineer’s toolbox for several decades. In the years since the technology’s inception, it has evolved from being an occasional (and expensive) novelty to an invaluable part of the engineering workflow. Two factors have driven this change. The first is the improvement in the equipment used for laser scanning. Scanners today are faster and more accurate than ever before. Laser scanning and photogrammetry have condensed days or weeks of data-gathering effort into hours.
The second evolution is in the software used to visualize the resulting data and to utilize it with structural analysis software. Point cloud data must be processed and analyzed before it can become truly useful information, and thanks to rapid advancements in 3D CAD, BIM, and structural analysis software, this is also faster and more accurate every day.
Our firm has successfully used point clouds and 3D models to:
· Access inaccessible spaces
· Map deformations in walls and floors
· Retrieve additional field measurements
· Investigate intricate structures
· Create quick and easy building visualizations
· Map multiple layers of building information into complex 3D structural models
· Identify areas of a building’s structural systems that need intervention (renovation, repair, replacement, reinforcement)
· Plan the course of Construction Phase activities based on the 3D information and modeling created during the Investigation, Analysis, and Design phases.
In order to have a successful point cloud workflow, of course, you must have effective communication between members of the project team—architects, engineers, contractors, and owners. Over our long history with laser scanning, we have honed our ability to identify important parameters for scanning criteria to meet the overlapping needs of architectural and structural point clouds. We have also seen
what kind of language is essential in writing contracts and requests for proposals to ensure that all parties have a clear understanding of the tasks at hand, and to ensure that the deliverables meet the stated requirements.
This presentation will cover lessons learned for various steps of a project, from working with surveyors to define accuracy & coverage requirements, through processing data into useful model information, to incorporating existing conditions into working documentation. Specific emphasis within our presentation will be how information and analysis results from 3D structural modeling have facilitated the maintenance and preservation of large amount of historic fabric. Our research efforts have included improving the integration of non-destructive testing (NDT) results, material testing, and electronic monitoring results into the overall 3D models that we’ve created. Our experience creating advanced and detailed structural models of existing buildings has pushed the envelope on such modeling for the past 20 years. Our most recent efforts have included collaboration with contractors, builders, and owners to help make the Construction Phase work proceed as smoothly as possible—utilizing 3D modeling.
Nathan Hicks joined Silman in 2009, and is an associate with the firm. In 2010, Nathan was awarded the Silman Fellowship, spending six months at the National Trust for Historic Preservation working with the Historic Sites Department. Prior to Silman he received a BS in Architectural Engineering and MS in Architecture from California Polytechnic State University. Nathan is a member of APT DC, National Trust for Historic Preservation, and the DC Preservation League. Since joining Silman he has worked on the St. Elizabeth’s West Campus in DC, Washington Union Station, Mount Vernon and the Lincoln Memorial; among other projects.
Kate Morrical joined Silman’s DC office as an engineer and CAD specialist in 2003. She worked on new projects and renovations, contributing both to the engineering design and construction documents. After working for several years with Autodesk, where she worked closely with the AutoCAD and Revit teams, Kate returned to Silman as the firm’s Digital Design Manager. She is tasked with developing an integrated approach to modeling, design, and documentation across multiple software platforms. Kate is involved with every level of BIM at Silman, including in-house training and technical support, production of construction documents, and internal and external model coordination.