[00:00:00] H.J. James Choo, PhD: My name is James Choo. I’m a member of the PPI Technical Committee. This morning I have the honor of introducing our next esteemed speaker, professor Iris D. Tommelein. She is a professor of engineering and project management in the Civil and Environmental Engineering Department and directs the Project Production Systems Laboratory at UC Berkeley.
[00:00:23] H.J. James Choo, PhD: She has been studying, developing and applying principles and methods of project-based production management for the architecture-engineering-construction (AEC) industry. This approach is what’s actually termed “lean construction.” She’s the recipient of the 2012 ASCE Walter L. Huber Civil Engineering Prize, 2014 ASCE Peurifoy Construction Research Award, and 2015 LCI Lean Pioneer Award.
[00:00:55] H.J. James Choo, PhD: She brings to us a topic that is highly relevant to what’s going on in our industry currently. Mainly driven by those that compete on time to market. Many organizations are attempting to or are in the process of moving work offsite; this approach is referred to, by many, as “industrialized construction.”
[00:01:18] H.J. James Choo, PhD: As this is done, it brings forth some risks that may jeopardize the benefit of moving work offsite. Professor Tommelein is working with PPI in preparing a survey to understand how organizations are tackling this issue. And we are forecasting this to be sent out early next year.
[00:01:41] H.J. James Choo, PhD: We look forward to your participation. Today she’ll discuss some challenges specifically associated with managing tolerances related to product and processes, and introduce some methods to address them, as well as highlight some further research questions. Please join us in welcoming Professor Iris Tommelein. Iris, over to you.
[00:02:05] Iris D. Tommelein, PhD: Great thank you so much, James. Good morning, everyone. Good day, everyone. I’m actually here at PPI in San Francisco looking at the beautiful view of the San Francisco Bay Bridge. And it’s a pleasure to have a slot in this very exciting symposium talking about so many different topics related to project production systems.
[00:02:26] Iris D. Tommelein, PhD: Design is, as James mentioned, tolerances are an old topic, and yet I think of it as perhaps as current as it ever has been because it’s so relevant to everything that’s going on as we’re moving to production, using robots and so forth, which work to very different tolerances than we’ve been used to.
[00:02:51] Iris D. Tommelein, PhD: And let me switch to sharing my screen. Bear with me here. Always good to start with the definition. A working definition, we can argue about it, but the basic understanding is that there will always be variation in whatever we do. And tolerance specifies the range of variation that is permitted.
[00:03:17] Iris D. Tommelein, PhD: And specifically today we’ll talk about dimensional variations. We also have location variations and so forth. So there’s a range of variation that’s permitted. Without impacting the structural integrity, the performance of the product, the operating capability, or any of those components. And with that, I’ll give you a simple example of where tolerance problems manifest themselves.
[00:03:42] Iris D. Tommelein, PhD: This was actually a case study that we did here on the retrofit of the Bay Bridge several years ago. Let’s start in the upper right. We’re talking about bolted connections, and the challenge with getting the nut on the bolt is that there needs to be sufficient engagement of that nut on the thread of the bolt.
[00:04:03] Iris D. Tommelein, PhD: If the bolt is too line, then the nuts may hit the shank and not have enough purchase. Likewise, if the bolt is too short, it’s not going to have enough purchase, it’s not gonna work very well. And you see on the left-hand side here, a picture where in fact the bolt is too short, and therefore this is not a satisfactory condition.
[00:04:23] Iris D. Tommelein, PhD: So we may wonder, “How does this condition occur?” Well, of course it has to do with dimensional variation. First of all, the metal components that you see here have some tolerance associated with them. And then if the product is painted or galvanized, of course there’s some additional sickness that occurs, perhaps not evenly.
[00:04:43] Iris D. Tommelein, PhD: And then if you put the pieces together, of course it could be that if one piece is a little thicker and another piece is a little thicker then ultimately your bolt is not long enough to actually fully engage. The problem gets exacerbated, of course. If you have more layers, more plies of steel, as you see in the slower right drawing, which is actually a sketch of a condition on the Bay Bridge where the contractor encountered many, many problems with bolts not being satisfactorily placed and then therefore not meeting inspection requirements. And literally hundreds of bolts had to be taken out and replaced and taken out again and replaced until everybody understood that, in fact, the inspection requirements were not satisfactory and something had to change in regards to the length of the bolt in order to meet the performance requirements as specified by the engineer.
[00:05:40] Iris D. Tommelein, PhD: Right, so we see tolerance problems within an individual specialization. Of course, tolerance problems also occur across different specialties. And here we should talk about an important concept, which is process capability. When we do the work, we will find out that at one time we realized the product with certain dimensions.
[00:06:04] Iris D. Tommelein, PhD: If we repeat our process, the outcome may have slightly different dimensions. If we repeat the process again, the outcome may yet again have slightly different dimensions. And so as we repeat our process many, many different times, we collect all these dimensions. We can plot a histogram, like a bell curve perhaps, although the distribution doesn’t have to be symmetrical around a mean, but let’s assume it is a bell shape distribution.
[00:06:29] Iris D. Tommelein, PhD: The mean value is what we would call the denominal dimension of the part, and then there will be some dimensional variation around the mean. And of course, some extreme outliers perhaps are possible in exceptional circumstances. The situation that you see here is a metal door frame that is put in precast concrete.
[00:06:51] Iris D. Tommelein, PhD: These are manufactured parts and manufactured in different factories, and yes, two different tolerances. The metal door frame has a much tighter tolerance. You can see a much more peaked bell shape distribution. Then the precast concrete, which has the flatter distribution on the right-hand side in the middle of your figure.
[00:07:13] Iris D. Tommelein, PhD: As long as the door frame fits within the opening, that is not a problem. In fact, we want the doorframe to be slightly smaller, and then we will fill the cracks. We will fill the gaps with caulking, as you can see on the right-hand side. However, as you see in the middle of the figure, it could be that these two probability distributions overlap with one another.
[00:07:34] Iris D. Tommelein, PhD: That is that the door frame may be a little bit bigger than the mean, and that the opening is maybe a little bit smaller than the mean and therefore that the door frame doesn’t fit, in which case, of course we have the problem and it’s that kind of dimensional variation that we need to better understand and develop management methods for.
[00:07:59] Iris D. Tommelein, PhD: To add another dimension of complexity to managing tolerances is, of course, that the dimensions of the product will vary over time. They vary based on environmental conditions, temperature and humidity or major influence factors. Of course dimensions also vary during handling of the components, whether, for example, it’s in a factory or whether it’s through transportation or whether it’s final handling for installation.
[00:08:29] Iris D. Tommelein, PhD: And then, of course, dimensions will vary when the unit is put in operating conditions. When you have a pipe in a power plant facility. When hot fluids run through the pipe, obviously the pipe will expand, and when the plant shuts down, the pipe will pool again. And so we need to allow for all that dimensional variation.
[00:08:47] Iris D. Tommelein, PhD: So it isn’t just about fabrication and construction, it’s also about operating the system when it’s in use. About 20 years or so ago, but Todd, you can correct me if I’m wrong. Todd told me that buildings leak at the intersection of contracts. And unfortunately that is still all too true today, and tolerances are a culprit.
[00:09:14] Iris D. Tommelein, PhD: So what are the kinds of problems that we encounter and what can we do about it? So a big problem that we have is a problem that comes from tolerance incompatibility. I already suggested the problem between fabricated parts like doors and precast concrete. But the problem perhaps is more deeply rooted in the way we specify tolerances and the expectations that we put on our project production system.
[00:09:44] Iris D. Tommelein, PhD: Tolerance incompatibility unfortunately occurs when we look at industry standards. And I will read here this little article that talks about anchor bolt tolerances. Where we have the interface of the contracts between structural steel and the foundation concrete, so to read on the right-hand side, industry standard tolerance.
[00:10:09] Iris D. Tommelein, PhD: The concrete and steel industry both provide anchor bolt tolerances. Unfortunately, the tolerances are not compatible. Tolerances provided by the American Institute of Steel Construction are more restrictive than anchor bolt tolerances provided by the American Concrete Institute. And this is an article that dates from 2016.
[00:10:31] Iris D. Tommelein, PhD: The good news is that action was taken. The concrete contractors ACI and AIC got together and actually came to an agreement as to how they should specify the anchor bolts that are very much at the interface of their two specializations. But a key lesson learned here is that, of course, the designer is responsible for specifying tolerances and they need very much to pay attention to what the contractors can actually realize, because many times we find that specifications are unrealistic or unnecessarily costly. So we don’t want to over specify tolerances. We don’t want to make tolerances tighter than they really need to be, to have a system that is fit for purpose. On the contractor end, of course, the contractor needs to have process capability, needs to be able to meet the tolerance specifications, and there are many ways to do that.
[00:11:30] Iris D. Tommelein, PhD: I just want to provide you with one example of practice that we call “mistake proofing.” Many times contractors will make some kind of a jig. They will measure the dimensions that are required and then spend a lot of time making sure that the jig is in the right location, and then they can simply replicate.
[00:11:46] Iris D. Tommelein, PhD: The use of the jig from one location to another in order to avoid that similar mistake would be repeated over and over again. Right, so now that you have some somewhat better insight as to the real importance for us to still pay attention to tolerance management in this day and age.
[00:12:04] Iris D. Tommelein, PhD: Even though it is an old problem, I’d like to go over a couple of methods that we can use for tolerance management and we can think about these as reactive methods. Things that we can do after a tolerance issue has arisen. But we also have preventive methods that we can apply to avoid tolerance issues from arising.
[00:12:26] Iris D. Tommelein, PhD: And we have identified nine of these, and I’m going to highlight just a few of them with some very graphical examples. So hopefully they will be easy to remember. Alright, windows and cast in place concrete. A challenging condition. It’s a butt joint. Butt joints are noticeably harder to manage properly than overlapping joints, but yet we see many butt joints, architects and designers often like it for aesthetics.
[00:12:57] Iris D. Tommelein, PhD: And then, as in this case, this specified cock joints with very restrictive requirements. Three eighths of an inch as a minimum to guarantee the cocking performance and three quarter of an inch maximum, for aesthetic reasons – they didn’t want it to be too large. So add to that now, the tolerance specification for the windows and the tolerance specification for cast in place concrete, and you appreciate that this is not going to be easy to manage.
[00:13:23] Iris D. Tommelein, PhD: So, of course, what happens in the field is that the number of windows does not fit in the opening. As you can see from the picture of a field document that says the window or the panel size changed and make this work, “make this work” means either you adjust the concrete, you get out your grinder or pat it out where needed, or you fix the window, you trim the window, you shim the window, or, you know, you custom size and fit.
[00:13:57] Iris D. Tommelein, PhD: Or maybe you have to do both. So those are very traditional methods that are used. Of course on the reactive end we have a problem. Now how do we fix it? Another method is to try to figure out if parts can match. So if you have many openings of varying sizes, you have various windows of various sizes, perhaps you can identify some of the bigger openings and find some of the bigger windows and match those up.
[00:14:26] Iris D. Tommelein, PhD: And perhaps you can find some of the smaller windows and match and put those in the smaller opening if you’re lucky. Of course, there’s no guarantee that this will be feasible, but I think the key thing to realize here is this is very much a reactive method and we can take an enormous amount of time to find the right matches.
[00:14:46] Iris D. Tommelein, PhD: But matching problems are very common to our industry. I want to diverge here. As an aside, maybe talk about not just tolerances, not just dimensional variation being a challenge, but also product variety being a challenge. So this was a picture that one of my students took many years ago for their course project and in this specific setting.
[00:15:07] Iris D. Tommelein, PhD: Sure enough they found seven different window types. Now, this may all be very useful from a functional perspective, but of course if you have much product variety, then that means that you have to get the right window in the right location, and there is no substitute ability. So the matching problem, if I can simply depict it, would be shown as, as you see on the left-hand side, the expectation is that we have an activity that delivers components 1, 2, 3, in that order we have another activity that delivers components A, B, C, in that order, and that, at time of installation, we can match 1 with A and 2 with B and 3 with C, and so forth. Of course, this doesn’t always happen.
[00:16:01] Iris D. Tommelein, PhD: Things get out of sequence for whatever reason, and therefore we end up with extra inventory and a lot of time spent on site trying to find matches and unfortunately sometimes making mistakes, getting things in the wrong location sort of as we rework, as well. So there’s a number of lean practices that we can apply in this regard to alleviate this problem. Obviously standardization deferring customization, if there’s more, less uniqueness to the components than the matching problem will be alleviated.
[00:16:24] Iris D. Tommelein, PhD: The other thing that we can do, of course, is to allow for flexibility in the system, such as allowing for rapid change order or what we call “single minute exchange of a die.” When that is possible, of course we have flexibility to move things around from one location to another, to switch over rapidly, to work in small batches and to apply load leveling, and all that helps with the system.
[00:16:49] Iris D. Tommelein, PhD: Another method that we can apply is direct control and process design. So if we know that there is a dimension that really, really matters, we might just pay all our attention on site to controlling that dimension. The anchor ball jigs, we’re already one example of that. Here’s an example of a window block out.
[00:17:11] Iris D. Tommelein, PhD: If you need to have a minimum opening for the window, that will be installed later. Well, why don’t you guarantee that opening with a block out so that your concrete will be cast to the right dimension. Another tolerance management method is to use adjustable parts. This is very common in modular construction as well.
[00:17:35] Iris D. Tommelein, PhD: It’s where we have centering cones on individual modules. So that when the next module gets placed on top of the lower one, it is easy to create the perfect alignment between the pieces and there’s very little positioning, measurement to be done. So this is another very nice mistake-proofing practice. Adjustable parts also come in other shapes and forms.
[00:18:01] Iris D. Tommelein, PhD: Fixing jigs are very common, for example, to attach exterior clouding to structural frames. This one shows three different axes of flexibility. You can make things work that way, but of course it takes time and effort to get all the coordinates of all the parts and pieces aligned in the way that we’ll meet the performance and the product require.
[00:18:24] Iris D. Tommelein, PhD: And yet another example is the use of flexible parts, also very much used for long term operation. For example, in seismic conditions, as you see in the right-hand side for seismic joints. But of course in construction these types of fixtures are very helpful as well because they allow you to adjust things a little bit when the deviation is a little larger than expected.
[00:18:48] Iris D. Tommelein, PhD: Right, and then the last method I wanted to expand. The really important one, as far as I’m concerned, because it very much brings us in the design world, in the world of virtual design and construction, where, for so many of us, everything is just perfect.
[00:19:12] Iris D. Tommelein, PhD: Everything is plumb, everything is straight. Everything is exactly of the dimension as we want it to be. And lo and behold, what we find is when we get to the field it is unfortunately not the case, that the world is so perfect because we have the dimensional variation. So back to our windows in concrete as a sketch of a situation from the drawings of this specific project, on the left-hand side here, you see that we have a concrete column.
[00:19:40] Iris D. Tommelein, PhD: And the concrete column is dimensional. So we’ll start from the center line. We have one foot to the right, then we have a ten-foot window, and then we have a four-foot part of the casting place concrete. So that makes for a total of 15 feet. Above that on this drawing, the designer also has specified that this mention between centerlines of columns needs to be 15 feet.
[00:20:05] Iris D. Tommelein, PhD: This forms what we call a “tolerance loop.” A tolerance loop is in, if we start from the centerline of, let’s say the left column, we walk to the right one foot, we walk to the right 10 feet, we walk to the right four feet, and we return 15 feet. We came back to the same place. Every time you have a tolerance loop, you know you have a tolerance problem.
[00:20:29] Iris D. Tommelein, PhD: You can see these problems in the drawings. You need to address them. It is rather a fallacy to think that in product specification you do not have to think about means and methods. Means and methods are really important. The sequence in which you put things together really, really matters in terms of how tolerances will accumulate.
[00:20:50] Iris D. Tommelein, PhD: What can you do about this? To give you one simple example, you don’t close the loop. So in this example, you see one foot and then five feet for the window, and then no dimension specified for the concrete in between the two windows. This gives an indication that perhaps that dimension could be used to absorb some of the dimensional variation that will occur as you try to build as close as possible as you can to meet the other dimensions that are given in the system so we can communicate in design drawings and specifications what our intent is as to which dimensions really matter, as in like the window dimensions really, really matter and then make sure that you cast the concrete so that the windows will.
[00:21:37] Iris D. Tommelein, PhD: Right, so keeping this short, coming back to the framework that I presented with reactive methods and preventive methods I mentioned, we have nine methods identified that are commonly used in tolerance management and have expanded on a few, some of these methods. The ones I started with are the reactive methods that you leave it to the field to figure out how to make do, how to actually fix the problems that have been created upstream in the production system.
[00:22:12] Iris D. Tommelein, PhD: Ultimately these methods work, but they also are very costly to use. And therefore the argument is maybe people, designers and builders together need to better understand what methods are available that they can use further upstream and be preventive in preventing tolerance issues from arising.
[00:22:35] Iris D. Tommelein, PhD: So a quick recap. A big question that should come up is who specifies the tolerances, who does the work to meet them and who inspects? As we saw in the bridge bolts case, we need to coordinate and manage, of course, intra-trade and inter-trade tolerances. I talked about what I call the BIM illusion.
[00:22:59] Iris D. Tommelein, PhD: Nothing is in fact exact when you get to the field. We don’t want to dimension everything with deterministic values. Even that’s only a mouse-click away on your virtual design models. And the thing I haven’t amplified perhaps enough, is that we should not be over specifying tolerances. It’s easy for designers to think that if they have tighter tolerances, that it will be easier for field execution to match.
[00:23:26] Iris D. Tommelein, PhD: But the reality is that tighter tolerance also demands certain process capability of the builders, and the builders may or may not have that process capability. So the real question is, “What tolerance values will result in acceptable performance?” And then to match that, “How capable is your construction process?”
[00:23:49] Iris D. Tommelein, PhD: What can you do and what does it cost? But I should highlight: tighter tolerances may cost more money, but they don’t necessarily have to cost more money. It’s the same as with quality. Better quality does not necessarily cost more. I also think getting the right tolerances in place does not necessarily cost more.
[00:24:14] Iris D. Tommelein, PhD: In fact, it will save you a lot of money because you wouldn’t have all the grinding and chipping that may come. Right, so then the last two bullets, what process capability is required to meet the tolerance specification? And finally, what level of product customization is desired and how does it impact process performance?
[00:24:34] Iris D. Tommelein, PhD: So these are, I think, very fundamental project production, system design questions. And I hope that you now will add tolerances to your list of vocabulary as concerns to address this again as we move to offsite fabrication and are changing the world of designing and making. So with that, I’ll pass it back to James and see if you have any questions.
[00:25:01] H.J. James Choo, PhD: Thank you, Iris, for that provoking presentation. And one thing that’s actually interesting is, as you actually said, that you’ve been actually working on this for a very, very long time. And there’s actually numerous examples that you actually had to accumulate over the years that we only actually have seen the glimpse of.
[00:25:22] H.J. James Choo, PhD: But one of the things that, as you actually just pointed out, is as the more and more work is actually moved off site, the chances of this actually occurring, not only is at the work phase, but it’s actually two work phases now where it could be actually happening in the pre-assembly area and as well as the installation and the special tolerances that you actually touched up on actually has huge implications.
[00:25:49] H.J. James Choo, PhD: But as the products become more, let’s say, unique. Once it gets preassembled, it actually has to go into specific locations, right?
[00:25:59] Iris D. Tommelein, PhD: Mm-hmm.
[00:26:01] H.J. James Choo, PhD: So can you tell us a little bit about what you see happening in terms of the temporal tolerances as work actually moves more offsite? How the matching problem actually plays out?
[00:26:11] Iris D. Tommelein, PhD: Yeah, well the matching problem, so I’ve looked at a couple of projects in San Francisco of multi-unit residential housing. Typically we have, you know, a concrete podium of one or two stories that is built on site, and then we have these modules that need to be mounted on them. Well, that interface between the modules and the concrete podium is, I think, something that is certainly worth studying a little bit more and maybe standardizing on.
[00:26:42] Iris D. Tommelein, PhD: I can’t see that there would be so much desire to have so much product variety. I think the performance requirements on that connection should be pretty well understood by now. So if we can, if we can standardize in that regard, that would be very helpful. Other things that we see is that modules are prefabricated offsite.
[00:27:04] Iris D. Tommelein, PhD: They have a lot of somewhat flexible components, timber and drywall, for example. Some components are perhaps a bit more rigid than others, drywall and tile, and what we find, of course, is in the process of handling these modules and getting them maneuvered and put into their final location is that a lot of the inside components crack.
[00:27:26] Iris D. Tommelein, PhD: And so therefore there now is a lot of rework on the construction site to the point where one really needs to think about how much, and exactly what should we do offsite and how much, and exactly what should we be doing onsite, it’s not because we move the work offsite that it’s necessarily saving us everything we can save in terms of time and work at a construction site.
[00:27:50] H.J. James Choo, PhD: Yeah, it’d be actually really interesting to figure out how each organization is tackling these issues with our survey that we’re planning to send out early next year.
[00:28:02] Iris D. Tommelein, PhD: Yeah, it’ll be interesting.
[00:28:05] H.J. James Choo, PhD: Well, I think that’s all we actually had the time for. And thank you, Iris, for your presentation. And with this, I’ll turn it over to Gary.
[00:28:17] Iris D. Tommelein, PhD: Sure, thank you very much. Thank you.