WEBVTT Kind: captions Language: en-US 00:00:01.579 --> 00:00:04.560 Hello, everybody. 00:00:04.560 --> 00:00:07.440 Happy new year, and welcome back from the shutdown. 00:00:07.440 --> 00:00:11.260 So this is our first seminar of the year here. 00:00:11.260 --> 00:00:15.020 And we’ll be back to having seminars regularly 00:00:15.020 --> 00:00:18.060 every Wednesday, barring another shutdown. 00:00:18.060 --> 00:00:22.539 And next week, we’ll be having Anne Hulsey from Stanford come to talk 00:00:22.539 --> 00:00:27.090 to us about tall buildings and cordoned- off areas following earthquakes. 00:00:27.090 --> 00:00:31.340 And now I will hand things over to Tom Hanks to introduce today’s speaker. 00:00:32.460 --> 00:00:34.460 Here you go, Tom. 00:00:38.000 --> 00:00:42.000 - Hi, Jack. [laughs] Good to see you again. 00:00:42.000 --> 00:00:46.240 Excuse me. I’m still getting over a cold. 00:00:46.240 --> 00:00:54.340 Let’s see. I’ve known Jack for, well, probably too long. [laughs] 00:00:55.720 --> 00:00:59.000 2001 or something. 00:00:59.000 --> 00:01:03.570 I met him as a graduate – when he was a graduate student up at Stanford. 00:01:03.570 --> 00:01:09.220 He and Dave Shelly got undergraduate degrees at Whitman College in 00:01:09.220 --> 00:01:15.780 mathematics and physics together. And so we’ve – that’s been a great 00:01:15.780 --> 00:01:20.820 connection for us with both of them. Then he went to Stanford and 00:01:20.820 --> 00:01:25.080 got a master’s degree in structural engineering and statistics. 00:01:25.080 --> 00:01:30.920 And then a Ph.D. with Allin Cornell. And I think that’s where I first met you. 00:01:30.920 --> 00:01:38.120 And then went to ETH and was there for a year and then came back to Stanford 00:01:38.120 --> 00:01:41.440 and has been on the faculty there since. And I guess you’re 00:01:41.440 --> 00:01:43.710 a full professor by now. - Not quite. 00:01:43.710 --> 00:01:46.090 - Not quite. Okay, well, you have something to 00:01:46.090 --> 00:01:51.300 look forward to other than, of course, giving us another talk here. 00:01:52.580 --> 00:01:57.640 I got to know Jack through the Extreme Ground Motion project 00:01:57.640 --> 00:02:02.860 at Yucca Mountain. And – oh, let’s see. Let me say – 00:02:02.860 --> 00:02:07.380 he’s won all kinds of engineering awards that I’m not going to tell you too much 00:02:07.380 --> 00:02:09.860 because, more importantly, at least to us, I think, 00:02:09.860 --> 00:02:12.770 are his connections to the Earth sciences. 00:02:12.770 --> 00:02:18.590 He’s been an editor for the Bulletin of the Seismological Society of America, 00:02:18.590 --> 00:02:22.680 and he sits on the planning committee of SCEC. 00:02:22.680 --> 00:02:26.440 And I got to know him really well through the Extreme Ground Motion 00:02:26.450 --> 00:02:31.170 project at Yucca Mountain. And, among other things, 00:02:31.170 --> 00:02:40.060 Jack made a lot of probabilistic sense out of fragile geologic structures. 00:02:40.060 --> 00:02:43.430 And I was thinking the other night – or, last night, when this – have you 00:02:43.430 --> 00:02:48.060 ever thought about thinking about fragile engineering structures in the 00:02:48.060 --> 00:02:53.420 same context? You know, in terms of a population of such things? 00:02:54.560 --> 00:02:58.680 And I think the answer is probably no, but – and that’s good because he’s 00:02:58.680 --> 00:03:03.690 going to talk about use of ground motion simulations in engineering practice. 00:03:03.690 --> 00:03:06.620 Okay, thanks for being here, Jack, right. 00:03:07.240 --> 00:03:10.500 Shall I give this to you? 00:03:14.100 --> 00:03:16.940 - Good morning. Is it – can you hear that? Okay, good. 00:03:16.950 --> 00:03:19.630 Well, thanks. Thank you, Tom. And thanks to all of you for the 00:03:19.630 --> 00:03:23.710 invitation to come back and speak. I’m very pleased to be here, 00:03:23.710 --> 00:03:26.440 and I’m pleased that you’re back open for business, 00:03:26.440 --> 00:03:28.520 and we can kind of continue our engagement, 00:03:28.520 --> 00:03:32.320 and you can continue your important work that you’re doing here. 00:03:32.320 --> 00:03:36.800 I brought a few slides this morning. I think the initial conversations 00:03:36.810 --> 00:03:41.099 were around kind of ground motion simulations and tall buildings 00:03:41.100 --> 00:03:43.220 and some various things that you were interested in. 00:03:43.220 --> 00:03:47.320 And I guess with the kind of – my understanding of some interest here, 00:03:47.330 --> 00:03:50.260 I brought a few slides to kind of just give you some general background 00:03:50.260 --> 00:03:54.550 about how structural engineers think about ground motions and how ground 00:03:54.550 --> 00:03:59.569 motion simulations that geophysicists are producing might be used, or will 00:03:59.569 --> 00:04:02.970 be used, by the engineering practice. And then a little bit of kind of research 00:04:02.970 --> 00:04:06.420 that we’ve done in that area to think about how we might ask questions 00:04:06.420 --> 00:04:08.500 and answer questions that we have from the engineering community 00:04:08.500 --> 00:04:11.900 about some of the work that’s going on in the Earth sciences. 00:04:12.720 --> 00:04:16.200 So I’ll thank – so Ganyu Teng and Lynne Burks are Ph.D. students 00:04:16.209 --> 00:04:18.180 and former – Lynne is a former Ph.D. student at Stanford. 00:04:18.180 --> 00:04:20.430 He’s done a lot of the work that I’ll show here. 00:04:20.430 --> 00:04:22.620 And a couple collaborators, and then Rob Graves, Kim Olsen, 00:04:22.620 --> 00:04:26.211 Scott Callaghan were really instrumental in kind of providing a lot of the 00:04:26.211 --> 00:04:30.599 data we’re looking at here. And SCEC has been a financial 00:04:30.599 --> 00:04:34.080 supporter of this, so definitely appreciate the collegiality and the 00:04:34.080 --> 00:04:36.520 support from the Earth sciences community as we kind of wander 00:04:36.520 --> 00:04:40.160 around asking questions as engineers in this – in this domain. 00:04:40.800 --> 00:04:45.879 Okay, so to get into things, to think about how we use ground 00:04:45.879 --> 00:04:48.940 motions in earthquake engineering. So I think the – you know, the easiest 00:04:48.940 --> 00:04:53.469 conceptual thing to think about, and kind of cartoon version, is to say, you know, 00:04:53.469 --> 00:04:57.260 we could think about, you know, earthquake source models and do, 00:04:57.260 --> 00:05:00.180 you know, rupture simulation, wave propagation, produce time series, 00:05:00.190 --> 00:05:02.819 and put those into buildings and kind of see what happens to the buildings. 00:05:02.819 --> 00:05:06.879 And that gets kind of colloquially called rupture-to-rafters simulations 00:05:06.880 --> 00:05:11.100 over – kind of the entire thing is this one pipeline of numerical, 00:05:11.100 --> 00:05:14.199 you know, calculations and simulations and things like that. 00:05:14.200 --> 00:05:18.640 And that’s certainly interesting and has been impactful in some of these 00:05:18.640 --> 00:05:22.680 big projects like the ShakeOut scenario or the HayWired scenario, in terms of 00:05:22.680 --> 00:05:25.259 kind of understanding what the – what a particular earthquake scenario 00:05:25.260 --> 00:05:31.700 might produce in terms of damage. But it – and so I don’t mean to be critical 00:05:31.700 --> 00:05:34.639 of any of those efforts, but only to point out, that’s in contrast to the 00:05:34.639 --> 00:05:37.749 way that most engineers really interface with this type of data in 00:05:37.749 --> 00:05:41.300 their day-to-day work and in the practice of designing buildings or 00:05:41.300 --> 00:05:44.909 analyzing individual buildings. And we take kind of a more circuitous route. 00:05:44.909 --> 00:05:48.659 And I’ll – to the extent that this is not familiar to you, hopefully this is 00:05:48.659 --> 00:05:52.710 something educational and – that I can both tell you a little bit about the process 00:05:52.710 --> 00:05:55.480 and kind of how this all came about. 00:05:55.480 --> 00:05:58.100 So really, as I think about it, we’ve got kind of a first stage 00:05:58.100 --> 00:05:59.919 of analysis, which is – which is hazard analysis. 00:05:59.919 --> 00:06:04.180 And that – the Survey is very familiar with as kind of the holder of the 00:06:04.180 --> 00:06:07.809 national seismic hazard assessments and a lot of research and other projects 00:06:07.809 --> 00:06:10.030 going on in that area. But we really go from kind of 00:06:10.030 --> 00:06:13.699 the earthquake ruptures and wave propagation and we do this probabilistic 00:06:13.699 --> 00:06:16.889 assessment of all possible ruptures and all possible, you know, 00:06:16.889 --> 00:06:19.409 permutations of ground motions that could result from these ruptures. 00:06:19.409 --> 00:06:22.629 And there’s that whole process that goes on. 00:06:22.629 --> 00:06:24.759 But the outcome of that, from an engineering perspective, 00:06:24.759 --> 00:06:29.300 is a design response spectrum. And so this is loosely based on 00:06:29.300 --> 00:06:32.660 ground motion amplitudes in terms of spectral accelerations. 00:06:32.660 --> 00:06:35.520 And it’s loosely based on – well, it is based on spectral accelerations. 00:06:35.520 --> 00:06:38.990 It’s loosely based on amplitudes that are exceeded with something like a, 00:06:38.990 --> 00:06:42.470 you know, target return period, although there’s modifiers for caps 00:06:42.470 --> 00:06:44.969 and risk adjustments and things like that. 00:06:44.969 --> 00:06:50.030 But it’s ultimately driven by kind of the frequency of simulations or calculations 00:06:50.030 --> 00:06:53.779 that predict a ground motion amplitude of some sort. 00:06:53.779 --> 00:06:57.160 And ground motion simulations play an important role here, increasingly. 00:06:57.160 --> 00:07:00.569 So, you know, traditionally the ground motion amplitude predictions 00:07:00.569 --> 00:07:05.469 have been based on kind of empirical models. 00:07:05.469 --> 00:07:08.370 But increasingly, we’re thinking about ways to use simulations there that, 00:07:08.370 --> 00:07:11.089 in Los Angeles, there’s activity going on around SCEC. 00:07:11.089 --> 00:07:13.809 In Seattle, Art Frankel has been, you know, pushing at these 00:07:13.809 --> 00:07:16.240 directions for a long time. I know there’s work in Memphis, 00:07:16.240 --> 00:07:20.599 and the Survey’s got their involvement in kind of all of those efforts. 00:07:20.599 --> 00:07:28.509 So certainly that target response spectrum could be computed 00:07:28.509 --> 00:07:31.110 with consideration of ground motion simulations and the particular insights 00:07:31.110 --> 00:07:33.500 they might provide about the amplitude of shaking that any 00:07:33.500 --> 00:07:36.529 particular site could see. And that’s important. 00:07:36.529 --> 00:07:40.539 That is kind of almost entirely an Earth science problem, and I’m interested in it, 00:07:40.539 --> 00:07:43.250 but that’s not kind of what I’ll focus on today. 00:07:43.250 --> 00:07:45.899 What I thought would be interesting to talk about with you today is 00:07:45.899 --> 00:07:48.800 really kind of a second step. And this is where the engineers – 00:07:48.800 --> 00:07:51.470 practicing engineers typically come into the picture. 00:07:51.470 --> 00:07:56.139 And so an engineer will come along, and whether they’re using the 00:07:56.139 --> 00:07:59.080 USGS hazard maps, or they’re using some site-specific study 00:07:59.080 --> 00:08:01.900 that’s been performed, they’ll start with that target response spectrum. 00:08:01.900 --> 00:08:04.800 And the engineer is not doing any sort of calculations themselves to get that. 00:08:04.800 --> 00:08:07.270 That’s coming from some sort of external group. 00:08:07.270 --> 00:08:10.740 And then they are going to do some analysis to assess the 00:08:10.740 --> 00:08:15.550 performance of their structure under ground shaking with that intensity. 00:08:15.550 --> 00:08:19.699 And I should say – so the focus of my talk will be about, you know, 00:08:19.700 --> 00:08:22.160 how might simulated ground motions play a role here? 00:08:22.160 --> 00:08:26.379 Because that’s another stage in which we will use ground motions. 00:08:26.379 --> 00:08:28.689 We typically use many fewer ground motions. 00:08:28.689 --> 00:08:32.520 So on the order of a dozen or so ground motions rather than, you know, 00:08:32.520 --> 00:08:34.280 thousands or hundreds of thousands that might be done 00:08:34.280 --> 00:08:36.620 in the hazard analysis stage. 00:08:37.340 --> 00:08:40.560 And actually, typically, we don’t use any ground motions at all. 00:08:40.570 --> 00:08:42.680 So I should note, and maybe some of you have heard this before, 00:08:42.680 --> 00:08:49.210 but 99% of structures are designed or analyzed using some sort of a 00:08:49.210 --> 00:08:53.320 static loading that’s deemed equivalent to the dynamic loading that would be 00:08:53.320 --> 00:08:56.050 associated with that response spectrum. So most of the time, we’re not building 00:08:56.050 --> 00:08:58.820 dynamic models or analyzing dynamic models of the way 00:08:58.820 --> 00:09:02.890 that structures perform. It’s just a – it’s a complicated and 00:09:02.890 --> 00:09:05.620 labor-intensive effort, and it’s deemed kind of not necessary 00:09:05.620 --> 00:09:08.560 for your Home Depots or for your residential houses 00:09:08.560 --> 00:09:11.770 or your three-story office buildings and things like that. 00:09:11.770 --> 00:09:15.310 For special structures – so high-rise buildings or base-isolated buildings 00:09:15.310 --> 00:09:20.080 or hospitals – you know, certain classes of structures that are special or have 00:09:20.080 --> 00:09:22.950 something unusual in their structural performance, or there’s something about 00:09:22.950 --> 00:09:26.010 the system that’s really – you need to understand the dynamic behavior, 00:09:26.010 --> 00:09:28.860 then we go to this domain. And those are the – you know, 00:09:28.860 --> 00:09:32.540 those are typically – you know, typically good engineers that have a, 00:09:32.540 --> 00:09:34.530 you know, understanding of advanced analysis techniques. 00:09:34.530 --> 00:09:36.670 There’s more resources available for the analysis. 00:09:36.670 --> 00:09:39.500 And that’s a place where simulated ground motions could play 00:09:39.500 --> 00:09:43.300 an important role, and that’s really kind of the piece where myself, as a 00:09:43.300 --> 00:09:47.110 trained structural engineer, I’m kind of interested in interfacing to. 00:09:47.110 --> 00:09:49.950 So hopefully – with that background, hopefully that’s, you know, 00:09:49.950 --> 00:09:51.800 educational or interesting to some of you. 00:09:51.800 --> 00:09:54.300 And I want – let’s dig into that a little more. 00:09:54.300 --> 00:09:57.620 So today, I’ll think about this kind of right-hand side of the picture. 00:09:57.620 --> 00:10:00.020 This passage from the target response spectrum 00:10:00.020 --> 00:10:02.190 on to assessing structural response. 00:10:02.190 --> 00:10:05.560 And I should say, kind of traditionally, this has been done using 00:10:05.560 --> 00:10:08.580 recorded ground motions. So we’ll take recordings from some 00:10:08.580 --> 00:10:11.950 sort of earthquake deemed comparable to the one that we’re thinking about 00:10:11.950 --> 00:10:13.840 that might affect our structure. 00:10:13.840 --> 00:10:18.760 Maybe we’ll do some processing of that ground motion and proceed. 00:10:18.760 --> 00:10:22.380 And so I’ll be interested in saying, well, you know, if we did this with recordings 00:10:22.390 --> 00:10:24.590 or if we did this with simulations, you know, what would be the 00:10:24.590 --> 00:10:26.170 differences and what would be the pros and cons? 00:10:26.170 --> 00:10:29.080 I want to talk about that a little bit through a few examples today. 00:10:29.780 --> 00:10:32.320 And I’ll tell you a little bit about the specific requirements 00:10:32.330 --> 00:10:34.960 that go into that as well. So here is some more background 00:10:34.960 --> 00:10:38.650 that, again, might be interesting or educational for some of you. 00:10:39.100 --> 00:10:41.720 There’s a few different documents that might guide an engineer in 00:10:41.720 --> 00:10:44.860 doing these types of assessments. In most cases, it would be 00:10:44.860 --> 00:10:48.860 this ASCE 7 standard. So this is a technical standard 00:10:48.860 --> 00:10:52.500 that’s produced by the – a few different agencies, 00:10:52.500 --> 00:10:54.580 but the American Society of Civil Engineers and Structural 00:10:54.580 --> 00:10:58.630 Engineering Institute publish this. And then a lot of building codes will 00:10:58.630 --> 00:11:02.560 then, by reference, adopt this and say, the loadings on your buildings will 00:11:02.560 --> 00:11:06.300 have to conform to this standard. And then that building code is going to 00:11:06.300 --> 00:11:08.380 drive the acceptance of a particular building design 00:11:08.380 --> 00:11:12.690 in a particular jurisdiction. So through kind of, you know, 00:11:12.690 --> 00:11:14.810 a few different stages, this is the document that ultimately is 00:11:14.810 --> 00:11:19.010 specifying what has to happen. And it’ll say, you know, 00:11:19.010 --> 00:11:22.630 under what circumstances you actually have to do dynamic analysis. 00:11:22.630 --> 00:11:26.320 And if you do choose to do dynamic analysis, it’s Chapter 16 of this thing, 00:11:26.320 --> 00:11:30.450 for those that are interested. And there’s, you know, a dozen pages 00:11:30.450 --> 00:11:32.750 or so kind of talking through what the requirements are. 00:11:32.750 --> 00:11:35.180 And I just wanted to grab a few that might be interesting. 00:11:35.180 --> 00:11:39.480 So it’s going to reference off to the hazard assessment and 00:11:39.480 --> 00:11:42.000 reference off to the target spectrum. That’s described elsewhere. 00:11:42.000 --> 00:11:44.390 And so we have this language. You’ve probably read it 00:11:44.390 --> 00:11:46.830 a couple minutes ago. So we’re going to have to select 00:11:46.830 --> 00:11:49.880 ground motions from events with the same tectonic regime, 00:11:49.880 --> 00:11:52.490 having generally consistent magnitude and fault distances and those controlling 00:11:52.490 --> 00:11:56.680 the target spectrum and have similar spectral shape to that target spectrum. 00:11:56.680 --> 00:12:00.380 So that’s that kind of spectrum that I drew on the previous page. 00:12:00.380 --> 00:12:04.150 So the engineer will go determine that spectrum from some source 00:12:04.150 --> 00:12:07.150 and then use something like de-aggregation to figure out 00:12:07.150 --> 00:12:10.110 magnitudes and distances that are interesting. 00:12:10.110 --> 00:12:12.830 And then you can kind of identify tectonic regime pretty quickly. 00:12:12.830 --> 00:12:15.620 So we would – you know, sitting here, if we were going to put a new building in 00:12:15.620 --> 00:12:18.670 this neighborhood, we’d go get our target spectrum, which would be pretty high. 00:12:18.670 --> 00:12:22.430 We’d see that the hazard is controlled by probably magnitude 8 San Andreas 00:12:22.430 --> 00:12:26.020 events at 15 kilometers away, or whatever it is, to here. 00:12:26.020 --> 00:12:30.340 And then we’ll go try to, you know, find some ground motions to conform to that. 00:12:30.340 --> 00:12:35.700 In terms of finding them, we – let’s see. Somewhere in here, we should 00:12:35.700 --> 00:12:38.380 have some sort of recordings. Oh, yeah. So there’s an implicit 00:12:38.380 --> 00:12:40.000 preference for recorded ground motions. 00:12:40.000 --> 00:12:42.910 And that’s just kind of reflecting historical practice rather than some sort 00:12:42.910 --> 00:12:46.290 of conscious decision, I think, talking to the code writers that are in this. 00:12:46.290 --> 00:12:48.040 So we’re going to go look for recordings. 00:12:48.040 --> 00:12:50.190 And then it says, when they’re not available, 00:12:50.190 --> 00:12:52.660 you can use – supplement those with simulated ground motions. 00:12:52.660 --> 00:12:56.810 And this is, I think, historically – some of this language, I think, 00:12:56.810 --> 00:12:59.740 dates back 20 or 30 years to where the balance between – 00:12:59.740 --> 00:13:02.290 well, there are many fewer recordings and then also many fewer 00:13:02.290 --> 00:13:05.060 simulations and much less-refined simulation technology. 00:13:05.060 --> 00:13:09.610 And I think this is largely applied in the eastern U.S., where there are 00:13:09.610 --> 00:13:13.740 just more recordings in stable continental regions appropriate for New Madrid-type 00:13:13.740 --> 00:13:16.240 events or something like that. And so they put in this thing. 00:13:16.240 --> 00:13:19.870 So, both in the – in the – in these words, as well as in professional practice, 00:13:19.870 --> 00:13:24.220 there is this preference toward recorded ground motions if present. 00:13:24.220 --> 00:13:28.560 And just one note that, you know, buildings are three-dimensional, and so 00:13:28.560 --> 00:13:33.060 we need – at a minimum, we need two horizontal components of ground motion. 00:13:33.060 --> 00:13:36.590 And you do some evaluations to decide if vertical shaking is going to 00:13:36.590 --> 00:13:39.660 be important for your building or not. In a lot of cases, it’s not really a big 00:13:39.660 --> 00:13:43.550 driver of demands in the structure. But, if you need a – if you think vertical 00:13:43.550 --> 00:13:46.300 shaking is important, then you have to go get the vertical component as well. 00:13:46.300 --> 00:13:48.920 So we need at least two component, maybe three component ground motions. 00:13:48.920 --> 00:13:51.880 So that’s just something that influences the suitability of 00:13:51.880 --> 00:13:53.640 some ground motion simulation algorithms. 00:13:53.640 --> 00:13:57.280 So that’s increasingly not an issue for simulators. 00:13:58.320 --> 00:14:04.160 Okay, so this document gets updated every five years or so – 00:14:04.160 --> 00:14:07.970 the latest round coming out in 2016. 00:14:07.970 --> 00:14:13.320 And so, in 2013 or ’14, when this was kind of undergoing revision, 00:14:13.320 --> 00:14:16.060 we did a few studies to kind of look at updates to the language 00:14:16.060 --> 00:14:19.010 and what impact it would have. And one is, we wanted to put a little 00:14:19.010 --> 00:14:22.710 more language in the commentary especially, but just do some experiments 00:14:22.710 --> 00:14:25.960 about kind of how simulations might work in these situations. 00:14:25.960 --> 00:14:28.960 And so, for a few slides here, I’ll show you one example of a study 00:14:28.960 --> 00:14:32.720 that kind of went along with that code update that Lynne Burks, the student 00:14:32.720 --> 00:14:36.529 at Stanford, and a practicing engineer and myself did a few years ago. 00:14:36.529 --> 00:14:41.520 So this practicing engineer, Reid Zimmerman, was one of the 00:14:41.520 --> 00:14:44.570 people who was tasked with saying, you know, as we update this code 00:14:44.570 --> 00:14:47.490 language, we want to make sure that we don’t accidentally kind of trigger 00:14:47.490 --> 00:14:51.290 some very – you know, relaxation of our rules or a big increase 00:14:51.290 --> 00:14:54.160 in the rules that was unintentional. And so we take a series of buildings 00:14:54.160 --> 00:14:57.490 that satisfied the previous code and kind of re-analyze them under the new 00:14:57.490 --> 00:15:00.730 building code just to try to understand, are we – do we think we’re doing 00:15:00.730 --> 00:15:03.370 a better job in terms of the requirements, or did we miss something? 00:15:03.370 --> 00:15:07.350 And so this was a building that he had recently designed and worked with. 00:15:07.350 --> 00:15:11.210 So it’s kind of a mid-rise building. It’s the type of building you wouldn’t 00:15:11.210 --> 00:15:13.930 probably typically analyze dynamically, but he had a dynamic model that he had 00:15:13.930 --> 00:15:18.970 built for other reasons consisting of kind of – so the top figure is a – 00:15:18.970 --> 00:15:22.140 kind of a picture of just the lateral force-resisting system, 00:15:22.140 --> 00:15:25.400 which is the important part of the building for earthquakes. 00:15:25.400 --> 00:15:27.860 And then the bottom picture is a plan view showing kind of the 00:15:27.860 --> 00:15:31.590 lines of the particular framing that would be resisting earthquake forces. 00:15:33.120 --> 00:15:36.160 So it’s kind of moment frames, braced frames – the typically structural 00:15:36.160 --> 00:15:39.210 systems we see kind of all over here for office buildings, residential buildings, 00:15:39.210 --> 00:15:43.860 and things. This is actually a dormitory on the Berkeley campus. 00:15:43.860 --> 00:15:46.600 So I think you all probably know the Bay Area picture pretty well. 00:15:46.600 --> 00:15:48.600 The building is kind of up in the upper left. 00:15:48.600 --> 00:15:51.100 And it’s about a kilometer away from the Hayward Fault, 00:15:51.100 --> 00:15:53.490 just right at the south end of the Berkeley campus. 00:15:53.490 --> 00:15:59.210 So in red, which may not be real clear there, but you – but you can 00:15:59.210 --> 00:16:00.740 see the Berkeley stadium with the Hayward Fault, you know, 00:16:00.740 --> 00:16:02.180 passing right through the middle of there. 00:16:02.180 --> 00:16:04.290 And we’re just a kilometer off to this building location. 00:16:04.290 --> 00:16:09.370 So this is an interesting, you know, example in that the – so Berkeley had 00:16:09.370 --> 00:16:12.690 kind of – when they built this dorm, had done some extra analysis 00:16:12.690 --> 00:16:14.500 the first time through. And this is a case where you might say, 00:16:14.500 --> 00:16:16.680 well, simulations would be great because I don’t have that many 00:16:16.680 --> 00:16:20.020 observations of these very near-fault ground motions strong intensities, 00:16:20.020 --> 00:16:21.520 but I can simulate some. 00:16:21.520 --> 00:16:24.350 And, you know, that would be convenient instead of having to 00:16:24.350 --> 00:16:28.560 scrape around in my recordings to try to find something this close to a fault. 00:16:29.110 --> 00:16:33.660 So what we did is we went through two parallel selection exercises. 00:16:33.660 --> 00:16:38.000 So, on the left, we used recordings from the PEER ground motion database. 00:16:38.000 --> 00:16:42.540 So this is a database, at the time, was a 3,000-or-so kind of 00:16:42.540 --> 00:16:45.700 strong ground motion recordings. Now somewhere between 8,000 00:16:45.700 --> 00:16:50.130 and 20,000, depending on kind of what you count as relevant for engineers. 00:16:50.130 --> 00:16:53.220 And that would be the traditional practice even today is you’d go to 00:16:53.220 --> 00:16:56.750 a database like this, and you’d go filter it and look for kind of 00:16:56.750 --> 00:17:00.550 matching magnitudes, distances. At the time, this was all shallow 00:17:00.550 --> 00:17:03.200 crustal earthquakes, so the tectonic regime matched naturally. 00:17:03.200 --> 00:17:05.420 And now they’ve got permutations in the database 00:17:05.420 --> 00:17:08.270 for other types of tectonic regimes. 00:17:08.270 --> 00:17:10.660 And then we also used the SCEC broadband platform. 00:17:10.660 --> 00:17:14.240 So this is the software platform that implements a few different 00:17:14.240 --> 00:17:19.860 ground motion simulation algorithms and in a way that is kind of intended to 00:17:19.860 --> 00:17:22.600 facilitate other types of studies where you don’t have to kind of 00:17:22.600 --> 00:17:26.340 go to the scientist themself and have them produce ground motions. 00:17:26.340 --> 00:17:28.640 And so we thought that was kind of well-suited that there’s some potential 00:17:28.659 --> 00:17:31.850 that, you know, engineering practice could adopt a tool like that or engage 00:17:31.850 --> 00:17:35.230 with people who are able to use tools like that to produce ground motions. 00:17:35.230 --> 00:17:38.870 So this particular study will use – there’s kind of multiple algorithms built into it, 00:17:38.870 --> 00:17:44.039 but we used – most of the pieces of it were from Graves and Pitarka with 00:17:44.039 --> 00:17:46.960 one component from San Diego State from Kim Olsen. 00:17:46.960 --> 00:17:51.679 And this is a – I should note this is a – it’s called kind of hybrid broadband 00:17:51.679 --> 00:17:54.960 ground motion simulations, for those of you in the – in the business. 00:17:54.960 --> 00:17:57.490 So there’s a – there’s kind of the low frequencies are produced 00:17:57.490 --> 00:18:03.610 using a deterministic kind of a rupture realization and wave propagation. 00:18:03.610 --> 00:18:06.369 But we also need high-frequency ground motions for these engineering analyses. 00:18:06.369 --> 00:18:10.419 We need kind of a full broadband set of energy input to the structure. 00:18:10.419 --> 00:18:12.999 And so that’s produced using a separate algorithm, 00:18:12.999 --> 00:18:16.539 which is the stochastic algorithm, and so those are stitched together. 00:18:16.539 --> 00:18:19.119 And so this is just referencing the particular algorithms, 00:18:19.120 --> 00:18:22.700 but the general workflow could work with other types of algorithms. 00:18:24.000 --> 00:18:28.480 Okay, so then a little hesitant to sometimes put up tables in PowerPoints, 00:18:28.490 --> 00:18:31.119 but to give you a sense of what would happen. 00:18:31.119 --> 00:18:34.740 So the current – the building code we were looking at – ASCE 7-16 – 00:18:34.740 --> 00:18:37.240 requires 11 ground motions as inputs to these analyses. 00:18:37.240 --> 00:18:39.830 So we’ve got kind of 11 rows in the table. 00:18:39.830 --> 00:18:43.169 On the left is what we would typically come up with with current practice. 00:18:43.169 --> 00:18:47.110 So these are recordings from kind of our favorite earthquakes – Imperial Valley, 00:18:47.110 --> 00:18:49.630 Loma Prieta, Northridge. We’ll go to, you know, other places 00:18:49.630 --> 00:18:54.100 in the world, like Chi-Chi, maybe to get a larger magnitude event, or a few. 00:18:54.100 --> 00:18:56.559 And then we’re looking for, you know, similar magnitudes. 00:18:56.559 --> 00:19:01.749 So we’ve got kind of 6-1/2, 6.7, 7.6 – and I should, sorry, back up that – 00:19:01.749 --> 00:19:05.679 so, given the location of that site, we’re looking for magnitude 7 00:19:05.679 --> 00:19:09.049 Hayward-type ruptures. So I’d like a magnitude 7 strike-slip, 00:19:09.049 --> 00:19:12.490 you know, rupture, and I’d like to have a recording a kilometer away from it. 00:19:12.490 --> 00:19:15.379 You know, these recordings on the left are close – you know, 00:19:15.379 --> 00:19:18.700 between 1 and 12 kilometers – but they’re not exactly 1. 00:19:18.700 --> 00:19:20.710 The site conditions are going to vary a little bit just based on 00:19:20.710 --> 00:19:25.200 where the instruments are. And then, because this design spectrum 00:19:25.200 --> 00:19:28.950 is very large – like, we’re designing for very rare ground motions. 00:19:29.680 --> 00:19:33.100 I may have alluded to it loosely. Again, it’s a little more 00:19:33.110 --> 00:19:34.730 complicated than this. We can think about ground motions 00:19:34.730 --> 00:19:37.440 that occur – or, ground motions that are exceeded once every 00:19:37.440 --> 00:19:42.539 2,500 years in terms of those amplitudes. And if the Hayward rupture is something 00:19:42.539 --> 00:19:45.950 that’s happening every 200 or 300 years, our design amplitudes are going to be 00:19:45.950 --> 00:19:48.779 much bigger than an average ground motion from a Hayward-type rupture. 00:19:48.779 --> 00:19:51.039 We’re thinking about, like, a Hayward rupture, but, you know, 00:19:51.039 --> 00:19:54.330 a factor of 2 or 3 larger in amplitude, given the variability from rupture to 00:19:54.330 --> 00:19:57.999 rupture we might see. So very large amplitudes we’re designing for. 00:19:57.999 --> 00:20:01.499 And so that final column of the recordings shows scale factors. 00:20:01.499 --> 00:20:03.190 We try to limit them, but, you know, some of these are 00:20:03.190 --> 00:20:07.549 doubled or tripled in amplitude relative to what was recorded as a way to 00:20:07.549 --> 00:20:10.460 bump up the ground motion amplitudes up to that design spectrum. 00:20:10.460 --> 00:20:14.120 And so, you know, this business of trying to kind of mix – you know, 00:20:14.120 --> 00:20:18.080 make these recordings work for the case we’re thinking about is not anything 00:20:18.080 --> 00:20:22.190 that’s – you know, we’re excited about, but that’s kind of the best we can do 00:20:22.190 --> 00:20:26.020 in terms of getting realistic dynamic loading for the structure at present. 00:20:26.020 --> 00:20:29.490 So then, over on the right, the story is a lot cleaner conceptually, right? 00:20:29.490 --> 00:20:34.149 I can just set up the broadband platform to generate ruptures of magnitude 7 00:20:34.149 --> 00:20:37.440 vertical strike-slip earthquakes, and I can put receivers a kilometer away 00:20:37.440 --> 00:20:41.249 from the simulation, and then we generate it on the order of, I think, 00:20:41.249 --> 00:20:45.399 150 simulations, or 200 simulations, under those conditions. 00:20:45.399 --> 00:20:48.540 And then the – you know, out of that, there was enough that we’re close to 00:20:48.540 --> 00:20:51.261 the design spectrum that we could take – we don’t have to scale them. 00:20:51.261 --> 00:20:54.389 We can just generate more simulations until we get suitable ones. 00:20:54.389 --> 00:20:55.860 So there’s no need to kind of post-process. 00:20:55.860 --> 00:20:59.470 We can specify the site conditions, and so on. 00:20:59.470 --> 00:21:02.679 And so something that’s kind of a much cleaner conceptual fit 00:21:02.679 --> 00:21:05.529 with what you’re trying to achieve, and that would be the – I think, 00:21:05.529 --> 00:21:08.580 where engineers would find a lot of value out of these simulations, 00:21:08.580 --> 00:21:12.059 particularly if – you know, if we could, you know, continue adding features 00:21:12.059 --> 00:21:14.499 about the particular region or the particular fault or the 00:21:14.499 --> 00:21:17.559 particular site that would, you know, produce realistic waveforms 00:21:17.560 --> 00:21:19.999 for what would anticipate in the future. 00:21:20.800 --> 00:21:24.580 Okay, so here’s a plot of – so these are response spectra. 00:21:25.540 --> 00:21:28.020 Let me point out a few things. So on the left is the set of recordings. 00:21:28.020 --> 00:21:30.860 On the right is the set of simulations. 00:21:31.420 --> 00:21:34.809 So – and so I should say, so on the horizontal axis, this is the period, 00:21:34.809 --> 00:21:36.910 or the reciprocal of the frequency. 00:21:36.910 --> 00:21:38.999 And on the vertical axis is the spectral acceleration. 00:21:38.999 --> 00:21:41.210 And this is – since we’re going to talk about response spectra 00:21:41.210 --> 00:21:44.179 for a few slides here, just to make sure everybody’s on the same page. 00:21:44.179 --> 00:21:47.830 So this is – we’re measuring kind of the amplitude of an elastic oscillator 00:21:47.830 --> 00:21:52.710 that has the specified period and has some – it’s kind of lightly damped. 00:21:52.710 --> 00:21:58.470 So it’s – in some senses, it’s analogous to, like, a smoothed Fourier amplitude. 00:21:58.470 --> 00:22:02.129 But it – but it’s not. It’s a – it’s a response of an oscillator. 00:22:02.129 --> 00:22:06.499 And so a couple things to point out here. So the dashed black line is – 00:22:06.499 --> 00:22:09.830 at the very bottom of all these plots – is what we would anticipate is the 00:22:09.830 --> 00:22:15.040 median shaking from a Hayward-like scenario at 1 kilometer away. 00:22:15.040 --> 00:22:17.320 And then the solid black line is our design spectrum. 00:22:17.320 --> 00:22:19.679 So this is what I was pointing out that this kind of factor of 00:22:19.679 --> 00:22:23.649 almost 2 bigger amplitude than the median response spectrum. 00:22:23.649 --> 00:22:25.899 So we’re designing for these very large ground motions 00:22:25.900 --> 00:22:29.180 that we expect to occur, but to occur very rarely. 00:22:29.180 --> 00:22:33.420 And that, you know, is the right thing to do but creates problems in terms of 00:22:33.420 --> 00:22:37.280 finding ground motions that are going to match that naturally. 00:22:37.940 --> 00:22:42.100 So then the light lines in both figures are the response spectra 00:22:42.110 --> 00:22:45.600 of individual ground motions. And then the heavy blue line is the 00:22:45.600 --> 00:22:48.309 mean of those ground motions, and that mean needs to be kind of 00:22:48.309 --> 00:22:52.340 equal to or above the design spectrum over the range of periods that’s 00:22:52.340 --> 00:22:54.580 kind of deemed important for the response of the building. 00:22:54.580 --> 00:22:56.660 And there’s various ways to determine that. 00:22:57.389 --> 00:23:00.220 So the point here was just to – we don’t need to worry too much about the 00:23:00.220 --> 00:23:03.350 subtleties of the shapes, but just to note that the recordings and the simulations 00:23:03.350 --> 00:23:08.360 have basically comparable amplitudes as measured via response spectra. 00:23:09.740 --> 00:23:12.480 Here are some time histories. Sometimes I’ll have people guess 00:23:12.480 --> 00:23:14.720 kind of which ones are the recordings or the simulations, but they’re actually 00:23:14.720 --> 00:23:16.960 in kind of the same order as the previous slide. 00:23:16.960 --> 00:23:20.279 So on the left is the velocity time histories of some of the recordings. 00:23:20.279 --> 00:23:22.009 On the right, comparable for the simulations. 00:23:22.009 --> 00:23:25.269 And so, you know, those look like real ground motions. 00:23:25.269 --> 00:23:27.309 They have the complexities of real ground motions. 00:23:27.309 --> 00:23:32.140 These aren’t some sort of kind of idealized white noise-type times series, 00:23:32.140 --> 00:23:34.059 but they’ve got, you know, kind of all the complexities 00:23:34.059 --> 00:23:36.820 that the simulators think are appropriate in here. 00:23:36.820 --> 00:23:40.970 So in terms of amplitudes, everything looks comparable from that response 00:23:40.970 --> 00:23:43.809 spectra in terms of time series. Certainly, those would be something 00:23:43.809 --> 00:23:47.340 that anybody would be comfortable with from an engineering perspective. 00:23:47.340 --> 00:23:51.029 And then we go put these into the dynamic model. 00:23:51.029 --> 00:23:53.950 And so here’s a couple videos of – since the model is dynamic, 00:23:53.950 --> 00:23:56.929 you can actually kind of measure what’s going on over time. 00:23:56.929 --> 00:24:00.320 And the engineer will then monitor kind of displacements of particular 00:24:00.320 --> 00:24:03.070 points in the building, measure the forces that individual components 00:24:03.070 --> 00:24:06.190 in the building are having to carry, and then we’ll check those against 00:24:06.190 --> 00:24:09.820 kind of allowable displacements and forces and accelerations. 00:24:09.820 --> 00:24:12.179 So just to make tangible that there is kind of this 00:24:12.179 --> 00:24:16.480 full dynamic analysis going on in the design office. 00:24:16.480 --> 00:24:24.059 So what we did – or, what Reid did, our collaborator – he ran the full, 00:24:24.059 --> 00:24:26.869 you know, set of dynamic analysis for each of the 11 recordings, 00:24:26.869 --> 00:24:30.059 for each of the 11 simulations, and then we’re typically interested 00:24:30.059 --> 00:24:32.989 in kind of peak displacements and peak forces up, you know, that are resulting 00:24:32.989 --> 00:24:37.480 from these. And then we’ll continue on to do the checking with those. 00:24:40.700 --> 00:24:43.820 Okay, so this is the type of things that we’ll look at out the 00:24:43.820 --> 00:24:48.680 back of those analyses. So these are plots of story drift ratio 00:24:48.690 --> 00:24:51.549 in the building, and this is in the fault normal direction that the building was 00:24:51.549 --> 00:24:55.120 roughly oriented kind of perpendicular and parallel to the fault. 00:24:55.120 --> 00:24:58.970 So in the fault normal direction. We’ve got plots of this story drift ratio, 00:24:58.970 --> 00:25:02.879 which is, at each story, the peak displacement of the top of the story 00:25:02.879 --> 00:25:05.970 relative to the bottom of the story divided by the story height. 00:25:05.970 --> 00:25:08.370 So it’s kind of like a rotation in radians. 00:25:08.370 --> 00:25:11.650 The biggest rotation we got kind of at that story. 00:25:11.650 --> 00:25:15.769 And things like, you know, 4% or so is – we can manage 00:25:15.769 --> 00:25:20.640 to design these types of frames to resist that type of deformation. 00:25:20.640 --> 00:25:24.720 And, you know, 1 or 2 or 3% is – we’re going to see damage to drywall 00:25:24.730 --> 00:25:27.440 and things like that, but the building will be safe for the occupants. 00:25:27.440 --> 00:25:29.980 And so that’s the type of thing we’re checking. 00:25:30.710 --> 00:25:34.900 The plots of the height of the building will look like this, in that, from the 00:25:34.910 --> 00:25:38.190 zero to 1, there was a concrete wall basement that’s very stiff and rigid, 00:25:38.190 --> 00:25:40.450 and the displacements aren’t a problem down there. 00:25:40.450 --> 00:25:42.669 And in the upper stories, we get these drifts. 00:25:42.669 --> 00:25:45.450 And so the light gray lines are the profiles from each 00:25:45.450 --> 00:25:49.590 individual dynamic analysis. And then the red line is the average 00:25:49.590 --> 00:25:52.730 of those kind of story by story, so the average in the horizontal direction. 00:25:52.730 --> 00:25:57.039 And it’s the average of the 11 that we’re checking against with the 00:25:57.040 --> 00:26:01.880 motivation being that we understand there will be variability from simulation 00:26:01.880 --> 00:26:04.900 to simulation, but with only 11 ground motions, which is 00:26:04.909 --> 00:26:07.320 kind of practically what’s doable in a design office. 00:26:07.320 --> 00:26:11.880 We’re not actually going to capture the full variability in the – in the – I’m going 00:26:11.880 --> 00:26:14.989 to say kind of the extreme large demand. So there’s kind of other factors built 00:26:14.989 --> 00:26:17.730 into the codes to account for that. And the engineer’s job is just to 00:26:17.730 --> 00:26:21.549 figure out, kind of on average, what would those deformations be. 00:26:21.549 --> 00:26:24.080 So we got a set of results from the recordings and the simulations. 00:26:24.080 --> 00:26:27.330 Hard to see anything dramatic there, so we can stack those on top of each 00:26:27.330 --> 00:26:31.659 other to do some more assessment. So on the left is that same data from the 00:26:31.660 --> 00:26:35.880 previous slide but just the mean values. And in blue are the recordings. 00:26:35.880 --> 00:26:37.760 And in red are the simulations. 00:26:37.760 --> 00:26:41.320 So we got 10% bigger demands out of the simulations. 00:26:41.320 --> 00:26:44.620 And on the right is the same thing but in the fault parallel direction. 00:26:44.620 --> 00:26:47.970 And the reverse happened there where the simulations produced less demands. 00:26:47.970 --> 00:26:51.190 So a couple comments here is that, because the structural systems were 00:26:51.190 --> 00:26:56.340 different in the two directions, that’s the primary reason why the story 00:26:56.340 --> 00:26:59.419 drift ratios on the right are smaller. So it’s a stiffer structural system in that 00:26:59.419 --> 00:27:02.049 fault parallel direction, and the system can just 00:27:02.049 --> 00:27:03.380 resist deformations quite a bit more. 00:27:03.380 --> 00:27:06.289 So the structural system plays a – you know, a much bigger role 00:27:06.289 --> 00:27:11.640 in the amount of deformations that are getting produced than the input source. 00:27:12.540 --> 00:27:15.240 And then, in terms of these differences on the order of 10% or so, 00:27:15.240 --> 00:27:19.000 kind of digging into it, it turns out it was just that the simulations were 00:27:19.000 --> 00:27:23.090 more strongly polarized with larger amplitudes in the fault normal direction. 00:27:23.090 --> 00:27:26.600 With the recordings, we had maintained normal fault parallel orientations of the 00:27:26.600 --> 00:27:28.919 recordings and kind of applied those appropriately to the building. 00:27:28.919 --> 00:27:33.409 But I think, because of the variety of earthquakes, 00:27:33.409 --> 00:27:35.669 we didn’t have all strike-slip earthquakes. 00:27:35.669 --> 00:27:37.360 We didn’t have all of them very close into the fault. 00:27:37.360 --> 00:27:40.960 We were probably unlikely to be capturing the near-fault polarization 00:27:40.960 --> 00:27:43.820 of the ground shaking, whereas the simulations can handle that 00:27:43.820 --> 00:27:46.429 quite naturally because we can get exactly the conditions we want. 00:27:46.429 --> 00:27:52.149 So I think this is a case where everybody found things pretty satisfying. 00:27:52.149 --> 00:27:58.059 So, to a first order, 10% is not something that we’re going to lose sleep over 00:27:58.059 --> 00:28:01.379 from an engineering design perspective. I had an engineering professor 00:28:01.379 --> 00:28:03.729 a long time ago say, you know, we’re not designing Steinways 00:28:03.729 --> 00:28:06.269 with these things. Like, there’s a lot of uncertainty in the analysis. 00:28:06.269 --> 00:28:07.700 There’s a lot of uncertainty in the loading. 00:28:07.700 --> 00:28:09.570 If we’re getting ourselves down to the last 10% whether things are 00:28:09.570 --> 00:28:13.269 going to work or not, you know, we’re overthinking it a little bit. 00:28:13.269 --> 00:28:15.369 So we wouldn’t – we wouldn’t, you know, dramatically change 00:28:15.369 --> 00:28:19.140 the design based on these numbers. So, in a sense, we’re getting kind of 00:28:19.140 --> 00:28:23.100 stable, you know, believable numbers out of both sources of ground motions. 00:28:23.100 --> 00:28:27.440 And, to the extent that the simulations are different, they’re different for 00:28:27.440 --> 00:28:30.350 a reason that we could physically, you know, anticipate to be real. 00:28:30.350 --> 00:28:33.129 We’re not – it’s not some artifact of the simulation, necessarily. 00:28:33.129 --> 00:28:34.859 And so we are getting some sort of insight 00:28:34.859 --> 00:28:36.779 that maybe we’re not getting from traditional methods. 00:28:36.780 --> 00:28:41.180 So I think everybody felt that was a success in that regard. 00:28:42.920 --> 00:28:45.499 I’m going to show you a couple more types of these figures. 00:28:45.499 --> 00:28:50.029 So these are plots of the story shear. So this is, story by story, if you took 00:28:50.029 --> 00:28:52.919 a cut through all of the columns, and you looked at how much force 00:28:52.919 --> 00:28:55.480 was getting transferred in the lateral direction by all of those columns – 00:28:55.480 --> 00:28:58.899 and, I should say, the braces and things at that story, you know, how much – 00:28:58.899 --> 00:29:02.080 how much lateral force is having to be carried through the building. 00:29:02.080 --> 00:29:05.739 So that increases as you go down the building because each lower story has 00:29:05.739 --> 00:29:09.769 to carry the entire building above it as it’s getting shaken laterally. 00:29:09.769 --> 00:29:11.989 Those results were almost identical. And that’s largely because of the 00:29:11.989 --> 00:29:14.919 way the structure is designed. So it’s designed to, you know, 00:29:14.919 --> 00:29:17.519 carry loads up to a certain point, and then yield and kind of distribute 00:29:17.519 --> 00:29:20.850 those loads in a predictable manner. And so, even with some variability 00:29:20.850 --> 00:29:25.160 in the loading conditions, the forces are going to be pretty well understood. 00:29:26.920 --> 00:29:31.420 And so – yeah, so I guess to kind of finish discussion of that example, we – 00:29:31.420 --> 00:29:34.879 yeah, we felt that things were – looked reasonable and that nobody 00:29:34.879 --> 00:29:36.950 had any concerns about using that particular set 00:29:36.950 --> 00:29:39.840 of simulations to think about analyzing a building. 00:29:41.100 --> 00:29:45.740 Okay, so let me look at a kind of related study, but somewhat distinct. 00:29:45.740 --> 00:29:48.380 So this idea of the – of the horizontal polarization 00:29:48.380 --> 00:29:51.700 of the ground motions was interesting to us. 00:29:51.700 --> 00:29:54.149 And because the buildings are 3D, and because the – so the direction 00:29:54.149 --> 00:29:57.049 of the large-amplitude shaking is important to us as we think about 00:29:57.049 --> 00:30:00.929 the behavior of these buildings, we were interested in polarization. 00:30:00.929 --> 00:30:03.029 And we had seen some hints that there might – things might not 00:30:03.029 --> 00:30:05.859 always be as smooth as they were in the previous study. 00:30:05.859 --> 00:30:08.690 So, in this case, we used the broadband platform again. 00:30:08.690 --> 00:30:12.020 Same software platform, but we looked at a few different algorithms. 00:30:12.020 --> 00:30:14.799 And, in this case, we looked at historical ground motions. 00:30:14.800 --> 00:30:17.440 So, you know, there’s various ways to do comparisons. 00:30:17.440 --> 00:30:20.020 So, in the previous case, we looked at kind of 00:30:20.020 --> 00:30:23.700 recordings selected using a building code procedure versus simulations. 00:30:23.700 --> 00:30:26.560 In this case, we’re going to look at recordings from a specific event – 00:30:26.560 --> 00:30:30.520 the Loma Prieta earthquake – versus simulations of that same event. 00:30:30.529 --> 00:30:33.600 So these are kind of recording stations that recorded the 00:30:33.600 --> 00:30:38.399 Loma Prieta earthquake. And so we’ve got – and then I guess 00:30:38.399 --> 00:30:39.999 we looked at the kind of nearby stations. 00:30:40.000 --> 00:30:43.960 So all the stations within 20 kilometers of that rupture projection. 00:30:43.960 --> 00:30:47.630 And so we’ve got response spectra over on the right from those. 00:30:47.630 --> 00:30:51.049 And then we took a comparable set of simulations from this broadband 00:30:51.049 --> 00:30:54.519 platform, and there was a big validation exercise going on at the time, 00:30:54.519 --> 00:30:59.039 where the simulators were participating, and a third-party group was kind of 00:30:59.040 --> 00:31:01.620 vetting the results and things. We used the same data. 00:31:01.620 --> 00:31:04.700 And we looked at three different algorithms. 00:31:04.700 --> 00:31:08.760 So this EXSIM that Gail Atkinson and Dave Boore have been involved in. 00:31:08.779 --> 00:31:11.440 This – GP is the Graves Pitarka algorithm – the same one 00:31:11.440 --> 00:31:15.640 from the previous slides. And then CSM is a composite source model, 00:31:15.640 --> 00:31:18.920 and John Anderson at Nevada-Reno – or, I guess the U.S. Geological Survey 00:31:18.920 --> 00:31:24.220 now, is – was involved with kind of coding up that algorithm. 00:31:24.220 --> 00:31:27.720 And so over on the right is an example of some of the response spectra at those 00:31:27.720 --> 00:31:31.399 same sites from the Graves and Pitarka algorithm. 00:31:31.399 --> 00:31:34.340 And so some others were looking at the absolute amplitudes of these things, 00:31:34.340 --> 00:31:37.529 but we were interested in the polarization in particular. 00:31:37.529 --> 00:31:41.549 And so, just to think about the way we polarize – or, we measure polarization, 00:31:41.549 --> 00:31:44.399 so this is actually – Dave Boore came up with these metrics. 00:31:44.399 --> 00:31:47.729 I understand he’s laid up at the moment, 00:31:47.729 --> 00:31:51.149 but I was thinking about him as I was putting together these slides. 00:31:51.149 --> 00:31:54.100 Sorry, I should – I should back up for a second. 00:31:54.100 --> 00:31:58.380 So, as we think about the kind of – the loading over multiple directions 00:31:58.389 --> 00:32:03.369 in this building, the – what I’ve got here is plotted on the horizontal axis here 00:32:03.369 --> 00:32:07.919 displacement traces from the Imperial Valley El Centro Differential Array 00:32:07.919 --> 00:32:10.179 recording, just as an example. And then we’ve got an elastic 00:32:10.179 --> 00:32:13.690 oscillator sitting on top of it. And this oscillator’s got the same period 00:32:13.690 --> 00:32:18.890 in kind of any axis, and so we can – we can let these oscillators move. 00:32:18.890 --> 00:32:20.649 So on the left is a 1-1/2-second oscillator. 00:32:20.649 --> 00:32:23.400 And on the right is a 3-second oscillator. 00:32:24.160 --> 00:32:27.460 And so let me let that roll. So on the – up in space is the 00:32:27.470 --> 00:32:30.970 differential displacement of the – of the oscillator relative to the ground. 00:32:30.970 --> 00:32:34.099 And those displacements are what we ultimately turn into the response spectra. 00:32:34.099 --> 00:32:36.679 But, as we look in different orientations, 00:32:36.679 --> 00:32:40.830 the peak displacement is going to be different. And so here we can 00:32:40.830 --> 00:32:43.460 kind of rotate it up and just look at the differential displacements. 00:32:43.460 --> 00:32:48.840 So the thing I like to use these things to point out is that, you know, we think about 00:32:48.840 --> 00:32:53.440 this loading in multiple directions and, you know, if you’re designing a flagpole, 00:32:53.440 --> 00:32:55.609 kind of the maximum displacement in any direction would be 00:32:55.609 --> 00:32:58.259 what you would be worried about. But in real buildings, we typically 00:32:58.259 --> 00:33:00.710 have kind of two planes where we have frames. 00:33:00.710 --> 00:33:02.309 And it’s going to be those frames worked individually 00:33:02.309 --> 00:33:05.862 to resist forces that’s important. And so the – kind of the orientation 00:33:05.862 --> 00:33:09.660 of large demands versus those frames is important to us. 00:33:10.360 --> 00:33:12.500 So this is the ground motion that’s relatively polarized. 00:33:12.509 --> 00:33:14.279 There’s particular orientations where the displacements 00:33:14.279 --> 00:33:16.980 are much bigger than others. 00:33:16.980 --> 00:33:19.929 But even for a particular ground motion, you know, 00:33:19.929 --> 00:33:24.049 polarization is frequency-dependent. So here you can see the 1-1/2-second 00:33:24.049 --> 00:33:26.529 oscillator peak displacements were almost orthogonal 00:33:26.529 --> 00:33:28.470 to the 3-second oscillator peak displacements, 00:33:28.470 --> 00:33:30.960 even though the input was identical in both cases. 00:33:30.960 --> 00:33:35.919 So, you know, from a – I guess for you all to think about the – 00:33:35.919 --> 00:33:39.519 kind of the insights that we can get about potential polarization of shaking 00:33:39.519 --> 00:33:43.700 and the inputs to simulations that are influencing these polarization 00:33:43.700 --> 00:33:45.960 are important to us. And then, from the engineer’s 00:33:45.960 --> 00:33:49.779 perspective, there’s – I think there’s sometimes a rush to think about, like, 00:33:49.779 --> 00:33:51.639 well, what’s the strong direction of the shaking? 00:33:51.639 --> 00:33:53.840 And there is no strong direction of shaking. 00:33:53.840 --> 00:33:58.979 It depends on kind of what the dynamic properties of your – of your system are. 00:33:58.979 --> 00:34:01.340 And we just need kind of realistic representations 00:34:01.340 --> 00:34:05.240 of the complexities that the real world is going to throw at us. 00:34:05.240 --> 00:34:07.510 Okay, so with that is kind of a little bit of motivation. 00:34:07.510 --> 00:34:09.630 And we think about measuring this polarization. 00:34:09.630 --> 00:34:13.000 And so – let me think. 00:34:13.420 --> 00:34:16.560 Let me remind myself kind of how far we go. 00:34:16.560 --> 00:34:18.860 So we’ve got kind of examples of these traces. 00:34:18.860 --> 00:34:21.780 So these are – these are traces from 0.2-second oscillators. 00:34:21.780 --> 00:34:25.100 And so I’ve got, on the left, a recording from Loma Prieta. 00:34:25.100 --> 00:34:26.880 On the right, a Graves Pitarka simulation. 00:34:26.880 --> 00:34:29.000 And on the – in the middle, Graves Pitarka. 00:34:29.000 --> 00:34:33.820 On the full – on the far right, this composite source method simulation. 00:34:33.820 --> 00:34:37.830 And these are kind of the medium – or, the median polarized out of the suite. 00:34:37.830 --> 00:34:39.720 So there’s kind of half of them that are more polarized than this 00:34:39.720 --> 00:34:42.480 and half that are less. And so – yeah, so Dave Boore 00:34:42.480 --> 00:34:46.390 was the one who kind of came up with our current parameterization of this, 00:34:46.390 --> 00:34:49.480 which is to say, if I’m thinking about how strong a multi-component ground 00:34:49.480 --> 00:34:52.600 motion is, I could look at the peak displacement demand 00:34:52.600 --> 00:34:55.190 over all possible orientations. And that’s what we now 00:34:55.190 --> 00:34:59.720 call the RotD100 metric. So the kind of 100th percentile 00:34:59.720 --> 00:35:01.980 displacement demand over all rotation angles. 00:35:01.980 --> 00:35:05.330 Or we could think about the median over the rotation angles, 00:35:05.330 --> 00:35:09.930 which we call the RotD50 for a 50th percentile amplitude. 00:35:09.930 --> 00:35:14.200 And so the – in a figure like on the left, the difference between those two 00:35:14.200 --> 00:35:16.750 would be relatively small. So the median amplitude would be, 00:35:16.750 --> 00:35:19.850 you know, similar to the 100th percentile amplitude. 00:35:19.850 --> 00:35:23.510 On a trace like on the far right, the ratio would be bigger. 00:35:23.510 --> 00:35:27.280 And the ratio can get as big as square root of 2, or 1.41, just from 00:35:27.280 --> 00:35:31.570 kind of geometry of these rotations. And the ratio can get as small as 1. 00:35:31.570 --> 00:35:34.030 If you had the same shaking amplitude in all directions, 00:35:34.030 --> 00:35:37.030 the median would be identical to the – to the maximum. 00:35:37.030 --> 00:35:40.540 And so that’s an easy way to kind of look at this polarization. 00:35:40.540 --> 00:35:43.820 And then the – we see a typical pattern that, at longer periods, 00:35:43.820 --> 00:35:46.660 we see more strong polarization in the ground motions. 00:35:46.660 --> 00:35:48.780 And some of that’s just duration-dependent. 00:35:48.780 --> 00:35:52.620 I think you have less time to get through a lot of cycles of an oscillator when the – 00:35:52.620 --> 00:35:54.880 when the period of the oscillator is longer compared to the 00:35:54.880 --> 00:35:57.160 duration of shaking. And then some of that has – 00:35:57.170 --> 00:36:01.180 I think is kind of real polarization as well, that there’s just more 00:36:01.180 --> 00:36:05.180 kind of directional scattering maybe at lower frequencies. 00:36:05.180 --> 00:36:07.660 Okay, so here’s these ratios I was talking about. 00:36:07.660 --> 00:36:11.410 So the horizontal axis I’ve got periods plotted, and then on 00:36:11.410 --> 00:36:14.860 the vertical axis is this ratio. So we can look at the RotD100 00:36:14.860 --> 00:36:18.441 divided by RotD50, and the scale is going to go from – the smallest it can 00:36:18.441 --> 00:36:22.990 be is 1 at the bottom of the plot, up to square root of 2 at the top of the plot. 00:36:22.990 --> 00:36:25.240 And you can compute these things for empirical ground motions, 00:36:25.240 --> 00:36:27.930 and they’re really stable – large magnitudes and small magnitudes 00:36:27.930 --> 00:36:31.750 and distant recordings or close recordings or soft soils or hard soils. 00:36:31.750 --> 00:36:34.360 The ratios really seem quite stable on average. 00:36:34.360 --> 00:36:38.620 So here’s a couple empirical models that have been developed over the years. 00:36:38.620 --> 00:36:42.910 And so you see kind of 1.2 at short periods up to 1.25 or 1.3 00:36:42.910 --> 00:36:47.140 maybe at long periods. That’s really a pretty reliable kind of 00:36:47.140 --> 00:36:49.470 median estimate, and that’s not to say there aren’t circumstances 00:36:49.470 --> 00:36:52.550 where things could change, such as, you know, near-fault directivity 00:36:52.550 --> 00:36:56.000 or something like that. But, on average, that’s kind of what we expect. 00:36:56.860 --> 00:37:00.560 There in black is the set of recordings from Loma Prieta. 00:37:00.570 --> 00:37:03.160 So there’s 20 recordings, and you can see it’s – you know, 00:37:03.160 --> 00:37:05.660 it’s bouncing around a little bit because there is some small-sample variability, 00:37:05.660 --> 00:37:08.310 but it’s in the neighborhood of those models, and that’s kind of a benchmark 00:37:08.310 --> 00:37:11.030 of, you know, what a stable answer might be. 00:37:11.030 --> 00:37:12.590 And then here’s a couple set of simulations. 00:37:12.590 --> 00:37:16.060 So in blue is this Graves and Pitarka simulations, and they’re, you know, 00:37:16.060 --> 00:37:20.620 a little more variable around that target ratio. 00:37:20.620 --> 00:37:24.030 This is, you know, a couple iterations ago. I think probably the later iterations 00:37:24.030 --> 00:37:27.220 of the algorithm are even a little more stable in terms of this. 00:37:27.220 --> 00:37:30.550 And then up at the very top is this composite source method. 00:37:30.550 --> 00:37:33.960 And in talking to John Anderson, he said that there’s kind of a radiation 00:37:33.960 --> 00:37:37.390 term associated with this as he steps along the fault, and that was not kind of 00:37:37.390 --> 00:37:40.630 randomized reflecting kind of geometrical complexities and things 00:37:40.630 --> 00:37:42.290 and that he thought, if it was randomized, that that would 00:37:42.290 --> 00:37:45.940 really break up that polarization. So that was a case where we saw 00:37:45.940 --> 00:37:49.130 something in the ground motions that doesn’t show up in kind of traditional 00:37:49.130 --> 00:37:52.200 response spectra metrics. But this directional polarization 00:37:52.200 --> 00:37:55.540 is important for buildings, and it is something that’s not physical, 00:37:55.540 --> 00:37:58.710 but it’s something that could be easily kind of addressed and, you know, 00:37:58.710 --> 00:38:01.280 rolled into future iterations of these simulation algorithms. 00:38:01.280 --> 00:38:04.240 So I think that’s also the role that we hope to play as engineers is 00:38:04.240 --> 00:38:07.990 kind of pointing out some of these features that we’d like further refined. 00:38:07.990 --> 00:38:10.150 And without us pulling it out, if – you know, there’s no need 00:38:10.150 --> 00:38:12.820 to be thinking about things like this. 00:38:15.080 --> 00:38:17.560 - [inaudible] - Sure. 00:38:17.560 --> 00:38:24.880 - [inaudible] 00:38:24.880 --> 00:38:27.800 - TS? - [inaudible] 00:38:27.800 --> 00:38:30.840 - Ah, yeah, okay. Thank you. - Oscillator period. 00:38:30.859 --> 00:38:33.510 - Got it. Got it. Yeah, so – thank you. So the question was about what 00:38:33.510 --> 00:38:37.560 the period on the horizontal axis is. So the period stays the same 00:38:37.560 --> 00:38:40.491 in all horizontal directions. So I would take a 1-1/2-second 00:38:40.491 --> 00:38:44.220 oscillator, say from the previous slide, and I would look at a 1-1/2-second 00:38:44.220 --> 00:38:46.700 oscillator at all possible orientations. 00:38:46.700 --> 00:38:50.500 And then I would take the maximum and the median of those. So, yeah. 00:38:51.140 --> 00:38:52.880 - Thank you. - Yep. 00:38:54.040 --> 00:38:58.960 Okay. So then, to pick a kind of a final example, and I’ll go through 00:38:58.960 --> 00:39:01.200 this one a little more quickly. We’re not quite done with it, but just 00:39:01.200 --> 00:39:03.000 to show kind of where we’re going. So now here we’ve moved on 00:39:03.000 --> 00:39:06.550 to southern California, and we’re using the CyberShake simulation. 00:39:06.550 --> 00:39:10.940 So this is another simulation infrastructure that’s run by Southern 00:39:10.940 --> 00:39:13.680 California Earthquake Center. And I think this is a case that’s 00:39:13.680 --> 00:39:16.340 getting a lot of attention on the hazard analysis side. 00:39:16.340 --> 00:39:18.490 So Los Angeles is thinking about using these simulations 00:39:18.490 --> 00:39:20.160 to update their hazard analysis. 00:39:20.160 --> 00:39:25.060 And we were interested in the – in the engineering side – the dynamic analysis. 00:39:25.060 --> 00:39:27.390 This one’s also interesting relative to the broadband platform because 00:39:27.390 --> 00:39:32.310 there’s a full 3D velocity model that’s used for the wave propagation. 00:39:32.310 --> 00:39:36.020 And so things like sedimentary basins and other kind of geometric complexities 00:39:36.020 --> 00:39:41.860 that we think are important down there, those are very hard features to represent 00:39:41.860 --> 00:39:44.160 in recorded ground motions when we’re taking recorded ground 00:39:44.160 --> 00:39:46.571 motions from all over the world. So if we could get simulations that 00:39:46.571 --> 00:39:50.160 are reflecting those complexities, and we felt comfortable using them, 00:39:50.160 --> 00:39:52.540 I think that would be a great opportunity for engineers designing, say, 00:39:52.540 --> 00:39:55.220 the tall buildings in that area. 00:39:55.220 --> 00:39:58.740 So we thought about a site in L.A. downtown – that’s that LADT site. 00:39:58.740 --> 00:40:02.600 And so that’s kind of partially in this basin – the L.A. Basin. 00:40:02.600 --> 00:40:04.260 And it’s a place where we’re designing tall buildings and 00:40:04.260 --> 00:40:07.520 building tall buildings that are using dynamic analysis, and there are 00:40:07.520 --> 00:40:10.610 sophisticated engineers who would be open to these types of things. 00:40:10.610 --> 00:40:12.410 So we wanted to show them kind of what these ground motions 00:40:12.410 --> 00:40:15.820 would look like. So I’ll go through a little more quickly. 00:40:15.820 --> 00:40:18.280 So the dashed black line here is that target spectrum 00:40:18.280 --> 00:40:22.300 from ASCE 7-16 – the same one we talked about before. 00:40:22.300 --> 00:40:24.380 And then what we did is, we went through the CyberShake database 00:40:24.390 --> 00:40:27.030 and went and selected ground motions, and so here those simulations 00:40:27.030 --> 00:40:30.070 were already available. And we just went to the LADT site 00:40:30.070 --> 00:40:34.140 and said, what are the ground motions most closely matching that 00:40:34.140 --> 00:40:36.410 target spectrum? And so those are shown in green here. 00:40:36.410 --> 00:40:39.220 There were 400,000 ground motions to choose from, so it was pretty easy 00:40:39.220 --> 00:40:42.620 to find a few that were closely matching that target spectrum. 00:40:43.940 --> 00:40:46.380 And kind of the same game. So in the top row is this 00:40:46.380 --> 00:40:48.930 L.A. downtown site. On the left are the simulations, 00:40:48.930 --> 00:40:52.390 and on the right are the – we did a comparable set of selections from 00:40:52.390 --> 00:40:55.800 recordings, again, to do the comparison. And we did the whole exercise at a 00:40:55.800 --> 00:41:00.320 second site in Pasadena outside of the basin, and that’s in the bottom row. 00:41:00.320 --> 00:41:02.540 So I’ll kind of flash at you a couple of the same types of things. 00:41:02.540 --> 00:41:07.460 So on the – on the left here, I’ll show you – so the tectonics are a little more 00:41:07.460 --> 00:41:11.660 complicated in southern California than they were at that Berkeley site. 00:41:11.660 --> 00:41:14.370 So de-aggregation now shows contributions to shaking from 00:41:14.370 --> 00:41:17.090 a lot of different sources. So the far left column – 00:41:17.090 --> 00:41:21.010 and so the top row is L.A. downtown. The bottom row is Pasadena. 00:41:21.010 --> 00:41:24.690 In the top, we’ve got a de-aggregation at a 1-second period. 00:41:24.690 --> 00:41:28.770 So we see lots of contributions from kind of zero to 20 kilometers on the far right. 00:41:28.770 --> 00:41:31.890 And kind of magnitudes 6 to 7-1/2, something like that. 00:41:31.890 --> 00:41:36.600 Puente Hills is a big contributor and kind of other [inaudible]. 00:41:36.600 --> 00:41:40.160 If we go to the middle column, in that top row, now we can see – 00:41:40.160 --> 00:41:42.790 I don’t know if this – does this mouse show? 00:41:42.790 --> 00:41:45.320 We get this contribution from the San Andreas Fault that pops up. 00:41:45.320 --> 00:41:51.210 So magnitude 8s out at 50 kilometers. So one thing that I always try to kind of 00:41:51.210 --> 00:41:54.350 communicate to engineers, if we go back to that language in the building code 00:41:54.350 --> 00:41:58.390 about, we need time series that match the magnitudes and the distances of the – 00:41:58.390 --> 00:42:01.740 of the faults, you know, causing the ground motions, if you look across 00:42:01.740 --> 00:42:04.760 a wide range of periods, there is no unique set of faults kind of 00:42:04.760 --> 00:42:06.600 causing the ground shaking at all periods. 00:42:06.600 --> 00:42:10.060 Yet, we need time series that kind of represent kind of all periods. 00:42:10.060 --> 00:42:13.040 So we’re always in a little bit of a conundrum there. 00:42:13.040 --> 00:42:18.870 But – so but anyways, with these de-aggregations as motivation, 00:42:18.870 --> 00:42:23.100 we went and selected ground motions. So with the NGA-West recordings – 00:42:23.100 --> 00:42:25.850 so these are the empirical recordings – we manually tried to fit those – 00:42:25.850 --> 00:42:27.880 you know, tried to match those distributions. 00:42:27.880 --> 00:42:31.570 Although it’s pretty hard to find kind of large-magnitude recordings that have 00:42:31.570 --> 00:42:34.450 kind of the right response spectra. So we didn’t do well there. 00:42:34.450 --> 00:42:36.750 With CyberShake, we didn’t have to try – we didn’t constrain 00:42:36.750 --> 00:42:39.250 the faults or the magnitudes and distances very much. 00:42:39.250 --> 00:42:41.790 We primarily were just searching off the response spectrum. 00:42:41.790 --> 00:42:46.450 And we could do a better job of kind of getting a few San Andreas ruptures 00:42:46.450 --> 00:42:51.320 and things into the record set without messing around too much. 00:42:51.320 --> 00:42:54.820 So we get kind of 11 ground motions and 11 ground motions again. 00:42:54.820 --> 00:42:56.860 These are some of the CyberShake ground motions. 00:42:56.860 --> 00:42:59.470 Just a flash that – say, on the left, there’s a couple Puente Hills 00:42:59.470 --> 00:43:03.230 simulations that are kind of 10 or 15 seconds of shaking – 00:43:03.230 --> 00:43:06.860 kind of very intense shaking, you know, close by to the site. 00:43:06.860 --> 00:43:09.080 And then, on the right, you know, those fonts are maybe a little small, 00:43:09.080 --> 00:43:10.800 but the time scales changed. 00:43:10.810 --> 00:43:13.440 So these are much longer kind of time windows on the right. 00:43:13.440 --> 00:43:17.760 These are some kind of big – there’s a Newport/Inglewood magnitude 7-1/2 00:43:17.760 --> 00:43:21.260 and a San Andreas in the lower right kind of shaking for a minute or two. 00:43:21.260 --> 00:43:23.760 You can maybe see some kind of basin resonance in that upper-right one, 00:43:23.760 --> 00:43:27.120 and you can see some kind of static offset in the lower-right one. 00:43:27.120 --> 00:43:31.400 So we see kind of the complexities that we would anticipate from some of these 00:43:31.400 --> 00:43:33.870 rupture scenarios, which would be interesting for engineers to probe. 00:43:33.870 --> 00:43:39.420 We’d like to see kind of this kind of dynamic feature diversity. 00:43:39.420 --> 00:43:41.560 Because we’re trying to probe these buildings and see, you know, 00:43:41.560 --> 00:43:45.040 under different types of circumstances, how is it going perform. 00:43:46.740 --> 00:43:49.510 We checked this polarization. So as long as we defined that plot 00:43:49.510 --> 00:43:50.990 before, I’ll show you another one. So this is that 00:43:50.990 --> 00:43:55.870 RotD100-over-RotD50 plot. And the red line is showing an 00:43:55.870 --> 00:43:59.760 empirical model. And the green and blue are showing their record set. 00:43:59.760 --> 00:44:01.520 So everything looks good in terms of the polarization. 00:44:01.520 --> 00:44:04.800 Kind of no concerns about that. 00:44:04.800 --> 00:44:07.690 One thing that we did pause on, and we’re kind of continuing to talk to 00:44:07.690 --> 00:44:10.440 Rob Graves about, is we did see some circumstances where we 00:44:10.440 --> 00:44:14.980 saw strong polarization. Try to do this a little quickly. 00:44:14.980 --> 00:44:17.070 So we – there were a few ground motions that came in 00:44:17.070 --> 00:44:20.021 kind of unusually polarized. And so we kind of tried to probe 00:44:20.021 --> 00:44:23.970 a little bit, what are the circumstances. So there are some situations – let me try 00:44:23.970 --> 00:44:28.940 to not spend too long on these figures, but – so this is – we tried to pick some 00:44:28.940 --> 00:44:33.500 different types of scenarios where we were – trying to get the mouse up 00:44:33.500 --> 00:44:36.700 on the screen – where we were looking at this polarization. 00:44:36.700 --> 00:44:42.240 So the maps show kind of a rupture and an epicenter and a station 00:44:42.240 --> 00:44:46.350 where we took a recording from. And then these traces show velocity 00:44:46.350 --> 00:44:48.920 versus velocity in the north-south-east-west directions. 00:44:48.920 --> 00:44:50.900 So we can kind of get a sense of how strong 00:44:50.900 --> 00:44:53.630 the shaking is in different orientations. 00:44:53.630 --> 00:44:57.341 And then we’ve got a similar map – and we tried to look at a – 00:44:57.341 --> 00:45:01.610 kind of a backwards directivity and a forwards directivity type of case. 00:45:01.610 --> 00:45:05.530 So these are magnitude 7-1/2 or so ruptures on the San Andreas Fault. 00:45:05.530 --> 00:45:07.990 The station is about 50 kilometers away. 00:45:07.990 --> 00:45:10.400 And over all the different ruptures, there’s about 100 realizations, I think, 00:45:10.400 --> 00:45:14.400 of this rupture with different epicenters, different slip distributions. 00:45:14.400 --> 00:45:17.020 And over here is the – in kind of a dashed line, 00:45:17.020 --> 00:45:20.990 is this referenced empirical model. Oop – went too far. 00:45:20.990 --> 00:45:23.220 This reference empirical model for this polarization – 00:45:23.220 --> 00:45:25.140 the RotD100 over RotD50. 00:45:25.140 --> 00:45:28.780 And then the solid line is the – is the median of the – all the simulations. 00:45:28.780 --> 00:45:30.890 And so we – usually we see things that look kind of 00:45:30.890 --> 00:45:33.740 like the right, which is what we would expect to see. 00:45:34.660 --> 00:45:36.220 So here’s a second case. I’m bouncing around. 00:45:36.220 --> 00:45:39.620 So we moved in a little closer. This is 30 kilometers away from the rupture. 00:45:39.620 --> 00:45:42.430 Same types of patterns. We see kind of good agreement. 00:45:42.430 --> 00:45:45.570 Or now I’m lost of which one I was looking at. 00:45:45.570 --> 00:45:48.980 So Case 1, Case 2 – these are 50 kilometers, 30 kilometers away. 00:45:48.980 --> 00:45:53.120 Oh, man. I think this mouse is running backwards to what I expect. 00:45:53.120 --> 00:45:56.240 Okay, so those look fine. But then there are some cases – 00:45:56.240 --> 00:46:00.640 so this is a case kind of off the end of the fault where you might expect 00:46:00.640 --> 00:46:04.570 a little more directivity, but it’s 50 kilometers away, 00:46:04.570 --> 00:46:08.030 which is a pretty long distance for kind of coherent polarization 00:46:08.030 --> 00:46:10.770 of the ground motions as we see it in empirical ground motions. 00:46:10.770 --> 00:46:14.370 And, in this case, almost every ground motion was somewhat polarized in 00:46:14.370 --> 00:46:18.940 the fault normal direction, roughly. And we get these plots where we 00:46:18.940 --> 00:46:21.630 switch to kind of very strong polarization at long periods 00:46:21.630 --> 00:46:26.720 in basically 100% of the ruptures. And so, to me, there’s some mix of 00:46:26.720 --> 00:46:31.820 either kind of stronger – some sort of kind of radiation pattern or directivity 00:46:31.820 --> 00:46:34.730 or something that’s really strongly polarizing these in a way that we 00:46:34.730 --> 00:46:37.180 don’t have observed in our recorded ground motions. 00:46:37.180 --> 00:46:42.060 Or there’s maybe something in the – you know, in the rupture generation or 00:46:42.060 --> 00:46:45.220 some piece of the simulation algorithm. And so this is where – again, I’ll say 00:46:45.220 --> 00:46:48.060 Rob Graves has been tremendously helpful in kind of explaining lots of 00:46:48.060 --> 00:46:51.220 things through here, and we’re kind of continuing to talk with him about kind of 00:46:51.220 --> 00:46:54.240 how pervasive this type of situation is, and is it real, or is it – 00:46:54.240 --> 00:46:56.830 you know, is it an artifact of something in the simulation. 00:46:56.830 --> 00:47:00.370 So, to me, I find those conversations really productive in trying to 00:47:00.370 --> 00:47:03.600 move forward these things to engineering practice. 00:47:03.600 --> 00:47:06.200 So these ground motions are all selected. They’re currently getting run 00:47:06.200 --> 00:47:08.960 through a couple of building models. Again, working with practicing 00:47:08.960 --> 00:47:12.870 structural engineers. So they’re kind of 3D dynamic building models. 00:47:12.870 --> 00:47:16.000 The left building – this is a 40-story building. 00:47:16.000 --> 00:47:19.020 Kind of one of these condo buildings that are going up in San Francisco 00:47:19.020 --> 00:47:23.740 and Los Angeles that an engineering firm is analyzing for us. 00:47:23.740 --> 00:47:25.740 That’s a real building that’s being built. 00:47:25.740 --> 00:47:27.690 On the right is a building from Christchurch. 00:47:27.690 --> 00:47:30.460 We found a – kind of an engaged structural engineer in Christchurch, 00:47:30.460 --> 00:47:32.730 New Zealand, that was interested in kind of probing these things 00:47:32.730 --> 00:47:35.110 and is running it through a model that they had available as well. 00:47:35.110 --> 00:47:39.080 And so we’re kind of collecting those results and processing, interpreting, 00:47:39.080 --> 00:47:40.830 to try to see, you know, what are we seeing as differences 00:47:40.830 --> 00:47:43.060 between the recordings and simulations. 00:47:46.120 --> 00:47:47.820 I think I’m going to go by this one. 00:47:47.820 --> 00:47:51.560 I think this one will be kind of more interesting. 00:47:51.560 --> 00:47:53.820 So, as we think through these exercises, and we’re trying to do these 00:47:53.820 --> 00:47:58.100 comparisons, you know, on one hand, it’s a – it’s a relatively standard set 00:47:58.100 --> 00:48:01.220 of analyses that we’re doing. The ground motion selection, 00:48:01.220 --> 00:48:02.950 the structural analysis, is not groundbreaking 00:48:02.950 --> 00:48:06.120 in any sort of academic sense. But what we really need to do is 00:48:06.120 --> 00:48:08.790 kind of set the stage thinking about how these things hit the ground 00:48:08.790 --> 00:48:11.540 in the real world. And so the things I think about is – 00:48:11.540 --> 00:48:14.860 you know, as I’ve kind of worked with engineers who are designing buildings 00:48:14.860 --> 00:48:18.271 or worked on some of these projects, you know, typically, there’s a – 00:48:18.271 --> 00:48:21.930 there’s a very tight timeline and not a lot of slop in the budget. 00:48:21.930 --> 00:48:24.290 I’m thinking about, you know, kind of commercial building projects. 00:48:24.290 --> 00:48:27.110 You know, things get a little different if it’s a – kind of some sort of 00:48:27.110 --> 00:48:29.940 unusual defense facility or a nuclear power plant or something like that. 00:48:29.940 --> 00:48:32.590 But if it’s a – if it’s a condo building or something, there’s a developer 00:48:32.590 --> 00:48:35.000 who wants that thing open as soon as possible. 00:48:35.000 --> 00:48:37.590 And they don’t want to wait around for extra weeks or months for 00:48:37.590 --> 00:48:42.060 kind of curiosity-driven kind of analysis of the building. 00:48:42.060 --> 00:48:44.460 And the – you know, the engineers who are doing these are bidding 00:48:44.460 --> 00:48:47.760 competitively on these projects, so they don’t put in, kind of, you know, 00:48:47.760 --> 00:48:51.980 lots and lots of extra time to be kind of curiosity-driven analysis. 00:48:51.980 --> 00:48:54.380 And, you know, especially on these developer projects, 00:48:54.390 --> 00:48:57.710 that’s a lot of the stuff that goes up in downtowns. 00:48:57.710 --> 00:48:59.510 These real estate developers really think about the earthquake 00:48:59.510 --> 00:49:02.740 engineering as a commodity. Hopefully that’s changing as, you know, 00:49:02.740 --> 00:49:06.710 we get this kind of nice run of public awareness of earthquake risks. 00:49:06.710 --> 00:49:08.690 But I think generally, they say, you know, I want this building, 00:49:08.690 --> 00:49:11.750 and what’s going to sell it is the beautiful floors and the beautiful views. 00:49:11.750 --> 00:49:15.850 And whatever engineer can do this job for me on time and at a reasonable 00:49:15.850 --> 00:49:18.270 budget, like, that’s who I’m hiring. I’m not hiring somebody because 00:49:18.270 --> 00:49:21.130 they’re doing, you know, cutting-edge engineering, 00:49:21.130 --> 00:49:24.930 you know, thinking. That’s – you know, each project 00:49:24.930 --> 00:49:28.540 is different, but I would say that’s a pretty common situation. 00:49:28.540 --> 00:49:31.460 And the other thing is, these unusual buildings, almost always, 00:49:31.470 --> 00:49:35.200 when you have a dynamic analysis in the – somewhere in the loop 00:49:35.200 --> 00:49:38.140 in the analysis, there’s a – there’s a lot more complexity that goes into that 00:49:38.140 --> 00:49:41.540 structural analysis, and there’s a lot of decisions that the engineer has to make. 00:49:41.540 --> 00:49:44.740 And those decisions can influence the – you know, the answers that 00:49:44.740 --> 00:49:46.750 you’re getting out of your analysis, for good or for bad. 00:49:46.750 --> 00:49:49.680 And so almost always, there’s a peer review team involved that’s kind of 00:49:49.680 --> 00:49:52.560 watching over the shoulder and saying, did you think about this, you know, 00:49:52.560 --> 00:49:54.080 what’s the basis for your assumption here? 00:49:54.080 --> 00:49:56.440 Did you, you know, consider that this could go wrong? 00:49:56.440 --> 00:49:59.640 And the idea is to kind of build in some robustness into the system that an 00:49:59.640 --> 00:50:02.910 engineer doesn’t accidentally make an assumption that leads them to believe a 00:50:02.910 --> 00:50:07.790 design is satisfactory when there was some overlooked flaw in their design. 00:50:07.790 --> 00:50:10.740 But – so the good thing is that I think it builds some good robustness 00:50:10.740 --> 00:50:14.020 into the system. The bad thing is it disincentivizes innovation, right? 00:50:14.020 --> 00:50:16.570 So if I – if I’m the design engineer, and I walk in, and I say, I’ve got these 00:50:16.570 --> 00:50:20.240 CyberShake ground motions, and I’m going to use these instead of recordings, 00:50:20.240 --> 00:50:22.170 my peer reviewer may or may not be comfortable with that if they 00:50:22.170 --> 00:50:25.220 haven’t seen some evidence that this is a reasonable way to proceed. 00:50:25.220 --> 00:50:28.810 And so, you know, having, you know, literature available, having kind of prior 00:50:28.810 --> 00:50:31.930 experience with these things is really important so that the overall team is 00:50:31.930 --> 00:50:34.550 all comfortable with the path forward, that nobody wants to take a chance on – 00:50:34.550 --> 00:50:38.721 you know, on some step in the analysis that, you know, drives the design of 00:50:38.721 --> 00:50:41.240 a building that is hopefully going to be there for a hundred years and going to – 00:50:41.240 --> 00:50:44.420 you know, influences the safety of the public. 00:50:45.480 --> 00:50:47.220 And then the final one is that, you know, it’s pretty easy to 00:50:47.220 --> 00:50:49.490 go get recorded ground motions. There’s these PEER websites. 00:50:49.490 --> 00:50:52.640 There’s other software platforms. I could go get recorded ground motions 00:50:52.650 --> 00:50:56.560 myself in, you know, five or 10 minutes, and I can go off and do my analysis. 00:50:56.560 --> 00:50:59.400 And all the consultants that work in this space are comfortable with that. 00:50:59.400 --> 00:51:01.400 If I want to use simulated ground motions, the story’s 00:51:01.400 --> 00:51:05.310 a little more complicated, right? So 10 years ago, I had to, you know, 00:51:05.310 --> 00:51:07.710 know the simulator and have some sort of plan. 00:51:07.710 --> 00:51:11.490 You know, with some exceptions. Brad Aagaard had simulations up on his 00:51:11.490 --> 00:51:14.300 website for many, many years, I know. You know, not everybody’s kind of 00:51:14.300 --> 00:51:17.880 got those things out in public. Even if they are, you know, there’s still 00:51:17.880 --> 00:51:20.970 questions about kind of the assumptions, and the engineer should have some 00:51:20.970 --> 00:51:23.180 awareness of what’s going on. You know, some of these public 00:51:23.180 --> 00:51:26.200 platforms for simulating ground motions have gotten better, but you still need, 00:51:26.200 --> 00:51:28.850 you know, some understanding of how to get into those systems, and, 00:51:28.850 --> 00:51:31.910 you know, it’s SQL queries to get at CyberShake ground motions and things. 00:51:31.910 --> 00:51:34.920 And, again, given the tight budgets, tight timelines, to kind of stop and 00:51:34.920 --> 00:51:40.420 do something out of the ordinary is just a friction in the system that’s not – 00:51:40.420 --> 00:51:43.940 you know, not helpful in terms of getting to kind of a project delivery. 00:51:43.940 --> 00:51:46.030 And so I raise these just to say, you know, none of these are 00:51:46.030 --> 00:51:50.120 kind of irrational things. But they are real frictions that kind of 00:51:50.120 --> 00:51:54.700 limit, you know, our ability to kind of innovate even if we have kind of pretty 00:51:54.700 --> 00:51:58.710 good confidence that there are good ways to proceed. 00:51:58.710 --> 00:52:01.160 And so, you know, if I kind of reflect on this, and I was putting these together last 00:52:01.160 --> 00:52:04.290 night, I think, you know, a lot of these engineers, and the engineers that are – 00:52:04.290 --> 00:52:05.570 that are running those structural models, 00:52:05.570 --> 00:52:07.660 they’re all doing it on their own time, just volunteers. 00:52:07.660 --> 00:52:11.820 They’re curious. They want to do a good job. They want to build good buildings. 00:52:11.820 --> 00:52:14.120 And I think many of them say, well, you know, if you can tell me about, 00:52:14.130 --> 00:52:16.770 you know, simulated ground motions and something new – 00:52:16.770 --> 00:52:20.440 something that would make my design better, I’m all for it. 00:52:20.440 --> 00:52:24.730 But, you know, looking at those above incentives, like, none of them incentivize 00:52:24.730 --> 00:52:27.640 kind of, you know, trying to change the scheme there. 00:52:27.640 --> 00:52:30.020 And so you really want to come in with a pretty good understanding of, 00:52:30.020 --> 00:52:33.070 like, if I – if I make this change, and I don’t use my traditional method – 00:52:33.070 --> 00:52:36.710 I use something new – I use simulations, what is the implications for the 00:52:36.710 --> 00:52:40.760 performance of my building? What is the benefit for the project? 00:52:40.760 --> 00:52:42.100 Because it’s going to be a risk. 00:52:42.100 --> 00:52:45.020 It’s going to involve some kind of cost and time, potentially. 00:52:45.020 --> 00:52:47.270 And so I think, you know, the more we can kind of put literature out there 00:52:47.270 --> 00:52:51.740 showing that results are as expected or giving guidance on what features to be 00:52:51.740 --> 00:52:55.190 thinking about when heading down this path, you know, the easier using 00:52:55.190 --> 00:52:57.580 these simulations is going to be. And the other thing, kind of to that 00:52:57.580 --> 00:53:00.860 last bullet point is, the easier it is to get a hold of appropriate simulations, 00:53:00.860 --> 00:53:03.600 you know, the more likely they are to be used. 00:53:03.600 --> 00:53:06.500 And that’s not science. It’s – you know, unfortunately it’s – 00:53:06.500 --> 00:53:08.500 but it’s hard. It still takes time and resources 00:53:08.500 --> 00:53:12.900 to kind of produce those portals or availability of data and things. 00:53:12.900 --> 00:53:17.220 And so, you know, there are people at – you know, various people involved with 00:53:17.220 --> 00:53:20.980 these simulations thinking about that. And I think that’ll be, like, another step 00:53:20.980 --> 00:53:23.510 in this process of making these more mainstreamed. 00:53:24.400 --> 00:53:27.180 Okay, so kind of wrap things up. 00:53:28.540 --> 00:53:30.960 I’d say, you know, there’s a number of ways in which we could think about 00:53:30.970 --> 00:53:33.810 using ground motion simulations or validating ground motion simulations. 00:53:33.810 --> 00:53:35.900 Hopefully this was, you know, a little bit of a new way of thinking 00:53:35.900 --> 00:53:38.470 for some of you, that really, it’s – you know, from the design spectrum 00:53:38.470 --> 00:53:43.440 to the dynamic structural analysis, it’s kind of a distinct, you know, step or 00:53:43.440 --> 00:53:48.540 distinct type of analyses relative to using these simulations for hazard analysis. 00:53:48.550 --> 00:53:51.530 And actually, you know, the amplitude of these ground motions is not a – 00:53:51.530 --> 00:53:54.190 kind of a first-order concern in the – in the talk today. 00:53:54.190 --> 00:53:56.310 Like, the amplitude of the ground motions was specified before 00:53:56.310 --> 00:54:00.750 I started any of these calculations. So if your simulations are a little low 00:54:00.750 --> 00:54:02.820 in amplitude, I’m just going to produce more simulations until 00:54:02.820 --> 00:54:04.070 I get some that are big enough, right? 00:54:04.070 --> 00:54:07.580 So there’s a different set of questions involved with this step of the process. 00:54:08.460 --> 00:54:10.710 And so we’re conditioned on that design shaking amplitude 00:54:10.710 --> 00:54:13.880 and thought about kind of dynamic analysis and structures. 00:54:14.560 --> 00:54:17.720 I would say, you know, big picture, the simulations we were looking at 00:54:17.720 --> 00:54:19.690 were generally suitable for these types of purposes. 00:54:19.690 --> 00:54:24.180 I think you’d get good designs, good assessments out of these. 00:54:24.180 --> 00:54:28.940 And I think, in particular, if we think about situations like near-fault sites or 00:54:28.940 --> 00:54:32.011 sites close to very large-magnitude – or, you know, faults that could produce 00:54:32.011 --> 00:54:36.130 very large-magnitude earthquakes or sites in basins or some sort of unusual, 00:54:36.130 --> 00:54:39.930 you know, geometric or geologic condition, you know, we could really 00:54:39.930 --> 00:54:42.690 move beyond this traditional thinking of saying, well, I got to get a recording 00:54:42.690 --> 00:54:45.540 from somewhere else, and if that doesn’t reflect all the, you know, 00:54:45.540 --> 00:54:48.700 unique circumstances of my site, you know, that’s just what I got to do. 00:54:48.700 --> 00:54:53.380 I think we could get around that, which would be really exciting and beneficial. 00:54:53.380 --> 00:54:55.880 And so, you know, this work has all been kind of 00:54:55.880 --> 00:54:58.520 in pretty close discussion with practicing engineers. 00:54:58.520 --> 00:55:01.190 And I think we’re – kind of the next steps is to the continue kind of showing 00:55:01.190 --> 00:55:04.540 these results to engineers, getting their questions, showing these results to the 00:55:04.540 --> 00:55:06.970 people who do peer reviews and say, you know, if somebody brought these 00:55:06.970 --> 00:55:09.580 to a project, what questions or concerns would you have 00:55:09.580 --> 00:55:12.680 so that we can have them kind of addressed ahead of time and not kind of 00:55:12.680 --> 00:55:15.980 get stuck in weeks and weeks of meetings to try to resolve them. 00:55:15.980 --> 00:55:18.340 And then also think about, can we – can we produce some kind of standard 00:55:18.340 --> 00:55:22.520 sets of simulations that might be available to be kind of more – 00:55:22.520 --> 00:55:24.160 grabbed more quickly? 00:55:24.160 --> 00:55:27.490 And all of that is certainly going on with the ground motion simulators. 00:55:27.490 --> 00:55:31.360 And for me, personally, that’s happening a lot through SCEC. 00:55:31.360 --> 00:55:33.920 And there’s other people around the world doing kind of similar things. 00:55:33.920 --> 00:55:38.320 So those are a few thoughts of kind of where I think the state of engineering 00:55:38.320 --> 00:55:41.250 practice is with use of these. I think the future is bright. 00:55:41.250 --> 00:55:44.220 And hopefully these are kind of some food for thought about ways we might 00:55:44.220 --> 00:55:48.210 be able to make progress in kind of mainstreaming these tools more 00:55:48.210 --> 00:55:50.720 broadly in the engineering community. So thanks for your attention this 00:55:50.720 --> 00:55:53.000 morning, and I’m happy to take questions for a while. 00:55:53.000 --> 00:55:59.120 [Applause] 00:56:04.800 --> 00:56:08.880 - [inaudible] your three-dimensional building model. 00:56:08.980 --> 00:56:09.980 - Yeah. 00:56:10.700 --> 00:56:16.760 - Maybe it’s a typo. You have two designations – BRBF. 00:56:16.760 --> 00:56:19.740 - Yep. - And then BRRF. 00:56:19.740 --> 00:56:23.000 Is that a typo, or … - It could have been. I don’t remember. 00:56:23.010 --> 00:56:27.410 - Because what is BRRF? - Let’s make sure I got the 00:56:27.410 --> 00:56:31.180 right slide first, and then let’s … - Yeah. 00:56:31.540 --> 00:56:34.960 - We’ve got a buckling restrained brace frame. 00:56:35.480 --> 00:56:37.280 Yeah. I think that’s a BRBF, that should be the acronym. 00:56:37.290 --> 00:56:39.970 - Okay. - So that’s – these are these diagonal- 00:56:39.970 --> 00:56:42.580 braced frames that have the kind of concrete 00:56:42.580 --> 00:56:46.310 encasement around a steel brace so that they can kind of yield in compression 00:56:46.310 --> 00:56:48.690 and tension. You see them kind of on lots of the projects around here. 00:56:48.690 --> 00:56:52.400 So, yeah, it’s buckling restrain brace frames is all of the red portions. 00:56:52.400 --> 00:56:54.360 And then the moment-resisting frames are these – the portals 00:56:54.360 --> 00:56:56.880 without the diagonal brace in them. 00:56:59.640 --> 00:57:04.600 [Silence] 00:57:05.400 --> 00:57:09.840 - Thanks, Jack. That was a really great talk, and I learned a lot, for sure. 00:57:09.850 --> 00:57:12.230 So I have a question about the scaling. And I know you’re not really 00:57:12.230 --> 00:57:15.180 talking about that, but I think you do other work on scaling? 00:57:15.180 --> 00:57:16.800 - Yeah. - Could you maybe go to 00:57:16.800 --> 00:57:23.480 the table that I liked. [chuckles] So I guess maybe can you add some 00:57:23.480 --> 00:57:28.160 insight – this, and then – and then – I guess my question is that, 00:57:28.160 --> 00:57:31.280 in the simulations, you don’t have to scale because you’re specifying 00:57:31.280 --> 00:57:35.170 the exact magnitude and distance and target that you want. 00:57:35.170 --> 00:57:36.930 But on the recordings, you have to scale them. 00:57:36.930 --> 00:57:39.690 And, as you said, you try not to scale them too much, which is great. 00:57:39.690 --> 00:57:41.860 But aren’t you inherently then changing the frequency 00:57:41.860 --> 00:57:45.920 content when you’re scaling them? And then, if you go to the next slide, 00:57:45.920 --> 00:57:48.640 is that resulting in – oh, maybe it’s not as obvious. 00:57:48.640 --> 00:57:53.760 But it seems like you get more – a wider range of periods on the recordings 00:57:53.760 --> 00:57:57.500 than you do on the simulations. So is that … 00:57:57.500 --> 00:57:59.611 - Yeah. - … better? Does that imply the 00:57:59.611 --> 00:58:02.330 simulations are better in that sense? Or are you missing some … 00:58:02.330 --> 00:58:05.080 - Yeah, that’s … - … variability? Or, inherently in the 00:58:05.080 --> 00:58:08.060 recordings, you’re changing the stress drop of the events, right? 00:58:08.060 --> 00:58:10.940 Because you’re scaling the time series but not the frequency content? 00:58:10.940 --> 00:58:12.100 - Right. Yeah. 00:58:12.100 --> 00:58:13.630 - I don’t know if there was a question there. [laughs] My understanding. 00:58:13.630 --> 00:58:17.800 - Yeah, no, no. I was going to say, that’s a great topic to probe a little bit – 00:58:17.800 --> 00:58:22.640 this ground motion scaling. So it is pretty standard – I’m trying 00:58:22.640 --> 00:58:24.800 to think if I’ve ever seen a project where it didn’t happen. 00:58:24.800 --> 00:58:27.490 It’s pretty universal that we’re going to amplitude-scale these ground motions. 00:58:27.490 --> 00:58:29.520 Because we’re looking at these high amplitudes where we just 00:58:29.520 --> 00:58:32.260 don’t have a lot of recordings. And that’s one of these – so that, 00:58:32.260 --> 00:58:35.340 I would say, is one of the big – one of the big question marks on 00:58:35.350 --> 00:58:38.520 the use of recorded ground motions. It’s just been going on for 30 years. 00:58:38.520 --> 00:58:40.790 And so we kind of quit asking the question as freshly as we ask the 00:58:40.790 --> 00:58:44.910 questions about the new – the new alternative in town. 00:58:44.910 --> 00:58:48.190 So I would say – so, when we do that, we basically take all the accelerations 00:58:48.190 --> 00:58:51.400 and just multiply them by some constant. So this – in terms of these response 00:58:51.400 --> 00:58:54.710 spectra, because it’s the response of a linear oscillator, if we – if we 00:58:54.710 --> 00:58:57.800 double all the amplitudes, the response of the linear oscillator will just double. 00:58:57.800 --> 00:59:02.160 So these response spectra will just move up and down vertically as we scale. 00:59:02.160 --> 00:59:03.660 So there’s not any change in the – 00:59:03.660 --> 00:59:06.340 like, the relative amplitudes with frequency or anything. 00:59:06.980 --> 00:59:09.600 And so it’s convenient, actually, for finding ground motions 00:59:09.600 --> 00:59:11.960 because I just have to go find a shape, and then I can pick the scale factor, 00:59:11.960 --> 00:59:14.900 then kind of dial it in. It’s not physical. 00:59:15.580 --> 00:59:20.420 And the – I think the – you know, it went out when the competitor was, 00:59:20.420 --> 00:59:24.880 take a real recording and amplitude-scale it, or modulate white noise. 00:59:24.880 --> 00:59:27.230 Because the real recording at least has some sort of non-stationarities 00:59:27.230 --> 00:59:29.410 and some other complexities in it. 00:59:29.410 --> 00:59:33.700 Although it’s no longer kind of the recording that came from an earthquake. 00:59:34.600 --> 00:59:38.240 So, yeah, so that’s very much a flaw in the current scheme, or a limitation 00:59:38.240 --> 00:59:42.270 of the – of the current practice. It’s just – it’s been done so many times 00:59:42.270 --> 00:59:47.360 that there’s kind of more comfort with it. You know, and I – to a first order, 00:59:47.360 --> 00:59:49.920 these response spectrum – and maybe I should have made that more explicit – 00:59:49.920 --> 00:59:52.700 these response spectra are, like, the indicator of how the building 00:59:52.700 --> 00:59:55.500 is going to perform. And so basically, when you see 00:59:55.500 --> 00:59:58.330 these response spectra matching up strongly from left to right, you’ve got 00:59:58.330 --> 01:00:01.340 a pretty strong sense that the building response is going to be the same. 01:00:01.340 --> 01:00:04.020 So a building is more complex than a linear oscillator, 01:00:04.020 --> 01:00:08.800 but these things are pretty predictive. And so, you know, some studies have 01:00:08.800 --> 01:00:11.950 tried to use kind of unscaled ground motions versus ground motions scaled to 01:00:11.950 --> 01:00:14.580 the comparable spectra and then look at the demands on the structure. 01:00:14.580 --> 01:00:16.380 And, you know, in a lot of circumstances, 01:00:16.380 --> 01:00:20.800 they’re pretty comparable. So I don’t – I don’t lose sleep over it, 01:00:20.800 --> 01:00:24.880 but I think you’re very right to point this out as, that’s clearly 01:00:24.880 --> 01:00:28.540 an inferiority of using recordings versus using simulations. 01:00:28.540 --> 01:00:32.680 - So do you think there’s any advantage in that sense to the simulations, then? 01:00:32.680 --> 01:00:34.320 - Yeah. Definitely. - And are you saying that 01:00:34.320 --> 01:00:36.940 this is just one example? Is that something you’re 01:00:36.940 --> 01:00:39.500 sort of advocating or seeing? - Yeah. So I think, you know, 01:00:39.510 --> 01:00:42.820 in terms of this table, like, all – every one of these columns kind of 01:00:42.820 --> 01:00:44.460 comes down on the side of the simulations, right? 01:00:44.460 --> 01:00:48.340 I’ve got those – I’ve got the seismic, you know, event pinned down 01:00:48.340 --> 01:00:50.510 exactly what I want. I’ve got my site pinned down exactly 01:00:50.510 --> 01:00:56.090 what I want, especially if I’m doing kind of 3D, you know, crustal modeling. 01:00:56.090 --> 01:00:59.500 And I’ve got the amplitudes kind of that are naturally resulting. 01:01:00.200 --> 01:01:03.060 And so, yeah, conceptually, it’s far superior, right? 01:01:03.060 --> 01:01:07.700 So I’m not – I’m not here to – yeah, I mean, I think it’s the future. 01:01:07.710 --> 01:01:10.530 And I’m really just – I’m trying to think through, like, what are the steps we 01:01:10.530 --> 01:01:14.940 need to – you know, to get through so that that future is realized? 01:01:17.380 --> 01:01:21.540 - Keeping on the same theme, I think what Annemarie’s point – 01:01:21.540 --> 01:01:26.130 when she was getting at the frequency is that the simulations for, like, 01:01:26.130 --> 01:01:30.870 a magnitude 7 or, like, a real magnitude 7 may have different 01:01:30.870 --> 01:01:34.860 frequency content than a magnitude 6.5 scaled up. 01:01:34.860 --> 01:01:36.420 - Yeah. - So it’s not necessarily that the 01:01:36.420 --> 01:01:40.720 scaling changes the frequency, but you actually have the wrong frequencies 01:01:40.720 --> 01:01:43.420 when you’re scaling the recordings. - Sure. 01:01:43.420 --> 01:01:48.980 - And so my question is, you showed the example for a five-story building. 01:01:48.980 --> 01:01:53.260 And so it’s primarily going to be responding to higher frequencies. 01:01:53.260 --> 01:01:57.280 Ground motions are a little more stable at, say, higher frequencies than they are 01:01:57.280 --> 01:02:02.040 at sort of long periods, and especially when you’re scaling and you’re – if the 01:02:02.040 --> 01:02:06.350 longer periods are more interesting, in the NGA-West2 database where 01:02:06.350 --> 01:02:11.460 they filtered out a lot of long periods, we may see a bigger discrepancy. 01:02:11.460 --> 01:02:16.240 And so I’m curious, are you aware of sort of systematic studies that have 01:02:16.250 --> 01:02:22.590 looked at cases where you’re in in-close large-magnitude events where we think 01:02:22.590 --> 01:02:26.869 PGV and PGD are going to have a greater importance, 01:02:26.869 --> 01:02:30.720 and the scaling may be inaccurate showing that we may have issues 01:02:30.720 --> 01:02:33.900 there when really – when we use that approach for design. 01:02:34.620 --> 01:02:38.100 - Yeah. Let me think of – couple thoughts on that. 01:02:38.100 --> 01:02:42.390 So, yeah, so I guess I see your point about the – when I’m taking some 01:02:42.390 --> 01:02:46.300 of these events that are from kind of some different magnitudes, certainly it 01:02:46.300 --> 01:02:50.380 relates to the frequency content. The site conditions, similarly. 01:02:50.380 --> 01:02:52.460 And so there is – those things are playing a role in the – 01:02:52.460 --> 01:02:54.730 in the frequency content present in the recordings. 01:02:54.730 --> 01:02:58.110 I would say, in this particular step of the analysis, we mitigate that a lot because 01:02:58.110 --> 01:03:01.220 we’ve specified a design spectrum. And so it may be that most 01:03:01.220 --> 01:03:06.520 6-1/2-magnitude earthquakes have less, you know, low-frequency energy 01:03:06.520 --> 01:03:10.060 than the one that I want, but because I’m specifying a design spectrum, 01:03:10.060 --> 01:03:12.270 I’m only going to pick the 6-1/2-magnitude recordings 01:03:12.270 --> 01:03:16.270 that had a lot of low-frequency energy consistent with what the 01:03:16.270 --> 01:03:18.400 hazard analysis said. So it’s really the hazard analysis 01:03:18.400 --> 01:03:21.720 that has to catch the appropriate kind of amplitudes. 01:03:21.720 --> 01:03:23.950 And then here, we’re picking a kind of small subset of 01:03:23.950 --> 01:03:28.520 recordings that match that. So I would say we’re somewhat robust 01:03:28.520 --> 01:03:33.160 to that issue, but it is a – it is a real issue in this – again, in this manipulation here. 01:03:33.920 --> 01:03:36.920 Yeah. So the one particular building I’ve got is not the – it’s not the 01:03:36.930 --> 01:03:40.330 poster child for using simulations because it’s got a shorter period, 01:03:40.330 --> 01:03:42.440 and that’s kind of where things are less interesting. 01:03:42.440 --> 01:03:45.280 One of the buildings that’s currently getting analyzed is a 40-story building 01:03:45.280 --> 01:03:48.420 with a 5-second period, and I think that’ll be more interesting in kind of 01:03:48.420 --> 01:03:51.800 getting at some of those longer-period issues. 01:03:52.460 --> 01:03:54.420 Yeah, there – I mean, there is some literature on kind of trying to look at 01:03:54.420 --> 01:03:58.220 sensitivity of these record properties and what’s predictive of response. 01:03:58.780 --> 01:04:03.640 And – but, you know, these giant data sets of simulations are going to let us 01:04:03.640 --> 01:04:06.700 take a much more refined look at that, I think, than kind of the more limited 01:04:06.700 --> 01:04:09.970 studies we’ve – to some degree, we’re limited when we use empirical 01:04:09.970 --> 01:04:13.820 recordings to do this because we’ve got hundreds or thousands of permutations 01:04:13.820 --> 01:04:17.300 we can try rather than, you know, many, many thousands. 01:04:18.740 --> 01:04:22.840 - I have a more general question. You said that this kind of a dynamic 01:04:22.840 --> 01:04:25.560 analysis is used for designing new buildings. 01:04:25.560 --> 01:04:29.300 Is there any effort underway to look at the existing inventory of buildings 01:04:29.300 --> 01:04:33.190 that weren’t designed with the sophistication now available to us 01:04:33.190 --> 01:04:35.090 to see which ones are at risk? 01:04:35.090 --> 01:04:40.210 - Yeah. Sure. So there’s – I think there’s two issues. 01:04:40.210 --> 01:04:43.820 There’s kind of the – yeah, the sophistication of the building designs. 01:04:43.820 --> 01:04:46.230 And then it’s – there’s a question of whether the building was designed 01:04:46.230 --> 01:04:48.590 or analyzed using an explicit dynamic analysis 01:04:48.590 --> 01:04:52.510 or some sort of an equivalent static analysis. 01:04:52.510 --> 01:04:55.910 So the – I guess the broader question of kind of existing buildings 01:04:55.910 --> 01:05:00.740 and their safety is a huge, you know, issue for us in the engineering 01:05:00.740 --> 01:05:05.570 community and just as a society. And we have, you know, all sorts of 01:05:05.570 --> 01:05:08.870 efforts going on in the research world and in the – you know, in practice to 01:05:08.870 --> 01:05:11.280 upgrade older buildings and in the policy world, you know, 01:05:11.280 --> 01:05:14.040 San Francisco and Los Angeles have been kind of leaders 01:05:14.040 --> 01:05:17.220 in trying to mandate some of these upgrades. 01:05:17.220 --> 01:05:22.740 In terms of the way that those buildings are assessed or the way that they are 01:05:22.740 --> 01:05:27.720 designed, really the split is not so much on new and old as much as the 01:05:27.720 --> 01:05:31.100 complexity of the response and whether we think kind of explicitly understanding 01:05:31.110 --> 01:05:35.400 the dynamic behavior is important. So I might have a – you know, 01:05:35.400 --> 01:05:38.050 a three-story concrete building, and it could be a brand-new building 01:05:38.050 --> 01:05:40.870 that I think is going to perform excellently, or a 1960s building 01:05:40.870 --> 01:05:44.530 that I think is really dangerous. It’s a – it’s a relatively simple problem 01:05:44.530 --> 01:05:46.670 in terms of the way it’s going to respond dynamically. 01:05:46.670 --> 01:05:51.110 And so generally, we think we can get a pretty good understanding of the 01:05:51.110 --> 01:05:54.610 performance of that building, whether good or bad, by analyzing it statically. 01:05:54.610 --> 01:05:59.480 And it’s just a much – it’s a much simpler problem to do a static analysis 01:05:59.480 --> 01:06:01.610 and then to say, well, under this amount of static load, 01:06:01.610 --> 01:06:04.320 I’m getting bad performance. So I can then infer that, 01:06:04.320 --> 01:06:08.180 under strong earthquake shaking, I know I’m going to get bad performance. 01:06:08.180 --> 01:06:13.160 So the two are related but somewhat independent questions. 01:06:13.160 --> 01:06:18.200 And sometimes with these older buildings, it’s pretty clear-cut that the 01:06:18.210 --> 01:06:21.160 performance is lacking in some way, and there’s maybe some particular detail, 01:06:21.160 --> 01:06:23.560 like a beam-column connection that we know is lacking. 01:06:23.560 --> 01:06:25.660 And building a sophisticated computer model isn’t going to 01:06:25.660 --> 01:06:29.280 bring us more information. We know we need to improve it in some way. 01:06:29.280 --> 01:06:32.120 And really, the question is about how to improve it rather than really getting a 01:06:32.120 --> 01:06:36.120 detailed understanding of the dynamic behavior, if that makes sense. 01:06:37.880 --> 01:06:45.240 - Yeah, Jack. For the 11 simulations, it was like you 01:06:45.240 --> 01:06:48.280 got 11 earthquakes that are all the same. - Yep. 01:06:48.280 --> 01:06:51.220 - And, of course, they aren’t. Or I hope they aren’t. 01:06:51.220 --> 01:06:52.200 - No. 01:06:52.200 --> 01:06:55.960 - And they aren’t, as showed by the response spectra. 01:06:55.960 --> 01:07:01.320 Just what is the variability from one event to the other? 01:07:01.320 --> 01:07:08.500 And is there any rupture propagation here if something starts off to the south? 01:07:08.500 --> 01:07:09.980 - Yeah. 01:07:09.990 --> 01:07:15.400 - These are kind of maybe 30-kilometer- long ruptures, just for example. 01:07:15.400 --> 01:07:20.100 And it starts off to the south and ruptures toward you. 01:07:20.100 --> 01:07:24.700 Your polarization is going to be one thing, but as it ruptures past you, 01:07:24.700 --> 01:07:28.740 it will be very different. - Right. 01:07:28.740 --> 01:07:33.260 - And, as it goes past you, it’ll be different yet again. 01:07:33.260 --> 01:07:38.440 And then, as Annemarie was saying, the stress drop can vary. 01:07:38.440 --> 01:07:39.760 - Yep. 01:07:40.580 --> 01:07:44.560 - All kinds of things can vary, so yeah, what is the range of 01:07:44.560 --> 01:07:47.680 variability in those models? - Yeah. So these are – yeah, 01:07:47.690 --> 01:07:50.610 good question. I should anticipate that in this community. 01:07:50.610 --> 01:07:55.370 So these are all kinematic rupture simulations, and the – I think the rupture 01:07:55.370 --> 01:07:58.470 geometry was pinned down – kind of, if you pin the geometry 01:07:58.470 --> 01:08:02.500 versus pinning the magnitude. I think the geometry was specified, 01:08:02.500 --> 01:08:05.480 and there was 10 hypocenters placed, you know, a couple depths, and then a 01:08:05.480 --> 01:08:07.230 long rupture in a few different places. 01:08:07.230 --> 01:08:11.930 And then there’s a kinematic rupture generator, and I think we ran 10 ruptures 01:08:11.930 --> 01:08:16.060 for each hypocenter location of just the slip distribution over time. 01:08:16.060 --> 01:08:20.000 And so that’s – so it’s almost all rupture variability. 01:08:20.000 --> 01:08:22.660 So there’s no path variability. It’s the same fault. 01:08:22.660 --> 01:08:26.140 It’s the same site every time. And it’s just different realizations 01:08:26.140 --> 01:08:29.750 of the – of the slip that are generating the variability in the ground motion and 01:08:29.750 --> 01:08:33.640 the variability in the response spectra. - So that’s the closest distance because 01:08:33.640 --> 01:08:37.210 then these different hypocenters could be – would be different 01:08:37.210 --> 01:08:39.190 distances from the building. - Right. So some of them the hypocenter 01:08:39.190 --> 01:08:42.500 is close to the site, and it’s mostly rupturing away, and some are bilateral. 01:08:42.500 --> 01:08:44.310 Some are mostly rupturing south to north. 01:08:44.310 --> 01:08:47.200 This is up pretty close to the north end of the fault, the site. 01:08:47.200 --> 01:08:50.420 - Okay. And so, in the ... 01:08:53.900 --> 01:08:58.240 What is the principle – is directivity an important source 01:08:58.240 --> 01:09:02.460 in the variability that we’ll see on the next slide? 01:09:03.080 --> 01:09:07.840 - Well, so the rupture – the variability is driving all of this. 01:09:07.840 --> 01:09:11.060 Kind of how we – how much we want to attribute to directivity. 01:09:11.060 --> 01:09:12.150 - Yeah. 01:09:12.150 --> 01:09:16.170 - I would say – I mean, the interesting things is, because we’re – you know, 01:09:16.170 --> 01:09:18.339 we’re specifying ground motion amplitudes that are bigger 01:09:18.339 --> 01:09:21.410 than the median amplitude. And these simulations are 01:09:21.410 --> 01:09:23.970 pretty well pinned that kind of median amplitudes look like 01:09:23.970 --> 01:09:25.260 the medians we see in recordings. 01:09:25.260 --> 01:09:28.910 But these high amplitudes are generally – kind of rupture towards 01:09:28.910 --> 01:09:32.360 the site – kind of a classic forward-directivity-type situation. 01:09:32.360 --> 01:09:34.830 Because the backwards directivity situation generally gives you 01:09:34.830 --> 01:09:37.430 lower-amplitude shaking. And then we wouldn’t pick them, right? 01:09:37.430 --> 01:09:40.850 So this is only a subset of all the simulations, and it’s basically the – 01:09:40.850 --> 01:09:44.400 loosely speaking, it’s the 11 strongest simulations. 01:09:44.400 --> 01:09:48.630 So – and that’s always a – you know, the question is not really explicit in our 01:09:48.630 --> 01:09:52.529 hazard analysis or in our ground motion selection, but, you know, usually we’re 01:09:52.529 --> 01:09:56.230 thinking about a magnitude and a distance, which is a pretty simple story. 01:09:56.230 --> 01:09:59.450 But implicitly, I think there is an assumption in the forward directivity 01:09:59.450 --> 01:10:03.170 when we’re saying we want a stronger- than-average ground motion amplitude. 01:10:03.170 --> 01:10:06.730 Or whatever the rupture and wave propagation circumstances give us the 01:10:06.730 --> 01:10:10.870 biggest amplitudes, more generally. - So one final question, then. 01:10:10.870 --> 01:10:19.560 The short period – at short periods, the variability looks to be significantly less 01:10:19.560 --> 01:10:23.590 for the simulations than the recordings. - Yeah. 01:10:23.590 --> 01:10:28.230 - Do you have any thoughts on that? - I do. Yeah. 01:10:28.230 --> 01:10:34.900 So this is the subtlety of this whole game that I didn’t bring up explicitly. 01:10:34.900 --> 01:10:39.670 So when we – when we go find these time series, we have to – so there was 01:10:39.670 --> 01:10:43.660 the language back on the – when I had the picture of the ASCE 7 book that said, 01:10:43.660 --> 01:10:47.530 you have to find ground motions consistent with the response spectrum. 01:10:47.530 --> 01:10:50.700 And it’s over a particular period range. So we do this – we establish ahead of 01:10:50.700 --> 01:10:54.060 time what the period range that we think is important for the building. 01:10:54.060 --> 01:10:56.290 So loosely speaking, we’ve got, like, kind of a first-mode period 01:10:56.290 --> 01:10:58.520 of vibration. We double that. 01:10:58.520 --> 01:11:00.230 And then we would go down to, like, a second- and a 01:11:00.230 --> 01:11:01.640 third-mode period of vibration. 01:11:01.640 --> 01:11:04.760 And say, over that whole range, I need to get a pretty good match. 01:11:04.760 --> 01:11:07.180 And then outside of that range, I’m less concerned because 01:11:07.180 --> 01:11:10.780 I don’t think that’s really indicating what’s going to happen to my building. 01:11:10.780 --> 01:11:14.880 So when I work with recordings, that – as soon as I shut off that 01:11:14.880 --> 01:11:19.240 kind of window, things can go a little bit more wild. 01:11:19.240 --> 01:11:22.200 Because that’s where the record scaling, for instance, can be an issue. 01:11:22.200 --> 01:11:27.080 If I’m not pinning the spectrum at these other frequencies, and if I’m bringing a 01:11:27.080 --> 01:11:29.830 low-magnitude earthquake up to bigger amplitudes, those high frequencies 01:11:29.830 --> 01:11:33.040 are coming up maybe too high. Whereas, in the simulations, 01:11:33.040 --> 01:11:35.270 that’s less of an issue because I’m not doing the processing. 01:11:35.270 --> 01:11:38.989 So it’s – that’s another thing that – and in these high-rise buildings, 01:11:38.989 --> 01:11:43.450 these – you know, the short frequencies, you know, that might be a hertz still. 01:11:43.450 --> 01:11:46.470 And that can be important. You know, the good news is, 01:11:46.470 --> 01:11:50.450 our kind of – our scaling stuff that’s, you know, fingernails on a chalkboard 01:11:50.450 --> 01:11:54.400 for many of you, I think, it only overloads the structures because 01:11:54.400 --> 01:11:57.200 you get these big kind of high amplitudes over at the far left. 01:11:57.200 --> 01:11:59.910 In the simulations. That’s a little bit less of a problem. 01:11:59.910 --> 01:12:02.370 So it is an artifact of these complicated rules, 01:12:02.370 --> 01:12:06.210 but I kind of dodged explaining all the details in them. 01:12:06.210 --> 01:12:08.280 So good eyes to catch that. 01:12:09.560 --> 01:12:19.720 - So I – you – I find it interesting that, on your drift plots, you have only 11 or 01:12:19.730 --> 01:12:24.710 so records, and you choose the median, and it looks fine, and you move on. 01:12:24.710 --> 01:12:26.210 - Yeah. - The building’s good. 01:12:26.210 --> 01:12:29.160 But two or three of the records show the building is damaged. 01:12:29.680 --> 01:12:34.940 - Yeah. Well, so yeah. That’s a – it’s an interesting thing, 01:12:34.940 --> 01:12:38.630 and I sometimes talk with – so talk with engineers about this. 01:12:38.630 --> 01:12:42.120 But certainly, in the code development – oh, I overshot, didn’t I? 01:12:42.120 --> 01:12:45.090 In the code development world, we talk a lot about this stuff too. 01:12:45.090 --> 01:12:48.760 So the idea – there we go. 01:12:48.760 --> 01:12:51.610 So these figures were – we just kind of looked at the heavy red lines, 01:12:51.610 --> 01:12:54.210 and we didn’t look at all those gray lines. But some of the gray lines can be, 01:12:54.210 --> 01:12:56.360 you know, double the red line in terms of amplitudes. 01:12:56.360 --> 01:13:01.070 So it’s not that we ignore those things. It’s that we don’t think we can measure 01:13:01.070 --> 01:13:06.610 them from 11 dynamic analyses. So if you let me do 40 of these analyses, 01:13:06.610 --> 01:13:09.790 or, you know, 50 or something, then I would start to feel more confident 01:13:09.790 --> 01:13:12.370 about being able to estimate a probability distribution. 01:13:12.370 --> 01:13:16.310 So what – but that is not – that’s kind of a deal-breaker from the people in these 01:13:16.310 --> 01:13:19.000 design offices trying to get these designs to crank through. 01:13:19.000 --> 01:13:22.710 So what we’ve done is we have – and building codes in general, 01:13:22.710 --> 01:13:25.290 for other types of loads as well – live loads and wind loads and 01:13:25.290 --> 01:13:28.160 everything, we do the same thing. That we really – we pin ourselves 01:13:28.160 --> 01:13:33.050 off these nominal values. So I’ll say the average of my, you know, 01:13:33.050 --> 01:13:36.280 drift ratios – or, story drift ratios has to be less than 4%. 01:13:36.280 --> 01:13:40.820 And if that’s the case, and if I have some – as a code calibration exercise, 01:13:40.820 --> 01:13:44.380 I’ve got an idea of how much variability I’m going to see in a typical building 01:13:44.380 --> 01:13:47.630 of this type. There’s other factors in there that builds in – safety in. 01:13:47.630 --> 01:13:51.830 So it’s not that if any one of these gets past 4%, the building is in – 01:13:51.830 --> 01:13:55.640 you know, is not going to perform. It’s – I might think I’ve got a 6 or 7% 01:13:55.640 --> 01:14:00.680 capacity, but I can – based on kind of uncertainty in what that capacity is, and 01:14:00.680 --> 01:14:04.050 based on the variability from record to record in my responses, I can back off 01:14:04.050 --> 01:14:09.340 that an average of 4% will get me kind of an acceptable kind of margin in here. 01:14:09.340 --> 01:14:11.040 And there’s a whole bunch of factors like that. 01:14:11.040 --> 01:14:13.700 There’s – some of you have probably heard about kind of R factors. 01:14:13.700 --> 01:14:15.770 That’s kind of a big, famous one. 01:14:15.770 --> 01:14:19.440 There’s factors around kind of the means versus the tails. 01:14:19.440 --> 01:14:21.810 When we design – when we choose the return period of the ground motions 01:14:21.810 --> 01:14:24.140 we’re designing for, there’s another factor, right? 01:14:24.140 --> 01:14:26.920 And none of that is to say that the building is indestructible. 01:14:26.920 --> 01:14:29.780 But we’re calibrating towards this kind of low probability that the 01:14:29.780 --> 01:14:32.760 building could have a collapse. And all those decisions are somewhat 01:14:32.760 --> 01:14:36.050 hidden from the designer so that the designer has got kind of a more limited 01:14:36.050 --> 01:14:38.560 set of calculations they have to think about. 01:14:38.560 --> 01:14:41.400 And then the code, hopefully, has built some robustness that, 01:14:41.410 --> 01:14:45.070 if those limited checks are sufficient, then we’ve got confidence that the 01:14:45.070 --> 01:14:49.960 real reliability of the system is okay. So that’s the philosophy behind that. 01:14:49.960 --> 01:14:52.860 - Okay, we have maybe one last question from Mehmet, and then 01:14:52.860 --> 01:14:55.460 we should probably wrap this up. - Okay. 01:14:55.460 --> 01:15:02.400 Well, I hope this won’t take too long. But I was a little bit concerned about 01:15:02.400 --> 01:15:09.080 the simulated – simulation table where you have Vs30, I guess, 01:15:09.080 --> 01:15:13.120 because it’s not clear to see. You have all 500. 01:15:13.120 --> 01:15:20.530 Now, how does that relate when you consider basin effects, 01:15:20.530 --> 01:15:23.600 for example, as you mentioned in Los Angeles? 01:15:23.600 --> 01:15:29.520 That we know from Tohoku, what happened in Osaka and Tokyo. 01:15:29.520 --> 01:15:34.700 Then there’s Dubai – tall buildings, earthquakes from Iran. 01:15:34.700 --> 01:15:46.700 And when you go to depths of 2,000, 3,000 meters, then obviously the Vs30 01:15:46.700 --> 01:15:53.800 being to 500 meters per second doesn’t bode well at all. 01:15:53.800 --> 01:15:58.000 I mean, we have evidence of this, and we have databases. 01:15:58.000 --> 01:16:04.080 And writing – Brad can talk about this database and all this. 01:16:04.080 --> 01:16:07.380 - Right. - Yeah, so in a sense, is there – 01:16:07.380 --> 01:16:10.560 there’s no connection. I don’t see the connection. 01:16:10.560 --> 01:16:13.540 Do you have some suggestion on this? 01:16:13.540 --> 01:16:18.280 - Sure. Yeah. So for this particular site in Berkeley, they had geotechnical 01:16:18.290 --> 01:16:20.880 investigations, and they had a site profile, and then this 500 … 01:16:20.880 --> 01:16:22.160 - Oh, this is only for Berkeley. 01:16:22.160 --> 01:16:25.740 - Yeah. This is for that Berkeley site. So basically, we specified the site 01:16:25.740 --> 01:16:29.350 conditions, and then, in the broadband platform, you’ve got some kind of 01:16:29.350 --> 01:16:33.220 limited ability to address that condition. So – but the basins – so more broadly, 01:16:33.220 --> 01:16:37.310 thinking about the role of simulations, I think sites in basins is one place where, 01:16:37.310 --> 01:16:39.830 again, simulations would – that’s the future. 01:16:39.830 --> 01:16:42.690 Like, we just – we can’t make progress on that problem 01:16:42.690 --> 01:16:45.640 with kind of traditional techniques. And so, to the extent that these 01:16:45.640 --> 01:16:50.770 simulations can reflect basin effects, that’s going to be a great asset for us. 01:16:50.770 --> 01:16:54.560 I will say, you know, kind of, as we think about this path forward, so the – 01:16:54.560 --> 01:16:57.840 you know, the Los Angeles basin has been kind of a very, you know, 01:16:57.840 --> 01:17:01.130 well-studied basin in the simulation world by some of the people 01:17:01.130 --> 01:17:03.920 that I work with. You know, the amplitudes associated 01:17:03.920 --> 01:17:06.900 with kind of waves, you know, getting channeled into that basin 01:17:06.900 --> 01:17:11.370 and things have – they’re not, you know, necessarily fully stable across kind of 01:17:11.370 --> 01:17:13.930 simulation algorithms or iterations of these things. 01:17:13.930 --> 01:17:16.970 And then that Vs30, that near-surface site conditions, you know, when we 01:17:16.970 --> 01:17:20.520 have very soft soils and we have nonlinear amplification, I think there’s 01:17:20.520 --> 01:17:23.440 also still opportunities for the simulations to further improve 01:17:23.440 --> 01:17:26.010 in some of those regards. So these are those complex situations 01:17:26.010 --> 01:17:28.620 where we could really benefit from further insights. 01:17:28.620 --> 01:17:31.750 But they’re also complex problems from the scientific perspective 01:17:31.750 --> 01:17:36.599 to kind of get them right. And so I – yeah. 01:17:36.599 --> 01:17:40.390 But even as we sit here today, there’s really big opportunities to get, you know, 01:17:40.390 --> 01:17:42.460 more insights out of the data that’s existing today. 01:17:42.460 --> 01:17:47.360 And I think that’ll only get better over time as you all continue your science. 01:17:47.360 --> 01:17:49.840 Thanks. - All right. Well, great. 01:17:49.840 --> 01:17:51.240 Thank you very much. 01:17:51.240 --> 01:17:53.780 Please join me in giving our speaker another round of applause. 01:17:53.780 --> 01:17:54.780 - Thank you. 01:17:54.780 --> 01:17:57.080 [Applause] 01:17:57.080 --> 01:18:00.380 - And if you’d like to join us for lunch, we’re going to leave pretty soon, so 01:18:00.380 --> 01:18:05.200 please come on up, and we can continue this conversation over lunch. Thanks. 01:18:12.880 --> 01:18:14.240 - Thank you. 01:18:15.940 --> 01:18:17.720 - That was fantastic. - Thank you.