WEBVTT

ef52afb0-7151-4b8e-a5a5-25f7cf8e64db-0
00:00:01.320 --> 00:00:02.000
Thanks, Catherine.

dd49ce94-ce57-4eaf-b0e5-ebe04c0bb0f3-0
00:00:02.200 --> 00:00:05.480
So it's my pleasure to introduce
our speaker today Omar Issa.

cbf8d153-8d7e-4d19-a68d-fae8eaf1115d-0
00:00:05.880 --> 00:00:10.240
He is the co-founder and CEO of
the start-up ResiQuant.

80972c94-f0a8-42e4-8b6e-d2bf6fd1f266-0
00:00:10.520 --> 00:00:14.268
Omar is a recent graduate from
Stanford, where he earned a PhD

80972c94-f0a8-42e4-8b6e-d2bf6fd1f266-1
00:00:14.268 --> 00:00:17.541
in structural engineering
working with Jack Baker, and

80972c94-f0a8-42e4-8b6e-d2bf6fd1f266-2
00:00:17.541 --> 00:00:20.755
before that a Bachelor of
Science in Engineering from

80972c94-f0a8-42e4-8b6e-d2bf6fd1f266-3
00:00:20.755 --> 00:00:22.600
UCLA, which is where I met him.

a82e7094-51d9-4da6-8f34-72e0b7140ea3-0
00:00:23.760 --> 00:00:27.546
Throughout his career, he's been
very involved with student

a82e7094-51d9-4da6-8f34-72e0b7140ea3-1
00:00:27.546 --> 00:00:29.440
chapters of both PRI and PEER.

6c0d55c6-a558-427b-8ebe-b974ed430d2f-0
00:00:29.560 --> 00:00:33.273
He's also served on the BSSC
Provisions Update Committee, so

6c0d55c6-a558-427b-8ebe-b974ed430d2f-1
00:00:33.273 --> 00:00:35.160
involved in our building codes.

6ea245b2-756f-47f4-af13-884c47167a68-0
00:00:35.680 --> 00:00:38.917
He's won several awards and
scholarships, including the Shaw

6ea245b2-756f-47f4-af13-884c47167a68-1
00:00:38.917 --> 00:00:41.040
Family Fellowship on
Catastrophic risk.

48d2dda6-4df4-4ca6-b896-2d3e399c2de5-0
00:00:42.840 --> 00:00:47.661
And I'm really excited to hear
about what he's been working on

48d2dda6-4df4-4ca6-b896-2d3e399c2de5-1
00:00:47.661 --> 00:00:50.800
at Stanford since we were last
in touch.

a4563352-60b3-497b-901a-355f1d779157-0
00:00:51.120 --> 00:00:52.720
So with that, I'll hand it over.

35041c0f-a467-409b-b8a3-5eb30e96d383-0
00:00:53.200 --> 00:00:56.640
Yeah, and hopefully the camera
picks him up.

e019006f-5c67-46d2-9f7a-b312b1979ffc-0
00:00:56.640 --> 00:00:56.960
Awesome.

56dc7cf0-dbde-41ae-9769-8c96d795e4ec-0
00:00:56.960 --> 00:00:58.080
No, thank you so much Grace.

8e291175-37c4-4f1e-a61c-9fbc7a063d31-0
00:00:58.080 --> 00:00:58.800
This is great!

8ff84586-0ff5-4a5c-8ebb-d57c0ef0b05e-0
00:00:59.160 --> 00:01:00.200
I'm very honored.

431c6107-3a18-4249-999b-1c9e09a43ab9-0
00:01:00.200 --> 00:01:03.670
I've heard that I am one of the
First in this room to be giving

431c6107-3a18-4249-999b-1c9e09a43ab9-1
00:01:03.670 --> 00:01:04.240
this talk.
  
e12d3918-f4ad-4ef6-9191-651e73f0fc64-0
00:01:04.240 --> 00:01:08.360
So there is no bar today, which
is great, very relieving.

bd23af3c-731e-459b-81d4-09b1ee368f81-0
00:01:09.320 --> 00:01:10.600
It's a pleasure to be here.

bafdd6f0-45b8-4030-a935-103d2907a447-0
00:01:10.640 --> 00:01:14.678
I've never been on site and on
the campus, so I've never had a

bafdd6f0-45b8-4030-a935-103d2907a447-1
00:01:14.678 --> 00:01:15.320
reason to.

1b74cdf0-860e-4db4-a50a-98b2541ae3b4-0
00:01:15.320 --> 00:01:16.840
So I really appreciate the
invitation.

c0106196-d9c4-4463-b417-9b67400542fd-0
00:01:16.840 --> 00:01:18.520
It's a pleasure to meet all of
you.

e4ef1085-3d5f-4dc9-b73c-7e8ec97c619f-0
00:01:19.640 --> 00:01:23.273
So I thought about what, you
know, maybe to talk about today,

e4ef1085-3d5f-4dc9-b73c-7e8ec97c619f-1
00:01:23.273 --> 00:01:26.964
and I put together, I think, a
list of things that I think are

e4ef1085-3d5f-4dc9-b73c-7e8ec97c619f-2
00:01:26.964 --> 00:01:28.839
meaningful learnings for my PhD.

b7649370-396a-4214-a071-761c027d5de5-0
00:01:29.160 --> 00:01:31.040
But I also want to introduce you
to myself.

a02a1070-4383-479a-8b92-d1653486d350-0
00:01:31.240 --> 00:01:32.360
You know, how did I get here?

d5607f7f-ee83-4527-813d-83fbedd18ab2-0
00:01:32.360 --> 00:01:35.979
How do I start thinking about
catastrophic risk disasters and

d5607f7f-ee83-4527-813d-83fbedd18ab2-1
00:01:35.979 --> 00:01:36.680
engineering?

a77ca6fa-ad3f-4cd3-9f9e-2e8f9ba19b53-0
00:01:37.360 --> 00:01:38.880
So I want to dial back the
clock.

28231a58-be61-43a3-a6eb-d365c276df20-0
00:01:39.160 --> 00:01:42.502
Let's take it back maybe to the
time I was born, maybe a little

28231a58-be61-43a3-a6eb-d365c276df20-1
00:01:42.502 --> 00:01:42.920
earlier.

b3f6a626-efb7-4b8a-ac52-51e2ef29e464-0
00:01:43.160 --> 00:01:46.035
I was actually, I grew up in a
town not too far from here in

b3f6a626-efb7-4b8a-ac52-51e2ef29e464-1
00:01:46.035 --> 00:01:47.120
Danville, the East Bay.

eef474e1-5759-471d-acb2-4d024f1588c3-0
00:01:47.760 --> 00:01:51.800
My dad is a structural engineer,
graduated from Berkeley in 1989.

7fc64561-193b-453b-9040-4ffbb459b6e4-0
00:01:51.880 --> 00:01:55.120
His whole career was off the
tailwinds of Loma Prieta.

37c4d72f-87f8-460b-a5c3-9b3b6a65b439-0
00:01:55.320 --> 00:01:58.277
He served, you know, designing,
you know, schools, hospitals for

37c4d72f-87f8-460b-a5c3-9b3b6a65b439-1
00:01:58.277 --> 00:02:00.916
his career and actually was
Professor Filippo's first PhD

37c4d72f-87f8-460b-a5c3-9b3b6a65b439-2
00:02:00.916 --> 00:02:01.280
student.

c5814e7c-096a-45c6-8e58-baea1b9030ad-0
00:02:02.760 --> 00:02:04.863
I come from a big family, but
none of my siblings did

c5814e7c-096a-45c6-8e58-baea1b9030ad-1
00:02:04.863 --> 00:02:05.760
structural engineering.

ee52373d-fbde-4c83-a0f7-afb8e71b65fa-0
00:02:05.760 --> 00:02:10.108
So I was really the chosen
one to continue and kind of

ee52373d-fbde-4c83-a0f7-afb8e71b65fa-1
00:02:10.108 --> 00:02:14.253
like, and I just followed my
dad's career, discovered my own

ee52373d-fbde-4c83-a0f7-afb8e71b65fa-2
00:02:14.253 --> 00:02:16.360
path in structural engineering.

f9b0f570-9138-48ea-9397-b1845cb27acd-0
00:02:17.240 --> 00:02:20.703
I went to Jordan in 2009, did
high school there and then kind

f9b0f570-9138-48ea-9397-b1845cb27acd-1
00:02:20.703 --> 00:02:24.167
of took a lap around the world
where I went to UCLA, as Grace

f9b0f570-9138-48ea-9397-b1845cb27acd-2
00:02:24.167 --> 00:02:27.519
mentioned in 2015, had a great
time in Southern California.

f193c075-63b5-4e52-b637-f2dac574ab5c-0
00:02:27.520 --> 00:02:31.353
It's very sunny, very fun, but
it's also where I discovered my

f193c075-63b5-4e52-b637-f2dac574ab5c-1
00:02:31.353 --> 00:02:31.840
passion.

be7086de-b5ae-49e4-addd-26620f8fd6c6-0
00:02:32.320 --> 00:02:35.409
Around 2019, Ridgecrest
earthquake happened, you know,

be7086de-b5ae-49e4-addd-26620f8fd6c6-1
00:02:35.409 --> 00:02:37.600
hundreds of miles east of
downtown LA.

a3630ccc-98d6-4c80-8b10-483cc7f2f33f-0
00:02:38.240 --> 00:02:40.424
Of course, they send the
undergraduates to do the

a3630ccc-98d6-4c80-8b10-483cc7f2f33f-1
00:02:40.424 --> 00:02:41.080
reconnaissance.

3e4a5ce0-359d-412e-a590-e50e1d94b100-0
00:02:41.160 --> 00:02:45.414
So I was out there in 100° heat
inspecting buildings, but also

3e4a5ce0-359d-412e-a590-e50e1d94b100-1
00:02:45.414 --> 00:02:47.440
the surface fault projections 

766eac29-2885-46a4-82e7-354d87c9fcff-0
00:02:47.680 --> 00:02:48.640
following the disaster.

fdc2cfe4-81d6-4e36-9a4b-fbd7af161ad5-0
00:02:49.800 --> 00:02:52.652
I expected buildings in
Ridgecrest and nearby Trona, but

fdc2cfe4-81d6-4e36-9a4b-fbd7af161ad5-1
00:02:52.652 --> 00:02:55.755
my principal responsibility at
the beginning of that time was

fdc2cfe4-81d6-4e36-9a4b-fbd7af161ad5-2
00:02:55.755 --> 00:02:58.358
actually to help with the
photogrammetry efforts of

fdc2cfe4-81d6-4e36-9a4b-fbd7af161ad5-3
00:02:58.358 --> 00:02:59.960
preserving those fault features.

dcc3bee3-396d-4df9-a2cc-9a7de5c7a5a4-0
00:02:59.960 --> 00:03:03.397
So as part of a team, it is my
first time doing reconnaissance

dcc3bee3-396d-4df9-a2cc-9a7de5c7a5a4-1
00:03:03.397 --> 00:03:06.835
that was extremely proud of this
was joined by folks at the 

dcc3bee3-396d-4df9-a2cc-9a7de5c7a5a4-2
00:03:06.835 --> 00:03:10.327
Near East Center and other PhD students at

dcc3bee3-396d-4df9-a2cc-9a7de5c7a5a4-3
00:03:10.327 --> 00:03:10.600
UCLA.

db28f088-0222-44c2-b9ec-09fa6afd1fb8-0
00:03:11.080 --> 00:03:13.360
The default surface rupture was
impressive.

76d9fc3a-85db-4d87-932a-42781d4a1204-0
00:03:13.440 --> 00:03:16.530
I was assisted with flying
drones and in the end we came up

76d9fc3a-85db-4d87-932a-42781d4a1204-1
00:03:16.530 --> 00:03:19.240
with a very valuable data set
that helped the community.

3d4d2b0d-07f0-40a1-944a-5998695a3eee-0
00:03:19.760 --> 00:03:23.400
But it was really looking at the
buildings that inspired me.

18b62968-2b0a-4480-a994-c69bae5fd130-0
00:03:23.680 --> 00:03:26.812
There are obviously types of
buildings and taxonomies that we

18b62968-2b0a-4480-a994-c69bae5fd130-1
00:03:26.812 --> 00:03:29.743
expect to suffer some type of
damage in these events, but

18b62968-2b0a-4480-a994-c69bae5fd130-2
00:03:29.743 --> 00:03:31.360
there are others that we do not.

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-0
00:03:31.600 --> 00:03:35.889
And I think at the time when I
was in Ridgecrest, hearing about

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-1
00:03:35.889 --> 00:03:40.111
how hospitals, in this case one
of the hospitals in Ridgecrest

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-2
00:03:40.111 --> 00:03:44.467
had to evacuate patients who are
inside due to, you know, non, I

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-3
00:03:44.467 --> 00:03:48.690
guess not really so severe, non
structural damage kind of made

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-4
00:03:48.690 --> 00:03:52.912
me think about this discrepancy
between our building codes and

dd26c9bf-c55e-47c9-90a1-1a85107c0f96-5
00:03:52.912 --> 00:03:54.320
what society expects.

87b67825-6a63-4352-ab3d-ce028e934e81-0
00:03:54.720 --> 00:03:57.910
And this discrepancy ultimately
is what inspired me to go to

87b67825-6a63-4352-ab3d-ce028e934e81-1
00:03:57.910 --> 00:04:00.630
Stanford and spend the past
several years trying to

87b67825-6a63-4352-ab3d-ce028e934e81-2
00:04:00.630 --> 00:04:03.560
understand how do we design
buildings not just for life

87b67825-6a63-4352-ab3d-ce028e934e81-3
00:04:03.560 --> 00:04:06.280
safety, but ultimately for
recovery and resilience.

80753dd4-775e-45b0-b3f2-133487149757-0
00:04:06.640 --> 00:04:08.793
So today I'm going to share a
little bit about those

80753dd4-775e-45b0-b3f2-133487149757-1
00:04:08.793 --> 00:04:09.200
learnings.

a273fe99-0a1c-40d0-a9ef-1cdc5abc8e83-0
00:04:09.400 --> 00:04:12.008
I'm fresh off the press, as
Grace mentioned, so they're

a273fe99-0a1c-40d0-a9ef-1cdc5abc8e83-1
00:04:12.008 --> 00:04:13.080
still fresh in my head.

bca5d8f0-f73e-4f16-bde5-ff78fa8bd1b5-0
00:04:13.800 --> 00:04:17.379
And the question I'll lead with
is how do we design buildings

bca5d8f0-f73e-4f16-bde5-ff78fa8bd1b5-1
00:04:17.379 --> 00:04:18.880
for earthquake resilience?

694d4705-c18b-45ef-bea3-eb7badc3aba5-0
00:04:19.400 --> 00:04:23.184
So let's begin by a reckoning
that there is a missing link

694d4705-c18b-45ef-bea3-eb7badc3aba5-1
00:04:23.184 --> 00:04:26.840
between building codes and
community resilience schools.

b09a1d01-35c4-4de0-a5ec-756d9ab78fc0-0
00:04:27.640 --> 00:04:30.719
Our building codes like ASCE 716
are specified and we are

b09a1d01-35c4-4de0-a5ec-756d9ab78fc0-1
00:04:30.719 --> 00:04:33.480
designing buildings in
accordance with life safety.

181e1a81-ac5a-46d0-9136-51100dd78ad9-0
00:04:33.480 --> 00:04:37.090
So Omar and others can obviously
leave these buildings alive, but

181e1a81-ac5a-46d0-9136-51100dd78ad9-1
00:04:37.090 --> 00:04:40.208
it is in stark contrast with
community resilience goals,

181e1a81-ac5a-46d0-9136-51100dd78ad9-2
00:04:40.208 --> 00:04:43.600
which are expressed in terms of
time to functionally recover.

7fb51bfb-b278-4b9a-bb0e-d32ea9d5d106-0
00:04:44.760 --> 00:04:48.314
This discrepancy has been
studied, i.e., the implications of

7fb51bfb-b278-4b9a-bb0e-d32ea9d5d106-1
00:04:48.314 --> 00:04:51.440
designing for life safety in a
resilience context.

a9cb0a39-d726-42c3-9463-5f28f911aa15-0
00:04:51.440 --> 00:04:55.365
And studies like FEMA P-58,
volume 5 found that 20 to 40% of

a9cb0a39-d726-42c3-9463-5f28f911aa15-1
00:04:55.365 --> 00:04:59.613
modern code conforming buildings
designed to the code today would

a9cb0a39-d726-42c3-9463-5f28f911aa15-2
00:04:59.613 --> 00:05:03.345
lose occupancy and 15 to 20%
would be irreparable under a

a9cb0a39-d726-42c3-9463-5f28f911aa15-3
00:05:03.345 --> 00:05:04.440
major earthquake.

660836c0-0f1f-4543-b927-b131713d87bb-0
00:05:05.080 --> 00:05:08.194
So this consequence is what
inspired many of us in the

660836c0-0f1f-4543-b927-b131713d87bb-1
00:05:08.194 --> 00:05:11.534
community to go above and see
and explore a paradigm shift

660836c0-0f1f-4543-b927-b131713d87bb-2
00:05:11.534 --> 00:05:15.271
where maybe we don't think about
life safety, but something great

660836c0-0f1f-4543-b927-b131713d87bb-3
00:05:15.271 --> 00:05:16.800
or something like recovery.

b3239feb-a7a8-4dfe-a4fa-d38315f16531-0
00:05:17.200 --> 00:05:20.057
And so there has been a
significant bottom-up push for

b3239feb-a7a8-4dfe-a4fa-d38315f16531-1
00:05:20.057 --> 00:05:21.200
recovery based design.

999063d0-929b-45a0-a098-9d560d125631-0
00:05:21.600 --> 00:05:24.385
And we cannot talk about this
push without talking about the

999063d0-929b-45a0-a098-9d560d125631-1
00:05:24.385 --> 00:05:27.307
20 years of research done by the
Pacific Earthquake Engineering

999063d0-929b-45a0-a098-9d560d125631-2
00:05:27.307 --> 00:05:30.092
Research Center on the topic of
performance based earthquake

999063d0-929b-45a0-a098-9d560d125631-3
00:05:30.092 --> 00:05:30.640
engineering.

044d3d97-4727-4800-b877-b1c2ed3321cc-0
00:05:31.120 --> 00:05:34.371
These approaches which are
probabilistic in nature

044d3d97-4727-4800-b877-b1c2ed3321cc-1
00:05:34.371 --> 00:05:38.324
characterizing, can quantify
building performance in terms of

044d3d97-4727-4800-b877-b1c2ed3321cc-2
00:05:38.324 --> 00:05:41.575
decision variables of
stakeholder interest, things

044d3d97-4727-4800-b877-b1c2ed3321cc-3
00:05:41.575 --> 00:05:45.273
that have been codified in FEMA
P-58, both the first and second

044d3d97-4727-4800-b877-b1c2ed3321cc-4
00:05:45.273 --> 00:05:47.760
edition and now more recently
ATC 138.

5e781632-0c54-4f40-8a39-9dd2e0d371b3-0
00:05:48.080 --> 00:05:52.226
You can think of FEMA P-58 as a
method that translates site

5e781632-0c54-4f40-8a39-9dd2e0d371b3-1
00:05:52.226 --> 00:05:56.788
specific hazard through building
very specific buildings that are

5e781632-0c54-4f40-8a39-9dd2e0d371b3-2
00:05:56.788 --> 00:05:59.000
subject to these ground motions.

b4380384-66b7-4b1e-aaf2-af6e6b4d8e31-0
00:05:59.000 --> 00:06:02.622
And then those responses are
translated into damage functions

b4380384-66b7-4b1e-aaf2-af6e6b4d8e31-1
00:06:02.622 --> 00:06:03.440
and then lost. [no audio]

4c98eebc-266f-467b-8a21-b0dd1ad88578-0
00:06:18.400 --> 00:06:21.634
software programs like SP3,
which are used by the industry

4c98eebc-266f-467b-8a21-b0dd1ad88578-1
00:06:21.634 --> 00:06:25.197
practitioners today and Pelican
that are used by researchers and

4c98eebc-266f-467b-8a21-b0dd1ad88578-2
00:06:25.197 --> 00:06:28.322
very impressive examples of
recovery based design are in

4c98eebc-266f-467b-8a21-b0dd1ad88578-3
00:06:28.322 --> 00:06:31.666
practice right now, like the
Oregon State Treasury building,

4c98eebc-266f-467b-8a21-b0dd1ad88578-4
00:06:31.666 --> 00:06:34.901
which I'll talk about later,
which achieved very ambitious

4c98eebc-266f-467b-8a21-b0dd1ad88578-5
00:06:34.901 --> 00:06:37.916
resilience objectives through
non structural component

4c98eebc-266f-467b-8a21-b0dd1ad88578-6
00:06:37.916 --> 00:06:40.931
detailing and structural
structural upgrades like base

4c98eebc-266f-467b-8a21-b0dd1ad88578-7
00:06:40.931 --> 00:06:41.479
isolation.

461cbe4e-a70c-434d-bc59-5dbc70b5d254-0
00:06:41.920 --> 00:06:45.006
And then there are others that
are less fancy, like Casa

461cbe4e-a70c-434d-bc59-5dbc70b5d254-1
00:06:45.006 --> 00:06:48.254
Adelante, a senior housing
facility designed by David Maher

461cbe4e-a70c-434d-bc59-5dbc70b5d254-2
00:06:48.254 --> 00:06:51.340
and Justine at Berkeley that
achieved at cost resilience

461cbe4e-a70c-434d-bc59-5dbc70b5d254-3
00:06:51.340 --> 00:06:54.751
through very innovative and very
careful structural detailing,

461cbe4e-a70c-434d-bc59-5dbc70b5d254-4
00:06:54.751 --> 00:06:57.080
like these dampers in the Matt
Foundation.

a3409146-9d5e-44cc-b861-794549ce94b1-0
00:06:57.960 --> 00:07:00.600
This push that we're seeing by
the private sector.

db304816-55d9-4800-90c9-f9b48f32ec0c-0
00:07:00.600 --> 00:07:03.944
And you know, I think more
bottoms up by engineers, more

db304816-55d9-4800-90c9-f9b48f32ec0c-1
00:07:03.944 --> 00:07:07.581
grass roots is complemented by
an equal and opposite top down

db304816-55d9-4800-90c9-f9b48f32ec0c-2
00:07:07.581 --> 00:07:08.520
push from above.

fb61b862-eaa1-49b4-a34d-71683b2d8553-0
00:07:08.600 --> 00:07:12.058
Starting with this white paper
from the Earthquake Engineering

fb61b862-eaa1-49b4-a34d-71683b2d8553-1
00:07:12.058 --> 00:07:15.077
Research Institute in 2019,
specifying what functional

fb61b862-eaa1-49b4-a34d-71683b2d8553-2
00:07:15.077 --> 00:07:18.041
recovery as a conceptual
framework and policy options

fb61b862-eaa1-49b4-a34d-71683b2d8553-3
00:07:18.041 --> 00:07:18.920
might look like.

cfd20e2b-912f-4672-a50a-8674ecd68177-0
00:07:18.960 --> 00:07:23.018
And we can think of functional
recovery as this limit state

cfd20e2b-912f-4672-a50a-8674ecd68177-1
00:07:23.018 --> 00:07:26.806
above life safety where we are
not looking for collapse

cfd20e2b-912f-4672-a50a-8674ecd68177-2
00:07:26.806 --> 00:07:30.526
prevention, but rather the
minimum set of repairs that

cfd20e2b-912f-4672-a50a-8674ecd68177-3
00:07:30.526 --> 00:07:34.314
restore a building's, I guess like minimum

cfd20e2b-912f-4672-a50a-8674ecd68177-4
00:07:34.314 --> 00:07:35.600
intended functions.

5df0a130-1719-427f-940d-4325ee9d8b69-0
00:07:35.600 --> 00:07:38.578
So things like chips and
wallboard, things like very

5df0a130-1719-427f-940d-4325ee9d8b69-1
00:07:38.578 --> 00:07:42.120
basic cosmetic damages are not
considered in this limit state.

e7a6e017-6406-489d-be22-fafe241595ac-0
00:07:42.160 --> 00:07:44.425
It's basically how do we get us
back in the building and doing

e7a6e017-6406-489d-be22-fafe241595ac-1
00:07:44.425 --> 00:07:45.360
what we were doing before.

2085d58c-cabf-4517-b0f0-baedfac2d967-0
00:07:45.800 --> 00:07:49.771
Since then, 2021, the NIST FEMA
report to Congress released, and

2085d58c-cabf-4517-b0f0-baedfac2d967-1
00:07:49.771 --> 00:07:53.681
then more recently in 2022, the
Building Seismic Safety Council

2085d58c-cabf-4517-b0f0-baedfac2d967-2
00:07:53.681 --> 00:07:57.224
forming the 2026 PUC committee
to explore what functional

2085d58c-cabf-4517-b0f0-baedfac2d967-3
00:07:57.224 --> 00:08:01.134
recovery might look like in the
next generation of the building

2085d58c-cabf-4517-b0f0-baedfac2d967-4
00:08:01.134 --> 00:08:01.440
code.

92bf7c6b-e41d-4100-a918-e2cb3f1044bf-0
00:08:01.920 --> 00:08:04.908
Despite all this progress that
I'm very optimistic about the

92bf7c6b-e41d-4100-a918-e2cb3f1044bf-1
00:08:04.908 --> 00:08:07.946
strategies for recovery
based design like how do we get

92bf7c6b-e41d-4100-a918-e2cb3f1044bf-2
00:08:07.946 --> 00:08:10.200
to resilience are still not well
established.

7ea08da2-1105-4116-94f5-5860f089681e-0
00:08:10.560 --> 00:08:13.179
And the way we can think about
this is first looking at life

7ea08da2-1105-4116-94f5-5860f089681e-1
00:08:13.179 --> 00:08:13.480
safety.

3c655e87-6e32-4f40-80ef-320296b056e1-0
00:08:13.800 --> 00:08:17.140
In the world of life safety, we
have a good handle on effective

3c655e87-6e32-4f40-80ef-320296b056e1-1
00:08:17.140 --> 00:08:18.080
design strategies.

cd9b113c-b5af-40cc-bc68-c0b6ed01dd66-0
00:08:18.360 --> 00:08:20.120
These strategies are generally
additive.

5d06ed03-3ab6-41ba-afbd-d5e51f08aa5b-0
00:08:20.120 --> 00:08:23.736
We know that we can confidently
dampen here, stiffen the

5d06ed03-3ab6-41ba-afbd-d5e51f08aa5b-1
00:08:23.736 --> 00:08:27.288
structure there and generally
speaking we are achieving

5d06ed03-3ab6-41ba-afbd-d5e51f08aa5b-2
00:08:27.288 --> 00:08:31.158
additive benefits in terms of
collapse safety or life safety

5d06ed03-3ab6-41ba-afbd-d5e51f08aa5b-3
00:08:31.158 --> 00:08:31.920
performance.

0e527a2d-d587-4b50-9d7f-d77193d72f7e-0
00:08:32.320 --> 00:08:36.352
And then finally, if we look at
performance objectives, they're

0e527a2d-d587-4b50-9d7f-d77193d72f7e-1
00:08:36.352 --> 00:08:36.920
in place.

aa05cb6f-93ab-4f22-ac73-75e3a53acd84-0
00:08:36.920 --> 00:08:38.360
We know what we're, we know what
we're looking for.

a3c4c519-4323-4d72-ac08-b4a5ad5c34bc-0
00:08:38.360 --> 00:08:41.480
We know what we're optimizing
for and the work and and the

a3c4c519-4323-4d72-ac08-b4a5ad5c34bc-1
00:08:41.480 --> 00:08:44.760
strategy or the design space is
generally very, very limited.

6fa8c4b4-53d4-4e6f-b231-b24c284d4570-0
00:08:45.120 --> 00:08:46.908
When you look at functional
recovery, things are very

6fa8c4b4-53d4-4e6f-b231-b24c284d4570-1
00:08:46.908 --> 00:08:47.240
different.

d8f5e1b3-4fb1-4d8b-b885-2b3cbeaeb265-0
00:08:47.760 --> 00:08:50.800
There is a lack of intuition on
effective design strategies.

9493c434-1af6-482f-b551-0cdb603306a1-0
00:08:51.120 --> 00:08:53.600
These strategies can also lead
to counteractive effects.

5ebcc1bd-610b-4a9b-b112-6ed54788630b-0
00:08:53.920 --> 00:08:56.770
When we stiffen a structure, a
lot of the times what we're

5ebcc1bd-610b-4a9b-b112-6ed54788630b-1
00:08:56.770 --> 00:08:59.814
doing is we are increasing or
amplifying accelerations in part

5ebcc1bd-610b-4a9b-b112-6ed54788630b-2
00:08:59.814 --> 00:09:02.520
of the buildings that may
attenuate the benefits of non

5ebcc1bd-610b-4a9b-b112-6ed54788630b-3
00:09:02.520 --> 00:09:03.680
structural improvements.

b633c1b5-cd79-48af-8a15-69bc3afc8a54-0
00:09:04.000 --> 00:09:06.323
And then finally, performance
objectives are not well

b633c1b5-cd79-48af-8a15-69bc3afc8a54-1
00:09:06.323 --> 00:09:06.840
established.

f0afd487-ea23-41a5-8057-0df5566b5a95-0
00:09:06.840 --> 00:09:07.880
It's pretty complicated.

2f74df52-62ce-4868-ac43-309982e3545a-0
00:09:07.880 --> 00:09:11.003
We don't necessarily know as a
society what we are looking for

2f74df52-62ce-4868-ac43-309982e3545a-1
00:09:11.003 --> 00:09:14.077
in the context of functional
recovery and the design space is

2f74df52-62ce-4868-ac43-309982e3545a-2
00:09:14.077 --> 00:09:14.920
very, very large.

c7cb61c9-1a61-4bdb-9cf3-0e4f7546b583-0
00:09:15.200 --> 00:09:18.623
Not only do we have structural
improvements, we can improve non

c7cb61c9-1a61-4bdb-9cf3-0e4f7546b583-1
00:09:18.623 --> 00:09:19.800
structural components.

71e0b801-93f1-4050-a47d-0140af774f52-0
00:09:19.800 --> 00:09:22.536
We can look at utility backup
and redundancy actions and

71e0b801-93f1-4050-a47d-0140af774f52-1
00:09:22.536 --> 00:09:23.400
recovery planning.

dc2cb61b-1129-407c-be42-323d63d8eb98-0
00:09:23.640 --> 00:09:26.966
And so this becomes a very high
dimensional space and it's

dc2cb61b-1129-407c-be42-323d63d8eb98-1
00:09:26.966 --> 00:09:30.405
highly nonlinear, meaning that
these individual improvements

dc2cb61b-1129-407c-be42-323d63d8eb98-2
00:09:30.405 --> 00:09:34.014
are not necessarily going to add
on one another in an intuitive

dc2cb61b-1129-407c-be42-323d63d8eb98-3
00:09:34.014 --> 00:09:34.239
way.

81f39dbe-9f24-4e58-abdd-2b1be7d4947f-0
00:09:34.520 --> 00:09:37.459
And so it can become very
difficult to scan and search for

81f39dbe-9f24-4e58-abdd-2b1be7d4947f-1
00:09:37.459 --> 00:09:40.000
resource efficient design
strategies for recovery.

28afe1e8-fcfd-4f87-be97-499d2e6de595-0
00:09:40.560 --> 00:09:43.496
And So what I propose in my PhD
is a multi faceted approach

28afe1e8-fcfd-4f87-be97-499d2e6de595-1
00:09:43.496 --> 00:09:45.160
focusing on two important
things.

3778535b-3333-49fd-b03a-78bb776320cb-0
00:09:45.160 --> 00:09:48.114
The first is how do we
rigorously select these

3778535b-3333-49fd-b03a-78bb776320cb-1
00:09:48.114 --> 00:09:50.440
performance objectives for
recovery?

ed5c4d0f-f4ba-44f1-89d8-c40ff5c93d43-0
00:09:50.880 --> 00:09:54.679
And then once we know what we
want, how do we rapidly extract

ed5c4d0f-f4ba-44f1-89d8-c40ff5c93d43-1
00:09:54.679 --> 00:09:57.560
efficient designs to achieve
those objectives?

83b21f56-b6e3-4cb5-aa4f-d4b747e310d8-0
00:09:58.040 --> 00:10:01.095
And if we can do those two
things and we can repeat them

83b21f56-b6e3-4cb5-aa4f-d4b747e310d8-1
00:10:01.095 --> 00:10:03.936
hundreds, thousands of times
across different sites,

83b21f56-b6e3-4cb5-aa4f-d4b747e310d8-2
00:10:03.936 --> 00:10:06.670
different building
configurations, only then in my

83b21f56-b6e3-4cb5-aa4f-d4b747e310d8-3
00:10:06.670 --> 00:10:09.887
belief, I guess here will we be
able to accelerate recovery

83b21f56-b6e3-4cb5-aa4f-d4b747e310d8-4
00:10:09.887 --> 00:10:11.119
based design intuition.

b16c36c7-d46d-4971-b05b-2a704e8df489-0
00:10:11.120 --> 00:10:12.000
We need more examples.

1deab694-0c35-4911-a0a5-d3d1becc1911-0
00:10:12.000 --> 00:10:13.600
We need to be able to search the
design space.

4c4fe84b-71e1-4a69-9f1d-6cf14c2d648c-0
00:10:14.040 --> 00:10:16.360
So today I'm going to talk about
a few research objectives.

32799a27-214c-4bc3-b821-3da14c4add56-0
00:10:16.360 --> 00:10:19.485
I won't go through everything
here, but I think there are a

32799a27-214c-4bc3-b821-3da14c4add56-1
00:10:19.485 --> 00:10:21.360
couple that are important to
share.

ef2f9ef0-7517-4c59-ba30-5878526d0c99-0
00:10:21.600 --> 00:10:24.646
The first objective is the
recovery based performance,

ef2f9ef0-7517-4c59-ba30-5878526d0c99-1
00:10:24.646 --> 00:10:26.640
objective selection and
evaluation.

a4b961ba-db1e-4121-8c4b-c27794ca225d-0
00:10:26.640 --> 00:10:29.180
I'm assuming many of you in the
audience and online have will

a4b961ba-db1e-4121-8c4b-c27794ca225d-1
00:10:29.180 --> 00:10:31.640
find some of this familiar in
the world of collapse safety.

6eeba557-f730-4941-a667-e35b95bbb93f-0
00:10:31.640 --> 00:10:33.960
And I think it's it's a relevant
topic for today.

5fac5148-4c26-4bf5-bf51-df28cb7278da-0
00:10:33.960 --> 00:10:37.431
And then the second is that
other piece I was mentioning of

5fac5148-4c26-4bf5-bf51-df28cb7278da-1
00:10:37.431 --> 00:10:39.920
how do we isolate those design
strategies.

aa576ba7-0bd5-4af1-a04c-286aa05ba3cf-0
00:10:40.080 --> 00:10:43.577
So I'm going to talk about those
two in my PHDI did a little bit

aa576ba7-0bd5-4af1-a04c-286aa05ba3cf-1
00:10:43.577 --> 00:10:47.074
more work on the topic of how do
we optimize design for lifetime

aa576ba7-0bd5-4af1-a04c-286aa05ba3cf-2
00:10:47.074 --> 00:10:50.303
performance considering all the
seismic intensities and all

aa576ba7-0bd5-4af1-a04c-286aa05ba3cf-3
00:10:50.303 --> 00:10:52.240
their relative rates of
occurrence.

6021df6c-2e30-403f-b36e-023143cd8bd2-0
00:10:52.240 --> 00:10:55.232
I won't talk about that today
since just to kind of leave time

6021df6c-2e30-403f-b36e-023143cd8bd2-1
00:10:55.232 --> 00:10:57.845
for discussion and I'll kick
things off with the first

6021df6c-2e30-403f-b36e-023143cd8bd2-2
00:10:57.845 --> 00:10:58.320
objective.

eb049556-6e83-40d9-ac86-d9b1097956e4-0
00:10:59.360 --> 00:11:05.032
This first study I did alongside
Doctor Dustin Cook with NIST and

eb049556-6e83-40d9-ac86-d9b1097956e4-1
00:11:05.032 --> 00:11:05.720
then Dr.

4b82254b-5f2b-4f5d-b200-b7321dca4139-0
00:11:05.720 --> 00:11:07.280
Nicholas Luto from the USGS.

85643da2-2e90-4677-8376-5db83848fb10-0
00:11:08.120 --> 00:11:12.191
So the motivation here and just
context setting is that recovery

85643da2-2e90-4677-8376-5db83848fb10-1
00:11:12.191 --> 00:11:16.200
based performance objectives are
an area of active exploration.

ff444d0a-d200-47e3-ac4a-2bae4c7264c5-0
00:11:16.200 --> 00:11:17.360
We're still figuring things out.

37706edf-3982-4270-a4cc-538949904046-0
00:11:17.720 --> 00:11:20.872
And as I mentioned, the VSSC
formed the provision update

37706edf-3982-4270-a4cc-538949904046-1
00:11:20.872 --> 00:11:24.357
committee to explore this and
they formed what they're calling

37706edf-3982-4270-a4cc-538949904046-2
00:11:24.357 --> 00:11:27.786
the FRTC, lots of acronyms here,
the functional recovery Task

37706edf-3982-4270-a4cc-538949904046-3
00:11:27.786 --> 00:11:31.160
committee that is split into, of
course, more subcommittees.

59a1aa2b-5119-4595-86bc-88abd94c4812-0
00:11:31.160 --> 00:11:32.600
So there's five subcommittees
here.

7611f8a4-e5fc-4e80-88d4-c3e4ed8a5167-0
00:11:32.880 --> 00:11:35.866
The first is looking at key
terms, performance metrics,

7611f8a4-e5fc-4e80-88d4-c3e4ed8a5167-1
00:11:35.866 --> 00:11:38.960
target times, prescriptive
provisions, and hazard levels.

b5c8689d-2ecf-4185-890b-4eed0ed4b0e4-0
00:11:39.520 --> 00:11:42.313
And in my conversations with
folks here, which is a very

b5c8689d-2ecf-4185-890b-4eed0ed4b0e4-1
00:11:42.313 --> 00:11:45.351
talented group of of individuals
who are diving into this and

b5c8689d-2ecf-4185-890b-4eed0ed4b0e4-2
00:11:45.351 --> 00:11:48.144
trying to see how we can fit
functional recovery in many

b5c8689d-2ecf-4185-890b-4eed0ed4b0e4-3
00:11:48.144 --> 00:11:51.133
times this idea of what is a
good performance objective came

b5c8689d-2ecf-4185-890b-4eed0ed4b0e4-4
00:11:51.133 --> 00:11:51.280
up.

56fd6774-5c66-41c0-83ea-1ff6132a8b58-0
00:11:51.760 --> 00:11:55.243
And so in spirit of trying to
provide a more rigorous basis to

56fd6774-5c66-41c0-83ea-1ff6132a8b58-1
00:11:55.243 --> 00:11:58.727
select and evaluate performance
objectives outside even the US

56fd6774-5c66-41c0-83ea-1ff6132a8b58-2
00:11:58.727 --> 00:12:01.768
context or outside of the
provisions context, I define

56fd6774-5c66-41c0-83ea-1ff6132a8b58-3
00:12:01.768 --> 00:12:05.252
performance objectives in this
study in terms of two important

56fd6774-5c66-41c0-83ea-1ff6132a8b58-4
00:12:05.252 --> 00:12:07.520
pieces, a goal and a checking
procedure.

788f26bc-fe2b-4f05-8e45-bcb4bfc3ddb7-0
00:12:08.040 --> 00:12:11.640
The goal is setting the lifetime
risk target.

d9eb12ad-45f0-4dc2-9b91-bd52a60e7075-0
00:12:11.920 --> 00:12:15.080
What do we want to achieve over
the lifetime of the building?

d608ea29-8915-4cb2-b0ff-8675b3f1b33f-0
00:12:15.560 --> 00:12:18.698
And we are doing so by
expressing this lifetime

d608ea29-8915-4cb2-b0ff-8675b3f1b33f-1
00:12:18.698 --> 00:12:23.014
performance metric, For example,
the probability of failure in 50

d608ea29-8915-4cb2-b0ff-8675b3f1b33f-2
00:12:23.014 --> 00:12:26.807
years to be less than some
threshold Z, where failure can

d608ea29-8915-4cb2-b0ff-8675b3f1b33f-3
00:12:26.807 --> 00:12:30.992
be a limit state, where maybe we
define a target time exceeding

d608ea29-8915-4cb2-b0ff-8675b3f1b33f-4
00:12:30.992 --> 00:12:34.720
some threshold, for example,
exceeding 30 days, 60 days.

41b48bd3-a92c-4c9a-9a53-8dc3a0a40c03-0
00:12:34.720 --> 00:12:37.480
That's something that society or
different stakeholders will set.

6cad9b1a-b421-4d14-84ad-efd8d3dfe471-0
00:12:37.680 --> 00:12:39.800
But let's keep it simple and
define it as failure.

a75a5045-8f37-403d-89a3-d15b92a3b0c6-0
00:12:40.440 --> 00:12:44.100
If we do so, then what we'll end
up with here is probably a site

a75a5045-8f37-403d-89a3-d15b92a3b0c6-1
00:12:44.100 --> 00:12:47.310
specific hazard curve for a
building that quantifies the

a75a5045-8f37-403d-89a3-d15b92a3b0c6-2
00:12:47.310 --> 00:12:50.801
annual rate of exceeding some
intensity on the Y axis with an

a75a5045-8f37-403d-89a3-d15b92a3b0c6-3
00:12:50.801 --> 00:12:53.786
intensity on the X axis
superimposed with a building

a75a5045-8f37-403d-89a3-d15b92a3b0c6-4
00:12:53.786 --> 00:12:57.052
fragility that represents the
probability of that failure

a75a5045-8f37-403d-89a3-d15b92a3b0c6-5
00:12:57.052 --> 00:12:59.080
state conditioned on the
intensity.

e8e4c163-5b3c-4345-9d5d-51d34363e473-0
00:12:59.480 --> 00:13:03.155
And if we integrate these, the
fragility with the slope of the

e8e4c163-5b3c-4345-9d5d-51d34363e473-1
00:13:03.155 --> 00:13:06.772
hazard curve, we can come up
with a rate of failure and using

e8e4c163-5b3c-4345-9d5d-51d34363e473-2
00:13:06.772 --> 00:13:09.806
that rate of failure and
assuming a Poisson process

e8e4c163-5b3c-4345-9d5d-51d34363e473-3
00:13:09.806 --> 00:13:12.840
calculating the probability of
failure in 50 years.

a2028955-1b9c-4f3e-bed1-a898cbc25b3b-0
00:13:13.240 --> 00:13:16.659
This is one way of expressing
our goal in design, though we

a2028955-1b9c-4f3e-bed1-a898cbc25b3b-1
00:13:16.659 --> 00:13:19.680
will never really do this for
every single building.

106c70a2-401a-4b64-bcc2-6b8a31602172-0
00:13:19.680 --> 00:13:21.160
We will never look at lifetime
risk.

d7bdb911-9236-43a7-8f62-47b463e47bcc-0
00:13:21.400 --> 00:13:25.217
So in typically what we do is we
perform a check, the check

d7bdb911-9236-43a7-8f62-47b463e47bcc-1
00:13:25.217 --> 00:13:28.080
gauges whether or not the goal
has been met.

c75f8157-e26d-4828-98fe-f3acf4a18685-0
00:13:28.080 --> 00:13:29.760
And I'll show some examples in a
bit.

46aef0b3-d5e9-4a11-a8f5-b2744e1b7c1d-0
00:13:30.240 --> 00:13:33.426
The way that we specify the
check is gauging the failure

46aef0b3-d5e9-4a11-a8f5-b2744e1b7c1d-1
00:13:33.426 --> 00:13:36.779
probability or checking the
failure probability at a single

46aef0b3-d5e9-4a11-a8f5-b2744e1b7c1d-2
00:13:36.779 --> 00:13:39.853
intensity and ensuring that that
intense, that failure

46aef0b3-d5e9-4a11-a8f5-b2744e1b7c1d-3
00:13:39.853 --> 00:13:43.040
probability is less than this
user selected threshold Y.

901b172e-e371-4bdc-a79f-52778182190d-0
00:13:43.640 --> 00:13:46.367
Taking a look at that same
fragility, what we are

901b172e-e371-4bdc-a79f-52778182190d-1
00:13:46.367 --> 00:13:49.859
effectively doing is trying to
make sure that one point on that

901b172e-e371-4bdc-a79f-52778182190d-2
00:13:49.859 --> 00:13:53.351
fragility is less than Y, IE we
are guaranteeing performance on

901b172e-e371-4bdc-a79f-52778182190d-3
00:13:53.351 --> 00:13:56.897
one point of the curve and we're
using that as a proxy to say we

901b172e-e371-4bdc-a79f-52778182190d-4
00:13:56.897 --> 00:13:57.880
have met the goal.

e6951a91-2296-4558-a6ec-15ac82bdb6f1-0
00:13:58.360 --> 00:14:00.659
The problem with this and the
challenge is that we're only

e6951a91-2296-4558-a6ec-15ac82bdb6f1-1
00:14:00.659 --> 00:14:01.400
checking one point.

de8b8335-deb9-4cd1-9283-ed226691ef53-0
00:14:01.680 --> 00:14:04.964
The fragility can look different
in different parts of the curve,

de8b8335-deb9-4cd1-9283-ed226691ef53-1
00:14:04.964 --> 00:14:07.951
and that is where this research
comes in to say what is the

de8b8335-deb9-4cd1-9283-ed226691ef53-2
00:14:07.951 --> 00:14:08.200
what?

d6983e26-8c77-4e1b-8d2d-0a5b499fa40a-0
00:14:08.200 --> 00:14:10.974
What are the implications of
that selection of why on our

d6983e26-8c77-4e1b-8d2d-0a5b499fa40a-1
00:14:10.974 --> 00:14:12.840
confidence that the goal has
been met.

55009351-fa02-448f-a7e4-5e667c478662-0
00:14:13.480 --> 00:14:17.674
So taking a dive deeper, let's
look at how this looks like in

55009351-fa02-448f-a7e4-5e667c478662-1
00:14:17.674 --> 00:14:18.960
different contexts.

eecc72b1-fa91-4e99-a75f-6da03d4623e1-0
00:14:18.960 --> 00:14:21.444
I'm going to start with what
we're familiar with, which is

eecc72b1-fa91-4e99-a75f-6da03d4623e1-1
00:14:21.444 --> 00:14:22.960
collapse, Collapse and life
safety.

d01e8b67-63e3-4c01-aeec-04ff691dd6dc-0
00:14:23.360 --> 00:14:26.120
So in the world of life safety,
the limit state is binary.

48404b0c-bd73-46b9-8e3f-c86279966923-0
00:14:26.120 --> 00:14:27.240
You collapse or you don't.

3f9c4fc4-0993-45b1-bf18-16fd4af75d08-0
00:14:27.720 --> 00:14:30.618
The goal that we use in terms,
you know in the building code

3f9c4fc4-0993-45b1-bf18-16fd4af75d08-1
00:14:30.618 --> 00:14:33.184
today is we're specifying a
failure as collapse and a

3f9c4fc4-0993-45b1-bf18-16fd4af75d08-2
00:14:33.184 --> 00:14:35.560
probability of collapse in 50
years less than 1%.

51b38035-8101-4694-a82f-c711d693bee7-0
00:14:36.040 --> 00:14:38.389
But we never really actually
check this for every single

51b38035-8101-4694-a82f-c711d693bee7-1
00:14:38.389 --> 00:14:40.080
building we can get into
technicalities.

a6989f29-bbb7-417a-8e2d-cbf7f57638b7-0
00:14:40.400 --> 00:14:43.642
I'd say the better way to frame
it is we are checking quoting

a6989f29-bbb7-417a-8e2d-cbf7f57638b7-1
00:14:43.642 --> 00:14:46.517
quote that collapse at the M at
the maximum considered

a6989f29-bbb7-417a-8e2d-cbf7f57638b7-2
00:14:46.517 --> 00:14:49.550
earthquake ground motion
intensity at a site is less than

a6989f29-bbb7-417a-8e2d-cbf7f57638b7-3
00:14:49.550 --> 00:14:49.760
10%.

4ab7c9da-3e3d-4433-8166-98d071d2a1b4-0
00:14:50.080 --> 00:14:52.110
Of course, what we're really
doing is we're not even really

4ab7c9da-3e3d-4433-8166-98d071d2a1b4-1
00:14:52.110 --> 00:14:53.160
checking this either in design.

a2247dcb-d5eb-4677-bb8a-b9de9d1cd03c-0
00:14:53.160 --> 00:14:55.801
We're actually following the
equivalent lateral force

a2247dcb-d5eb-4677-bb8a-b9de9d1cd03c-1
00:14:55.801 --> 00:14:58.835
procedure in the building code
that is a proxy for getting to

a2247dcb-d5eb-4677-bb8a-b9de9d1cd03c-2
00:14:58.835 --> 00:14:59.080
this.

c2f9d244-0350-4f0f-99bc-131e9e714d35-0
00:14:59.400 --> 00:15:02.202
But my point here is that we're
only checking one point on that

c2f9d244-0350-4f0f-99bc-131e9e714d35-1
00:15:02.202 --> 00:15:02.640
fragility.

3e2c4c46-9925-483c-91f1-ff533dfc3d1c-0
00:15:03.440 --> 00:15:06.795
Now, in the world of functional
recovery, the limit stays a bit

3e2c4c46-9925-483c-91f1-ff533dfc3d1c-1
00:15:06.795 --> 00:15:07.320
different.

eac76163-484b-4a1c-8d5c-da0517cde0ed-0
00:15:07.600 --> 00:15:11.365
In this case, we are concerned
about exceeding unacceptable

eac76163-484b-4a1c-8d5c-da0517cde0ed-1
00:15:11.365 --> 00:15:14.440
downtime, unacceptable
functional recovery time.

0edb2bb7-7c2c-4d33-a156-e291d98029ac-0
00:15:14.720 --> 00:15:16.360
How does that look like in terms
of a goal?

4a93699b-8870-49ba-8b6b-5123cd52ac9f-0
00:15:16.800 --> 00:15:17.880
Well, it's not very clear.

2fc7558c-18ad-4f29-a90d-7d2b533ebaac-0
00:15:18.200 --> 00:15:19.720
And how does that look like in
terms of a check?

9da69030-0b42-4472-897c-e97c60177d1f-0
00:15:20.000 --> 00:15:20.960
Well, that's also not clear.

8e27a4ca-d70f-49b1-857a-2303a11a7a9d-0
00:15:21.080 --> 00:15:24.164
So as part of this research, we
really tried to dive in and try

8e27a4ca-d70f-49b1-857a-2303a11a7a9d-1
00:15:24.164 --> 00:15:26.862
to see if we can help provide
some clarity on these two

8e27a4ca-d70f-49b1-857a-2303a11a7a9d-2
00:15:26.862 --> 00:15:27.200
pieces.

2d5eb75d-4ffc-4678-af23-93c5cffa9153-0
00:15:27.560 --> 00:15:31.446
I'm going to start with the goal
and in shedding light on it, I

2d5eb75d-4ffc-4678-af23-93c5cffa9153-1
00:15:31.446 --> 00:15:35.029
want to start by showing you
some results where we analyze

2d5eb75d-4ffc-4678-af23-93c5cffa9153-2
00:15:35.029 --> 00:15:38.248
trends across different
performance goal metrics, IE

2d5eb75d-4ffc-4678-af23-93c5cffa9153-3
00:15:38.248 --> 00:15:40.920
different ways of expressing
lifetime risk.

9e3e439c-cbdd-4a90-8280-2973e4ab4077-0
00:15:41.840 --> 00:15:45.766
So to start this study, we begin
by looking at trends across

9e3e439c-cbdd-4a90-8280-2973e4ab4077-1
00:15:45.766 --> 00:15:47.440
performance, full metrics.

e27baa81-256b-4ee8-8bdd-97c6b736fcf2-0
00:15:47.440 --> 00:15:50.520
There's different ways of
characterizing lifetime risks.

807cf56e-9966-4394-a9a1-0deaec9906e3-0
00:15:50.960 --> 00:15:54.049
And So what we did is we
analyzed a large set of code

807cf56e-9966-4394-a9a1-0deaec9906e3-1
00:15:54.049 --> 00:15:57.425
conforming archetype buildings
that that we can easily put

807cf56e-9966-4394-a9a1-0deaec9906e3-2
00:15:57.425 --> 00:15:59.599
together using some numerical
models.

994e8725-cedb-491d-a193-5cd13bae44f0-0
00:16:00.040 --> 00:16:03.652
So what we did is we started
with a simple three story

994e8725-cedb-491d-a193-5cd13bae44f0-1
00:16:03.652 --> 00:16:07.200
building design designed in
accordance with ASCE 716.

2aff3d29-9fb5-486e-84a3-0074b3492a12-0
00:16:07.200 --> 00:16:10.424
Henry Burton and his team at
UCLA have pulled together a very

2aff3d29-9fb5-486e-84a3-0074b3492a12-1
00:16:10.424 --> 00:16:12.920
helpful Python module for this
called Auto SDA.

b0afb73e-60de-4418-8978-ba5cb59e4f94-0
00:16:13.360 --> 00:16:16.769
We designed it in Oakland with a
commercial occupancy type using

b0afb73e-60de-4418-8978-ba5cb59e4f94-1
00:16:16.769 --> 00:16:18.920
a special moment resisting frame
system.

fa31d39b-e473-4c09-b8ab-6596142b9f85-0
00:16:19.360 --> 00:16:20.600
And we designed it in two ways.

c661b1d6-1b2f-4f65-b18f-9493e1281ce2-0
00:16:20.920 --> 00:16:24.520
First using a baseline case with
2.5% drift.

76f23cca-5a85-444a-91d3-a766effabdc5-0
00:16:24.560 --> 00:16:26.940
Then we cut that drift limit in
half to structurally stiff in

76f23cca-5a85-444a-91d3-a766effabdc5-1
00:16:26.940 --> 00:16:27.440
the building.

332dd6eb-d73d-4677-908d-a126213a7e32-0
00:16:28.040 --> 00:16:30.656
And then what we did is we
randomly hardened a bunch of non

332dd6eb-d73d-4677-908d-a126213a7e32-1
00:16:30.656 --> 00:16:32.400
structural components in the
buildings.

8195fe60-3f87-409c-af64-80630255d791-0
00:16:32.400 --> 00:16:35.545
That is, we went in the model
and for say a list of 36

8195fe60-3f87-409c-af64-80630255d791-1
00:16:35.545 --> 00:16:39.205
components, maybe it's the, you
know, the HVAC units, it's the,

8195fe60-3f87-409c-af64-80630255d791-2
00:16:39.205 --> 00:16:42.694
you know, the facade elements,
it's the all of the different

8195fe60-3f87-409c-af64-80630255d791-3
00:16:42.694 --> 00:16:45.440
components that make up a, a
performance model.

3174efff-679e-432b-a423-9d141b232b0d-0
00:16:45.760 --> 00:16:48.377
In the FEMA P 58 assessment that
I mentioned to you at the

3174efff-679e-432b-a423-9d141b232b0d-1
00:16:48.377 --> 00:16:51.128
beginning, we just went in and
tried to create a large set of

3174efff-679e-432b-a423-9d141b232b0d-2
00:16:51.128 --> 00:16:53.879
designs by randomly picking and
choosing different components

3174efff-679e-432b-a423-9d141b232b0d-3
00:16:53.879 --> 00:16:55.920
and strengthening them or making
them weaker.

95fdcbd3-f31f-4b4f-81bb-939c957006b1-0
00:16:56.280 --> 00:16:58.920
And that gave us a very large
data set of building signs.

c7dd029f-e9f6-4390-ba9e-054937008c2c-0
00:16:59.240 --> 00:17:02.619
Then what we did is we ran a
time base, which is one way of

c7dd029f-e9f6-4390-ba9e-054937008c2c-1
00:17:02.619 --> 00:17:05.999
saying an assessment using the
performance based earthquake

c7dd029f-e9f6-4390-ba9e-054937008c2c-2
00:17:05.999 --> 00:17:09.435
engineering methodology that
considers lifetime risk, IE not

c7dd029f-e9f6-4390-ba9e-054937008c2c-3
00:17:09.435 --> 00:17:13.096
just risk at a single intensity,
but across all intensities that

c7dd029f-e9f6-4390-ba9e-054937008c2c-4
00:17:13.096 --> 00:17:15.800
could happen over the lifetime
of the building.

b6d2b061-afee-4e79-8cfd-5d4d9c7bd47c-0
00:17:15.920 --> 00:17:18.740
And we're always assuming in
this case a lifetime or a time

b6d2b061-afee-4e79-8cfd-5d4d9c7bd47c-1
00:17:18.740 --> 00:17:19.680
horizon of 50 years.

a085125b-11b3-49ac-b8c8-1a8cff9a05fc-0
00:17:19.960 --> 00:17:23.931
And what we did after generating
10,000 of these designs under

a085125b-11b3-49ac-b8c8-1a8cff9a05fc-1
00:17:23.931 --> 00:17:27.650
these circumstances is we
recorded the lifetime risk using

a085125b-11b3-49ac-b8c8-1a8cff9a05fc-2
00:17:27.650 --> 00:17:31.370
a few different performance flow
metrics to show how maybe

a085125b-11b3-49ac-b8c8-1a8cff9a05fc-3
00:17:31.370 --> 00:17:35.215
success in one metric may not
necessarily lead to success in

a085125b-11b3-49ac-b8c8-1a8cff9a05fc-4
00:17:35.215 --> 00:17:35.720
another.

baa635d5-bfd5-4dcc-9c9e-52bee7be41ac-0
00:17:36.560 --> 00:17:38.920
So I want to show you a couple
of interesting results.

c8b6a59b-85a8-4447-ad90-fde8b2d3cea0-0
00:17:39.160 --> 00:17:42.178
The 1st is the story of two
buildings, I like to call it at

c8b6a59b-85a8-4447-ad90-fde8b2d3cea0-1
00:17:42.178 --> 00:17:42.480
least.

0a48aa52-2ad8-4c55-bd7d-882e21c56575-0
00:17:42.520 --> 00:17:46.205
So we have two buildings, one
building, again designed using

0a48aa52-2ad8-4c55-bd7d-882e21c56575-1
00:17:46.205 --> 00:17:49.648
the code prescribed, two 2.5%
drift limit with a few non

0a48aa52-2ad8-4c55-bd7d-882e21c56575-2
00:17:49.648 --> 00:17:53.394
structural upgrades, and then
one where we just cut the drift

0a48aa52-2ad8-4c55-bd7d-882e21c56575-3
00:17:53.394 --> 00:17:54.240
limit in half.

80b38729-e8b6-4165-b2a4-33d53b34bbb6-0
00:17:55.240 --> 00:17:58.905
What's interesting about this
case, and I'm showing here a

80b38729-e8b6-4165-b2a4-33d53b34bbb6-1
00:17:58.905 --> 00:18:02.509
plot when the X axis is the
target time and the Y axis is

80b38729-e8b6-4165-b2a4-33d53b34bbb6-2
00:18:02.509 --> 00:18:05.926
the probability of exceeding
that target time over the

80b38729-e8b6-4165-b2a4-33d53b34bbb6-3
00:18:05.926 --> 00:18:07.480
lifetime of the building.

719822e6-039d-4467-a595-bbcc1a17ac1b-0
00:18:07.800 --> 00:18:12.134
And you can think of this X axis
as showing us a variety or a

719822e6-039d-4467-a595-bbcc1a17ac1b-1
00:18:12.134 --> 00:18:15.560
continuum of different
performance flow metrics.

29cccce4-afc3-4dc1-8604-1f22b7318e2d-0
00:18:16.200 --> 00:18:20.457
To give you some context here,
some stakeholders might say that

29cccce4-afc3-4dc1-8604-1f22b7318e2d-1
00:18:20.457 --> 00:18:24.049
my goal for this building is to
to make sure that the

29cccce4-afc3-4dc1-8604-1f22b7318e2d-2
00:18:24.049 --> 00:18:28.240
probability of exceeding 7 days
in 50 years is very, very low.

636da85d-9a2b-468f-9026-10647ce0f182-0
00:18:28.600 --> 00:18:31.634
And those might be folks who are
working in public safety or

636da85d-9a2b-468f-9026-10647ce0f182-1
00:18:31.634 --> 00:18:32.480
hospital systems.

3a708736-318c-416f-8993-7eafef885603-0
00:18:32.480 --> 00:18:34.520
They're very concerned about
those low times.

8cd5221e-f099-4b23-b784-a1a368eeed15-0
00:18:35.080 --> 00:18:38.713
Whereas maybe grocery stores,
maybe more, maybe grocery stores

8cd5221e-f099-4b23-b784-a1a368eeed15-1
00:18:38.713 --> 00:18:40.040
is a less good example.

f43bb1ef-9130-48c9-9e8b-3195edac4025-0
00:18:40.040 --> 00:18:41.960
I think COVID kind of showed us
that they're pretty important.

2f90bef0-f51b-462f-9c94-8681e5a84a37-0
00:18:41.960 --> 00:18:45.617
But maybe like, let's say a not
so important building, maybe not

2f90bef0-f51b-462f-9c94-8681e5a84a37-1
00:18:45.617 --> 00:18:49.330
necessarily this building, maybe
some other building that is far,

2f90bef0-f51b-462f-9c94-8681e5a84a37-2
00:18:49.330 --> 00:18:50.400
far less important.

d38caeb8-0661-4fb6-adc7-2e667665c075-0
00:18:51.400 --> 00:18:51.560
OK.

90b5c6ca-f5e9-4664-b061-246c78689ff9-0
00:18:52.040 --> 00:18:55.869
That building maybe somewhere
here is acceptable for society,

90b5c6ca-f5e9-4664-b061-246c78689ff9-1
00:18:55.869 --> 00:18:56.240
right?

9f0c041b-ef57-482f-86ae-f00b24de768f-0
00:18:56.680 --> 00:19:00.280
But those are two different ways
of expressing lifetime risk.

352d465b-1099-4aab-990e-d61c257ad20a-0
00:19:00.600 --> 00:19:03.934
And what we saw in this
experiment is that at 120 days,

352d465b-1099-4aab-990e-d61c257ad20a-1
00:19:03.934 --> 00:19:07.448
if you look at that plot, the
probability of exceeding 120

352d465b-1099-4aab-990e-d61c257ad20a-2
00:19:07.448 --> 00:19:11.021
days for both of these buildings
that have been improved in

352d465b-1099-4aab-990e-d61c257ad20a-3
00:19:11.021 --> 00:19:13.880
different ways is identical, is
limited to 20%.

7612add6-bd92-4669-8685-df78c2ecb8c4-0
00:19:14.400 --> 00:19:17.836
Now what is interesting is that
depending on the intervention,

7612add6-bd92-4669-8685-df78c2ecb8c4-1
00:19:17.836 --> 00:19:21.000
the performance either less than
or greater than changes.

e3a7ea9f-09da-4fc3-84a2-f4edbc9600d6-0
00:19:21.440 --> 00:19:25.046
So that building that has been
improved non structurally

e3a7ea9f-09da-4fc3-84a2-f4edbc9600d6-1
00:19:25.046 --> 00:19:28.652
outperforms the stiffened
building at lower target times

e3a7ea9f-09da-4fc3-84a2-f4edbc9600d6-2
00:19:28.652 --> 00:19:31.120
and the trend reverses after 120
days.

6ae46b52-4678-4371-8ef3-5d240a824ebb-0
00:19:31.120 --> 00:19:33.940
And you can think of this
intuitively as at lower target

6ae46b52-4678-4371-8ef3-5d240a824ebb-1
00:19:33.940 --> 00:19:36.612
times things that impede
function, things that impede

6ae46b52-4678-4371-8ef3-5d240a824ebb-2
00:19:36.612 --> 00:19:37.800
reoccupancy matter more.

b8860bee-c7cf-46e3-baf9-8c9a0fd57f2b-0
00:19:38.080 --> 00:19:41.920
For example, can can inhabitants
come back into the building?

d206cb8a-b2dc-4a29-b631-be6a194a09a1-0
00:19:42.160 --> 00:19:45.120
Is the envelope, is the envelope
intact?

410a134d-c82b-4d26-a3a7-d08538a1f864-0
00:19:45.400 --> 00:19:47.884
Whereas at longer times we're
looking at collapse or

410a134d-c82b-4d26-a3a7-d08538a1f864-1
00:19:47.884 --> 00:19:50.275
irreparable damage that are
really governed by, by

410a134d-c82b-4d26-a3a7-d08538a1f864-2
00:19:50.275 --> 00:19:51.400
structural enhancements.

356399c1-322c-4f52-b2ba-388b688dd63c-0
00:19:51.400 --> 00:19:54.120
And that's why that stiffen
building outperforms there.

4ecb2fb9-b4a9-4089-950f-1716ba99f38d-0
00:19:54.680 --> 00:19:56.360
But we didn't stop at 2
buildings.

84ee5a25-db32-4daf-9cb7-96ad87cd1525-0
00:19:56.360 --> 00:19:59.132
We repeated this for 10,000
buildings to see is this an edge

84ee5a25-db32-4daf-9cb7-96ad87cd1525-1
00:19:59.132 --> 00:19:59.360
case?

2578f3de-0e71-45ce-9aba-5288a21e8cf3-0
00:19:59.360 --> 00:20:00.520
You know, is there a trend here?

b184e7ff-25e3-4af6-8e75-8192433a80d5-0
00:20:01.040 --> 00:20:04.851
And so we went in and we, we
took the entire set of results

b184e7ff-25e3-4af6-8e75-8192433a80d5-1
00:20:04.851 --> 00:20:08.600
and we explored correlations
between two types of metrics.

737b9ff1-54fe-40bc-9b2b-335185dc023a-0
00:20:08.600 --> 00:20:12.764
So the first 2 are the same that
we saw before, probability of

737b9ff1-54fe-40bc-9b2b-335185dc023a-1
00:20:12.764 --> 00:20:14.880
exceeding 120 days and 50 years.

b466e24d-4d15-4ecc-b212-e9b0f6c7e955-0
00:20:15.240 --> 00:20:19.227
And then on the Y axis, another
metric denoted as the expected

b466e24d-4d15-4ecc-b212-e9b0f6c7e955-1
00:20:19.227 --> 00:20:20.240
annual downtime.

c32068ea-cb72-4959-b713-e771f5ea6a32-0
00:20:20.240 --> 00:20:23.320
This is an annualized metric
similar to expected annual loss.

da1ce455-486e-443e-b755-ef8c70f1c82e-0
00:20:23.720 --> 00:20:27.138
That's meant to be a holistic
metric that kind of takes

da1ce455-486e-443e-b755-ef8c70f1c82e-1
00:20:27.138 --> 00:20:28.360
everything together.

5daeff67-a014-4517-a84f-d780a35e2394-0
00:20:28.760 --> 00:20:31.800
And what we found is that
they're very well correlated.

562d381b-fbd8-425d-a82d-030bee357623-0
00:20:32.160 --> 00:20:35.890
And it was, it's, it's not a a
coincidence that if we look at

562d381b-fbd8-425d-a82d-030bee357623-1
00:20:35.890 --> 00:20:39.199
this 20% mark, which is
essentially that 20% we saw on

562d381b-fbd8-425d-a82d-030bee357623-2
00:20:39.199 --> 00:20:42.568
the last slide and these two
icons representing the two

562d381b-fbd8-425d-a82d-030bee357623-3
00:20:42.568 --> 00:20:46.359
archetypes that I showed you in
the last slide, they're pretty

562d381b-fbd8-425d-a82d-030bee357623-4
00:20:46.359 --> 00:20:46.719
close.

4939dad7-eee2-41af-a151-debefc2487e1-0
00:20:47.040 --> 00:20:50.927
But the trend is that in general
using one metric and using the

4939dad7-eee2-41af-a151-debefc2487e1-1
00:20:50.927 --> 00:20:53.600
other could be potentially
interchangeable.

b6228b4a-505e-447a-85af-d1df9933c767-0
00:20:54.360 --> 00:20:57.848
But if we take a look at another
example, which I kind of hinted

b6228b4a-505e-447a-85af-d1df9933c767-1
00:20:57.848 --> 00:21:01.014
at earlier, if we look at say
probability of exceeding 120

b6228b4a-505e-447a-85af-d1df9933c767-2
00:21:01.014 --> 00:21:03.911
days in 50 years and the
probably of exceeding just a

b6228b4a-505e-447a-85af-d1df9933c767-3
00:21:03.911 --> 00:21:07.399
week, we find that actually they
are not very correlated at all.

e27bd215-132c-4478-97e3-8fac36f5e745-0
00:21:07.960 --> 00:21:11.235
And what is interesting from
this, and I think the key take

e27bd215-132c-4478-97e3-8fac36f5e745-1
00:21:11.235 --> 00:21:14.675
away is that we need to be very
careful about where we set the

e27bd215-132c-4478-97e3-8fac36f5e745-2
00:21:14.675 --> 00:21:18.005
goal posts because success in
one metric may not necessarily

e27bd215-132c-4478-97e3-8fac36f5e745-3
00:21:18.005 --> 00:21:19.480
lead to success in another.

14246e19-e839-47aa-97ed-de37cb477b4c-0
00:21:19.720 --> 00:21:22.444
And if we are optimizing for
one, we're going to lead to

14246e19-e839-47aa-97ed-de37cb477b4c-1
00:21:22.444 --> 00:21:25.217
interventions that are totally
ensuring we guarantee that

14246e19-e839-47aa-97ed-de37cb477b4c-2
00:21:25.217 --> 00:21:25.599
outcome.

d414f327-8477-4dfa-af84-84a577831b2e-0
00:21:25.600 --> 00:21:29.132
And we want to make sure that we
are not forgetting blind spots

d414f327-8477-4dfa-af84-84a577831b2e-1
00:21:29.132 --> 00:21:29.960
that may occur.

755cff1a-cc39-4ac4-962d-c5922a8d80dc-0
00:21:30.520 --> 00:21:33.744
So overall, the study gave us a
few important insights, but

755cff1a-cc39-4ac4-962d-c5922a8d80dc-1
00:21:33.744 --> 00:21:35.840
those are all geared towards the
goal.

b56df4ef-0717-44d5-add3-e13a739f5b03-0
00:21:35.960 --> 00:21:37.840
Now I want to shift focus
towards the check.

6aaa7ef4-6cb8-4e37-9ed3-9f69dc04bb3f-0
00:21:37.840 --> 00:21:41.054
And if you guys remember, what
we're thinking about are how we

6aaa7ef4-6cb8-4e37-9ed3-9f69dc04bb3f-1
00:21:41.054 --> 00:21:44.268
design checking procedures that
lead to the highest confidence

6aaa7ef4-6cb8-4e37-9ed3-9f69dc04bb3f-2
00:21:44.268 --> 00:21:47.075
that the goal is being met,
because when we're doing a

6aaa7ef4-6cb8-4e37-9ed3-9f69dc04bb3f-3
00:21:47.075 --> 00:21:49.320
check, we're only looking at one
intensity.

0275e2e1-4d41-4567-9991-98d5d4a41afa-0
00:21:49.720 --> 00:21:53.530
And so the question is, in light
of this issue that we're only

0275e2e1-4d41-4567-9991-98d5d4a41afa-1
00:21:53.530 --> 00:21:57.160
checking one point, how do we
design or calibrated a check?

22c9738d-5c3e-472f-8f7c-7414e244c219-0
00:21:57.160 --> 00:22:00.759
Specifically, how do we choose
this point on the curve to to

22c9738d-5c3e-472f-8f7c-7414e244c219-1
00:22:00.759 --> 00:22:03.945
check that will lead to the
highest confidence that a

22c9738d-5c3e-472f-8f7c-7414e244c219-2
00:22:03.945 --> 00:22:05.480
generic goal is being met.

eec868e7-f37f-4961-8ad4-ebd82391987c-0
00:22:06.040 --> 00:22:09.013
So we did this in a slightly
different way to gauge the

eec868e7-f37f-4961-8ad4-ebd82391987c-1
00:22:09.013 --> 00:22:12.200
efficacy, I'm calling it, of
different checking procedures.

da8da5bf-2e77-4166-b234-ef49bf82bbc1-0
00:22:12.440 --> 00:22:15.360
We performed a large set of
fragility based assessments.

5d95fd44-b72b-47ab-9cdc-db19ba03c74a-0
00:22:15.360 --> 00:22:18.198
So we're no longer doing E 58,
we're no longer doing building

5d95fd44-b72b-47ab-9cdc-db19ba03c74a-1
00:22:18.198 --> 00:22:19.160
specific assessments.

3b544913-6087-47df-a378-b65e36e623fa-0
00:22:19.160 --> 00:22:22.899
We're just using fragility
curves that can accommodate a

3b544913-6087-47df-a378-b65e36e623fa-1
00:22:22.899 --> 00:22:26.704
large variety of building
taxonomies and, and I guess you

3b544913-6087-47df-a378-b65e36e623fa-2
00:22:26.704 --> 00:22:29.000
could say building
configurations.

173070fb-cb26-45aa-b234-b913c798aa86-0
00:22:29.400 --> 00:22:32.717
And so we begin this by starting
with what we would, you know,

173070fb-cb26-45aa-b234-b913c798aa86-1
00:22:32.717 --> 00:22:34.560
realistically change with a
check.

4abcde53-6598-4eb0-a3c2-d97bb07cd4f2-0
00:22:34.840 --> 00:22:37.575
And you can think of it just to
kind of provide some context

4abcde53-6598-4eb0-a3c2-d97bb07cd4f2-1
00:22:37.575 --> 00:22:37.800
here.

98ee69db-8eda-401c-8b8d-a98929323280-0
00:22:38.200 --> 00:22:41.709
There are conversations in the
structural engineering community

98ee69db-8eda-401c-8b8d-a98929323280-1
00:22:41.709 --> 00:22:45.165
as to whether or not we should
be checking performance using a

98ee69db-8eda-401c-8b8d-a98929323280-2
00:22:45.165 --> 00:22:46.920
median functional recovery time.

7f8dc804-d15e-4fbe-ac25-eabc6243ea73-0
00:22:47.280 --> 00:22:51.073
A median functional recovery
time, you can think of it as a,

7f8dc804-d15e-4fbe-ac25-eabc6243ea73-1
00:22:51.073 --> 00:22:54.805
a threshold or a point on the
fragility curve equivalent to

7f8dc804-d15e-4fbe-ac25-eabc6243ea73-2
00:22:54.805 --> 00:22:57.479
50%, like a 50% probability of
exceedance.

f978891d-da5d-4968-95c5-1ffb88aec10a-0
00:22:57.800 --> 00:23:00.599
There are other engineers who
say I want to be checking that

f978891d-da5d-4968-95c5-1ffb88aec10a-1
00:23:00.599 --> 00:23:03.536
we have exceeded a target time
using the 90th percentile, which

f978891d-da5d-4968-95c5-1ffb88aec10a-2
00:23:03.536 --> 00:23:06.105
would get you down to the 10th
percentile point on that

f978891d-da5d-4968-95c5-1ffb88aec10a-3
00:23:06.105 --> 00:23:06.840
fertility curve.

05cba82f-e656-4919-be44-37565e3b8a2e-0
00:23:07.080 --> 00:23:08.560
That's where all of this is
coming from.

27d4d502-d7c5-407b-b41d-988962472790-0
00:23:08.920 --> 00:23:13.579
And So what we did is we started
with just the checking procedure

27d4d502-d7c5-407b-b41d-988962472790-1
00:23:13.579 --> 00:23:17.745
and varied this this threshold Y
that we saw on the on the

27d4d502-d7c5-407b-b41d-988962472790-2
00:23:17.745 --> 00:23:19.440
fertility curve earlier.

b543f19a-ef45-4952-8f4e-7cbc02a89724-0
00:23:20.000 --> 00:23:24.652
Then we said, OK, let's specify
different values of Z that is

b543f19a-ef45-4952-8f4e-7cbc02a89724-1
00:23:24.652 --> 00:23:28.480
the acceptable probability of
failure in 50 years.

7b7599af-09ef-4eff-8e8b-c5af04fbcd93-0
00:23:28.680 --> 00:23:30.440
We can think of this also as a
risk target.

82f17222-d029-4dcc-9c34-b77bd6930e00-0
00:23:30.480 --> 00:23:31.360
Let's change that.

36b181eb-205d-4431-acd1-3dbf0b1d11d0-0
00:23:32.160 --> 00:23:35.239
And then let's also change the
site at which we are doing the

36b181eb-205d-4431-acd1-3dbf0b1d11d0-1
00:23:35.239 --> 00:23:36.680
risk targeting or the design.

37808e3f-a3d9-483e-a3d8-287b3a0eb2c6-0
00:23:37.520 --> 00:23:42.360
Then what we do is using LUCO
2007 simple risk targeting, we

37808e3f-a3d9-483e-a3d8-287b3a0eb2c6-1
00:23:42.360 --> 00:23:47.439
go in and we figure out what is
the, what is the intensity that

37808e3f-a3d9-483e-a3d8-287b3a0eb2c6-2
00:23:47.439 --> 00:23:52.360
would yield the performance goal
and based on the selected Y.

86320a18-ccc6-4555-bed5-d52a2b8cabd2-0
00:23:52.360 --> 00:23:55.760
So this is giving you a sense of
what that risk targeting does.

727218e2-a8b4-470e-88c4-a2a9a6f40795-0
00:23:55.760 --> 00:23:56.840
It gives you this intensity.

8cc6f43d-c265-4683-a441-4c6b79ab92cb-0
00:23:57.160 --> 00:24:00.352
And then we're using a beta for
risk targeting of 0.6, I

8cc6f43d-c265-4683-a441-4c6b79ab92cb-1
00:24:00.352 --> 00:24:00.800
believe.

f94db733-c73b-4ffb-963f-6a7a5b04f319-0
00:24:01.120 --> 00:24:02.960
And then what we're doing is
just varying the shape.

0404ac16-ab37-42f4-bb8e-5aac9022b918-0
00:24:03.280 --> 00:24:06.159
So we're saying, OK, let's pinch
this part of the curve, and then

0404ac16-ab37-42f4-bb8e-5aac9022b918-1
00:24:06.159 --> 00:24:08.819
let's change the curve in other
locations to accommodate the

0404ac16-ab37-42f4-bb8e-5aac9022b918-2
00:24:08.819 --> 00:24:11.611
fact that if you're checking in
one place, ultimately there may

0404ac16-ab37-42f4-bb8e-5aac9022b918-3
00:24:11.611 --> 00:24:13.879
be differences in other parts of
the agility curve.

13a638c2-8118-4eb5-adc2-772b74f272af-0
00:24:14.440 --> 00:24:18.034
Then we go in and we compute the
coefficient of variation on the

13a638c2-8118-4eb5-adc2-772b74f272af-1
00:24:18.034 --> 00:24:21.462
probability of failure in 50
years considering the difference

13a638c2-8118-4eb5-adc2-772b74f272af-2
00:24:21.462 --> 00:24:21.960
in shape.

b21ce2eb-d84b-49aa-a178-97ebb2c13d61-0
00:24:22.920 --> 00:24:26.582
And so if we do this across many
different sites, many different

b21ce2eb-d84b-49aa-a178-97ebb2c13d61-1
00:24:26.582 --> 00:24:29.963
values of YIE, different checks
and different goals, we can

b21ce2eb-d84b-49aa-a178-97ebb2c13d61-2
00:24:29.963 --> 00:24:32.780
start to develop some
relationships about what it

b21ce2eb-d84b-49aa-a178-97ebb2c13d61-3
00:24:32.780 --> 00:24:35.880
means to have a good check in
the context of recovery.

3b63b376-e946-4864-8114-67df979206b7-0
00:24:36.360 --> 00:24:38.980
So I'll show you an example here
that is relevant for the

3b63b376-e946-4864-8114-67df979206b7-1
00:24:38.980 --> 00:24:41.600
conversations I was mentioning
with structural engineers.

11efc530-a9d4-44f2-998b-0a881371cd1b-0
00:24:42.080 --> 00:24:43.120
Let's start with the goal.

a620259d-0fe9-4c36-a29f-ced897dc066d-0
00:24:43.560 --> 00:24:46.953
The goal, which is one that has
been pitched in the BSSC, for

a620259d-0fe9-4c36-a29f-ced897dc066d-1
00:24:46.953 --> 00:24:50.292
example, at one point is to
limit the probability of failure

a620259d-0fe9-4c36-a29f-ced897dc066d-2
00:24:50.292 --> 00:24:51.880
in 50 years to less than 20%.

b264e26f-83cf-4b79-8ee0-f3b1f73bbb05-0
00:24:52.400 --> 00:24:55.280
OK, now let's say what is it?

6d40a5e3-598b-4ea6-b5df-14adaea75f6a-0
00:24:55.320 --> 00:24:58.432
Let's ask ourselves what could
be a good check to fulfill this

6d40a5e3-598b-4ea6-b5df-14adaea75f6a-1
00:24:58.432 --> 00:24:58.680
goal?

273e1cd8-e4d6-4760-8112-f720abcad386-0
00:24:59.040 --> 00:25:02.889
Well, some of you might say,
let's use the median and some of

273e1cd8-e4d6-4760-8112-f720abcad386-1
00:25:02.889 --> 00:25:06.739
you might say, let's use the
90th percentile or limit the the

273e1cd8-e4d6-4760-8112-f720abcad386-2
00:25:06.739 --> 00:25:09.719
probability, the failure
probability to to 10%.

3a7ddede-df7b-44c4-b7dd-61f2948f604c-0
00:25:10.280 --> 00:25:13.140
So you can see that depending on
that decision, fragility curves

3a7ddede-df7b-44c4-b7dd-61f2948f604c-1
00:25:13.140 --> 00:25:13.800
look different.

f037310d-8458-43f6-a06b-09255a4e14ca-0
00:25:14.240 --> 00:25:17.774
And we repeat this experiment
for a variety of sites and we

f037310d-8458-43f6-a06b-09255a4e14ca-1
00:25:17.774 --> 00:25:20.720
assess the coefficient of
variation at each site.

3bc9d157-830a-46d5-b5e3-856793fa2243-0
00:25:21.000 --> 00:25:22.400
So I'm going to show you guys
the result.

962a948e-1663-495f-ba6b-7431bbb24bd3-0
00:25:22.800 --> 00:25:25.489
But what I'll say is, is that if
you use a higher conditional

962a948e-1663-495f-ba6b-7431bbb24bd3-1
00:25:25.489 --> 00:25:28.265
probability of failure, you in
general get less confidence that

962a948e-1663-495f-ba6b-7431bbb24bd3-2
00:25:28.265 --> 00:25:29.480
the performance will is met.

b03ba437-007a-4e07-bfc5-fd458971e988-0
00:25:29.800 --> 00:25:32.590
I want you to pay attention to
this orange curve that

b03ba437-007a-4e07-bfc5-fd458971e988-1
00:25:32.590 --> 00:25:35.896
represents using the median and
this blue curve that represents

b03ba437-007a-4e07-bfc5-fd458971e988-2
00:25:35.896 --> 00:25:37.240
using the 90th percentile.

350285b7-e691-4467-b6f8-abc060afeaab-0
00:25:37.680 --> 00:25:38.880
And then I'll show you the
results.

05e78cbb-e326-4a9a-bdce-1537f40e933f-0
00:25:38.880 --> 00:25:42.559
Here the Y axis is the
coefficient of variation on the

05e78cbb-e326-4a9a-bdce-1537f40e933f-1
00:25:42.559 --> 00:25:46.640
probability of failure in 50
years, IE our performance goal.

7240e5ff-cf58-497a-87ee-d4e76cf5ede9-0
00:25:46.920 --> 00:25:48.880
And then here on the X axis, I'm
showing the site.

a0fd6c01-2b22-47cf-af16-e03857069826-0
00:25:49.120 --> 00:25:52.783
These ZS are the same 34 sites
that were used in E HERP 2015,

a0fd6c01-2b22-47cf-af16-e03857069826-1
00:25:52.783 --> 00:25:54.320
and these are the results.

cb2b9663-82af-46fb-b9e9-980d8e19e131-0
00:25:54.720 --> 00:25:59.628
So you can see that for the 34
sites, for the blue curve, we've

cb2b9663-82af-46fb-b9e9-980d8e19e131-1
00:25:59.628 --> 00:26:04.152
limited the coefficient of
variation to roughly on average

cb2b9663-82af-46fb-b9e9-980d8e19e131-2
00:26:04.152 --> 00:26:06.760
10% no matter what site you
pick.

184e5671-12a7-434f-9632-753e06a51db1-0
00:26:07.080 --> 00:26:10.617
Whereas there is significant
variation when you're using AY

184e5671-12a7-434f-9632-753e06a51db1-1
00:26:10.617 --> 00:26:14.096
of 0.5 across the different
sites, the average is actually

184e5671-12a7-434f-9632-753e06a51db1-2
00:26:14.096 --> 00:26:15.040
more than two X.

0c64ca4d-4872-41d7-ac7d-c4d958167b34-0
00:26:15.400 --> 00:26:18.950
And what you can really, I think
the meaningful take away here is

0c64ca4d-4872-41d7-ac7d-c4d958167b34-1
00:26:18.950 --> 00:26:22.501
that if we are going to pick 1.2
on the fragility curve to make a

0c64ca4d-4872-41d7-ac7d-c4d958167b34-2
00:26:22.501 --> 00:26:25.783
check for recovery, we probably
don't want to be picking the

0c64ca4d-4872-41d7-ac7d-c4d958167b34-3
00:26:25.783 --> 00:26:26.160
medium.

118dfb32-3748-436f-8607-37e303e220a6-0
00:26:26.720 --> 00:26:30.520
And we repeated this for a
variety of different cases.

bf60b641-3672-4301-bc6b-140fb4bb5a9c-0
00:26:30.720 --> 00:26:33.617
And then we actually took the
average across all of the sites

bf60b641-3672-4301-bc6b-140fb4bb5a9c-1
00:26:33.617 --> 00:26:34.880
and these were the results.

2f90aa8f-ce0d-41fc-9d3d-f4794b890d44-0
00:26:35.480 --> 00:26:38.206
The results on this Y axis are
showing the point of the

2f90aa8f-ce0d-41fc-9d3d-f4794b890d44-1
00:26:38.206 --> 00:26:41.128
fragility we're checking, so the
conditional probability of

2f90aa8f-ce0d-41fc-9d3d-f4794b890d44-2
00:26:41.128 --> 00:26:42.199
failure for the check.

838ef2a6-f861-4014-9c37-b9935e1a285c-0
00:26:42.640 --> 00:26:45.711
And then the Y axis is showing
mean coefficient of variation

838ef2a6-f861-4014-9c37-b9935e1a285c-1
00:26:45.711 --> 00:26:46.920
across all of the sites.

f0b7b277-19f2-42c6-a207-1b6a74f64be0-0
00:26:47.360 --> 00:26:51.111
And what we find is that there's
kind of this magical location

f0b7b277-19f2-42c6-a207-1b6a74f64be0-1
00:26:51.111 --> 00:26:54.684
where basically if we're using
AY of 0.1, it doesn't really

f0b7b277-19f2-42c6-a207-1b6a74f64be0-2
00:26:54.684 --> 00:26:57.840
matter what the performance goal
is, what that C is.

81ca4b91-bb19-4926-a935-f28c627f53bb-0
00:26:58.000 --> 00:27:01.329
We are consistently getting a,
an average coefficient of

81ca4b91-bb19-4926-a935-f28c627f53bb-1
00:27:01.329 --> 00:27:05.009
variation that is lower than, I
would say, many points on this

81ca4b91-bb19-4926-a935-f28c627f53bb-2
00:27:05.009 --> 00:27:05.360
curve.

306c8e04-09a0-435f-ae72-594ce594bdde-0
00:27:05.760 --> 00:27:09.568
And interestingly, if we look at
the second most popular place

306c8e04-09a0-435f-ae72-594ce594bdde-1
00:27:09.568 --> 00:27:13.073
that engineers are discussing,
which is the a conditional

306c8e04-09a0-435f-ae72-594ce594bdde-2
00:27:13.073 --> 00:27:17.063
probability of failure of 50% or
using the median to check, there

306c8e04-09a0-435f-ae72-594ce594bdde-3
00:27:17.063 --> 00:27:20.810
is a significant dependency on
the efficacy of the check as a

306c8e04-09a0-435f-ae72-594ce594bdde-4
00:27:20.810 --> 00:27:24.014
function of the risk target,
where the more the less

306c8e04-09a0-435f-ae72-594ce594bdde-5
00:27:24.014 --> 00:27:27.520
ambitious you go, the lower the
coefficient of variation.

bd132eda-6b15-4780-9f83-83100e26f506-0
00:27:27.520 --> 00:27:31.034
But obviously many values of Z
that are being discussed are,

bd132eda-6b15-4780-9f83-83100e26f506-1
00:27:31.034 --> 00:27:32.360
you know, .2, maybe .1.

ece2f30e-b8f9-44af-a650-cd49ebafcf63-0
00:27:32.680 --> 00:27:35.536
And so you can see that
coefficient of variation, you

ece2f30e-b8f9-44af-a650-cd49ebafcf63-1
00:27:35.536 --> 00:27:37.600
know, increased by a significant
deal.

93af898c-4028-43ae-a3ae-62aa8849f4d0-0
00:27:37.960 --> 00:27:41.082
So actually right now at the
BSSC for context, I think that

93af898c-4028-43ae-a3ae-62aa8849f4d0-1
00:27:41.082 --> 00:27:44.100
they're, they're currently
looking for a place between 10

93af898c-4028-43ae-a3ae-62aa8849f4d0-2
00:27:44.100 --> 00:27:47.431
and 25% that represent, I think
some of the more stable regions

93af898c-4028-43ae-a3ae-62aa8849f4d0-3
00:27:47.431 --> 00:27:48.159
of this curve.

c718b087-603c-47ac-85db-a12c28cc06d1-0
00:27:48.600 --> 00:27:51.418
So in terms of research
contributions, we talked about

c718b087-603c-47ac-85db-a12c28cc06d1-1
00:27:51.418 --> 00:27:51.880
the goal.

cac6ac3e-072f-49e4-a1b5-078fa6bb3fd7-0
00:27:52.120 --> 00:27:54.366
We're expressing the goal in
terms of a probability of

cac6ac3e-072f-49e4-a1b5-078fa6bb3fd7-1
00:27:54.366 --> 00:27:55.960
failure in 50 years to be less
than Z.

782c7f31-f542-4b3e-9359-a68cb5e75da0-0
00:27:56.240 --> 00:27:58.992
We talked about a check how that
is looking at one point on the

782c7f31-f542-4b3e-9359-a68cb5e75da0-1
00:27:58.992 --> 00:27:59.680
fragility curve.

406ec575-a9c3-476c-a6e9-20c4c5b6a1a8-0
00:28:00.000 --> 00:28:03.457
And in general, the performance
objectives for recovery should

406ec575-a9c3-476c-a6e9-20c4c5b6a1a8-1
00:28:03.457 --> 00:28:06.640
be defined in terms of the goal
and a checking procedure.

6a450213-f8fa-4c06-a477-7b7a9cb7c6fd-0
00:28:06.920 --> 00:28:09.685
We saw a correlation study that
showed not all performance full

6a450213-f8fa-4c06-a477-7b7a9cb7c6fd-1
00:28:09.685 --> 00:28:10.680
metrics are correlated.

b24c85aa-3d91-41ab-a51e-4b55c1e0d10c-0
00:28:10.960 --> 00:28:13.860
And then finally, using
fragility assessments that use

b24c85aa-3d91-41ab-a51e-4b55c1e0d10c-1
00:28:13.860 --> 00:28:17.181
the conditional probability of
failure between 10 and 20% that

b24c85aa-3d91-41ab-a51e-4b55c1e0d10c-2
00:28:17.181 --> 00:28:20.134
we just saw in the previous
slide, we can significantly

b24c85aa-3d91-41ab-a51e-4b55c1e0d10c-3
00:28:20.134 --> 00:28:23.614
improve the efficacy of a design
check regardless of the selected

b24c85aa-3d91-41ab-a51e-4b55c1e0d10c-4
00:28:23.614 --> 00:28:24.880
performance goal target.

6833b94e-ef02-4fbc-be45-633d586997a2-0
00:28:25.400 --> 00:28:27.320
So that was all about the goal.

c0954e31-2e91-4cf0-9539-dcce8138e3ee-0
00:28:27.880 --> 00:28:30.297
In the next part of the
presentation, I want to switch

c0954e31-2e91-4cf0-9539-dcce8138e3ee-1
00:28:30.297 --> 00:28:31.440
gears to how we get there.

b8520b18-a3b3-4a56-a451-9abe60718ba4-0
00:28:31.680 --> 00:28:33.440
You might ask me, well, Omar,
that's all great.

fbfec762-a5c0-4c13-897d-74e97724e2db-0
00:28:33.440 --> 00:28:35.920
And those are a lot of curves,
but like what do we have to do

fbfec762-a5c0-4c13-897d-74e97724e2db-1
00:28:35.920 --> 00:28:38.240
with these buildings to actually
achieve achieve recovery

fbfec762-a5c0-4c13-897d-74e97724e2db-2
00:28:38.240 --> 00:28:38.720
performance?

1b0f42c6-9d6d-48ec-870e-f7f95ccb81f0-0
00:28:39.040 --> 00:28:42.845
So this part of the study I
actually did with Doctor Rodrigo

1b0f42c6-9d6d-48ec-870e-f7f95ccb81f0-1
00:28:42.845 --> 00:28:45.840
Silva Lopez and Professor Henry
Burton at UCLA.

38be8532-a9fc-45f4-b978-f1d606db684b-0
00:28:46.920 --> 00:28:49.880
And so to begin this, I just
want us to take a field trip.

3003fda3-2ee7-4e2e-bf08-76016e9c864a-0
00:28:49.880 --> 00:28:52.950
So let's take a look at the
front cover of the New York

3003fda3-2ee7-4e2e-bf08-76016e9c864a-1
00:28:52.950 --> 00:28:53.280
Times.

fafc2f39-9d82-45ae-975c-1e81c3a706e3-0
00:28:53.280 --> 00:28:56.196
I absolutely love this this this
cover of the New York Times

fafc2f39-9d82-45ae-975c-1e81c3a706e3-1
00:28:56.196 --> 00:28:59.113
because it's very rare they talk
about, you know, earthquake

fafc2f39-9d82-45ae-975c-1e81c3a706e3-2
00:28:59.113 --> 00:28:59.640
resilience.

cba64851-88f7-49a8-8758-476fb421328f-0
00:28:59.640 --> 00:29:00.760
So I thought it was pretty cool.

9c1c4441-ef88-4b1a-8415-c8b5f971b2d2-0
00:29:01.120 --> 00:29:03.416
Here they're showing the Oregon
State Treasury building that I

9c1c4441-ef88-4b1a-8415-c8b5f971b2d2-1
00:29:03.416 --> 00:29:04.400
mentioned you guys earlier.

acef22c2-2d52-48f7-a1aa-de9bb4df226c-0
00:29:04.760 --> 00:29:05.920
It's a very beautiful building.

585106e1-3102-4396-9a7a-770733c74d89-0
00:29:06.280 --> 00:29:09.547
It's base isolated with up to 18
inches of horizontal movement in

585106e1-3102-4396-9a7a-770733c74d89-1
00:29:09.547 --> 00:29:10.240
any direction.

89ec80c9-f763-44ba-a299-b1b573671c1c-0
00:29:10.280 --> 00:29:13.639
And if we take a cut in the
middle, we can start to see some

89ec80c9-f763-44ba-a299-b1b573671c1c-1
00:29:13.639 --> 00:29:16.724
examples of innovative
structural detailing that led to

89ec80c9-f763-44ba-a299-b1b573671c1c-2
00:29:16.724 --> 00:29:17.440
this outcome.

f6c2b497-0e99-443f-af93-198e67feda66-0
00:29:17.920 --> 00:29:20.446
First, obviously is the base
isolation that's a structural

f6c2b497-0e99-443f-af93-198e67feda66-1
00:29:20.446 --> 00:29:20.960
improvement.

3cb4f08a-20cf-4b7f-89d6-855bb0a69b36-0
00:29:21.360 --> 00:29:24.675
The second is the high
performance glass and facade

3cb4f08a-20cf-4b7f-89d6-855bb0a69b36-1
00:29:24.675 --> 00:29:28.627
elements on the exterior that
are resisting, I mean excessive

3cb4f08a-20cf-4b7f-89d6-855bb0a69b36-2
00:29:28.627 --> 00:29:29.520
storage rates.

d5261f7d-83c6-4291-82dd-adb2bb7c938c-0
00:29:29.920 --> 00:29:32.440
And then the third is you notice
these are very clean ceilings.

449ae419-e125-4489-b940-83ebf123b3e6-0
00:29:32.680 --> 00:29:35.634
There's not a whole lot of
things that can break, which is

449ae419-e125-4489-b940-83ebf123b3e6-1
00:29:35.634 --> 00:29:38.738
often one of the things that
impedes return to function after

449ae419-e125-4489-b940-83ebf123b3e6-2
00:29:38.738 --> 00:29:39.840
a after an earthquake.

67e9698e-455e-4d86-ba1c-cb5d941a0ecb-0
00:29:39.840 --> 00:29:41.400
I know you guys are looking at
the ceiling right now.

710d3b8e-2366-40f1-bf13-e28cfebcb7de-0
00:29:42.000 --> 00:29:45.920
There are some things here I
wouldn't need to worry though.

4097626e-03d6-4b49-84f1-3468540ab46d-0
00:29:45.920 --> 00:29:49.202
But yeah, so this building is a
great example of of how we

4097626e-03d6-4b49-84f1-3468540ab46d-1
00:29:49.202 --> 00:29:50.760
achieve functional recovery.

51f915cb-3031-4d92-a2e9-dfeb5d5f0c6d-0
00:29:52.080 --> 00:29:54.735
Now in terms of the actual work,
the question is how do we

51f915cb-3031-4d92-a2e9-dfeb5d5f0c6d-1
00:29:54.735 --> 00:29:56.400
generalize this across any
building?

f82eeb15-9dd4-4f6a-82a5-8978762017cc-0
00:29:56.400 --> 00:29:59.476
How do we know what is the
efficient design strategy for

f82eeb15-9dd4-4f6a-82a5-8978762017cc-1
00:29:59.476 --> 00:30:00.880
recovery for any building?

a7a55999-6f8d-47d8-8ed1-51edd165adfc-0
00:30:00.960 --> 00:30:05.949
So what we did as we started
with a check for a given check

a7a55999-6f8d-47d8-8ed1-51edd165adfc-1
00:30:05.949 --> 00:30:10.690
and a given building at a given
site, how do we find the

a7a55999-6f8d-47d8-8ed1-51edd165adfc-2
00:30:10.690 --> 00:30:15.846
strategies, IE how do we find
cost optimal resource efficient

a7a55999-6f8d-47d8-8ed1-51edd165adfc-3
00:30:15.846 --> 00:30:21.251
strategies to achieve earthquake
recovery, objective development

a7a55999-6f8d-47d8-8ed1-51edd165adfc-4
00:30:21.251 --> 00:30:26.241
and provisions and engineering
intuition and really support

a7a55999-6f8d-47d8-8ed1-51edd165adfc-5
00:30:26.241 --> 00:30:30.400
those who are trying to do
benefit cost analysis.

2ee516e7-cc56-41bc-8538-46b0620f12fc-0
00:30:30.720 --> 00:30:33.354
Anyone who's doing benefit cost
analysis, their first question

2ee516e7-cc56-41bc-8538-46b0620f12fc-1
00:30:33.354 --> 00:30:34.400
is how much does it cost?

b401a04c-af6b-48c8-85e0-b3814baa4bae-0
00:30:34.720 --> 00:30:37.229
Well, we can't do that if we
don't have an understanding of

b401a04c-af6b-48c8-85e0-b3814baa4bae-1
00:30:37.229 --> 00:30:38.400
what is an effective design.

46c841b4-158d-46ee-aabf-eb1b1253ff84-0
00:30:38.680 --> 00:30:40.480
Otherwise it undermines that
analysis.

fbb72529-2f25-415d-adb0-fee66ed140c1-0
00:30:40.760 --> 00:30:44.003
So I would say the big problem
with today's approach is if you

fbb72529-2f25-415d-adb0-fee66ed140c1-1
00:30:44.003 --> 00:30:46.681
were trying to do an
optimization is that they lack

fbb72529-2f25-415d-adb0-fee66ed140c1-2
00:30:46.681 --> 00:30:49.410
the computational efficiency to
perform this type of

fbb72529-2f25-415d-adb0-fee66ed140c1-3
00:30:49.410 --> 00:30:50.080
optimization.

123bd815-9a01-4a0b-a81c-1a9f2d6084d2-0
00:30:50.520 --> 00:30:53.320
I mentioned to you that the
design space is pretty large.

6f323af9-0590-4d30-8c9b-f4faf056c446-0
00:30:53.400 --> 00:30:56.415
If we wanted to actually do this
using today's approaches, we'd

6f323af9-0590-4d30-8c9b-f4faf056c446-1
00:30:56.415 --> 00:30:59.147
go and we'd start, OK, how do we
improve the the building

6f323af9-0590-4d30-8c9b-f4faf056c446-2
00:30:59.147 --> 00:30:59.759
structurally?

d5f3f529-fa9c-4a69-b90a-2c81a5adf362-0
00:30:59.960 --> 00:31:01.840
That's going to be maybe what's
the lateral system?

af326053-1ab3-43ac-b13d-be18f9a6a8be-0
00:31:02.160 --> 00:31:03.520
What's the dampers we're going
to put in?

ae340eab-8793-4f18-a77e-6cc5a136372d-0
00:31:03.520 --> 00:31:04.440
What's the importance factor?

dc651485-b0e4-4cbe-b1aa-8a7659ced668-0
00:31:04.440 --> 00:31:05.200
What's the drift limit?

09753c0f-b6e2-40f9-be85-1251329299c9-0
00:31:05.680 --> 00:31:07.781
Then you might say, well, Omar,
I want some non structural in

09753c0f-b6e2-40f9-be85-1251329299c9-1
00:31:07.781 --> 00:31:08.120
there too.

b9d7b0ce-012a-4146-9e7e-eba9b96afbd6-0
00:31:08.120 --> 00:31:10.784
So we'll probably say, OK, maybe
we improved a few different

b9d7b0ce-012a-4146-9e7e-eba9b96afbd6-1
00:31:10.784 --> 00:31:11.920
features non structurally.

46255fda-3e0b-4ddb-ab2e-24670814100b-0
00:31:12.200 --> 00:31:14.800
Some of you might be interested
in utility backup actions,

46255fda-3e0b-4ddb-ab2e-24670814100b-1
00:31:14.800 --> 00:31:16.960
Others might be interested in
recovery planning.

4b962f55-48a8-4416-99f1-70a4d024a467-0
00:31:17.240 --> 00:31:19.953
For all intents and purposes,
let's just put it in an array

4b962f55-48a8-4416-99f1-70a4d024a467-1
00:31:19.953 --> 00:31:20.360
called X.

83cbfd0c-d122-49ec-8393-0c1fbf105647-0
00:31:20.840 --> 00:31:22.840
So here it's in one array X.

03cafb69-f2d0-4cfb-9451-23ffc1228437-0
00:31:23.200 --> 00:31:26.703
Let's take that array and
propagate it into this very

03cafb69-f2d0-4cfb-9451-23ffc1228437-1
00:31:26.703 --> 00:31:30.596
constrained needs, I guess
representation of all the things

03cafb69-f2d0-4cfb-9451-23ffc1228437-2
00:31:30.596 --> 00:31:31.440
you could do.

77cc6af7-667b-48d5-a744-9f26eacbfdad-0
00:31:31.880 --> 00:31:35.104
What I would do if I were to do
this optimization using today's

77cc6af7-667b-48d5-a744-9f26eacbfdad-1
00:31:35.104 --> 00:31:37.925
approaches is run a building
specific performance based

77cc6af7-667b-48d5-a744-9f26eacbfdad-2
00:31:37.925 --> 00:31:41.200
assessment using the procedures
I mentioned at the introduction.

b9000a78-028d-4e54-97e5-fcac0c4e955f-0
00:31:41.200 --> 00:31:44.973
These are your FEMA P58 damage
and loss assessment, your AATC

b9000a78-028d-4e54-97e5-fcac0c4e955f-1
00:31:44.973 --> 00:31:46.800
138, the downtime assessments.

1a6b5963-4b0d-4ac1-8cd6-543c92e8d62b-0
00:31:47.120 --> 00:31:50.894
The reason why I keep mentioning
ATC 138 is that FEMA P 58 alone

1a6b5963-4b0d-4ac1-8cd6-543c92e8d62b-1
00:31:50.894 --> 00:31:54.494
is only giving you component
level repair times, but is in no

1a6b5963-4b0d-4ac1-8cd6-543c92e8d62b-2
00:31:54.494 --> 00:31:57.804
way accounting for realistic
repair sequencing and those

1a6b5963-4b0d-4ac1-8cd6-543c92e8d62b-3
00:31:57.804 --> 00:32:01.347
impeding factors that would
prevent initiation of repairs in

1a6b5963-4b0d-4ac1-8cd6-543c92e8d62b-4
00:32:01.347 --> 00:32:02.160
the 1st place.

588bafca-977c-4737-9134-e64c2ca0a0db-0
00:32:02.160 --> 00:32:05.242
These are your permitting,
financing and and other things

588bafca-977c-4737-9134-e64c2ca0a0db-1
00:32:05.242 --> 00:32:08.696
that from past disasters we know
impact functional recovery time

588bafca-977c-4737-9134-e64c2ca0a0db-2
00:32:08.696 --> 00:32:09.440
significantly.

81283152-17cf-4dff-b5bd-bc78119ee1a2-0
00:32:10.000 --> 00:32:12.040
So let's say we, you know, we
have this all set up.

558f652e-fc58-41fe-8971-5a613c876789-0
00:32:12.400 --> 00:32:15.906
Now you would have to do this
may run this maybe 10,000 times

558f652e-fc58-41fe-8971-5a613c876789-1
00:32:15.906 --> 00:32:18.960
using an objective function that
you that you choose.

e46962cb-cd1c-43ed-a9fe-62112a56fae4-0
00:32:18.960 --> 00:32:21.948
Some of you will be interested
in cost, some of you will be

e46962cb-cd1c-43ed-a9fe-62112a56fae4-1
00:32:21.948 --> 00:32:24.936
interested in materials, IE how
do I achieve say a month of

e46962cb-cd1c-43ed-a9fe-62112a56fae4-2
00:32:24.936 --> 00:32:27.625
downtime or a month of
functional recovery time using

e46962cb-cd1c-43ed-a9fe-62112a56fae4-3
00:32:27.625 --> 00:32:29.120
the lowest resources possible.

64a4a05e-9fcb-47ec-929e-6a65ff36219c-0
00:32:29.600 --> 00:32:32.886
This probably would take you
maybe 2-3 days at most to do

64a4a05e-9fcb-47ec-929e-6a65ff36219c-1
00:32:32.886 --> 00:32:35.040
for, for one building it's very
slow.

d9d22213-fc57-493f-aa24-dd74f0a5b59c-0
00:32:35.400 --> 00:32:38.804
So what we do is we replace the
entire middle part with a neural

d9d22213-fc57-493f-aa24-dd74f0a5b59c-1
00:32:38.804 --> 00:32:42.000
network based surrogate model
that can achieve computational

d9d22213-fc57-493f-aa24-dd74f0a5b59c-2
00:32:42.000 --> 00:32:44.200
efficiency gains and orders of
magnitude.

dfaaaa90-ef49-4a03-a744-77288ddc39a6-0
00:32:45.120 --> 00:32:48.620
So the study which explored this
idea of using a surrogate model

dfaaaa90-ef49-4a03-a744-77288ddc39a6-1
00:32:48.620 --> 00:32:50.560
consisted of three core
components.

9ddfb646-786e-4940-82ff-ab74eeefc69c-0
00:32:50.880 --> 00:32:53.983
The first is the optimization
itself, the second is the

9ddfb646-786e-4940-82ff-ab74eeefc69c-1
00:32:53.983 --> 00:32:57.586
surrogate modeling, and then the
third is what do the outputs of

9ddfb646-786e-4940-82ff-ab74eeefc69c-2
00:32:57.586 --> 00:32:59.360
such an analysis even look like?

4b45d9ea-f1ce-417a-8e2e-bb5767e8706f-0
00:32:59.800 --> 00:33:02.781
So I'll start with the 1st and
I'll spare you all of the

4b45d9ea-f1ce-417a-8e2e-bb5767e8706f-1
00:33:02.781 --> 00:33:03.200
details.

bf32f0d3-ade5-4a18-bdcd-a901f4dc9e76-0
00:33:03.320 --> 00:33:05.639
I will only summarize this in
one slide and of course, feel

bf32f0d3-ade5-4a18-bdcd-a901f4dc9e76-1
00:33:05.639 --> 00:33:07.920
free to ask me after this or
dive into the paper for more.

3934a775-78c0-495e-83e7-0adfd11329fb-0
00:33:08.400 --> 00:33:12.150
But that the general framework
is to start with how we design

3934a775-78c0-495e-83e7-0adfd11329fb-1
00:33:12.150 --> 00:33:15.900
buildings today at a specific
site using ASE 716 and say, for

3934a775-78c0-495e-83e7-0adfd11329fb-2
00:33:15.900 --> 00:33:19.710
example, this very nice three
story building achieved a median

3934a775-78c0-495e-83e7-0adfd11329fb-3
00:33:19.710 --> 00:33:23.400
functional recovery time of 155
days at the 475 year return.

1c3d86f6-e100-4f88-a679-a30c69c8738f-0
00:33:24.080 --> 00:33:27.320
Then what we would do is say,
OK, I want I, let's say you have

1c3d86f6-e100-4f88-a679-a30c69c8738f-1
00:33:27.320 --> 00:33:29.480
a check in mind or a goal in
mind, right?

c8ec99f5-1bf8-49c0-8839-e962a37287e3-0
00:33:29.720 --> 00:33:32.320
And we've, we've defined those
per the previous section.

493699cd-ca22-4a9a-9e0b-594e9f0459a2-0
00:33:32.320 --> 00:33:34.160
So we're saying, OK, my check.

c1682c2a-c0e5-4981-bdca-b43da84402c6-0
00:33:34.160 --> 00:33:36.792
Now I know I told you guys don't
use the median, but we're using

c1682c2a-c0e5-4981-bdca-b43da84402c6-1
00:33:36.792 --> 00:33:37.440
the median here.

5b1e4127-6340-42a9-b10e-9f4ea1953735-0
00:33:37.800 --> 00:33:41.627
OK, let's achieve 30 days of
functional recovery time under

5b1e4127-6340-42a9-b10e-9f4ea1953735-1
00:33:41.627 --> 00:33:42.840
the 475 year event.

50b6785e-ad01-448f-bab9-5df89a417e54-0
00:33:43.240 --> 00:33:46.611
Let's limit the probability of
exceedance to 50%, IE let's use,

50b6785e-ad01-448f-bab9-5df89a417e54-1
00:33:46.611 --> 00:33:49.720
let's compare 30 with the median
functional recovery time.

b2b92f58-5d09-47d1-9751-d73700127065-0
00:33:50.200 --> 00:33:53.280
The next question is what is the
strategy in this case?

a4454e2a-d793-4e92-a8af-459385f9daff-0
00:33:53.280 --> 00:33:54.440
Let's just go non structural.

7411e766-b850-4808-b03c-2df0de129f2e-0
00:33:54.440 --> 00:33:55.840
We're not going to do anything
to the structure.

6bd25590-9c12-461e-928c-484bd7e295a8-0
00:33:55.840 --> 00:33:56.760
We're not going to stiffen it.

bef98f33-abd8-450b-9350-8b88bfb44f0f-0
00:33:57.040 --> 00:33:59.341
We're just going to harden
specific non structural

bef98f33-abd8-450b-9350-8b88bfb44f0f-1
00:33:59.341 --> 00:34:00.560
components in the building.

2335f425-85f4-4a2c-9224-950a5d80f919-0
00:34:00.560 --> 00:34:03.120
And then we're going to ask
ourselves, is this feasible?

60b8551f-b4cf-438a-917a-f554dfcc2846-0
00:34:03.280 --> 00:34:06.088
There are certain interventions
that simply will never get us

60b8551f-b4cf-438a-917a-f554dfcc2846-1
00:34:06.088 --> 00:34:07.040
get us to the target.

d604ca09-b570-441c-8975-1f9deb1b3699-0
00:34:07.400 --> 00:34:10.235
You know, there it for specific
targets, especially those that

d604ca09-b570-441c-8975-1f9deb1b3699-1
00:34:10.235 --> 00:34:12.935
are ambitious, like 30 days,
you're getting a lot to get to

d604ca09-b570-441c-8975-1f9deb1b3699-2
00:34:12.935 --> 00:34:13.880
where you want to go.

ae22aaa4-9cb7-4219-8309-e3ccbc6830b7-0
00:34:14.240 --> 00:34:16.398
And this is just to make sure
that the optimization will

ae22aaa4-9cb7-4219-8309-e3ccbc6830b7-1
00:34:16.398 --> 00:34:17.080
actually converge.

dae6bf17-2df4-4293-be4a-01f32713ce2f-0
00:34:17.080 --> 00:34:19.880
If you if it's not feasible, you
won't have a viable solution.

050a2479-dc22-45c8-963a-98d95ee3c074-0
00:34:20.360 --> 00:34:22.691
And then finally, we're going to
do my favorite part, which is

050a2479-dc22-45c8-963a-98d95ee3c074-1
00:34:22.691 --> 00:34:23.320
the optimization.

8df7defb-7b52-4c8f-a789-1dce2cca9961-0
00:34:23.640 --> 00:34:25.983
And for the optimization,
there's two key ingredients

8df7defb-7b52-4c8f-a789-1dce2cca9961-1
00:34:25.983 --> 00:34:26.200
here.

d9ff5ecb-747e-4154-85e1-ae028ec92e2b-0
00:34:26.480 --> 00:34:29.305
The first is the objective
function, which I mentioned last

d9ff5ecb-747e-4154-85e1-ae028ec92e2b-1
00:34:29.305 --> 00:34:31.000
time is what are we optimizing
for?

9781561a-9eda-42e1-9a6d-d5408e3c765b-0
00:34:31.280 --> 00:34:33.960
Are we optimizing for materials
labor engineering?

b8958c26-0b54-4efa-9df6-a147e9c0cd6e-0
00:34:34.480 --> 00:34:37.579
In this case, I unfortunately
don't have an extensive database

b8958c26-0b54-4efa-9df6-a147e9c0cd6e-1
00:34:37.579 --> 00:34:40.580
of cost data, which is very
unfortunate because I think that

b8958c26-0b54-4efa-9df6-a147e9c0cd6e-2
00:34:40.580 --> 00:34:43.039
is ultimately the thing we all
are interested in.

6bb82a8d-650c-435a-a18d-191da2339018-0
00:34:43.440 --> 00:34:47.077
Instead, I have the second best
thing, which is a variable

6bb82a8d-650c-435a-a18d-191da2339018-1
00:34:47.077 --> 00:34:50.653
called XI that represents the
average or the shift in the

6bb82a8d-650c-435a-a18d-191da2339018-2
00:34:50.653 --> 00:34:53.921
fragility curves that we are
making to each of these

6bb82a8d-650c-435a-a18d-191da2339018-3
00:34:53.921 --> 00:34:54.600
components.

7147a43c-f177-4f87-b218-82155bf87d6b-0
00:34:54.600 --> 00:34:58.761
So when I mention or I, when I
say hardening in this context,

7147a43c-f177-4f87-b218-82155bf87d6b-1
00:34:58.761 --> 00:35:02.587
what I mean is that for a
specific component in the FEMA

7147a43c-f177-4f87-b218-82155bf87d6b-2
00:35:02.587 --> 00:35:06.681
P58 analysis, let's say your
HVAC duct, let's say your doors

7147a43c-f177-4f87-b218-82155bf87d6b-3
00:35:06.681 --> 00:35:10.440
that shift, let's say it's
1X2X3X is represented by XI.

86a3cc9e-7f75-4c67-87a9-4dc86c143acd-0
00:35:10.440 --> 00:35:13.411
So like one would mean we're
doing nothing to this component,

86a3cc9e-7f75-4c67-87a9-4dc86c143acd-1
00:35:13.411 --> 00:35:15.760
3 would mean we're strengthening
it three times.

202a7ce9-82d3-499d-bbf9-33f000200b98-0
00:35:16.080 --> 00:35:18.720
And then N represents the number
of components in the building.

deec2c0a-64e8-49e0-b620-490600ff7512-0
00:35:18.720 --> 00:35:22.320
In this case, we have 36 in this
model, 36 unique components.

da9a40c2-f542-4cea-bb59-5cf5f40bceb3-0
00:35:23.080 --> 00:35:25.995
And then 1 / N is basically
saying, let's take the average

da9a40c2-f542-4cea-bb59-5cf5f40bceb3-1
00:35:25.995 --> 00:35:27.280
of all these improvements.

ab561574-0e56-4e6d-a75b-bcabe370f1e9-0
00:35:27.640 --> 00:35:31.350
You can think of it also as like
an importance factor for these

ab561574-0e56-4e6d-a75b-bcabe370f1e9-1
00:35:31.350 --> 00:35:35.002
built for these variables like
what is what is a variable that

ab561574-0e56-4e6d-a75b-bcabe370f1e9-2
00:35:35.002 --> 00:35:38.423
is more important to recovery
agnostic to cost essentially

ab561574-0e56-4e6d-a75b-bcabe370f1e9-3
00:35:38.423 --> 00:35:40.800
similar to like a lasso type
assessment.

9ad5ec15-ee1c-49f8-9268-7e45f1d1a67e-0
00:35:41.560 --> 00:35:45.036
Then the next question would be,
OK, how do we actually estimate

9ad5ec15-ee1c-49f8-9268-7e45f1d1a67e-1
00:35:45.036 --> 00:35:48.405
the median functional recovery
time to evaluate whether or not

9ad5ec15-ee1c-49f8-9268-7e45f1d1a67e-2
00:35:48.405 --> 00:35:51.240
our check criteria of 30 days
has actually been met?

1d161fc9-1a19-4e09-a932-ae0577684a4a-0
00:35:51.600 --> 00:35:54.964
So I mentioned to you that if we
were to do this in today's

1d161fc9-1a19-4e09-a932-ae0577684a4a-1
00:35:54.964 --> 00:35:58.552
world, this would probably be a
simulation based approach using

1d161fc9-1a19-4e09-a932-ae0577684a4a-2
00:35:58.552 --> 00:36:02.140
FEMA P58 and ATC138 for a high
number of realizations to ensure

1d161fc9-1a19-4e09-a932-ae0577684a4a-3
00:36:02.140 --> 00:36:04.720
stability since this is a
stochastic process.

3a489cb4-74c5-4181-a224-68462ffa3fa5-0
00:36:05.080 --> 00:36:07.661
And what we do is replace the
entire thing with a surrogate

3a489cb4-74c5-4181-a224-68462ffa3fa5-1
00:36:07.661 --> 00:36:07.920
model.

9c1e0430-6bae-4ffd-8e28-418f4390947a-0
00:36:08.600 --> 00:36:11.357
Now when we get into surrogate
models, there's a lot of people

9c1e0430-6bae-4ffd-8e28-418f4390947a-1
00:36:11.357 --> 00:36:13.984
in the community who are using
them right now and there are

9c1e0430-6bae-4ffd-8e28-418f4390947a-2
00:36:13.984 --> 00:36:15.560
opinions on how you architect
them.

7d5a4ae7-f1d4-4952-8200-614f8400a3a8-0
00:36:16.040 --> 00:36:17.160
I have a few views.

8c517660-390b-4b83-a282-d41ab005e2dc-0
00:36:17.720 --> 00:36:21.025
The first view is that a fully
coupled architecture may not be

8c517660-390b-4b83-a282-d41ab005e2dc-1
00:36:21.025 --> 00:36:21.760
the best idea.

b1e4390f-593d-4f47-9409-23bd895798bf-0
00:36:22.120 --> 00:36:24.561
So I'm showing you an
architecture here where we can

b1e4390f-593d-4f47-9409-23bd895798bf-1
00:36:24.561 --> 00:36:26.957
map different building
enhancements directly to the

b1e4390f-593d-4f47-9409-23bd895798bf-2
00:36:26.957 --> 00:36:27.280
answer.

c102c63e-c878-477b-801d-497e4081ffb7-0
00:36:27.920 --> 00:36:31.374
The the disadvantage of this
approach is it takes a long time

c102c63e-c878-477b-801d-497e4081ffb7-1
00:36:31.374 --> 00:36:32.600
to train these models.

86c31b6f-8753-4fd7-b818-3016adae71ad-0
00:36:32.640 --> 00:36:34.080
It might take 3-4 days.

7893e434-94c0-4dc8-8c6d-f7e4867a6fc3-0
00:36:34.400 --> 00:36:37.856
And let's say, you know, Grace
comes to me and says, oh, I want

7893e434-94c0-4dc8-8c6d-f7e4867a6fc3-1
00:36:37.856 --> 00:36:41.150
to, you know, I want to use
this, this surrogate model for a

7893e434-94c0-4dc8-8c6d-f7e4867a6fc3-2
00:36:41.150 --> 00:36:44.120
case that's in Minnesota or, you
know, somewhere else.

b51df7cc-6982-430b-be88-0e1ec04e40d0-0
00:36:44.560 --> 00:36:47.003
And if we can't accommodate that
in the architecture, now we have

b51df7cc-6982-430b-be88-0e1ec04e40d0-1
00:36:47.003 --> 00:36:48.040
to retrain the entire thing.

fd6665e7-58ea-4408-8e3a-77089a868d91-0
00:36:48.040 --> 00:36:50.720
And so it's not generalizable
and it wastes resources.

648e3d35-839e-4809-b35c-0e9240737b08-0
00:36:51.240 --> 00:36:54.859
What I have proposed in this
study is a version that totally

648e3d35-839e-4809-b35c-0e9240737b08-1
00:36:54.859 --> 00:36:58.598
decouples the engineering demand
parameters or the engineering

648e3d35-839e-4809-b35c-0e9240737b08-2
00:36:58.598 --> 00:37:02.039
demand parameter prediction from
the downtime assessment.

996f3275-40c8-4f95-96de-a2329866f7a0-0
00:37:02.040 --> 00:37:05.343
So for context, engineering
demand parameters are these

996f3275-40c8-4f95-96de-a2329866f7a0-1
00:37:05.343 --> 00:37:08.587
variables in the FEMA P 58
assessment that account for

996f3275-40c8-4f95-96de-a2329866f7a0-2
00:37:08.587 --> 00:37:11.360
demands like storage drifts and
accelerations.

e5aa22a8-68c1-4d1c-a32d-d537ddf5ca13-0
00:37:11.720 --> 00:37:14.804
You can get a different vector
of engineering demand parameters

e5aa22a8-68c1-4d1c-a32d-d537ddf5ca13-1
00:37:14.804 --> 00:37:17.841
if you change the site, if you
strengthen the building and can

e5aa22a8-68c1-4d1c-a32d-d537ddf5ca13-2
00:37:17.841 --> 00:37:19.480
accommodate all of these
changes.

af8b550b-0c60-4b95-9a15-008276e65414-0
00:37:19.800 --> 00:37:22.692
And So what we did is we
featurized them so that if say

af8b550b-0c60-4b95-9a15-008276e65414-1
00:37:22.692 --> 00:37:25.482
tomorrow someone in this
audience were to calibrate a

af8b550b-0c60-4b95-9a15-008276e65414-2
00:37:25.482 --> 00:37:28.787
model for a different structural
system or you know, maybe some

af8b550b-0c60-4b95-9a15-008276e65414-3
00:37:28.787 --> 00:37:32.197
type of structural improvements,
we can have that as an auxiliary

af8b550b-0c60-4b95-9a15-008276e65414-4
00:37:32.197 --> 00:37:34.160
model that can be handled
separately.

37f55628-7cf4-440b-bd1a-9258c5ce709e-0
00:37:34.480 --> 00:37:37.367
And then this model that we have
calibrated here that includes

37f55628-7cf4-440b-bd1a-9258c5ce709e-1
00:37:37.367 --> 00:37:39.750
engineering demand parameters
and in non structural

37f55628-7cf4-440b-bd1a-9258c5ce709e-2
00:37:39.750 --> 00:37:42.454
enhancements using those XI
variables, I mentioned earlier

37f55628-7cf4-440b-bd1a-9258c5ce709e-3
00:37:42.454 --> 00:37:45.020
that these could be solely
responsible for the downtime

37f55628-7cf4-440b-bd1a-9258c5ce709e-4
00:37:45.020 --> 00:37:45.799
assessment piece.

a6412a3e-61d5-425e-a962-6ad4d615cc64-0
00:37:46.400 --> 00:37:49.235
And So what we did is we, we
went ahead with this decoupled

a6412a3e-61d5-425e-a962-6ad4d615cc64-1
00:37:49.235 --> 00:37:51.882
architecture and it worked
really well in, in this case

a6412a3e-61d5-425e-a962-6ad4d615cc64-2
00:37:51.882 --> 00:37:54.812
with a three story building, we
have 3 variables or four that

a6412a3e-61d5-425e-a962-6ad4d615cc64-3
00:37:54.812 --> 00:37:57.175
represent peak floor
accelerations, including the

a6412a3e-61d5-425e-a962-6ad4d615cc64-4
00:37:57.175 --> 00:37:59.680
ground or the roof, depending on
how you look at it.

0beaaf62-5fbe-4960-a97b-623d050fa606-0
00:38:00.000 --> 00:38:02.680
Three story drift ratio
variables.

7fcf25a9-2964-4b00-bdc8-a5e7fa16c902-0
00:38:03.080 --> 00:38:06.075
And then a series of 36 non
structural improvement variables

7fcf25a9-2964-4b00-bdc8-a5e7fa16c902-1
00:38:06.075 --> 00:38:09.267
that account for things like did
we improve the elevator, did we

7fcf25a9-2964-4b00-bdc8-a5e7fa16c902-2
00:38:09.267 --> 00:38:10.199
improve the stairs?

75d62bd7-a103-4cb3-a3be-05dd28e8b7ca-0
00:38:10.600 --> 00:38:13.819
And map all of that directly to
the median functional recovery

75d62bd7-a103-4cb3-a3be-05dd28e8b7ca-1
00:38:13.819 --> 00:38:16.782
time since ultimately that is
what's going to allow us to

75d62bd7-a103-4cb3-a3be-05dd28e8b7ca-2
00:38:16.782 --> 00:38:19.440
check whether or not we hit that
target of 30 days.

3496c274-0d11-4c0d-82ee-fd6cc854a4cc-0
00:38:19.960 --> 00:38:23.029
And this achieved a very high
predictive performance based on

3496c274-0d11-4c0d-82ee-fd6cc854a4cc-1
00:38:23.029 --> 00:38:23.920
the training data.

a846f5a0-3e5d-4abf-b4ad-86d1663e219f-0
00:38:23.920 --> 00:38:26.572
We are definitely trying to
overfit it in this case because

a846f5a0-3e5d-4abf-b4ad-86d1663e219f-1
00:38:26.572 --> 00:38:28.120
this is an interpolation
exercise.

6677b7b5-e4de-4beb-87ee-48d4b1d5b929-0
00:38:28.720 --> 00:38:30.880
So I came to you guys today with
a demo.

5770129b-7cf1-483f-bf26-82e19e919ce8-0
00:38:30.880 --> 00:38:32.960
I'm going to show you how this
actually works.

b09fce43-b424-4cc4-8aaa-5e7226523912-0
00:38:33.160 --> 00:38:36.680
This is real time or pseudo real
time, I don't know, but I think

b09fce43-b424-4cc4-8aaa-5e7226523912-1
00:38:36.680 --> 00:38:39.280
you'll find that it's, it's
interesting to see.

6cf2adbb-bcff-4409-a4fa-7349f12020ae-0
00:38:39.280 --> 00:38:43.141
So the algorithm is looking
through, and this is a genetic

6cf2adbb-bcff-4409-a4fa-7349f12020ae-1
00:38:43.141 --> 00:38:46.806
algorithm, by the way, looking
through the series of 36

6cf2adbb-bcff-4409-a4fa-7349f12020ae-2
00:38:46.806 --> 00:38:51.125
different components and it very
quickly, I would say in a matter

6cf2adbb-bcff-4409-a4fa-7349f12020ae-3
00:38:51.125 --> 00:38:54.986
of seconds, sifted through the
design space and identified

6cf2adbb-bcff-4409-a4fa-7349f12020ae-4
00:38:54.986 --> 00:38:58.847
12345 that are ultimately the
best, I guess, components to

6cf2adbb-bcff-4409-a4fa-7349f12020ae-5
00:38:58.847 --> 00:39:02.905
actually change to achieve this
30 day target time to go from

6cf2adbb-bcff-4409-a4fa-7349f12020ae-6
00:39:02.905 --> 00:39:03.560
155 to 30.

9660b9af-0a9b-42bf-9a13-95018de13904-0
00:39:04.240 --> 00:39:06.360
And what's interesting is that
this is agnostic of cost.

e12d3ac6-08f0-48fe-bd92-30365bd85465-0
00:39:06.360 --> 00:39:09.657
These are the variables that
have the highest influence on

e12d3ac6-08f0-48fe-bd92-30365bd85465-1
00:39:09.657 --> 00:39:10.720
downtime reduction.

0735adcf-b409-47cf-855b-aeb7769bdb2b-0
00:39:11.560 --> 00:39:13.240
And so let's take a look at this
a little bit more.

be1cb5c6-da1a-457f-8d10-a088dd55f8dc-0
00:39:13.800 --> 00:39:16.920
So this is for a 30 day target.

54fa2c5a-80bf-415d-9d9b-ee556648b14b-0
00:39:17.280 --> 00:39:19.974
Now you can ask me, well, Omar,
what if I wanted to repeat this

54fa2c5a-80bf-415d-9d9b-ee556648b14b-1
00:39:19.974 --> 00:39:22.543
for 20 days, 10 days, like I
mean, like I might be exploring

54fa2c5a-80bf-415d-9d9b-ee556648b14b-2
00:39:22.543 --> 00:39:23.680
the design space right now.

97d77e64-9e5c-4b93-b9bc-b0db4086c4d5-0
00:39:24.240 --> 00:39:26.919
And I think that's where this
approach really shines is that

97d77e64-9e5c-4b93-b9bc-b0db4086c4d5-1
00:39:26.919 --> 00:39:29.511
we can, if we for a specific
check, if you're unsure about

97d77e64-9e5c-4b93-b9bc-b0db4086c4d5-2
00:39:29.511 --> 00:39:32.146
that 30 day marker, you want to,
you want to tweak things a

97d77e64-9e5c-4b93-b9bc-b0db4086c4d5-3
00:39:32.146 --> 00:39:34.782
little bit, you can actually
vary the target times, fix the

97d77e64-9e5c-4b93-b9bc-b0db4086c4d5-4
00:39:34.782 --> 00:39:36.320
seismic intensity and this
repeat.

a588de05-9345-4475-9e7b-d95e8b8d46bf-0
00:39:36.720 --> 00:39:39.791
And this entire graph took me
maybe more, no more than maybe

a588de05-9345-4475-9e7b-d95e8b8d46bf-1
00:39:39.791 --> 00:39:41.000
2-3 minutes to generate.

61e4dadc-b975-4642-bfd0-56484f7f02d8-0
00:39:41.320 --> 00:39:45.034
But it shows you as you get more
ambitious with the functional

61e4dadc-b975-4642-bfd0-56484f7f02d8-1
00:39:45.034 --> 00:39:48.513
recovery time targets at this
intensity, you can see which

61e4dadc-b975-4642-bfd0-56484f7f02d8-2
00:39:48.513 --> 00:39:50.400
components start to matter more.

352ac323-d19a-4948-bae7-a32049789141-0
00:39:50.720 --> 00:39:53.882
And if we're zooming in on this
example, we can see that stairs

352ac323-d19a-4948-bae7-a32049789141-1
00:39:53.882 --> 00:39:56.847
for this very flexible steel
building are important because

352ac323-d19a-4948-bae7-a32049789141-2
00:39:56.847 --> 00:39:59.811
stairs prevent if they are
damaged, prevent egress, prevent

352ac323-d19a-4948-bae7-a32049789141-3
00:39:59.811 --> 00:40:02.480
the occupancy, therefore prevent
functional recovery.

2875c5ec-9a62-47ad-93c8-a3e3a0541507-0
00:40:02.880 --> 00:40:05.658
You can also see the same with
the glass curtain walls and the

2875c5ec-9a62-47ad-93c8-a3e3a0541507-1
00:40:05.658 --> 00:40:06.320
precast panels.

f947adca-44ef-4dc3-8568-970201458b68-0
00:40:06.560 --> 00:40:09.722
A flexible building like this is
all ultimately going to have

f947adca-44ef-4dc3-8568-970201458b68-1
00:40:09.722 --> 00:40:10.080
drifts.

9ed21fde-dace-4bca-a32a-f9394fb95b2b-0
00:40:10.400 --> 00:40:13.411
And so those drifts are going to
damage certain components that

9ed21fde-dace-4bca-a32a-f9394fb95b2b-1
00:40:13.411 --> 00:40:14.400
affect functionality.

7dd0c16b-d3a2-4d78-9fc8-d3cea543536f-0
00:40:14.720 --> 00:40:17.377
You can see that in more
ambitious cases, things like the

7dd0c16b-d3a2-4d78-9fc8-d3cea543536f-1
00:40:17.377 --> 00:40:20.080
HVAC, things like the electric
components start to matter.

559d5a30-8f1a-4c1a-848e-604e6a2985a0-0
00:40:20.480 --> 00:40:23.440
But the the idea here is, is
that by using surrogate models,

559d5a30-8f1a-4c1a-848e-604e6a2985a0-1
00:40:23.440 --> 00:40:26.304
we can more efficiently scan the
design space and start to

559d5a30-8f1a-4c1a-848e-604e6a2985a0-2
00:40:26.304 --> 00:40:28.440
develop some of that engineering
intuition.

cfa2489f-082b-4ff3-a001-b94ec6699938-0
00:40:29.000 --> 00:40:31.686
You could also come to me and
say, oh, I like 30 days, but I

cfa2489f-082b-4ff3-a001-b94ec6699938-1
00:40:31.686 --> 00:40:33.360
want to change the seismic
intensity.

9e6d3762-2bd0-4041-8a67-f14bb2dc221f-0
00:40:33.560 --> 00:40:37.189
We can do the same exact thing,
repeat the analysis and then

9e6d3762-2bd0-4041-8a67-f14bb2dc221f-1
00:40:37.189 --> 00:40:40.760
parse out for a 30 day target
across different intensities.

c6413d1a-dc70-4685-a868-8bef77e41e7a-0
00:40:41.000 --> 00:40:42.640
What would be those changes we
need to make?

b71f63d3-5f40-4384-83af-ea9bcbb2ab79-0
00:40:42.880 --> 00:40:45.513
And if you're, if, if any of you
are asking me what do the

b71f63d3-5f40-4384-83af-ea9bcbb2ab79-1
00:40:45.513 --> 00:40:47.880
numbers mean, I apologize if I
explained it earlier.

3f0109f3-54fb-4775-89cc-e386dca76c8a-0
00:40:48.080 --> 00:40:50.360
A1 means we do nothing to that
component.

36f644c1-cb67-456c-8c6a-73a788b448cf-0
00:40:50.880 --> 00:40:54.661
A2 A 1.5, A 3 is basically
saying for every quantity of

36f644c1-cb67-456c-8c6a-73a788b448cf-1
00:40:54.661 --> 00:40:58.511
those components in the
building, we are 1.5 exiting the

36f644c1-cb67-456c-8c6a-73a788b448cf-2
00:40:58.511 --> 00:41:01.280
medium capacity of of all damage
states.

2e3fbbde-eab0-4db4-b349-b46dd2dadfee-0
00:41:01.280 --> 00:41:02.240
That's basically what it means.

86af7269-df30-47d2-8390-a65597ffacd9-0
00:41:03.960 --> 00:41:05.200
OK, so this is all great.

0faf46cf-1600-4f39-9f7a-ab5642e7414a-0
00:41:05.200 --> 00:41:08.718
We have some insights here one
of the things that I, I think is

0faf46cf-1600-4f39-9f7a-ab5642e7414a-1
00:41:08.718 --> 00:41:11.961
interesting is that we can also
unveil the influence of of

0faf46cf-1600-4f39-9f7a-ab5642e7414a-2
00:41:11.961 --> 00:41:14.875
different structural
interventions on an optimal non

0faf46cf-1600-4f39-9f7a-ab5642e7414a-3
00:41:14.875 --> 00:41:16.360
structural design solution.

a925019a-b153-4ed6-b438-b8f580804998-0
00:41:16.360 --> 00:41:19.321
What I mean by that is is like
let's see how this optimal

a925019a-b153-4ed6-b438-b8f580804998-1
00:41:19.321 --> 00:41:22.384
solution changes if we did
stiffen the building and I think

a925019a-b153-4ed6-b438-b8f580804998-2
00:41:22.384 --> 00:41:25.294
this is interesting because I
told you guys earlier that

a925019a-b153-4ed6-b438-b8f580804998-3
00:41:25.294 --> 00:41:28.255
sometimes stiffening the
building can be to the detriment

a925019a-b153-4ed6-b438-b8f580804998-4
00:41:28.255 --> 00:41:30.399
of of non structural component
hardening.

8258a294-eed1-43b0-a3d4-64136c54496f-0
00:41:30.800 --> 00:41:33.795
So let's go ahead and, and vary
the target times and fix the

8258a294-eed1-43b0-a3d4-64136c54496f-1
00:41:33.795 --> 00:41:34.680
seismic intensity.

315caead-13af-4198-b57c-c88f454028ab-0
00:41:35.160 --> 00:41:38.485
And let's go and actually
change, Let's have 3 versions of

315caead-13af-4198-b57c-c88f454028ab-1
00:41:38.485 --> 00:41:41.641
this building, 1 version where
the structural system is

315caead-13af-4198-b57c-c88f454028ab-2
00:41:41.641 --> 00:41:45.136
identical, 1 version where we
increase the importance factor,

315caead-13af-4198-b57c-c88f454028ab-3
00:41:45.136 --> 00:41:47.560
IE we make it stiffer and
stronger to 1.5.

3b3896d8-e3b1-4456-ae35-1c9739252f50-0
00:41:47.880 --> 00:41:50.738
And then this crazy, I would say
not so realistic importance

3b3896d8-e3b1-4456-ae35-1c9739252f50-1
00:41:50.738 --> 00:41:51.160
factor 2.

353bdddc-8718-41a9-8f78-2c16fec0f64c-0
00:41:51.160 --> 00:41:53.480
When we go all the way, we're
going to go strengthen this

353bdddc-8718-41a9-8f78-2c16fec0f64c-1
00:41:53.480 --> 00:41:54.840
thing to the to the limit,
right?

98698447-e369-4f30-99fa-0638de434fc2-0
00:41:56.000 --> 00:42:00.560
What we can do then is map the
optimal F of X.

1176116f-1327-493c-b4da-7d3ef9b27769-0
00:42:00.560 --> 00:42:05.056
This is a pre abstract metric,
but you can think of it as the

1176116f-1327-493c-b4da-7d3ef9b27769-1
00:42:05.056 --> 00:42:08.320
average XI, the average
improvement we have.

9f0a4600-86b7-46bb-bf11-e5c9fe49b94f-0
00:42:08.560 --> 00:42:13.159
We, we, we have, we have or the
optimal average XI shift for

9f0a4600-86b7-46bb-bf11-e5c9fe49b94f-1
00:42:13.159 --> 00:42:15.120
that specific target time.

68ab49af-760c-4d2a-929b-088345af45d5-0
00:42:15.120 --> 00:42:18.579
So you can see here average
shift is one because we're

68ab49af-760c-4d2a-929b-088345af45d5-1
00:42:18.579 --> 00:42:22.478
already there like the baseline
case is already achieving 100

68ab49af-760c-4d2a-929b-088345af45d5-2
00:42:22.478 --> 00:42:25.120
and 55155 days of functional
public time.

5cea5cc5-88cb-4eec-9587-529e19ffdefb-0
00:42:25.520 --> 00:42:28.633
It's as you go further that you
start to see, OK, this average

5cea5cc5-88cb-4eec-9587-529e19ffdefb-1
00:42:28.633 --> 00:42:31.698
shift is increasing because we
need to make some improvements

5cea5cc5-88cb-4eec-9587-529e19ffdefb-2
00:42:31.698 --> 00:42:33.280
to the building to to get there.

19df76f6-6a78-4be1-a963-a53901173376-0
00:42:33.280 --> 00:42:36.278
And I'll paint this more
tangibly by showing you the

19df76f6-6a78-4be1-a963-a53901173376-1
00:42:36.278 --> 00:42:37.240
optimal solution.

ece1c2c0-d12a-415a-8ffd-3fff12649389-0
00:42:37.480 --> 00:42:39.880
For an example case, let's look
at 120 days.

8e6865be-829f-4ef7-b422-1af93141ea66-0
00:42:39.880 --> 00:42:43.935
So we started right quote
conforming 155 days, pulling now

8e6865be-829f-4ef7-b422-1af93141ea66-1
00:42:43.935 --> 00:42:44.760
to 120 days.

a41f71b4-da94-4980-b463-610f06bb1d6b-0
00:42:45.040 --> 00:42:47.840
What is the solution if we did
nothing in the building?

caabe9b5-7908-418a-b4eb-3851a3ad216a-0
00:42:47.840 --> 00:42:49.520
Strengthen it a little bit,
strengthen it a lot.

dc0f953f-5301-4409-836f-72cacc057f9b-0
00:42:50.120 --> 00:42:54.562
You can see that when the target
was very close to the original,

dc0f953f-5301-4409-836f-72cacc057f9b-1
00:42:54.562 --> 00:42:58.389
these structurally improved
versions, you know, they're

dc0f953f-5301-4409-836f-72cacc057f9b-2
00:42:58.389 --> 00:43:00.440
already kind of getting there.

567449cf-e87e-45a5-a807-d2d173b4e7c1-0
00:43:00.440 --> 00:43:04.510
They've already they're already
achieving 120 days without any

567449cf-e87e-45a5-a807-d2d173b4e7c1-1
00:43:04.510 --> 00:43:06.320
non structural intervention.

74937264-3310-4968-87f4-4da08952178e-0
00:43:06.680 --> 00:43:09.539
And maybe this case where we
have a no structural

74937264-3310-4968-87f4-4da08952178e-1
00:43:09.539 --> 00:43:13.141
improvements will make a couple
of improvements or upgrades to

74937264-3310-4968-87f4-4da08952178e-2
00:43:13.141 --> 00:43:14.800
the facade and the staircase.

898dd0d5-9dda-4021-9780-3d0bee63e344-0
00:43:15.560 --> 00:43:20.720
But let's go in and push it to
the limit to explore 21 days.

7ca36ebe-7590-4153-9f69-5d4c624d998d-0
00:43:21.000 --> 00:43:24.481
What I think is interesting
about this case is that the

7ca36ebe-7590-4153-9f69-5d4c624d998d-1
00:43:24.481 --> 00:43:28.459
trend actually reverses because
at this point if we look at say

7ca36ebe-7590-4153-9f69-5d4c624d998d-2
00:43:28.459 --> 00:43:31.816
the building that had no
structural improvements, the

7ca36ebe-7590-4153-9f69-5d4c624d998d-3
00:43:31.816 --> 00:43:35.359
average XI or the optimal
solution is actually the best.

9c8c932a-f2c3-4c0b-b06b-9a62d1936ee6-0
00:43:35.840 --> 00:43:39.054
Because here there are no
significant acceleration

9c8c932a-f2c3-4c0b-b06b-9a62d1936ee6-1
00:43:39.054 --> 00:43:42.457
amplifications that would
require us to significantly

9c8c932a-f2c3-4c0b-b06b-9a62d1936ee6-2
00:43:42.457 --> 00:43:44.600
improve non structural down
here.

fe7191ca-77af-4623-be7a-fbf4a92f8222-0
00:43:45.000 --> 00:43:48.195
Whereas with the sniffing
buildings, we have to make more

fe7191ca-77af-4623-be7a-fbf4a92f8222-1
00:43:48.195 --> 00:43:50.840
extensive non structural
upgrades on this side.

e5f71a95-a32e-4130-b43d-f44640566441-0
00:43:51.480 --> 00:43:53.995
I've seen this plot many times
in my life, so I have a mental

e5f71a95-a32e-4130-b43d-f44640566441-1
00:43:53.995 --> 00:43:54.320
mapping.

f2d5161e-0ba6-489c-8f63-34d8c6e58724-0
00:43:54.560 --> 00:43:58.930
But you can generally imagine
that the that the top components

f2d5161e-0ba6-489c-8f63-34d8c6e58724-1
00:43:58.930 --> 00:44:02.746
like the panels, curtain walls,
gypsum are story drift

f2d5161e-0ba6-489c-8f63-34d8c6e58724-2
00:44:02.746 --> 00:44:03.440
sensitive.

a3838cb7-2001-4d96-bde6-9df75f347864-0
00:44:03.880 --> 00:44:06.800
And then these ones here are
more acceleration sensitive.

285a90b2-8e23-49e0-96bf-4d7dfab33e91-0
00:44:07.240 --> 00:44:11.148
And so you can see that for
these very stiffened buildings

285a90b2-8e23-49e0-96bf-4d7dfab33e91-1
00:44:11.148 --> 00:44:15.123
that have higher accelerations,
more extensive upgrades are

285a90b2-8e23-49e0-96bf-4d7dfab33e91-2
00:44:15.123 --> 00:44:19.296
required for these acceleration
sensitive components to get to

285a90b2-8e23-49e0-96bf-4d7dfab33e91-3
00:44:19.296 --> 00:44:23.072
21 days, whereas the original
structural system required

285a90b2-8e23-49e0-96bf-4d7dfab33e91-4
00:44:23.072 --> 00:44:24.000
slightly less.

0ccb2802-d923-4e33-9309-eeed8580501e-0
00:44:24.040 --> 00:44:28.132
And of course, this average XI
metric is, you know, it's, it's

0ccb2802-d923-4e33-9309-eeed8580501e-1
00:44:28.132 --> 00:44:32.095
just a way of kind of giving us
a sense of or a benchmark to

0ccb2802-d923-4e33-9309-eeed8580501e-2
00:44:32.095 --> 00:44:32.680
optimize.

a38b7323-b0d4-4088-a6ee-a1afcafd59cd-0
00:44:32.680 --> 00:44:35.971
But I think regardless the,
these, these results are

a38b7323-b0d4-4088-a6ee-a1afcafd59cd-1
00:44:35.971 --> 00:44:39.884
insightful because it can tell
us maybe, maybe we we should be

a38b7323-b0d4-4088-a6ee-a1afcafd59cd-2
00:44:39.884 --> 00:44:43.610
thinking about interventions
differently depending on where

a38b7323-b0d4-4088-a6ee-a1afcafd59cd-3
00:44:43.610 --> 00:44:44.480
our target is.

317a2df5-0c88-40e3-82f5-c064f9fa907b-0
00:44:44.480 --> 00:44:46.040
I think that's really the take
away here.

63f4b539-e1cc-410b-9fb5-7ed92be538d8-0
00:44:46.840 --> 00:44:49.791
And so in this research we
talked a little bit about this

63f4b539-e1cc-410b-9fb5-7ed92be538d8-1
00:44:49.791 --> 00:44:52.844
optimization framework, which is
really apoc applied to one

63f4b539-e1cc-410b-9fb5-7ed92be538d8-2
00:44:52.844 --> 00:44:56.000
building, but can be generically
applied to many other cases.

2f00eeba-469c-4fd7-9b8a-6e299104564b-0
00:44:56.400 --> 00:44:59.336
I presented a neural network
architecture that I believe is

2f00eeba-469c-4fd7-9b8a-6e299104564b-1
00:44:59.336 --> 00:45:02.125
is very well suited for a
problem like this and then how

2f00eeba-469c-4fd7-9b8a-6e299104564b-2
00:45:02.125 --> 00:45:05.208
the framework can be deployed
for a code conforming case study

2f00eeba-469c-4fd7-9b8a-6e299104564b-3
00:45:05.208 --> 00:45:05.600
example.

c684a1e0-c84b-4df4-b03a-d9e4bc4dafab-0
00:45:05.840 --> 00:45:08.840
You could also do this for an
existing building for retrofits.

3f09cb88-c099-4f59-9e0f-eff8f69ff5fa-0
00:45:09.120 --> 00:45:11.760
This type of framework can be
applied to many different cases.

c21e75f4-5d9b-4e58-827a-41bef916aa6c-0
00:45:12.400 --> 00:45:15.341
So today we talked about, I
would say a fruit salad here,

c21e75f4-5d9b-4e58-827a-41bef916aa6c-1
00:45:15.341 --> 00:45:17.320
but we really started with
objectives.

2a661150-2242-4e53-826c-8606f44f495d-0
00:45:17.840 --> 00:45:20.978
We talked about how performance
objectives are a goal plus a

2a661150-2242-4e53-826c-8606f44f495d-1
00:45:20.978 --> 00:45:23.910
checking procedure, How goal
metrics are not necessarily

2a661150-2242-4e53-826c-8606f44f495d-2
00:45:23.910 --> 00:45:27.100
correlated and how using a check
that relies on a conditional

2a661150-2242-4e53-826c-8606f44f495d-3
00:45:27.100 --> 00:45:30.392
probability of failure between
10 and 20% significantly and can

2a661150-2242-4e53-826c-8606f44f495d-4
00:45:30.392 --> 00:45:33.479
reliably increase confidence
that we're getting to our goal

2a661150-2242-4e53-826c-8606f44f495d-5
00:45:33.479 --> 00:45:36.360
if we're going to only check the
fragility in one spot.

83b9b6a8-58dd-45ab-935c-10e0e8eb2909-0
00:45:37.040 --> 00:45:39.719
The second part of this research
I showed you how a machine

83b9b6a8-58dd-45ab-935c-10e0e8eb2909-1
00:45:39.719 --> 00:45:42.309
learning based optimization
framework that can be used to

83b9b6a8-58dd-45ab-935c-10e0e8eb2909-2
00:45:42.309 --> 00:45:45.123
rapidly extract those strategies
based on the design check and

83b9b6a8-58dd-45ab-935c-10e0e8eb2909-3
00:45:45.123 --> 00:45:47.713
how we can calibrate the
surrogate model architectures to

83b9b6a8-58dd-45ab-935c-10e0e8eb2909-4
00:45:47.713 --> 00:45:48.160
get there.

4a073340-3588-4c2c-92a6-7dff17c314d7-0
00:45:49.080 --> 00:45:52.338
But I think in terms of the
research I mentioned you guys, I

4a073340-3588-4c2c-92a6-7dff17c314d7-1
00:45:52.338 --> 00:45:55.490
came to Stanford because I
wanted to understand how can we

4a073340-3588-4c2c-92a6-7dff17c314d7-2
00:45:55.490 --> 00:45:57.200
design buildings for resilience.

464c9273-4736-4c6c-866e-0ae4c871c221-0
00:45:57.200 --> 00:46:00.028
And I think that these
approaches are extremely

464c9273-4736-4c6c-866e-0ae4c871c221-1
00:46:00.028 --> 00:46:01.560
promising to get us there.

5092516d-01fe-4b00-a4b2-ecce4765f7e4-0
00:46:01.920 --> 00:46:05.881
But in terms of where they fit
in the, in the grand scheme of

5092516d-01fe-4b00-a4b2-ecce4765f7e4-1
00:46:05.881 --> 00:46:10.098
the built world today, there are
these different players that are

5092516d-01fe-4b00-a4b2-ecce4765f7e4-2
00:46:10.098 --> 00:46:13.421
influencing our built
environment and how a a given

5092516d-01fe-4b00-a4b2-ecce4765f7e4-3
00:46:13.421 --> 00:46:17.000
building may influence a
community recovery trajectory.

9094e8d2-b231-4891-a2be-37518544ab7e-0
00:46:17.480 --> 00:46:20.722
A tool like this, research like
this can really help engineers

9094e8d2-b231-4891-a2be-37518544ab7e-1
00:46:20.722 --> 00:46:23.707
who are doing this bottoms up
advocacy, who are designing

9094e8d2-b231-4891-a2be-37518544ab7e-2
00:46:23.707 --> 00:46:26.897
these case study examples and in
government who are trying to

9094e8d2-b231-4891-a2be-37518544ab7e-3
00:46:26.897 --> 00:46:30.088
understand how do we flex our
provisions to get us to a point

9094e8d2-b231-4891-a2be-37518544ab7e-4
00:46:30.088 --> 00:46:32.816
where we're considering
functional recovery, which I

9094e8d2-b231-4891-a2be-37518544ab7e-5
00:46:32.816 --> 00:46:33.999
think is very exciting.

f4484332-8262-4ff8-8bb3-3e9e5c006853-0
00:46:34.280 --> 00:46:37.165
And if we do this for enough
buildings, we can successfully

f4484332-8262-4ff8-8bb3-3e9e5c006853-1
00:46:37.165 --> 00:46:39.907
support tenants who are within
these different community

f4484332-8262-4ff8-8bb3-3e9e5c006853-2
00:46:39.907 --> 00:46:42.360
functions and create a more
resilient built world.

38d16ea5-be54-4dd7-9720-1391beda61f3-0
00:46:43.200 --> 00:46:46.740
But there is a caveat here, and
that is one thing that has

38d16ea5-be54-4dd7-9720-1391beda61f3-1
00:46:46.740 --> 00:46:50.460
shifted my attention at least in
recent months, that a lot of

38d16ea5-be54-4dd7-9720-1391beda61f3-2
00:46:50.460 --> 00:46:52.800
this work exists in the realm of
risk.

eb56c1f1-e013-40bd-b118-2051634369ee-0
00:46:53.240 --> 00:46:57.443
And these risk models are very
powerful and require a lot of

eb56c1f1-e013-40bd-b118-2051634369ee-1
00:46:57.443 --> 00:47:00.200
data and are necessary to this
problem.

10fddd7a-8bbb-4f35-8c27-f24d67a0499f-0
00:47:00.640 --> 00:47:03.688
But what I have seen in industry
is that a lot of the

10fddd7a-8bbb-4f35-8c27-f24d67a0499f-1
00:47:03.688 --> 00:47:07.301
stakeholders I have spoken to do
not feel confident about their

10fddd7a-8bbb-4f35-8c27-f24d67a0499f-2
00:47:07.301 --> 00:47:08.599
inputs to these models.

c32b7928-59d1-41cb-b7e8-0b63402dd194-0
00:47:08.800 --> 00:47:12.141
They don't feel confident about
various data points on exposure

c32b7928-59d1-41cb-b7e8-0b63402dd194-1
00:47:12.141 --> 00:47:15.379
that are required to give us an
accurate understanding of our

c32b7928-59d1-41cb-b7e8-0b63402dd194-2
00:47:15.379 --> 00:47:18.564
portfolios and our building
stock that would serve as inputs

c32b7928-59d1-41cb-b7e8-0b63402dd194-3
00:47:18.564 --> 00:47:19.400
to these models.

d31c2d8e-f906-4f94-90de-c7044b86bbfe-0
00:47:19.800 --> 00:47:23.589
And so in the next chapter of my
life, I have dedicated it to, as

d31c2d8e-f906-4f94-90de-c7044b86bbfe-1
00:47:23.589 --> 00:47:26.977
Grace mentioned, starting a
company that is solely focused

d31c2d8e-f906-4f94-90de-c7044b86bbfe-2
00:47:26.977 --> 00:47:30.250
on addressing this issue using
AI that's been trained on

d31c2d8e-f906-4f94-90de-c7044b86bbfe-3
00:47:30.250 --> 00:47:33.581
structural engineering domain
knowledge and trying to, at

d31c2d8e-f906-4f94-90de-c7044b86bbfe-4
00:47:33.581 --> 00:47:37.255
scale, try to discern how can we
uncover the maybe the top 1015

d31c2d8e-f906-4f94-90de-c7044b86bbfe-5
00:47:37.255 --> 00:47:40.988
features that influence risk to
build the exposure data layer to

d31c2d8e-f906-4f94-90de-c7044b86bbfe-6
00:47:40.988 --> 00:47:43.399
power the next generation of
risk models.

cabb3e86-8db3-4dff-8364-6fa4539eb492-0
00:47:43.720 --> 00:47:45.280
That is really what I'm focused
on today.

f6f3e7e7-6020-4679-b3ce-9aea84f05c22-0
00:47:45.280 --> 00:47:47.880
I spend lots of time sleepless
nights thinking about it.

bf89d552-8cd7-4d43-ab9c-7237bad10c32-0
00:47:48.200 --> 00:47:51.212
And the goal would be to
interface with people in the

bf89d552-8cd7-4d43-ab9c-7237bad10c32-1
00:47:51.212 --> 00:47:54.504
private sector, but also the
public sector and academia to

bf89d552-8cd7-4d43-ab9c-7237bad10c32-2
00:47:54.504 --> 00:47:56.959
support research that requires
better data.

d8235b4b-1579-494a-940d-fd4984538a72-0
00:47:57.360 --> 00:47:59.160
And we're already starting to
work with folks.

e010c4fb-8fe3-400f-84f1-ded7c3fc5505-0
00:47:59.400 --> 00:48:02.618
And my invitation to everyone
watching this today is that if

e010c4fb-8fe3-400f-84f1-ded7c3fc5505-1
00:48:02.618 --> 00:48:05.732
this type of data, high fidelity
and credible, can support

e010c4fb-8fe3-400f-84f1-ded7c3fc5505-2
00:48:05.732 --> 00:48:09.109
resilience research right now,
please let me know because we're

e010c4fb-8fe3-400f-84f1-ded7c3fc5505-3
00:48:09.109 --> 00:48:10.639
very interested in a partner.

81a70a53-0aef-4246-bf22-67b931cbebdf-0
00:48:11.000 --> 00:48:13.886
And of course, as you guys know
about me, I'm an optimist and I

81a70a53-0aef-4246-bf22-67b931cbebdf-1
00:48:13.886 --> 00:48:15.600
believe that the future is
resilient.

6adbc841-956f-488c-98c0-6e0454aef293-0
00:48:15.880 --> 00:48:17.240
Thank you very much for tuning
in today.

98fe7f46-1a30-46d3-9654-05f0b0e39cb0-0
00:48:23.990 --> 00:48:26.070
You any questions in the room?

44641b84-5ec8-44df-b437-a3ecdf8f7175-0
00:48:29.920 --> 00:48:32.868
Oh, yes, please, Steve, great
talk and I really like the way

44641b84-5ec8-44df-b437-a3ecdf8f7175-1
00:48:32.868 --> 00:48:35.720
you sort of decouple all these
really complicated systems.

9204da11-e9ad-4cba-9b00-29d62eca9498-0
00:48:36.080 --> 00:48:39.568
So and I gather from your final
comment, one of the problems is

9204da11-e9ad-4cba-9b00-29d62eca9498-1
00:48:39.568 --> 00:48:43.003
how do you train these machine
learning algorithms to actually

9204da11-e9ad-4cba-9b00-29d62eca9498-2
00:48:43.003 --> 00:48:46.328
understand the controls for a
particular kind of building on

9204da11-e9ad-4cba-9b00-29d62eca9498-3
00:48:46.328 --> 00:48:48.999
functional recovery like a
hospital for exactly.

1950770d-a97a-4904-9814-33968394695d-0
00:48:49.600 --> 00:48:51.520
So there's a lot of expert
knowledge embedded.

d9d6c3a4-2c02-4f0f-9b40-d9dc81fdc98d-0
00:48:51.680 --> 00:48:54.698
How does HVAC versus electrical
versus stairways, all these

d9d6c3a4-2c02-4f0f-9b40-d9dc81fdc98d-1
00:48:54.698 --> 00:48:56.560
things, how does it impact
coverage?

6f8e5850-ab3b-4fba-b968-8b6aa44c5c91-0
00:48:56.960 --> 00:48:59.480
Is that based on observations
after earthquakes?

ff29174d-4b19-452c-93c0-c85b047ff512-0
00:48:59.600 --> 00:49:00.600
How do you determine them?

ba8247a0-682d-4226-831b-f1643f0227ad-0
00:49:00.920 --> 00:49:02.000
No, this is a great question.

792ec01d-3331-46ae-8518-bdec0fac18ad-0
00:49:02.080 --> 00:49:06.595
And I think this if we go back
to some of these points I

792ec01d-3331-46ae-8518-bdec0fac18ad-1
00:49:06.595 --> 00:49:10.160
mentioned on how the models are
are trained.

fe49609a-2951-440e-acb5-60f7f179ecf4-0
00:49:10.520 --> 00:49:14.785
So the models today are learning
from simulations and these

fe49609a-2951-440e-acb5-60f7f179ecf4-1
00:49:14.785 --> 00:49:19.050
simulations, which are coming
from FEMA P58 and ATC138, are

fe49609a-2951-440e-acb5-60f7f179ecf4-2
00:49:19.050 --> 00:49:23.102
attempting to embed at the
component level some of these

fe49609a-2951-440e-acb5-60f7f179ecf4-3
00:49:23.102 --> 00:49:23.600
things.

fa57c8d8-6673-41ce-a7c8-e996ffd32f6b-0
00:49:23.920 --> 00:49:27.848
For example, with ATC 138 that
is more focused on the downtime

fa57c8d8-6673-41ce-a7c8-e996ffd32f6b-1
00:49:27.848 --> 00:49:28.160
side.

e2dd0ff6-fdbb-4114-9c03-9058cc4637f0-0
00:49:28.160 --> 00:49:31.297
I know for a fact that the
impeding factors themselves have

e2dd0ff6-fdbb-4114-9c03-9058cc4637f0-1
00:49:31.297 --> 00:49:34.592
been calibrated on past events
and those come from our limited

e2dd0ff6-fdbb-4114-9c03-9058cc4637f0-2
00:49:34.592 --> 00:49:37.886
sources of people actually going
out collecting these impeding

e2dd0ff6-fdbb-4114-9c03-9058cc4637f0-3
00:49:37.886 --> 00:49:38.880
factor data points.

1d8f956b-b062-4253-86ac-26fd597a30ff-0
00:49:39.280 --> 00:49:42.934
Unfortunately though, I feel
like a lot of that data isn't

1d8f956b-b062-4253-86ac-26fd597a30ff-1
00:49:42.934 --> 00:49:46.774
tethered to a specific or is
not, is not categorized in terms

1d8f956b-b062-4253-86ac-26fd597a30ff-2
00:49:46.774 --> 00:49:50.677
of occupancy type many of the
times or maybe it's like focused

1d8f956b-b062-4253-86ac-26fd597a30ff-3
00:49:50.677 --> 00:49:52.039
on one occupancy type.

a2a6306d-1fd2-4dd4-a4b1-560d45ece780-0
00:49:52.040 --> 00:49:55.055
Maybe it's focused on housing,
not so much, you know, looking

a2a6306d-1fd2-4dd4-a4b1-560d45ece780-1
00:49:55.055 --> 00:49:56.320
at hospitals specifically.

d12ecaa8-581e-470b-9f48-1b9fcc29e3de-0
00:49:56.720 --> 00:50:00.643
I think that one of the the
things that that this model does

d12ecaa8-581e-470b-9f48-1b9fcc29e3de-1
00:50:00.643 --> 00:50:04.439
well is it's basically like
whatever we do to improve FEMA

d12ecaa8-581e-470b-9f48-1b9fcc29e3de-2
00:50:04.439 --> 00:50:07.720
P58 and ATC138IS reflected in
the surrogate model.

75fee107-606d-4f52-9244-b44a63000c46-0
00:50:08.360 --> 00:50:11.762
But I'd say right now the 20
years of research that we put

75fee107-606d-4f52-9244-b44a63000c46-1
00:50:11.762 --> 00:50:15.221
into FEMA P 58 as a method are
there, but there is a lot of

75fee107-606d-4f52-9244-b44a63000c46-2
00:50:15.221 --> 00:50:17.240
work to do on specific
components.

7785c9c8-0c75-4eac-ad6b-80e545b6eff0-0
00:50:17.600 --> 00:50:20.800
The FEMA P58 database only has
700 components.

8e744bd3-1ac6-4d58-98f3-8d997baa094e-0
00:50:21.040 --> 00:50:24.006
That's not nearly enough to
represent the vast variety of

8e744bd3-1ac6-4d58-98f3-8d997baa094e-1
00:50:24.006 --> 00:50:27.280
seismically certified equipment
that exists in these hospitals.

5b297473-e8b2-4c36-8107-c56bd3d2d9d1-0
00:50:27.440 --> 00:50:31.171
And one of the limitations I've
seen is in this, in this demo I

5b297473-e8b2-4c36-8107-c56bd3d2d9d1-1
00:50:31.171 --> 00:50:34.320
showed you guys one of those
numbers ever exceeded 3.

b8e7014c-6c42-4c8f-a733-e86154dade07-0
00:50:34.640 --> 00:50:37.748
That's because I artificially
capped the constraint on the

b8e7014c-6c42-4c8f-a733-e86154dade07-1
00:50:37.748 --> 00:50:40.857
improvement at 3 based on my
knowledge and what the trends

b8e7014c-6c42-4c8f-a733-e86154dade07-2
00:50:40.857 --> 00:50:42.280
I've seen in this database.

f80505ee-2c51-430f-9288-4042eef7e9a5-0
00:50:42.480 --> 00:50:45.598
But there could be a scenario
where seismically certified

f80505ee-2c51-430f-9288-4042eef7e9a5-1
00:50:45.598 --> 00:50:48.823
equipment for certain, for
certain hospitals or, or certain

f80505ee-2c51-430f-9288-4042eef7e9a5-2
00:50:48.823 --> 00:50:51.834
types of very important
equipment could be 10XI mean it

f80505ee-2c51-430f-9288-4042eef7e9a5-3
00:50:51.834 --> 00:50:54.200
could be 11X if it's seismically
certified.

f0bf0f95-6f94-48c7-918d-376d3459d587-0
00:50:54.200 --> 00:50:55.280
And that's not reflected here.

509289fe-d65c-4e07-b01d-e07451e6554c-0
00:50:55.560 --> 00:50:58.143
So I think one of the things
that I wrote in my memo as I

509289fe-d65c-4e07-b01d-e07451e6554c-1
00:50:58.143 --> 00:50:59.880
departed Stanford, they make us
right.

b8826f63-f685-4519-bb80-c4b79c2da06f-0
00:50:59.880 --> 00:51:02.720
Basically, what do we leave for
the next students who come in?

a6c8e88d-d998-4f2c-a00a-d46da9504396-0
00:51:02.720 --> 00:51:05.285
I was like, please take a look
at all this seismically

a6c8e88d-d998-4f2c-a00a-d46da9504396-1
00:51:05.285 --> 00:51:08.269
simplified equipment and get us
better data on both the cost of

a6c8e88d-d998-4f2c-a00a-d46da9504396-2
00:51:08.269 --> 00:51:11.254
improvement, then it also all of
the other fragilities that are

a6c8e88d-d998-4f2c-a00a-d46da9504396-3
00:51:11.254 --> 00:51:13.120
out there that we're not
incorporating.

438bab47-d5f2-4671-a4e6-6d7ac436ead3-0
00:51:13.440 --> 00:51:17.413
The other thing is that you'll
notice that I never, I, I didn't

438bab47-d5f2-4671-a4e6-6d7ac436ead3-1
00:51:17.413 --> 00:51:19.400
make this a discretized problem.

2ef774e6-fc62-47bc-b4ea-a8e60ac1e3c7-0
00:51:19.600 --> 00:51:22.232
So I never said, oh, are we
choosing component A or

2ef774e6-fc62-47bc-b4ea-a8e60ac1e3c7-1
00:51:22.232 --> 00:51:22.840
component B?

53607911-1dd3-4b59-8d77-f2503b260e70-0
00:51:23.080 --> 00:51:26.013
That is because the I do not
believe that the FEMA P58

53607911-1dd3-4b59-8d77-f2503b260e70-1
00:51:26.013 --> 00:51:29.160
database represents all of the
scenarios that exist today.

1941dc8b-8f5e-4024-be96-c9d84de9d048-0
00:51:29.640 --> 00:51:32.972
And actually I would rather this
algorithm tell me the fragility

1941dc8b-8f5e-4024-be96-c9d84de9d048-1
00:51:32.972 --> 00:51:35.895
I need to create to make that
target happen than have to

1941dc8b-8f5e-4024-be96-c9d84de9d048-2
00:51:35.895 --> 00:51:38.818
choose from say 3 or 4 options
of HVAC that exist in the

1941dc8b-8f5e-4024-be96-c9d84de9d048-3
00:51:38.818 --> 00:51:39.280
database.

232000a2-8cbf-49f6-85b9-d0bb043fd1d4-0
00:51:39.520 --> 00:51:41.843
There are some very passionate
and great people working on

232000a2-8cbf-49f6-85b9-d0bb043fd1d4-1
00:51:41.843 --> 00:51:42.040
this.

5e9666aa-a650-4287-9a4f-41301f31c47d-0
00:51:42.040 --> 00:51:44.851
I know people at SP3 are looking
at fragilities all the time and

5e9666aa-a650-4287-9a4f-41301f31c47d-1
00:51:44.851 --> 00:51:47.447
they have partnerships with
manufacturers right now looking

5e9666aa-a650-4287-9a4f-41301f31c47d-2
00:51:47.447 --> 00:51:47.880
into this.

7b27390c-f36c-4af3-8a03-f97fb2b32168-0
00:51:48.280 --> 00:51:51.176
But I think one of the it's not
necessarily related, but one of

7b27390c-f36c-4af3-8a03-f97fb2b32168-1
00:51:51.176 --> 00:51:53.440
the things that is still a major
gap is the cost.

03a63709-5144-4d2d-9d5e-e76350423afc-0
00:51:53.840 --> 00:51:56.882
Like we still don't have a good
handle on how cost functions map

03a63709-5144-4d2d-9d5e-e76350423afc-1
00:51:56.882 --> 00:51:58.240
to any of these improvements.

6295baec-da38-4716-b3a9-082869738878-0
00:52:00.120 --> 00:52:01.280
Yeah, of course that's great.

1f21be2a-be66-4012-9a20-d358620b14a9-0
00:52:03.480 --> 00:52:04.360
Any other questions?

0ad517c9-3a46-4984-8b84-e1363812cdcf-0
00:52:04.920 --> 00:52:08.080
How does the like big
complicated optimization and

0ad517c9-3a46-4984-8b84-e1363812cdcf-1
00:52:08.080 --> 00:52:11.674
looking at recovery times
compared to just increasing the

0ad517c9-3a46-4984-8b84-e1363812cdcf-2
00:52:11.674 --> 00:52:15.020
factors of safety on what we
already do as structural

0ad517c9-3a46-4984-8b84-e1363812cdcf-3
00:52:15.020 --> 00:52:15.640
engineers?

11b18675-59e4-46d7-9c7e-ff2e13613263-0
00:52:16.360 --> 00:52:18.920
That's a good, I think that's a
really important point.

3e198c66-9ba9-41ad-b3d8-37af78325632-0
00:52:19.480 --> 00:52:23.327
There may be a scenario where
maybe increasing the importance

3e198c66-9ba9-41ad-b3d8-37af78325632-1
00:52:23.327 --> 00:52:27.051
factor in some areas of the US
yields acceptable functional

3e198c66-9ba9-41ad-b3d8-37af78325632-2
00:52:27.051 --> 00:52:27.920
recovery time.

cf64bb76-bec9-4747-a614-fb94c0b1e3cd-0
00:52:27.960 --> 00:52:31.250
And I think that's once there
have been studies in the past,

cf64bb76-bec9-4747-a614-fb94c0b1e3cd-1
00:52:31.250 --> 00:52:34.648
let's say a year or so that are
trying to do that benchmarking

cf64bb76-bec9-4747-a614-fb94c0b1e3cd-2
00:52:34.648 --> 00:52:37.400
activity to say, where does code
get us right now?

ac5be34d-8359-4b04-91e4-ef83b8784120-0
00:52:37.680 --> 00:52:40.970
Like there are probably some
areas of the US that maybe code

ac5be34d-8359-4b04-91e4-ef83b8784120-1
00:52:40.970 --> 00:52:44.476
is enough, especially if we are
deciding, hey, like we're trying

ac5be34d-8359-4b04-91e4-ef83b8784120-2
00:52:44.476 --> 00:52:47.713
to get to in, in, in many of
these conversations I've seen,

ac5be34d-8359-4b04-91e4-ef83b8784120-3
00:52:47.713 --> 00:52:50.680
by the way, it's not as granular
as like, oh, 30 days.

38ec85cb-41fb-4eed-92db-1853c0263154-0
00:52:51.000 --> 00:52:53.712
A lot of the times they're
saying days to weeks, weeks to

38ec85cb-41fb-4eed-92db-1853c0263154-1
00:52:53.712 --> 00:52:54.040
months.

6fc5e0e0-efe0-4daf-a9c4-f0238b62776f-0
00:52:54.040 --> 00:52:57.765
There's probably some areas of
the US where maybe reducing

6fc5e0e0-efe0-4daf-a9c4-f0238b62776f-1
00:52:57.765 --> 00:53:01.491
drift limit and importance
factor changes, like going risk

6fc5e0e0-efe0-4daf-a9c4-f0238b62776f-2
00:53:01.491 --> 00:53:04.080
category one to two to three
could help.

140ff190-da31-4717-a7f5-649dfe5275f7-0
00:53:04.280 --> 00:53:08.234
And I actually think that the
BSSC is looking at toggling

140ff190-da31-4717-a7f5-649dfe5275f7-1
00:53:08.234 --> 00:53:12.120
those factors as opposed to
doing like this type of high

140ff190-da31-4717-a7f5-649dfe5275f7-2
00:53:12.120 --> 00:53:13.280
dimensional work.

48030f90-9fbd-4f4d-bb8c-8abcf3c558fc-0
00:53:13.880 --> 00:53:16.705
Isn't that basically already
kind of built in, in cases where

48030f90-9fbd-4f4d-bb8c-8abcf3c558fc-1
00:53:16.705 --> 00:53:19.530
you have like the California
building code as opposed to like

48030f90-9fbd-4f4d-bb8c-8abcf3c558fc-2
00:53:19.530 --> 00:53:22.401
Kansas probably doesn't have its
own yeah, yeah, yeah, seismic

48030f90-9fbd-4f4d-bb8c-8abcf3c558fc-3
00:53:22.401 --> 00:53:23.040
building code.

5118a79e-c6c1-4cbe-9eb6-2911d06eb337-0
00:53:23.720 --> 00:53:24.600
That's a great point.

fcf9984e-8c0e-4761-b982-9baa9c73137b-0
00:53:25.120 --> 00:53:28.343
The way I think of it is we have
the template, like the

fcf9984e-8c0e-4761-b982-9baa9c73137b-1
00:53:28.343 --> 00:53:31.280
provisions are a template and
they have the knobs.

43f395e4-2dcb-4678-bdbf-6b13339ff8fe-0
00:53:31.800 --> 00:53:34.908
And then the question is how do
we turn the knobs to achieve

43f395e4-2dcb-4678-bdbf-6b13339ff8fe-1
00:53:34.908 --> 00:53:36.080
this effective outcome?

8acf93d4-97ec-42f1-a065-7299f1c14765-0
00:53:36.520 --> 00:53:40.220
And the question that people in
the BSc are asking is how, what

8acf93d4-97ec-42f1-a065-7299f1c14765-1
00:53:40.220 --> 00:53:43.515
are the places in the US where
the knob even needs to be

8acf93d4-97ec-42f1-a065-7299f1c14765-2
00:53:43.515 --> 00:53:43.920
turned?

39d2a2b8-e351-4509-a1de-633a6cddf310-0
00:53:44.160 --> 00:53:46.440
And like you're saying, I think
there are some regions where

39d2a2b8-e351-4509-a1de-633a6cddf310-1
00:53:46.440 --> 00:53:48.160
they probably don't even have to
do anything.

866624f8-0ba3-4fbd-92d0-2fecc08d391b-0
00:53:48.320 --> 00:53:51.438
And then others, I would say,
especially areas in the Bay Area

866624f8-0ba3-4fbd-92d0-2fecc08d391b-1
00:53:51.438 --> 00:53:54.507
where we have like these caps, I
think these, these like risk

866624f8-0ba3-4fbd-92d0-2fecc08d391b-2
00:53:54.507 --> 00:53:57.477
caps, like that's, I think those
are the places where we're

866624f8-0ba3-4fbd-92d0-2fecc08d391b-3
00:53:57.477 --> 00:54:00.200
probably going to need to turn
things off quite a bit.

b9418431-dfe9-4c54-8b03-7ea0a3c47b7d-0
00:54:00.200 --> 00:54:02.967
And, and there's actually
another point there, which is

b9418431-dfe9-4c54-8b03-7ea0a3c47b7d-1
00:54:02.967 --> 00:54:05.931
like who gets to decide what
the, what the performance goal

b9418431-dfe9-4c54-8b03-7ea0a3c47b7d-2
00:54:05.931 --> 00:54:06.080
is?

48a64853-9e55-4a97-81ce-500f2eed04b5-0
00:54:06.120 --> 00:54:07.320
And this is something I wrote
about.

53b02b6a-3d79-47c7-85e8-7f8f776d29e4-0
00:54:08.000 --> 00:54:10.735
There's a lot of, a lot of
debate in this area about

53b02b6a-3d79-47c7-85e8-7f8f776d29e4-1
00:54:10.735 --> 00:54:13.833
whether or not it should be
based on benefit cost analysis,

53b02b6a-3d79-47c7-85e8-7f8f776d29e4-2
00:54:13.833 --> 00:54:15.640
revealed or expressed
preferences.

fe6ee5d8-038e-4d1f-9589-488ab6c318bb-0
00:54:16.080 --> 00:54:18.920
And I think it's honestly a
fascinating topic.

6555396f-c12c-4387-9625-238e9a121ebe-0
00:54:19.400 --> 00:54:22.183
But what I, what I see is
missing a lot of times is

6555396f-c12c-4387-9625-238e9a121ebe-1
00:54:22.183 --> 00:54:25.555
societal elicitation and asking
people, what do you expect out

6555396f-c12c-4387-9625-238e9a121ebe-2
00:54:25.555 --> 00:54:26.679
of the building code?

124bb5dd-f9ed-400f-90b9-fd50a4f17ce5-0
00:54:26.880 --> 00:54:29.169
And I think the spur report that
I showed in the first slide is a

124bb5dd-f9ed-400f-90b9-fd50a4f17ce5-1
00:54:29.169 --> 00:54:30.800
good example of what happens
when you do that.

b8f31646-7a7a-40e1-928d-9a765d27b16a-0
00:54:31.120 --> 00:54:34.057
But I don't think it's being
done nearly enough in places in

b8f31646-7a7a-40e1-928d-9a765d27b16a-1
00:54:34.057 --> 00:54:37.139
the US where I think the public
is generally unaware about what

b8f31646-7a7a-40e1-928d-9a765d27b16a-2
00:54:37.139 --> 00:54:38.679
the building code even delivers.

116bb7bf-d0b7-48cc-878e-d3f8c543ff46-0
00:54:38.680 --> 00:54:43.720
Basically, great presentation.

de6a43d5-90cc-4060-a762-4cad84263657-0
00:54:43.720 --> 00:54:44.560
Omar, as always.

f9926b55-2eb2-4e36-a402-8cbb4b744452-0
00:54:46.320 --> 00:54:48.920
I want to go back to the first
part of your presentation.

5db53643-dc6a-4f07-9e5b-c86eabc9ce6e-0
00:54:49.480 --> 00:54:54.072
And you were talking about the
the goal for functional recovery

5db53643-dc6a-4f07-9e5b-c86eabc9ce6e-1
00:54:54.072 --> 00:54:58.522
and the check for so recovery
and basically the goal you were

5db53643-dc6a-4f07-9e5b-c86eabc9ce6e-2
00:54:58.522 --> 00:55:03.043
you you want to see all ground
motion intensity levels and the

5db53643-dc6a-4f07-9e5b-c86eabc9ce6e-3
00:55:03.043 --> 00:55:04.120
check just one.

c0d54c73-2de1-4e42-8ccf-ff8a90a4cbc3-0
00:55:05.600 --> 00:55:05.840
So what?

f7ab9b61-90f2-408b-b39d-659fb5e00a77-0
00:55:05.960 --> 00:55:11.079
So my question is, why should
we, why should we go for the

f7ab9b61-90f2-408b-b39d-659fb5e00a77-1
00:55:11.079 --> 00:55:11.600
check?

cc07c9e1-c676-4c92-858f-2cd7d1ccca28-0
00:55:13.640 --> 00:55:18.525
Is it reasonable to instead go
for the goal or or the goal will

cc07c9e1-c676-4c92-858f-2cd7d1ccca28-1
00:55:18.525 --> 00:55:22.800
be the check of check the goal
directly and not do that

cc07c9e1-c676-4c92-858f-2cd7d1ccca28-2
00:55:22.800 --> 00:55:24.480
simplified assessment?

b50434a9-45f8-43e5-bdda-060144533ad0-0
00:55:25.360 --> 00:55:29.754
Because I, I once did a FEMA P58
type analysis for a building and

b50434a9-45f8-43e5-bdda-060144533ad0-1
00:55:29.754 --> 00:55:33.882
most of the time that I spent
trying to do the thing was, was

b50434a9-45f8-43e5-bdda-060144533ad0-2
00:55:33.882 --> 00:55:35.680
not running the simulation.

87ea9c8d-6ab6-4a99-ad31-db8c6eb94ec2-0
00:55:35.680 --> 00:55:37.560
It was actually building or or.

0c01e6a3-376d-4a48-9bd9-df4ce71a61d4-0
00:55:37.920 --> 00:55:41.760
So the most of the time is not
in running the analysis.

98b7aa95-5a8e-46b5-9a52-8542209e46ff-0
00:55:41.760 --> 00:55:45.448
So, so, so basically if, well, I
mean, what I've seen is that if

98b7aa95-5a8e-46b5-9a52-8542209e46ff-1
00:55:45.448 --> 00:55:48.852
you check different almost
intensity measures, it's like go

98b7aa95-5a8e-46b5-9a52-8542209e46ff-2
00:55:48.852 --> 00:55:52.200
view do instead of 1, you do 10
almost intensity measures.

9c434120-c028-43e9-95e0-8e31953a20ba-0
00:55:52.440 --> 00:55:56.479
OK, whatever you're running is
going to take 10 more, 10 times

9c434120-c028-43e9-95e0-8e31953a20ba-1
00:55:56.479 --> 00:55:56.800
more.

f427b964-bfc5-4a5e-83e5-6999eb2bbf52-0
00:55:57.040 --> 00:56:00.030
But it's not, that's not where
you spend all your time when

f427b964-bfc5-4a5e-83e5-6999eb2bbf52-1
00:56:00.030 --> 00:56:02.871
you, when you're doing something
for a real building and

f427b964-bfc5-4a5e-83e5-6999eb2bbf52-2
00:56:02.871 --> 00:56:03.320
practice.

1fbe41c0-74c3-4290-98d3-aae34a899160-0
00:56:03.320 --> 00:56:04.760
So what's your opinion about
that?

925cede9-1482-484d-8ddc-d9501989938b-0
00:56:04.760 --> 00:56:10.502
And is it reasonable to to ask
like practicing engineers to

925cede9-1482-484d-8ddc-d9501989938b-1
00:56:10.502 --> 00:56:12.800
check directly the goal?

44f993b6-755d-4348-876a-10d7b85a2e84-0
00:56:13.680 --> 00:56:15.600
Yes, No, this is very important.

8bf4c63c-e732-416e-a197-1147ef6f60f6-0
00:56:15.600 --> 00:56:18.273
And if this is ever recorded,
just know that I'm not trying to

8bf4c63c-e732-416e-a197-1147ef6f60f6-1
00:56:18.273 --> 00:56:19.080
create controversy.

c32d68f8-3834-40fe-8519-ae36d2280f7e-0
00:56:19.680 --> 00:56:23.782
I would say that for this type
of to clarify this type of

c32d68f8-3834-40fe-8519-ae36d2280f7e-1
00:56:23.782 --> 00:56:27.885
context or framing feels a
little more appropriate in the

c32d68f8-3834-40fe-8519-ae36d2280f7e-2
00:56:27.885 --> 00:56:32.058
context of provisions where
maybe we're not running a FEMA

c32d68f8-3834-40fe-8519-ae36d2280f7e-3
00:56:32.058 --> 00:56:33.119
P50 assessment.

7521a7a9-e2c3-4350-9285-fa3b52bf1e74-0
00:56:33.360 --> 00:56:37.252
Instead, we're calibrating maybe
an ELF type procedure that would

7521a7a9-e2c3-4350-9285-fa3b52bf1e74-1
00:56:37.252 --> 00:56:41.027
by proxy achieve the check kind
of how we're doing it with life

7521a7a9-e2c3-4350-9285-fa3b52bf1e74-2
00:56:41.027 --> 00:56:41.439
safety.

2291a884-2539-48fd-90e9-77cd2611cf90-0
00:56:41.920 --> 00:56:44.720
Because you're right, if this
was a FEMA P 58 assessment, I

2291a884-2539-48fd-90e9-77cd2611cf90-1
00:56:44.720 --> 00:56:47.706
would totally just do the life
the the time based assessment or

2291a884-2539-48fd-90e9-77cd2611cf90-2
00:56:47.706 --> 00:56:49.760
the risk based assessment get to
my answer.

cecf9b7e-055e-4b89-8f5d-bcb4b357086e-0
00:56:50.840 --> 00:56:54.067
But if we are doing if we're
dealing with folks who may not

cecf9b7e-055e-4b89-8f5d-bcb4b357086e-1
00:56:54.067 --> 00:56:57.402
have access to the inputs and
the data necessary to running a

cecf9b7e-055e-4b89-8f5d-bcb4b357086e-2
00:56:57.402 --> 00:57:00.899
full blown FEMA P50 assessment,
then this would be this checking

cecf9b7e-055e-4b89-8f5d-bcb4b357086e-3
00:57:00.899 --> 00:57:03.320
procedure would be more well
suited to them.

b8345c2a-4748-4131-9dd3-8183f9ada73d-0
00:57:03.720 --> 00:57:08.327
Now you may ask me, well, Omar,
checking one point is is kind of

b8345c2a-4748-4131-9dd3-8183f9ada73d-1
00:57:08.327 --> 00:57:09.320
a gamble here.

00b86e28-9c1e-4dcb-8982-fc665bc5212f-0
00:57:09.320 --> 00:57:10.440
Why don't we just check 2?

bd34fb19-6a51-4fe8-adf0-5c07e67e0316-0
00:57:10.960 --> 00:57:14.810
And that is, I would say been a
much more of a debate than I

bd34fb19-6a51-4fe8-adf0-5c07e67e0316-1
00:57:14.810 --> 00:57:18.725
thought because in in the, I
would say structural engineering

bd34fb19-6a51-4fe8-adf0-5c07e67e0316-2
00:57:18.725 --> 00:57:22.323
community, there's a lot of
pushback on having to do the

bd34fb19-6a51-4fe8-adf0-5c07e67e0316-3
00:57:22.323 --> 00:57:25.480
work twice for a single building
on these checks.

19a5c64f-3a6c-429f-bab6-0d2ffd5921f0-0
00:57:25.480 --> 00:57:28.025
Even though some standards like
the tall building initiative,

19a5c64f-3a6c-429f-bab6-0d2ffd5921f0-1
00:57:28.025 --> 00:57:30.325
there are these guidelines for
tall buildings where you

19a5c64f-3a6c-429f-bab6-0d2ffd5921f0-2
00:57:30.325 --> 00:57:32.871
implicitly do have to check like
a design level and a service

19a5c64f-3a6c-429f-bab6-0d2ffd5921f0-3
00:57:32.871 --> 00:57:34.760
level, like you're checking 2
ground motions.

05565e18-ef05-4aca-975f-99e2ae7d33eb-0
00:57:35.240 --> 00:57:39.407
But for the average building,
the 99%, many people would push

05565e18-ef05-4aca-975f-99e2ae7d33eb-1
00:57:39.407 --> 00:57:43.036
back against doing getting
another pinch point on the

05565e18-ef05-4aca-975f-99e2ae7d33eb-2
00:57:43.036 --> 00:57:43.440
curve.

62d300c0-65fd-4636-be40-313f2fbac596-0
00:57:43.800 --> 00:57:47.218
I am a as a non practicing
structural engineer, I am very

62d300c0-65fd-4636-be40-313f2fbac596-1
00:57:47.218 --> 00:57:48.280
much Pro 2 points.

52df2048-b70f-457a-a434-6c381111df4f-0
00:57:48.560 --> 00:57:51.336
But I'd say many people have,
you know, strong feelings about

52df2048-b70f-457a-a434-6c381111df4f-1
00:57:51.336 --> 00:57:51.560
that.

688f3474-5c2c-4ea6-bf7d-716c5575763d-0
00:57:51.680 --> 00:57:54.525
So this research would at least
help those who insist on having

688f3474-5c2c-4ea6-bf7d-716c5575763d-1
00:57:54.525 --> 00:57:57.371
one point or if it is embedded
in the provisions would at least

688f3474-5c2c-4ea6-bf7d-716c5575763d-2
00:57:57.371 --> 00:58:00.083
give you some confidence about
if we're going to choose one,

688f3474-5c2c-4ea6-bf7d-716c5575763d-3
00:58:00.083 --> 00:58:01.640
what would be like the best
place?

9e91bd8e-39ef-46de-a323-251950dbfd95-0
00:58:05.520 --> 00:58:06.360
That's a very good question.

6805057b-6111-43a9-b8d3-81e837494abb-0
00:58:08.400 --> 00:58:09.000
Anybody else?

24b99616-ed3f-4af7-9e1c-9c87206e62bb-0
00:58:11.160 --> 00:58:12.000
Yes, Chris.

b6127f88-46c5-4776-ad88-ca04e905e841-0
00:58:12.240 --> 00:58:15.760
So I don't think about building
stiffness hardly ever.

f0617b55-6916-4823-9b14-f523c36506e7-0
00:58:16.080 --> 00:58:19.850
So this might be a silly
question, but you showed the

f0617b55-6916-4823-9b14-f523c36506e7-1
00:58:19.850 --> 00:58:23.482
dependence of some of your
results on the stiffness

f0617b55-6916-4823-9b14-f523c36506e7-2
00:58:23.482 --> 00:58:27.671
modifications that you might
take when you're designing for

f0617b55-6916-4823-9b14-f523c36506e7-3
00:58:27.671 --> 00:58:28.440
resilience.

e31697c4-ba3d-46dd-818b-6448d668e4f7-0
00:58:29.200 --> 00:58:32.994
How much do individual building
designs, as they have been

e31697c4-ba3d-46dd-818b-6448d668e4f7-1
00:58:32.994 --> 00:58:36.981
created, influence changes and
stiffness that are inherent to

e31697c4-ba3d-46dd-818b-6448d668e4f7-2
00:58:36.981 --> 00:58:40.839
any improvement you might make
before you've done anything?

cd56411a-8074-45ac-9910-a56bdca1422c-0
00:58:40.840 --> 00:58:43.040
I don't know if that made sense.

12818f58-1dc9-42b2-a5e7-cc5092d4ad84-0
00:58:43.040 --> 00:58:44.080
Let me try and restate that.

764ab7d4-0f72-423d-8900-9949c588cae1-0
00:58:44.080 --> 00:58:47.730
Like how much does an initial
building design control the

764ab7d4-0f72-423d-8900-9949c588cae1-1
00:58:47.730 --> 00:58:48.360
stiffness?

c68eabf5-552c-490c-a87c-8c24b9cffd0f-0
00:58:48.360 --> 00:58:52.689
And can you control for that
when you're trying to understand

c68eabf5-552c-490c-a87c-8c24b9cffd0f-1
00:58:52.689 --> 00:58:56.600
what components you need to
build into your new design?

8657679f-6478-4598-9130-b3f36d475f65-0
00:58:57.000 --> 00:58:57.720
Yeah, no, that's it.

51e414b3-7ee7-4d66-882c-062948c84094-0
00:58:57.720 --> 00:58:59.160
That's a great, a great
question.

f1d8c436-4e68-497d-8e76-864dd95c0ed7-0
00:58:59.160 --> 00:59:01.494
I think like there's an
interdependence between, for

f1d8c436-4e68-497d-8e76-864dd95c0ed7-1
00:59:01.494 --> 00:59:04.269
example, I think the first step
you would probably do when you

f1d8c436-4e68-497d-8e76-864dd95c0ed7-2
00:59:04.269 --> 00:59:06.120
design a new building, this is
very easy.

d3297a19-9f6f-44f8-9e56-2c4d4e477d99-0
00:59:06.120 --> 00:59:07.480
This is much easier for new
buildings.

fbf7cd62-fa1f-44ca-b094-8e95fabd54c5-0
00:59:07.840 --> 00:59:09.600
Like let's say we're about to
design some new.

b49dc4b4-647f-48c4-8005-54d0e0c7242e-0
00:59:09.840 --> 00:59:12.280
The first question is what's the
lateral force resistance system?

dfd104c5-0f71-4341-bf41-f8bad790c410-0
00:59:12.600 --> 00:59:15.669
Steel, very flexible concrete
shear wall would probably be

dfd104c5-0f71-4341-bf41-f8bad790c410-1
00:59:15.669 --> 00:59:18.894
more stiff and then that would
influence probably the type of

dfd104c5-0f71-4341-bf41-f8bad790c410-2
00:59:18.894 --> 00:59:21.651
as we saw the type of non
structural line structural

dfd104c5-0f71-4341-bf41-f8bad790c410-3
00:59:21.651 --> 00:59:23.160
components we would bring in.

739a912f-a39f-4558-9248-222c22f47a92-0
00:59:23.360 --> 00:59:26.259
But if we think about it, a lot
of those components would come

739a912f-a39f-4558-9248-222c22f47a92-1
00:59:26.259 --> 00:59:26.720
in anyway.

710daf15-f605-4c34-93ee-19d3afc3d0e9-0
00:59:26.920 --> 00:59:29.220
The question is when we do
something special to them to

710daf15-f605-4c34-93ee-19d3afc3d0e9-1
00:59:29.220 --> 00:59:31.520
achieve, you know, the target
where we're trying to go.

1ca18177-7246-456a-817c-3b8e7ba58771-0
00:59:32.160 --> 00:59:35.186
I would say that as of right
now, and this is I guess maybe

1ca18177-7246-456a-817c-3b8e7ba58771-1
00:59:35.186 --> 00:59:38.363
not necessarily answering the
question directly, a lot of this

1ca18177-7246-456a-817c-3b8e7ba58771-2
00:59:38.363 --> 00:59:40.280
work is done through trial and
error.

904ea6ca-ac04-46e5-a716-ae80d5008208-0
00:59:40.720 --> 00:59:44.038
Like we're trying to see, you
know, through some of the

904ea6ca-ac04-46e5-a716-ae80d5008208-1
00:59:44.038 --> 00:59:47.475
software solutions I was
mentioning, OK, like if I choose

904ea6ca-ac04-46e5-a716-ae80d5008208-2
00:59:47.475 --> 00:59:50.319
this type of system, am I
getting to my target?

7bbae41c-ed6c-4ddc-914d-a0bd355ebfd6-0
00:59:50.320 --> 00:59:52.153
OK, I'm going to change these
three non structural, I'm going

7bbae41c-ed6c-4ddc-914d-a0bd355ebfd6-1
00:59:52.153 --> 00:59:53.040
to am I getting to the target?

efe6d7d9-e520-4b4a-b8cf-6ed1e6f4b998-0
00:59:53.040 --> 00:59:55.323
Whereas the optimization
approach just does everything

efe6d7d9-e520-4b4a-b8cf-6ed1e6f4b998-1
00:59:55.323 --> 00:59:57.440
head to end all at once to give
you what you need.

dc3e791c-4d2a-4e13-a74e-62822bafa54b-0
00:59:57.880 --> 01:00:02.121
But to your observation and
point with these assessments

dc3e791c-4d2a-4e13-a74e-62822bafa54b-1
01:00:02.121 --> 01:00:06.735
here, and this is kind of like a
minor detail that I think is

dc3e791c-4d2a-4e13-a74e-62822bafa54b-2
01:00:06.735 --> 01:00:08.000
worth mentioning.

64a6fb13-3a95-4e8f-a899-6db0f473a82a-0
01:00:08.760 --> 01:00:12.760
The reason I showed you the
results in this format.

63e0adb8-4ab8-4475-9fdd-7a4563f0d518-0
01:00:13.880 --> 01:00:14.360
Here.

feecd7c0-bb5b-4b32-8216-bc81a8b9c4c5-0
01:00:14.400 --> 01:00:18.280
So the reason why I showed you
guys the results in this format

feecd7c0-bb5b-4b32-8216-bc81a8b9c4c5-1
01:00:18.280 --> 01:00:21.853
is because the optimization was
only doing things or only

feecd7c0-bb5b-4b32-8216-bc81a8b9c4c5-2
01:00:21.853 --> 01:00:24.440
considering non structural
interventions.

ceffc8bd-8150-44a2-a518-c296d54e864b-0
01:00:25.160 --> 01:00:28.318
I think what you're getting to,
which I think is a much more

ceffc8bd-8150-44a2-a518-c296d54e864b-1
01:00:28.318 --> 01:00:31.373
appropriate and efficient way of
doing this, is to include

ceffc8bd-8150-44a2-a518-c296d54e864b-2
01:00:31.373 --> 01:00:34.480
structural and non structural
interventions simultaneously.

0c464a14-7d2d-46b0-b97b-c2e550667870-0
01:00:34.640 --> 01:00:37.899
And then do a non structural
structural, non structural, like

0c464a14-7d2d-46b0-b97b-c2e550667870-1
01:00:37.899 --> 01:00:41.105
a fully integrated approach
where we can sift through all of

0c464a14-7d2d-46b0-b97b-c2e550667870-2
01:00:41.105 --> 01:00:44.101
these structural and non
structural improvements and say

0c464a14-7d2d-46b0-b97b-c2e550667870-3
01:00:44.101 --> 01:00:46.520
what is resource optimal and
from both sides.

a7f4835d-8b6f-4836-823a-c3ab803bc365-0
01:00:47.400 --> 01:00:50.448
The reason why that didn't
happen here is because the

a7f4835d-8b6f-4836-823a-c3ab803bc365-1
01:00:50.448 --> 01:00:53.552
surrogate model architecture
that I calibrated had the

a7f4835d-8b6f-4836-823a-c3ab803bc365-2
01:00:53.552 --> 01:00:56.826
engineering demand parameters
featurized and then the non

a7f4835d-8b6f-4836-823a-c3ab803bc365-3
01:00:56.826 --> 01:00:59.761
structural component
improvements here, but I never

a7f4835d-8b6f-4836-823a-c3ab803bc365-4
01:00:59.761 --> 01:01:02.640
had an auxiliary model for the
structural changes.

876031f0-7d7f-41cf-8c09-686506515b0c-0
01:01:02.880 --> 01:01:06.672
So to compensate for that, I
simply repeated the optimization

876031f0-7d7f-41cf-8c09-686506515b0c-1
01:01:06.672 --> 01:01:10.160
3 times using three different
structural configurations.

1dc54c02-9cce-44bd-8a46-22270f5ea2b7-0
01:01:10.520 --> 01:01:14.476
If I were to expand on this, I
would totally just put them all

1dc54c02-9cce-44bd-8a46-22270f5ea2b7-1
01:01:14.476 --> 01:01:16.360
together, if that makes sense.

e54037c2-9497-4f77-b071-53ee622122a8-0
01:01:17.720 --> 01:01:18.000
Awesome.

c5b0e006-b685-4cfa-b676-50fab45a1be9-0
01:01:18.000 --> 01:01:19.760
This is this is, this is a good
question.

ce009713-e93b-4e66-92e2-73965547453f-0
01:01:21.560 --> 01:01:26.010
OK, Well, any other questions,
please find Omar after or during

ce009713-e93b-4e66-92e2-73965547453f-1
01:01:26.010 --> 01:01:27.680
lunch or this afternoon.

89484921-8209-4e4c-abb5-a882fe83dc5e-0
01:01:27.680 --> 01:01:33.773
And for those of us joining for
lunch, please meet out by the

89484921-8209-4e4c-abb5-a882fe83dc5e-1
01:01:33.773 --> 01:01:38.000
elevators on the second floor
about 11:50.

20453604-21b7-4813-8dde-4cd6f3062f31-0
01:01:38.560 --> 01:01:40.880
And let's give Omar another
round of applause.