1
00:00:00,120 --> 00:00:02,460
The following content is
provided under a Creative

2
00:00:02,460 --> 00:00:03,880
Commons license.

3
00:00:03,880 --> 00:00:06,090
Your support will help
MIT OpenCourseWare

4
00:00:06,090 --> 00:00:10,180
continue to offer high quality
educational resources for free.

5
00:00:10,180 --> 00:00:12,720
To make a donation or to
view additional materials

6
00:00:12,720 --> 00:00:16,680
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:16,680 --> 00:00:17,880
at ocw.mit.edu.

8
00:00:25,180 --> 00:00:28,000
OLIVIER DE WECK: So
let's get to work.

9
00:00:28,000 --> 00:00:30,730
The V-Model, well,
believe it or not,

10
00:00:30,730 --> 00:00:33,430
but we've arrived today
at the bottom of it,

11
00:00:33,430 --> 00:00:37,660
the pointy part of the V. We're
arriving at the bottom of it,

12
00:00:37,660 --> 00:00:41,260
and the bottom of
it is basically you

13
00:00:41,260 --> 00:00:45,880
driving down the design in all
its glory and all its details.

14
00:00:45,880 --> 00:00:48,220
And we call this
design definition,

15
00:00:48,220 --> 00:00:52,090
and then I want to talk about
multidisciplinary optimization,

16
00:00:52,090 --> 00:00:55,780
including what happens at
the CDR, the critical design

17
00:00:55,780 --> 00:00:56,370
review.

18
00:00:56,370 --> 00:00:57,092
Yeah, go ahead.

19
00:00:57,092 --> 00:00:59,050
AUDIENCE: All right, I
do have a quick question

20
00:00:59,050 --> 00:01:00,910
about A4 based to this.

21
00:01:00,910 --> 00:01:04,629
Do you want a CAD drawing
or a hand drawing?

22
00:01:04,629 --> 00:01:08,790
OLIVIER DE WECK: I'm fine with
a nice dimension hand drawing,

23
00:01:08,790 --> 00:01:13,270
but we're gonna talk
about CAD next week.

24
00:01:13,270 --> 00:01:17,430
So I do think it would be
good to have a CAD drawing,

25
00:01:17,430 --> 00:01:20,950
but you know CAD drawings can
come in very different levels

26
00:01:20,950 --> 00:01:22,420
of detail, right?

27
00:01:22,420 --> 00:01:25,480
So I think it would be
good to have an idea,

28
00:01:25,480 --> 00:01:27,100
if the PDR package--

29
00:01:27,100 --> 00:01:28,960
most of the PDR
packages that you

30
00:01:28,960 --> 00:01:32,230
see that have been submitted
in prior years at CanSat,

31
00:01:32,230 --> 00:01:34,060
they do have a CAD
drawing in there

32
00:01:34,060 --> 00:01:37,037
for their sort of
suggested configuration.

33
00:01:37,037 --> 00:01:38,620
And you know, it is
useful because you

34
00:01:38,620 --> 00:01:42,380
get to see whether the
packaging concept really works.

35
00:01:42,380 --> 00:01:44,200
You know, the sketching,
you can get away

36
00:01:44,200 --> 00:01:46,700
with a lot just sketching.

37
00:01:46,700 --> 00:01:49,060
So my answer would
tend to be yes,

38
00:01:49,060 --> 00:01:51,670
but if absolutely,
you don't feel

39
00:01:51,670 --> 00:01:53,740
like you need it or you
don't want to do it,

40
00:01:53,740 --> 00:01:58,650
then a nice hand drawing with
dimensions will do as well.

41
00:01:58,650 --> 00:02:04,900
OK, so let's talk about
the bottom of the V.

42
00:02:04,900 --> 00:02:06,610
The other thing I'm
gonna cover today

43
00:02:06,610 --> 00:02:09,250
is this idea of
multidisciplinary design

44
00:02:09,250 --> 00:02:12,910
optimization, and so
where does that fit in?

45
00:02:12,910 --> 00:02:16,000
So you remember we talked about
concept synthesis, concept

46
00:02:16,000 --> 00:02:18,520
generation, screening,
and selection,

47
00:02:18,520 --> 00:02:21,430
and then at the PDR
stage, you know,

48
00:02:21,430 --> 00:02:23,590
we picked one architecture.

49
00:02:23,590 --> 00:02:25,510
So today's topic
is really how do

50
00:02:25,510 --> 00:02:29,020
you get this architecture,
which is shown as this square

51
00:02:29,020 --> 00:02:31,360
here to an optimal?

52
00:02:31,360 --> 00:02:34,060
And I put optimal in
quotes here because we're

53
00:02:34,060 --> 00:02:38,440
gonna argue what that means, but
you've refined the design, such

54
00:02:38,440 --> 00:02:40,510
that you can go
present it at the CDR.

55
00:02:40,510 --> 00:02:44,080
So you take this box here,
which is your concept,

56
00:02:44,080 --> 00:02:45,340
and you open the box--

57
00:02:45,340 --> 00:02:46,900
you can see this on the right--

58
00:02:46,900 --> 00:02:49,360
and you really dive
into the details

59
00:02:49,360 --> 00:02:51,880
of the various
subsystems and components

60
00:02:51,880 --> 00:02:53,440
and how they're connected.

61
00:02:53,440 --> 00:02:57,200
And what you're trying to
do is fine tune your design.

62
00:02:57,200 --> 00:02:58,970
So this is shown
here schematically.

63
00:02:58,970 --> 00:03:01,840
Let's say there's a
design variable X1

64
00:03:01,840 --> 00:03:04,090
and you want to find where
is the optimal point?

65
00:03:04,090 --> 00:03:07,090
This point here seems to
be optimal to minimize

66
00:03:07,090 --> 00:03:10,390
metric three and
that's the design

67
00:03:10,390 --> 00:03:11,980
that you're gonna propose.

68
00:03:11,980 --> 00:03:16,120
So today's topic really is
about fine tuning the design

69
00:03:16,120 --> 00:03:20,050
and coming up with a
CDR level design, that's

70
00:03:20,050 --> 00:03:21,790
what we want to talk about.

71
00:03:21,790 --> 00:03:25,040
And so what I'll cover
is different topics.

72
00:03:25,040 --> 00:03:27,580
So first of all, the
NASA view of this.

73
00:03:27,580 --> 00:03:30,280
It's called the design
definition process.

74
00:03:30,280 --> 00:03:33,190
Then I'll talk about MDO,
multidisciplinary design

75
00:03:33,190 --> 00:03:36,730
optimization, the use
of so-called CDFs,

76
00:03:36,730 --> 00:03:39,280
concurrent design
facilities, and then

77
00:03:39,280 --> 00:03:43,220
finally, what happens at CDR.

78
00:03:43,220 --> 00:03:47,120
So the general idea of the
design solution definition

79
00:03:47,120 --> 00:03:49,490
process is that you
start with something

80
00:03:49,490 --> 00:03:51,360
like this on the left.

81
00:03:51,360 --> 00:03:55,310
This is a product breakdown
structure, essentially

82
00:03:55,310 --> 00:03:58,850
your system architecture,
your system decomposition.

83
00:03:58,850 --> 00:04:02,600
And on the right side,
you then translate this

84
00:04:02,600 --> 00:04:04,790
to a detailed design, OK?

85
00:04:04,790 --> 00:04:07,924
So for a launch
vehicle, we have,

86
00:04:07,924 --> 00:04:10,340
you know, the flight segment,
this is what-- actually it's

87
00:04:10,340 --> 00:04:13,820
gonna fly and it has a payload
element, the spacecraft

88
00:04:13,820 --> 00:04:16,980
bus, and then the launch and
the launch accommodation.

89
00:04:16,980 --> 00:04:21,470
So that's the general idea is
you have high-level breakdown,

90
00:04:21,470 --> 00:04:23,630
but then you need to
design every piece.

91
00:04:23,630 --> 00:04:26,720
You need to have
chosen your materials.

92
00:04:26,720 --> 00:04:29,180
You need to have
chosen the dimensions.

93
00:04:29,180 --> 00:04:31,580
You need to understand
the interfaces.

94
00:04:31,580 --> 00:04:34,490
And for every part, you
also need to decide what?

95
00:04:34,490 --> 00:04:37,790
What's a big decision
that you need

96
00:04:37,790 --> 00:04:41,460
to make when you go to
the detailed design,

97
00:04:41,460 --> 00:04:45,850
beside, you know, material,
dimensions, software?

98
00:04:45,850 --> 00:04:50,580
What is always a
big decision to make

99
00:04:50,580 --> 00:04:52,000
when you do a detailed design?

100
00:04:55,860 --> 00:04:56,367
Go ahead.

101
00:04:56,367 --> 00:04:58,325
AUDIENCE: Maybe the
interfaces between the two.

102
00:04:58,325 --> 00:05:00,116
OLIVIER DE WECK: Yep,
interfaces, all that.

103
00:05:00,116 --> 00:05:03,410
What I'm trying
to get at is COTS.

104
00:05:03,410 --> 00:05:06,440
Who knows this acronym, COTS?

105
00:05:06,440 --> 00:05:09,140
Let's see if the, EPFL,
anybody know COTS?

106
00:05:09,140 --> 00:05:09,890
What does it mean?

107
00:05:17,030 --> 00:05:19,742
I think we can't
hear you right now.

108
00:05:19,742 --> 00:05:21,200
AUDIENCE: Commercial
off the shelf.

109
00:05:21,200 --> 00:05:23,330
OLIVIER DE WECK:
Commercial off the shelf.

110
00:05:23,330 --> 00:05:27,200
So you have to decide
in the detailed design

111
00:05:27,200 --> 00:05:32,030
are you going to pick
components from a catalog,

112
00:05:32,030 --> 00:05:36,290
or are you going to custom
design these components, right?

113
00:05:36,290 --> 00:05:39,350
And so one of the
things at the CDR, that

114
00:05:39,350 --> 00:05:41,150
I always ask the question--

115
00:05:41,150 --> 00:05:43,310
so you present your CDR design.

116
00:05:43,310 --> 00:05:47,510
What percentage, what fraction
of your parts at the lowest

117
00:05:47,510 --> 00:05:54,680
level are COTS and what
fraction did you custom design?

118
00:05:54,680 --> 00:05:59,030
Mike, you've done a lot of
these unmanned aerial vehicles

119
00:05:59,030 --> 00:05:59,600
and so forth.

120
00:05:59,600 --> 00:06:01,730
Can you talk a
little bit about, you

121
00:06:01,730 --> 00:06:03,890
know, what's the typical
fraction of COTS,

122
00:06:03,890 --> 00:06:07,480
and how do you make that
decision when you design,

123
00:06:07,480 --> 00:06:08,540
you know, a flying thing?

124
00:06:11,060 --> 00:06:14,510
AUDIENCE: It really depends
on the scale and how robust.

125
00:06:14,510 --> 00:06:16,490
The problem with COTS
in the unmanned world

126
00:06:16,490 --> 00:06:20,422
is that you can easily
go to hobby grade parts.

127
00:06:20,422 --> 00:06:22,880
And so if you want to get into
more of the commercial, kind

128
00:06:22,880 --> 00:06:26,510
of more robust, almost
more like defense

129
00:06:26,510 --> 00:06:29,510
kind of, like,
robustness and you know,

130
00:06:29,510 --> 00:06:31,010
durability, you
really have to start

131
00:06:31,010 --> 00:06:33,230
going to custom designed parts.

132
00:06:33,230 --> 00:06:36,050
Because you know, if you
can do a whole design

133
00:06:36,050 --> 00:06:39,219
space of, you know, all the
COTS parts, but you know,

134
00:06:39,219 --> 00:06:40,760
they're always just
kind of, you now,

135
00:06:40,760 --> 00:06:43,580
Chinese-made electric motors
and batteries, and none of them

136
00:06:43,580 --> 00:06:46,640
that will perform to,
like, commercial spec,

137
00:06:46,640 --> 00:06:50,702
but it just depends, really,
what you need for the vehicle.

138
00:06:50,702 --> 00:06:51,660
OLIVIER DE WECK: I see.

139
00:06:51,660 --> 00:06:54,410
So what you're saying
is there's a gap.

140
00:06:54,410 --> 00:06:58,440
You can get a lot of COTS stuff
at the hobby shop level, right?

141
00:06:58,440 --> 00:07:01,880
So they're pretty affordable
but low reliability.

142
00:07:01,880 --> 00:07:05,990
Or you have really high end,
very expensive, you know,

143
00:07:05,990 --> 00:07:07,120
low production--

144
00:07:07,120 --> 00:07:09,410
AUDIENCE: Like traditional
aerospace grade, yeah.

145
00:07:09,410 --> 00:07:11,160
OLIVIER DE WECK: But there's
not much in the middle.

146
00:07:11,160 --> 00:07:12,010
Is that what you're saying?

147
00:07:12,010 --> 00:07:12,950
AUDIENCE: Yeah, for sure.

148
00:07:12,950 --> 00:07:13,220
Yeah.

149
00:07:13,220 --> 00:07:14,303
OLIVIER DE WECK: OK, good.

150
00:07:14,303 --> 00:07:18,322
So you know, that's the
reality of detailed design

151
00:07:18,322 --> 00:07:19,280
that you're gonna face.

152
00:07:19,280 --> 00:07:23,000
And surprisingly, in kind of
the formal system engineering

153
00:07:23,000 --> 00:07:26,550
methodology, there's not
a lot of guidance on this.

154
00:07:26,550 --> 00:07:28,790
There's not a lot of,
you know, guidance

155
00:07:28,790 --> 00:07:31,010
to say what's the
optimal fraction

156
00:07:31,010 --> 00:07:33,260
of commercial off
the shelf parts?

157
00:07:33,260 --> 00:07:37,070
This is really something you
have to work out yourself,

158
00:07:37,070 --> 00:07:40,790
and depending on, you know,
whether you're in aerospace

159
00:07:40,790 --> 00:07:47,990
or medical devices or automotive
or making toothpaste, there's--

160
00:07:47,990 --> 00:07:49,580
actually, yesterday
there was also

161
00:07:49,580 --> 00:07:51,260
a big case study on diapers.

162
00:07:51,260 --> 00:07:54,740
Diapers is a $3 billion
business, believe it or not.

163
00:07:54,740 --> 00:07:57,275
And they were going
for-- their solution

164
00:07:57,275 --> 00:08:01,550
A was to go for a COTS solution,
you know, a lot of vendors.

165
00:08:01,550 --> 00:08:03,500
And then they said, no.

166
00:08:03,500 --> 00:08:05,990
It's just gonna give
us, sort of, average,

167
00:08:05,990 --> 00:08:07,340
mediocre performance.

168
00:08:07,340 --> 00:08:08,990
We want something much better.

169
00:08:08,990 --> 00:08:12,260
So they went for a much
higher custom design solution.

170
00:08:12,260 --> 00:08:14,650
Took them took them
a while to do it,

171
00:08:14,650 --> 00:08:16,290
but it was worth it in the end.

172
00:08:16,290 --> 00:08:19,160
So this is a big decision
you have to make.

173
00:08:19,160 --> 00:08:22,050
So the point here
being, at the end,

174
00:08:22,050 --> 00:08:23,517
you have a very detailed design.

175
00:08:23,517 --> 00:08:24,350
You have blueprints.

176
00:08:24,350 --> 00:08:26,906
You have something you
can actually build.

177
00:08:26,906 --> 00:08:32,840
So you know, this is a list of
all the things you need to do.

178
00:08:32,840 --> 00:08:36,841
You define the solution space,
develop your alternatives.

179
00:08:36,841 --> 00:08:38,299
All the things
we've already talked

180
00:08:38,299 --> 00:08:41,360
about at the conceptual level,
but you do this at every-- you

181
00:08:41,360 --> 00:08:43,549
do it at the subsystem
level, and even

182
00:08:43,549 --> 00:08:45,200
at the component level.

183
00:08:45,200 --> 00:08:46,670
What are the alternate designs?

184
00:08:46,670 --> 00:08:48,760
Cost performance and schedule.

185
00:08:48,760 --> 00:08:51,050
Select your design
solution, drive it down

186
00:08:51,050 --> 00:08:54,410
to the lowest level of
detail, and then identify

187
00:08:54,410 --> 00:08:57,890
the enabling products, which
would be like support equipment

188
00:08:57,890 --> 00:08:59,460
and things like this.

189
00:08:59,460 --> 00:09:02,690
So the best illustration
I can give you

190
00:09:02,690 --> 00:09:07,550
why this matters is these two
pictures I'm gonna show you.

191
00:09:07,550 --> 00:09:10,040
The top picture
that you see here,

192
00:09:10,040 --> 00:09:14,420
this is a picture that was drawn
in the '70s to sell the space

193
00:09:14,420 --> 00:09:16,640
shuttle program to US Congress.

194
00:09:16,640 --> 00:09:18,950
This was the vision, right?

195
00:09:18,950 --> 00:09:21,500
This is the orbiter,
and you can see

196
00:09:21,500 --> 00:09:25,250
it's sitting there in this
pristine hangar, right, very

197
00:09:25,250 --> 00:09:28,820
clean, shiny floors,
maybe four or five people.

198
00:09:28,820 --> 00:09:31,550
The payload bay is open.

199
00:09:31,550 --> 00:09:35,180
There's, like, one or two carts
of ground support equipment.

200
00:09:35,180 --> 00:09:37,970
You just load up your new
satellite, close the doors,

201
00:09:37,970 --> 00:09:40,250
and off you go for
your next mission.

202
00:09:40,250 --> 00:09:42,770
That's a reusable vehicle.

203
00:09:42,770 --> 00:09:49,420
That's what we wanted, and then
what we actually got is this.

204
00:09:49,420 --> 00:09:52,900
You can't even see
the orbiter, right?

205
00:09:52,900 --> 00:09:54,550
But I promise you,
it's in there.

206
00:09:54,550 --> 00:09:56,830
It's hidden in that scaffolding.

207
00:09:56,830 --> 00:09:58,570
So this was the--
it's retired now,

208
00:09:58,570 --> 00:10:01,750
but this was the orbiter
processing facility at the NASA

209
00:10:01,750 --> 00:10:03,640
Kennedy Space Center.

210
00:10:03,640 --> 00:10:07,750
And the reality was
that to turn around

211
00:10:07,750 --> 00:10:12,190
an orbiter between flights
was almost like rebuilding it.

212
00:10:12,190 --> 00:10:16,990
And the two subsystems that
required the most person hours

213
00:10:16,990 --> 00:10:19,780
of work were the main engine--

214
00:10:19,780 --> 00:10:23,200
the main engine basically had
to be disassembled and inspected

215
00:10:23,200 --> 00:10:24,045
and so forth--

216
00:10:24,045 --> 00:10:25,295
and what was the other system?

217
00:10:27,970 --> 00:10:28,974
What's that?

218
00:10:28,974 --> 00:10:29,890
AUDIENCE: Heat shield.

219
00:10:29,890 --> 00:10:31,880
OLIVIER DE WECK: The
heat shield, the TPS,

220
00:10:31,880 --> 00:10:33,890
thermal protection system--

221
00:10:33,890 --> 00:10:35,160
if you've ever seen--

222
00:10:35,160 --> 00:10:38,350
who's seen the shuttle close up?

223
00:10:38,350 --> 00:10:40,420
OK, about half, half here.

224
00:10:40,420 --> 00:10:42,700
At EPFL, who's had a
chance to see the space

225
00:10:42,700 --> 00:10:45,470
shuttle, the orbiter, close up?

226
00:10:45,470 --> 00:10:48,850
Anybody?

227
00:10:48,850 --> 00:10:51,730
Go ahead.

228
00:10:51,730 --> 00:10:53,174
AUDIENCE: One or two.

229
00:10:53,174 --> 00:10:54,340
OLIVIER DE WECK: One or two.

230
00:10:54,340 --> 00:10:55,900
So those of you
that have seen it,

231
00:10:55,900 --> 00:10:58,300
what did you notice
about the heat shield?

232
00:10:58,300 --> 00:11:00,520
Did you notice anything
about the protection,

233
00:11:00,520 --> 00:11:03,927
the thermal protection system?

234
00:11:03,927 --> 00:11:04,510
Who's seen it?

235
00:11:08,820 --> 00:11:12,182
AUDIENCE: We do not
hear very well here.

236
00:11:12,182 --> 00:11:13,640
OLIVIER DE WECK:
You can't hear me?

237
00:11:18,580 --> 00:11:21,680
So the point I want
to make here is

238
00:11:21,680 --> 00:11:24,710
that the thermal protection
system, every tile

239
00:11:24,710 --> 00:11:25,880
is different.

240
00:11:25,880 --> 00:11:30,480
Every tile has a different
shape, a different curvature,

241
00:11:30,480 --> 00:11:32,810
and there were good
thermodynamic reasons,

242
00:11:32,810 --> 00:11:36,060
aerothermodynamic
reasons for that,

243
00:11:36,060 --> 00:11:39,410
but in terms of actually
designing, manufacturing,

244
00:11:39,410 --> 00:11:43,340
and turning around the vehicle,
it was a nightmare, right?

245
00:11:43,340 --> 00:11:47,330
So that it turned out to be
not a very reusable vehicle

246
00:11:47,330 --> 00:11:48,350
in the end, and why?

247
00:11:48,350 --> 00:11:50,180
Because a lot of
the detailed design

248
00:11:50,180 --> 00:11:53,870
decisions ended
up being critical,

249
00:11:53,870 --> 00:11:58,820
in terms of not really
giving us what we wanted,

250
00:11:58,820 --> 00:12:01,980
in terms of macro level
properties of the vehicle.

251
00:12:01,980 --> 00:12:04,880
So detailed design
really matters.

252
00:12:04,880 --> 00:12:07,040
Did you get that point, at EPFL?

253
00:12:07,040 --> 00:12:07,610
Is it clear?

254
00:12:13,640 --> 00:12:14,309
AUDIENCE: It is.

255
00:12:14,309 --> 00:12:15,600
OLIVIER DE WECK: OK, thank you.

256
00:12:15,600 --> 00:12:17,680
So here's some--

257
00:12:17,680 --> 00:12:19,890
I'm not going to go
through this in detail,

258
00:12:19,890 --> 00:12:23,910
but here's some flow diagrams,
process flow diagrams,

259
00:12:23,910 --> 00:12:26,040
from the System
Engineering Handbook

260
00:12:26,040 --> 00:12:29,090
that essentially describe,
you know, how this works.

261
00:12:29,090 --> 00:12:34,530
So your inputs are your baseline
logical decomposition model,

262
00:12:34,530 --> 00:12:37,620
which is essentially your
system architecture, right?

263
00:12:37,620 --> 00:12:41,190
And then your baseline derived,
your detailed technical

264
00:12:41,190 --> 00:12:42,630
requirements.

265
00:12:42,630 --> 00:12:44,779
You go through all
these processes,

266
00:12:44,779 --> 00:12:46,320
through all these
steps, and then you

267
00:12:46,320 --> 00:12:48,360
end up with a whole
bunch of outputs, which

268
00:12:48,360 --> 00:12:52,620
are more detailed requirements.

269
00:12:52,620 --> 00:12:54,380
The specifications,
do you remember,

270
00:12:54,380 --> 00:12:58,200
we teased apart requirements
and specifications?

271
00:12:58,200 --> 00:13:00,930
Requirements are what
the system should do.

272
00:13:00,930 --> 00:13:04,470
Requirement specifications are
how the system's actually been

273
00:13:04,470 --> 00:13:06,600
designed, right, blueprints.

274
00:13:06,600 --> 00:13:10,270
So that's specifications
at the subsystem level.

275
00:13:10,270 --> 00:13:13,590
And then you get other things,
like a verification, validation

276
00:13:13,590 --> 00:13:16,870
plan, and then logistics
and operating procedures.

277
00:13:16,870 --> 00:13:19,970
So very, very detailed work.

278
00:13:19,970 --> 00:13:22,120
The other thing that's
important is all the design

279
00:13:22,120 --> 00:13:24,970
considerations that go
into making detailed design

280
00:13:24,970 --> 00:13:25,580
decisions.

281
00:13:25,580 --> 00:13:28,390
So here's a fairly
long list of them,

282
00:13:28,390 --> 00:13:32,290
and you can group them into
performance, availability.

283
00:13:32,290 --> 00:13:34,480
This is more related
to reliability.

284
00:13:34,480 --> 00:13:38,710
So you could have a system that
has high nominal performance

285
00:13:38,710 --> 00:13:41,650
but is poor reliability,
right, so it's not available

286
00:13:41,650 --> 00:13:42,190
very much.

287
00:13:42,190 --> 00:13:43,390
That's not good.

288
00:13:43,390 --> 00:13:46,420
We want high performance
and high availability.

289
00:13:46,420 --> 00:13:49,000
That gives you your
technical effectiveness.

290
00:13:49,000 --> 00:13:51,280
And then how do you
actually operate the system?

291
00:13:51,280 --> 00:13:54,820
So operations,
maintenance, logistics,

292
00:13:54,820 --> 00:13:56,320
so that's process efficiency.

293
00:13:56,320 --> 00:13:59,107
Together, that gives you
system effectiveness.

294
00:13:59,107 --> 00:14:00,940
And then, of course,
you have the total cost

295
00:14:00,940 --> 00:14:03,820
of operating the system--
lifecycle cost, total ownership

296
00:14:03,820 --> 00:14:04,660
cost.

297
00:14:04,660 --> 00:14:06,610
And then system
effectiveness and cost

298
00:14:06,610 --> 00:14:09,610
together, give you what's called
here affordable operational

299
00:14:09,610 --> 00:14:10,960
effectiveness.

300
00:14:10,960 --> 00:14:14,530
So I would argue, in terms
of the space shuttle program,

301
00:14:14,530 --> 00:14:17,170
it had high system performance.

302
00:14:17,170 --> 00:14:21,175
It had OK system availability.

303
00:14:21,175 --> 00:14:23,550
They had-- definitely there
were some reliability issues,

304
00:14:23,550 --> 00:14:26,980
but it was manageable, but
it had very poor process

305
00:14:26,980 --> 00:14:28,110
efficiency--

306
00:14:28,110 --> 00:14:31,750
operations, maintenance,
logistics very, very difficult.

307
00:14:31,750 --> 00:14:35,830
So system effectiveness wasn't
that high, and as a result,

308
00:14:35,830 --> 00:14:38,580
the lifecycle cost
was also very high.

309
00:14:38,580 --> 00:14:41,400
There's a very
famous paper, where

310
00:14:41,400 --> 00:14:43,740
after the shuttle
program was ended,

311
00:14:43,740 --> 00:14:46,680
a lifecycle cost
analysis was done,

312
00:14:46,680 --> 00:14:52,080
and it turned out that one
launch cost about $1.5 billion.

313
00:14:52,080 --> 00:14:53,880
So during the program,
people would say,

314
00:14:53,880 --> 00:14:56,640
oh, one launch is
maybe $500 million,

315
00:14:56,640 --> 00:15:00,510
you know, the variable cost,
and the rest is fixed cost.

316
00:15:00,510 --> 00:15:03,270
But once the program is done,
it doesn't matter anymore

317
00:15:03,270 --> 00:15:05,160
whether it was fixed
or variable cost.

318
00:15:05,160 --> 00:15:08,370
All you know is what was the
total cost of the program,

319
00:15:08,370 --> 00:15:10,560
how many launches
did we actually do,

320
00:15:10,560 --> 00:15:12,180
and you just take the ratio.

321
00:15:12,180 --> 00:15:16,200
And it was $1.5 billion
per launch of the shuttle.

322
00:15:16,200 --> 00:15:18,720
And so I would say
that's way off from what

323
00:15:18,720 --> 00:15:20,130
the original target was.

324
00:15:20,130 --> 00:15:22,110
Yeah, Veronica?

325
00:15:22,110 --> 00:15:24,550
AUDIENCE: Does that number
account for crew and crew

326
00:15:24,550 --> 00:15:25,050
training?

327
00:15:27,539 --> 00:15:28,830
OLIVIER DE WECK: Good question.

328
00:15:28,830 --> 00:15:29,670
I think so.

329
00:15:29,670 --> 00:15:33,060
I think it basically-- whatever
was appropriated by Congress

330
00:15:33,060 --> 00:15:35,580
for the shuttle
program, and I think

331
00:15:35,580 --> 00:15:41,190
that would include crew
operations, as well, yeah.

332
00:15:41,190 --> 00:15:43,790
So I'm happy to point you
to that paper, actually.

333
00:15:43,790 --> 00:15:44,330
Yes?

334
00:15:44,330 --> 00:15:46,719
AUDIENCE: It also accounted
for R&D costs as well, right?

335
00:15:46,719 --> 00:15:48,260
OLIVIER DE WECK:
Yes, that's correct.

336
00:15:48,260 --> 00:15:50,000
It includes everything, all in.

337
00:15:50,000 --> 00:15:54,400
R&D costs-- R&D
took about 10 years

338
00:15:54,400 --> 00:15:56,540
and then operations
was about 30 years.

339
00:15:59,230 --> 00:16:01,080
OK, so you see--

340
00:16:01,080 --> 00:16:04,570
and of course, within
these are budget limits.

341
00:16:04,570 --> 00:16:08,280
So part of the reason
why the shuttle didn't

342
00:16:08,280 --> 00:16:10,140
have that high
process efficiency

343
00:16:10,140 --> 00:16:14,520
was because the R&D budget
was capped pretty tightly.

344
00:16:14,520 --> 00:16:16,470
So rather than
optimizing, you know,

345
00:16:16,470 --> 00:16:19,470
for maintainability
and turnaround,

346
00:16:19,470 --> 00:16:22,560
they said, we need to have
the system performance.

347
00:16:22,560 --> 00:16:24,960
We need to have reasonable
system availability,

348
00:16:24,960 --> 00:16:29,230
and then the rest, basically, is
we're gonna deal with it later.

349
00:16:29,230 --> 00:16:32,160
So what effectively
happened is they capped

350
00:16:32,160 --> 00:16:36,900
R&D cost, and as a result,
the lifecycle cost ballooned.

351
00:16:36,900 --> 00:16:38,400
I would argue if
they had spent, you

352
00:16:38,400 --> 00:16:41,610
know, two or three
more years optimizing,

353
00:16:41,610 --> 00:16:43,200
you know, maintenance
and turn around,

354
00:16:43,200 --> 00:16:47,160
really making it reusable,
the lifecycle costs could

355
00:16:47,160 --> 00:16:52,190
have been much better, but
that's not how it worked.

356
00:16:52,190 --> 00:16:55,580
So you have a lot of these
competing requirements.

357
00:16:55,580 --> 00:16:57,770
And so I want to
just ask you this--

358
00:16:57,770 --> 00:17:00,800
this is the first concept
question today, OK?

359
00:17:00,800 --> 00:17:03,920
So imagine you're-- this
is a very detailed design.

360
00:17:03,920 --> 00:17:07,109
This is just a bracket here, OK?

361
00:17:07,109 --> 00:17:11,990
And this bracket was designed
so it has, like, a tip load.

362
00:17:11,990 --> 00:17:15,109
Think of this as
a cantilever beam.

363
00:17:15,109 --> 00:17:18,510
So it's attached on the
right side or left side here.

364
00:17:18,510 --> 00:17:21,650
You have a load, and then
this is your displacement

365
00:17:21,650 --> 00:17:24,140
under a given load.

366
00:17:24,140 --> 00:17:28,250
And these three brackets
all have the same mass.

367
00:17:28,250 --> 00:17:31,970
They use the same
amount of material, OK?

368
00:17:31,970 --> 00:17:36,980
So you can think of this as you
get from the very simple design

369
00:17:36,980 --> 00:17:39,830
to the more intricate design,
you have more design freedom.

370
00:17:39,830 --> 00:17:41,150
So you get better performance.

371
00:17:41,150 --> 00:17:44,660
You see the
displacement goes down,

372
00:17:44,660 --> 00:17:46,710
but you also have
more complexity.

373
00:17:46,710 --> 00:17:48,680
So it's more
difficult to optimize.

374
00:17:48,680 --> 00:17:51,900
It's more difficult to
manufacture and so forth.

375
00:17:51,900 --> 00:17:55,940
And so we can actually show
these three choices here

376
00:17:55,940 --> 00:17:58,820
on a manufacturing
cost, per unit cost,

377
00:17:58,820 --> 00:18:01,760
versus the displacement,
structural displacement

378
00:18:01,760 --> 00:18:03,290
under load chart.

379
00:18:03,290 --> 00:18:05,660
So our simple design,
our one bar design,

380
00:18:05,660 --> 00:18:09,440
has a displacement of 2.5
millimeters under load,

381
00:18:09,440 --> 00:18:13,090
and it costs about $2 to make.

382
00:18:13,090 --> 00:18:16,160
The two-bar design has
much better performance,

383
00:18:16,160 --> 00:18:20,990
like 0.75 or something like
that, costs about $3 to make.

384
00:18:20,990 --> 00:18:23,570
And then this 17-bar
intricate design

385
00:18:23,570 --> 00:18:28,670
has 0.6 millimeters, so the
best structural performance,

386
00:18:28,670 --> 00:18:31,760
but its cost is
about $8 per piece.

387
00:18:31,760 --> 00:18:34,910
So my question to you is
which of these three designs

388
00:18:34,910 --> 00:18:39,210
would you select and why?

389
00:18:39,210 --> 00:18:44,700
So if you could go to that
link, enter your answer,

390
00:18:44,700 --> 00:18:49,300
and then we'll look
at the distribution.

391
00:18:49,300 --> 00:18:51,180
So we have 39 responses.

392
00:18:51,180 --> 00:18:52,750
That's good.

393
00:18:52,750 --> 00:18:53,930
Ah, interesting.

394
00:18:53,930 --> 00:18:56,700
OK.

395
00:18:56,700 --> 00:18:59,430
So one of you said
you're gonna go

396
00:18:59,430 --> 00:19:03,940
for the super-duper
17-bar design.

397
00:19:03,940 --> 00:19:08,800
Most of you, 74%, said
the two-bar design.

398
00:19:08,800 --> 00:19:11,410
And one of you said one bar.

399
00:19:11,410 --> 00:19:15,040
And then eight of you,
20%, said you're not sure.

400
00:19:15,040 --> 00:19:17,830
I guess you need more
information, right?

401
00:19:17,830 --> 00:19:22,930
So who picked the one-bar
design, the simple design?

402
00:19:22,930 --> 00:19:28,312
OK, so please explain
why you picked that.

403
00:19:28,312 --> 00:19:29,770
AUDIENCE: Well,
for my application.

404
00:19:29,770 --> 00:19:31,570
I'm planning on
mass producing them,

405
00:19:31,570 --> 00:19:33,280
and it's for
something that doesn't

406
00:19:33,280 --> 00:19:38,110
need the degree of displacement
that the other ones are doing,

407
00:19:38,110 --> 00:19:43,180
and so I'm really only seeking
to minimize manufacturing cost,

408
00:19:43,180 --> 00:19:46,180
and the one bar is the
easiest to manufacture

409
00:19:46,180 --> 00:19:47,910
and also gives me
room down the road

410
00:19:47,910 --> 00:19:50,490
to iterate if it's not working.

411
00:19:50,490 --> 00:19:53,430
OLIVIER DE WECK: OK, so So what
would be an example, industry

412
00:19:53,430 --> 00:19:54,280
or application?

413
00:19:54,280 --> 00:19:56,380
AUDIENCE: Like a
children's toy, perhaps,

414
00:19:56,380 --> 00:20:04,019
where you don't need very strict
requirements on displacement,

415
00:20:04,019 --> 00:20:05,560
something that you're
mass producing.

416
00:20:05,560 --> 00:20:07,685
OLIVIER DE WECK: Something
mass produced, you know,

417
00:20:07,685 --> 00:20:10,740
for sort of lower requirements.

418
00:20:10,740 --> 00:20:14,170
OK, I think that's
a valid answer.

419
00:20:14,170 --> 00:20:16,390
Most of you said
the two-bar design.

420
00:20:16,390 --> 00:20:17,800
So who said two bars?

421
00:20:17,800 --> 00:20:21,190
Let's see at EPFL, who
picked the two-bar design?

422
00:20:25,700 --> 00:20:26,400
AUDIENCE: Sure.

423
00:20:26,400 --> 00:20:30,550
It gives you the biggest
biggest bang for your buck.

424
00:20:30,550 --> 00:20:35,600
For not that much more cost, you
get a much more sturdy design

425
00:20:35,600 --> 00:20:39,440
and that's compared
for the 17-bar one.

426
00:20:39,440 --> 00:20:41,420
Really, it's still
pretty sturdy,

427
00:20:41,420 --> 00:20:44,120
but you're paying much
less, so it seems optimal.

428
00:20:44,120 --> 00:20:47,390
OLIVIER DE WECK: Yeah, so if
you think about this in a Pareto

429
00:20:47,390 --> 00:20:50,870
sense, I mean, they're
all Pareto-optimal,

430
00:20:50,870 --> 00:20:53,330
but this one is at the
knee of the curve, right?

431
00:20:53,330 --> 00:20:57,310
So you're getting, you said
bang for your buck or value.

432
00:20:57,310 --> 00:21:01,970
If you did, like, a displacement
or one over displacement,

433
00:21:01,970 --> 00:21:07,040
I guess, per manufacturing cost,
this one would do pretty well.

434
00:21:07,040 --> 00:21:10,760
So that's also a
very sensible answer.

435
00:21:10,760 --> 00:21:16,740
Who picked the 17-bar design,
the really intricate one?

436
00:21:16,740 --> 00:21:17,640
Who picked that?

437
00:21:17,640 --> 00:21:20,910
Oh, I think I might have picked
that when I tried out the form.

438
00:21:20,910 --> 00:21:25,090
[LAUGHS] I think so.

439
00:21:25,090 --> 00:21:26,700
I can't remember now.

440
00:21:26,700 --> 00:21:30,710
It's been already a while ago.

441
00:21:30,710 --> 00:21:34,140
So if, indeed, I was the one,
so how would I justify this?

442
00:21:34,140 --> 00:21:37,830
OK, so here this
bracket will go--

443
00:21:41,360 --> 00:21:45,060
this will go on the
Mars 2020 Rover.

444
00:21:45,060 --> 00:21:50,280
And the Rover will see a lot
of different thermal gradients.

445
00:21:50,280 --> 00:21:53,160
You know, it has
to be very precise

446
00:21:53,160 --> 00:21:55,560
because it holds
scientific instruments that

447
00:21:55,560 --> 00:21:58,275
need to be, you know,
pointing very accurately.

448
00:22:01,020 --> 00:22:03,210
Even though the mass here
is identical, but I think

449
00:22:03,210 --> 00:22:04,110
we could shrink this.

450
00:22:04,110 --> 00:22:06,840
You know, we get the best
performance structurally

451
00:22:06,840 --> 00:22:09,870
for the mass constraint
that we were allocated.

452
00:22:09,870 --> 00:22:12,770
And cost, you know, this
is all taxpayer dollars,

453
00:22:12,770 --> 00:22:15,510
so this is a government program.

454
00:22:15,510 --> 00:22:17,220
So we can spend the money.

455
00:22:17,220 --> 00:22:19,860
Of course, I don't
mean that really,

456
00:22:19,860 --> 00:22:22,820
but it is a high-performance
application.

457
00:22:22,820 --> 00:22:26,120
Therefore, even though
the benefit is only small,

458
00:22:26,120 --> 00:22:28,330
that small benefit
is worth it when

459
00:22:28,330 --> 00:22:29,580
you look at the whole mission.

460
00:22:29,580 --> 00:22:31,950
So we're we're going
to pay for that,

461
00:22:31,950 --> 00:22:35,410
and we're going to go for
this high-performance design.

462
00:22:35,410 --> 00:22:36,900
Do you see how that works?

463
00:22:36,900 --> 00:22:38,900
So the right answer--

464
00:22:38,900 --> 00:22:41,780
and I think a lot of you said,
you know, you're not sure,

465
00:22:41,780 --> 00:22:43,250
and it's because--

466
00:22:43,250 --> 00:22:45,930
who said you're not sure?

467
00:22:45,930 --> 00:22:47,779
Nathan.

468
00:22:47,779 --> 00:22:49,320
AUDIENCE: I said
not sure because you

469
00:22:49,320 --> 00:22:50,790
don't know all
the specifications

470
00:22:50,790 --> 00:22:53,790
or what the tolerances are
or any other information that

471
00:22:53,790 --> 00:22:55,532
would lead you to
make that decision.

472
00:22:55,532 --> 00:22:57,240
OLIVIER DE WECK: Right,
so you know, what

473
00:22:57,240 --> 00:22:58,500
are the requirements, right?

474
00:22:58,500 --> 00:23:00,000
We don't know the
requirements here.

475
00:23:00,000 --> 00:23:02,520
Are the requirements shall
or should requirements?

476
00:23:02,520 --> 00:23:05,520
If it's a shall
requirement, what it means

477
00:23:05,520 --> 00:23:09,135
is there would be some
hard limit here, right?

478
00:23:09,135 --> 00:23:12,570
The displacement shall be
less than one millimeter.

479
00:23:12,570 --> 00:23:16,530
If that's a hard requirement,
then this solution falls out,

480
00:23:16,530 --> 00:23:17,850
does not satisfy.

481
00:23:17,850 --> 00:23:19,770
And if it's a
should requirement,

482
00:23:19,770 --> 00:23:22,410
then it's a goal that
you're gonna trade off.

483
00:23:22,410 --> 00:23:24,750
The point I want to
make here, this lecture

484
00:23:24,750 --> 00:23:26,460
is about detailed design.

485
00:23:26,460 --> 00:23:28,920
This is detailed design, right?

486
00:23:28,920 --> 00:23:32,400
And you're gonna make hundreds
or thousands of detailed design

487
00:23:32,400 --> 00:23:35,580
decisions just like this
bracket and together,

488
00:23:35,580 --> 00:23:37,890
they're gonna give you, you
know, the performance cost

489
00:23:37,890 --> 00:23:40,140
of your whole system.

490
00:23:40,140 --> 00:23:42,330
And so every detailed
design decision

491
00:23:42,330 --> 00:23:46,020
you make needs to be informed
by the requirements but also

492
00:23:46,020 --> 00:23:49,590
the trade-offs between
these properties.

493
00:23:49,590 --> 00:23:51,520
Does that make sense?

494
00:23:51,520 --> 00:23:52,980
OK.

495
00:23:52,980 --> 00:23:55,050
So and that leads
us-- and so you

496
00:23:55,050 --> 00:23:57,240
want to make sure you
optimize your system.

497
00:23:57,240 --> 00:23:58,680
You want to make sure you get--

498
00:23:58,680 --> 00:24:00,900
and I liked, Katya, the
way you said it-- get

499
00:24:00,900 --> 00:24:02,280
the bang for your buck, right?

500
00:24:02,280 --> 00:24:07,590
You want to make sure you don't
have a lot of inefficiencies

501
00:24:07,590 --> 00:24:08,470
in your design.

502
00:24:08,470 --> 00:24:10,290
So that's what MDO is about.

503
00:24:10,290 --> 00:24:14,860
So let me go to MDO
here, what it is.

504
00:24:14,860 --> 00:24:18,900
So there's a technical committee
in AIAA, the American Institute

505
00:24:18,900 --> 00:24:22,340
for Aeronautics
and Astronautics,

506
00:24:22,340 --> 00:24:25,620
that was founded
about 20 years ago.

507
00:24:25,620 --> 00:24:28,780
So it's defined as an evolving
methodology, a body of methods,

508
00:24:28,780 --> 00:24:32,430
techniques, algorithms,
and related application

509
00:24:32,430 --> 00:24:35,040
practices for the design
of engineering systems,

510
00:24:35,040 --> 00:24:37,650
coupled by physical
phenomena and involving

511
00:24:37,650 --> 00:24:40,740
many interacting
subsystems and parts.

512
00:24:40,740 --> 00:24:43,050
So the main things
you need to do MDO

513
00:24:43,050 --> 00:24:48,330
are a mathematical model of the
system, analysis capability,

514
00:24:48,330 --> 00:24:50,850
and in some cases,
approximation concepts.

515
00:24:50,850 --> 00:24:52,500
So what that means is not--

516
00:24:52,500 --> 00:24:55,500
if you put all this together,
it could be very, very slow

517
00:24:55,500 --> 00:24:56,780
and computationally heavy.

518
00:24:56,780 --> 00:24:59,250
So you might approximate
some of these things,

519
00:24:59,250 --> 00:25:01,560
and I'll explain
a couple examples.

520
00:25:01,560 --> 00:25:05,490
Sensitivity analysis,
optimization processes,

521
00:25:05,490 --> 00:25:08,670
and a human interface-- those
are the main ingredients

522
00:25:08,670 --> 00:25:09,240
of MDO.

523
00:25:12,250 --> 00:25:14,610
Who's seen this chart before?

524
00:25:14,610 --> 00:25:16,800
OK, about three or four of you.

525
00:25:16,800 --> 00:25:23,030
Who's seen this chart before at
EPFL, this particular graphic?

526
00:25:23,030 --> 00:25:26,510
Anybody seen this before?

527
00:25:26,510 --> 00:25:27,890
AUDIENCE: Apparently, no one.

528
00:25:27,890 --> 00:25:29,514
OLIVIER DE WECK: OK,
so let me explain.

529
00:25:29,514 --> 00:25:33,590
This is a cartoon, OK, and
I've experienced this myself

530
00:25:33,590 --> 00:25:36,440
working at McDonnell
Douglas in the '90s.

531
00:25:36,440 --> 00:25:40,430
So each of these
airplanes shown here

532
00:25:40,430 --> 00:25:42,500
is what the airplane
would look like

533
00:25:42,500 --> 00:25:46,400
if that particular
discipline or group could

534
00:25:46,400 --> 00:25:49,160
call all the shots, OK?

535
00:25:49,160 --> 00:25:52,720
So let's look at
a couple of these.

536
00:25:52,720 --> 00:25:56,890
So the controls group is
all about controllability.

537
00:25:56,890 --> 00:26:02,030
So you see the flight control
surfaces are very big.

538
00:26:02,030 --> 00:26:03,860
Now, the problem
is the linkages,

539
00:26:03,860 --> 00:26:07,010
you know, these control
linkages are on the outside.

540
00:26:07,010 --> 00:26:08,870
That's really bad
for aerodynamics,

541
00:26:08,870 --> 00:26:14,260
but the controls people say,
I don't care about drag.

542
00:26:14,260 --> 00:26:17,090
I need to make sure my
control surfaces are

543
00:26:17,090 --> 00:26:18,790
deflecting appropriately.

544
00:26:18,790 --> 00:26:20,480
Here's the hydraulics
group, right?

545
00:26:20,480 --> 00:26:24,470
It's all about the
hydraulic pump, the filters,

546
00:26:24,470 --> 00:26:26,480
and getting hydraulic
pressure to all

547
00:26:26,480 --> 00:26:28,040
the different subsystems.

548
00:26:28,040 --> 00:26:30,650
You know, the big engine
here, the power plant group,

549
00:26:30,650 --> 00:26:32,990
the engine is the
most important thing.

550
00:26:32,990 --> 00:26:35,090
The loft group-- loft means--

551
00:26:35,090 --> 00:26:37,100
lofting is surfaces.

552
00:26:37,100 --> 00:26:39,630
You're dealing with
the shape of surfaces.

553
00:26:39,630 --> 00:26:41,030
So let's keep this simple.

554
00:26:41,030 --> 00:26:43,310
So we have a little,
you know, balsa airplane

555
00:26:43,310 --> 00:26:48,500
with little straight
surfaces, very easy for loft.

556
00:26:48,500 --> 00:26:50,770
Production is kind of similar.

557
00:26:50,770 --> 00:26:55,490
The aerodynamics group wants
a very nice high aspect ratio

558
00:26:55,490 --> 00:26:59,120
wing, very thin, very low drag.

559
00:26:59,120 --> 00:27:02,810
Problem is, there's no room
for passengers or cargo here.

560
00:27:02,810 --> 00:27:05,360
You see this, the stress group?

561
00:27:05,360 --> 00:27:08,950
They have a nice wing here
in the shape of an I-beam.

562
00:27:08,950 --> 00:27:11,930
I-beams are great, you know,
this is good if you're a civil

563
00:27:11,930 --> 00:27:14,390
engineer but not so much--

564
00:27:14,390 --> 00:27:16,800
none of these things
would work on their own.

565
00:27:16,800 --> 00:27:19,400
So what we really
want is something

566
00:27:19,400 --> 00:27:23,540
like this, where all
these considerations are

567
00:27:23,540 --> 00:27:28,340
blended and optimized so we get
a high-performance system not

568
00:27:28,340 --> 00:27:30,620
just a high-performance
subsystem.

569
00:27:30,620 --> 00:27:34,250
Said another way, a
high-performance system

570
00:27:34,250 --> 00:27:38,720
is not the same as just a
sum of individually optimized

571
00:27:38,720 --> 00:27:40,100
subsystems.

572
00:27:40,100 --> 00:27:44,010
That's, in a nutshell,
what MDO is about.

573
00:27:44,010 --> 00:27:47,000
So one of the readings
for today, post readings,

574
00:27:47,000 --> 00:27:49,415
is this paper, 6A.

575
00:27:49,415 --> 00:27:53,420
And this is based on a big
workshop we had in Germany

576
00:27:53,420 --> 00:27:56,302
about five years ago.

577
00:27:56,302 --> 00:27:57,260
That's actually longer.

578
00:27:57,260 --> 00:27:58,850
It's probably seven
years ago by now,

579
00:27:58,850 --> 00:28:01,190
but the paper came
out five years ago

580
00:28:01,190 --> 00:28:05,074
called, "MDO, Assessment and
Direction for Advancement."

581
00:28:05,074 --> 00:28:06,740
And this is a pretty
international group

582
00:28:06,740 --> 00:28:08,480
that put this together.

583
00:28:08,480 --> 00:28:11,360
And what we do in the paper,
we describe the history

584
00:28:11,360 --> 00:28:14,690
of multidisciplinary
optimization, where

585
00:28:14,690 --> 00:28:18,330
it is today, where it's heading,
and some of the key milestones.

586
00:28:18,330 --> 00:28:23,240
So this chart is essentially
a kind of history of MDO.

587
00:28:23,240 --> 00:28:25,520
I'm just gonna point
out a few things.

588
00:28:25,520 --> 00:28:29,460
The roots of MDO are in
structural optimization.

589
00:28:29,460 --> 00:28:32,520
So you saw the bracket
with all those--

590
00:28:32,520 --> 00:28:35,840
so there's actually a field
called structural optimization

591
00:28:35,840 --> 00:28:39,590
or topology optimization,
and it's very mature now.

592
00:28:39,590 --> 00:28:43,290
So you can buy, in some
of the CAD packages,

593
00:28:43,290 --> 00:28:47,450
there's some specialized tools
for optimizing structures.

594
00:28:47,450 --> 00:28:50,600
And when you look at it,
it looks like Swiss cheese

595
00:28:50,600 --> 00:28:53,720
with big holes, an
Emmentaler, big holes.

596
00:28:53,720 --> 00:28:55,760
You don't need a
lot of structure

597
00:28:55,760 --> 00:28:57,180
that doesn't do work for you.

598
00:28:57,180 --> 00:29:01,020
So a lot of these
structures look like webs,

599
00:29:01,020 --> 00:29:02,930
and the reason is
because the load

600
00:29:02,930 --> 00:29:05,090
path has been highly optimized.

601
00:29:05,090 --> 00:29:09,302
For example, if you
look at the A380--

602
00:29:09,302 --> 00:29:10,760
I don't have a
picture right here--

603
00:29:10,760 --> 00:29:17,320
but you can look at an A380 wing
spars and different structures.

604
00:29:20,000 --> 00:29:22,390
They have more
holes than structure

605
00:29:22,390 --> 00:29:24,680
because they've been
highly optimized.

606
00:29:24,680 --> 00:29:29,930
So that's the roots of
MDO, starting in the 1960s.

607
00:29:29,930 --> 00:29:33,770
Then, I would say, in the '80s,
mid-'80s, we started having

608
00:29:33,770 --> 00:29:36,680
more complex
decomposition techniques,

609
00:29:36,680 --> 00:29:41,180
and I will describe one
of these to you today.

610
00:29:41,180 --> 00:29:43,220
Optimization
algorithms were started

611
00:29:43,220 --> 00:29:46,280
to be built into
mainstream programs

612
00:29:46,280 --> 00:29:50,610
like Excel, MATLAB, Mathematica.

613
00:29:50,610 --> 00:29:56,010
And then more recently,
you know, in the 2000s,

614
00:29:56,010 --> 00:29:59,655
we had commercialization
of multilevel algorithms,

615
00:29:59,655 --> 00:30:01,905
and the one that I'll describe
to you is called BLISS.

616
00:30:06,250 --> 00:30:10,040
So let's talk through a very
specific example, and then

617
00:30:10,040 --> 00:30:11,100
one of those techniques.

618
00:30:11,100 --> 00:30:13,380
So here's the example.

619
00:30:13,380 --> 00:30:17,690
You're designing a wing,
and in designing a wing,

620
00:30:17,690 --> 00:30:22,520
the two main considerations
are structural and aerodynamic.

621
00:30:22,520 --> 00:30:25,970
So here's a simplified wing.

622
00:30:25,970 --> 00:30:30,020
It has a wing area,
it has a sweep angle,

623
00:30:30,020 --> 00:30:33,120
it has an aspect
ratio, and so forth.

624
00:30:33,120 --> 00:30:36,110
And there's loads
acting on the wing

625
00:30:36,110 --> 00:30:38,510
as you're flying
because of the pressure

626
00:30:38,510 --> 00:30:40,770
differential between the
upper and lower surface.

627
00:30:40,770 --> 00:30:45,950
So there's some net force,
P, acting on this wing.

628
00:30:45,950 --> 00:30:49,700
And as a result of that,
the wing will warp.

629
00:30:49,700 --> 00:30:53,690
So you will have some
twist angle in the wing,

630
00:30:53,690 --> 00:30:56,570
and that's just due to the
structural deformation.

631
00:30:56,570 --> 00:31:01,640
So that twist angle and
the displacements, in turn,

632
00:31:01,640 --> 00:31:04,990
will change the drag that
the wing experiences.

633
00:31:04,990 --> 00:31:08,120
So the lift and drag
of the wing will change

634
00:31:08,120 --> 00:31:10,180
as a function of twist angle.

635
00:31:10,180 --> 00:31:15,800
But as the lift and drag
change, the loads, of course,

636
00:31:15,800 --> 00:31:20,190
then change as well, and it
feeds back into the structure.

637
00:31:20,190 --> 00:31:23,780
And in this case, what we're
looking for is maximum range.

638
00:31:23,780 --> 00:31:27,050
We want to maximize the
range of the airplane.

639
00:31:27,050 --> 00:31:30,410
And this is a sort
of simplified version

640
00:31:30,410 --> 00:31:33,260
of the range equation, the
so-called Breguet range

641
00:31:33,260 --> 00:31:34,530
equation.

642
00:31:34,530 --> 00:31:38,450
And you can see that
the structure influences

643
00:31:38,450 --> 00:31:41,450
the range in two ways--
directly by weight,

644
00:31:41,450 --> 00:31:43,430
the actual weight
of the structure

645
00:31:43,430 --> 00:31:48,170
enters here twice in the range
equation, and then indirectly

646
00:31:48,170 --> 00:31:50,510
through its stiffness, right?

647
00:31:50,510 --> 00:31:53,000
The softer it is, the more
displacements, the more

648
00:31:53,000 --> 00:31:54,780
you affect the drag.

649
00:31:54,780 --> 00:31:59,600
So it's not so clear then
what is an optimal wing?

650
00:31:59,600 --> 00:32:01,700
Is it the lightest
wing possible?

651
00:32:01,700 --> 00:32:05,330
But the lightest wing possible
will have a larger twist

652
00:32:05,330 --> 00:32:09,740
and may have more drag, which
will reduce your range, right?

653
00:32:09,740 --> 00:32:13,600
So should you optimize
the wing for lightness

654
00:32:13,600 --> 00:32:18,350
or for this minimum
displacement or some combination

655
00:32:18,350 --> 00:32:19,880
or mix of the two?

656
00:32:19,880 --> 00:32:22,670
And the answer, of course,
is it's a mix of the two.

657
00:32:22,670 --> 00:32:25,010
There will be a
sweet spot somewhere,

658
00:32:25,010 --> 00:32:27,470
you know, between a very
stiff and very heavy

659
00:32:27,470 --> 00:32:31,330
wing and a very, very
flexible, lightweight wing.

660
00:32:31,330 --> 00:32:34,250
There will be some
optimal mix of the two,

661
00:32:34,250 --> 00:32:36,860
and that's the key here.

662
00:32:36,860 --> 00:32:40,430
So let me discuss
with you briefly--

663
00:32:40,430 --> 00:32:41,820
this is just one of the methods.

664
00:32:41,820 --> 00:32:45,110
There are several methods for
doing multidisciplinary design

665
00:32:45,110 --> 00:32:46,880
optimization, but
this is one of them

666
00:32:46,880 --> 00:32:51,320
called BLISS, Bi-Level
Integrated Systems Synthesis.

667
00:32:51,320 --> 00:32:54,170
And the general idea
here in this method

668
00:32:54,170 --> 00:32:59,420
is that we have a lower level
optimization of each subsystem

669
00:32:59,420 --> 00:33:02,840
but it's coupled through a
system level optimization.

670
00:33:02,840 --> 00:33:05,600
And what you want is, you
know, an optimal design

671
00:33:05,600 --> 00:33:09,110
at the system level, not
just optimal subsystems.

672
00:33:09,110 --> 00:33:11,870
So the context
here is we're gonna

673
00:33:11,870 --> 00:33:15,740
design a high speed, a
supersonic business jet.

674
00:33:15,740 --> 00:33:19,110
It's kind of shown
in this gray picture.

675
00:33:19,110 --> 00:33:22,200
And we have four
major subsystems

676
00:33:22,200 --> 00:33:25,640
we're gonna consider--
propulsion, aerodynamics,

677
00:33:25,640 --> 00:33:28,100
structures, and then
the performance, which,

678
00:33:28,100 --> 00:33:30,890
in this case, is range, OK?

679
00:33:30,890 --> 00:33:35,120
And what's important here in
decomposing the problem is

680
00:33:35,120 --> 00:33:37,280
the x's and the y's.

681
00:33:37,280 --> 00:33:39,194
Now, the x's are the
design variables.

682
00:33:39,194 --> 00:33:41,360
Do you remember, we talk
about the design variables,

683
00:33:41,360 --> 00:33:44,090
the knobs you can
turn as a designer?

684
00:33:44,090 --> 00:33:46,940
And so these here
are-- the upper ones,

685
00:33:46,940 --> 00:33:49,730
we call them local design
variables, you know,

686
00:33:49,730 --> 00:33:53,360
like the throttle setting, the
tail sweep, the wing moment

687
00:33:53,360 --> 00:34:00,080
arm, the thickness of the
wings, the taper ratio.

688
00:34:00,080 --> 00:34:02,630
So the idea of a
local design variable

689
00:34:02,630 --> 00:34:07,010
is that design variable only is
used in one of the subsystems.

690
00:34:07,010 --> 00:34:10,610
It's local to that subsystem.

691
00:34:10,610 --> 00:34:16,429
So you basically can choose
that in each of the subsystems,

692
00:34:16,429 --> 00:34:19,159
and it doesn't really
affect directly

693
00:34:19,159 --> 00:34:21,620
any of the other
subsystems-- indirectly, yes,

694
00:34:21,620 --> 00:34:23,600
but not directly.

695
00:34:23,600 --> 00:34:25,130
And then we have the y's here.

696
00:34:25,130 --> 00:34:30,000
The y's are the outputs from
one subsystem to another.

697
00:34:30,000 --> 00:34:34,940
So drag, lift, NZ is
your maximum load factor,

698
00:34:34,940 --> 00:34:38,360
your range, your specific fuel
consumption, your wing twist,

699
00:34:38,360 --> 00:34:39,510
and so forth.

700
00:34:39,510 --> 00:34:43,250
So those are outputs from a
subsystem or discipline that

701
00:34:43,250 --> 00:34:45,170
either matter at
the system level

702
00:34:45,170 --> 00:34:49,550
or are required as inputs
by another subsystem.

703
00:34:49,550 --> 00:34:52,580
And then finally, the
last group here, these

704
00:34:52,580 --> 00:34:55,010
are known as the shared
design variables,

705
00:34:55,010 --> 00:34:57,470
so the wing aspect
ratio, the altitude,

706
00:34:57,470 --> 00:35:01,460
mach number, wing sweep.

707
00:35:01,460 --> 00:35:05,920
Those are basically required
by more than one subsystem, so

708
00:35:05,920 --> 00:35:09,050
two, three, or
four, in this case.

709
00:35:09,050 --> 00:35:11,980
Then they are considered
as shared variables,

710
00:35:11,980 --> 00:35:15,970
and they would be considered
system level variables

711
00:35:15,970 --> 00:35:19,540
because they matter for
more than one subsystem.

712
00:35:19,540 --> 00:35:22,750
So what this
particular method does

713
00:35:22,750 --> 00:35:25,450
is essentially
decompose the problem

714
00:35:25,450 --> 00:35:28,900
into the various subsystems,
so in this case, again,

715
00:35:28,900 --> 00:35:32,330
aerodynamics, propulsion,
structures, and range.

716
00:35:32,330 --> 00:35:36,610
And these subsystems are sending
this information to each other

717
00:35:36,610 --> 00:35:39,430
and are all tied together
at the highest level

718
00:35:39,430 --> 00:35:41,560
through the shared
variables, the shared design

719
00:35:41,560 --> 00:35:45,760
variables that are shared
by at least two subsystems.

720
00:35:45,760 --> 00:35:47,590
So again, local
variables are only

721
00:35:47,590 --> 00:35:51,490
unique to a specific subsystem,
and then for the wise,

722
00:35:51,490 --> 00:35:54,520
these outputs, internal
outputs, there's

723
00:35:54,520 --> 00:35:58,300
a distinction between
the Y stars, which

724
00:35:58,300 --> 00:36:01,750
are coupling variables that are
input to a particular subsystem

725
00:36:01,750 --> 00:36:04,360
and then the Y hats
are coupling variables

726
00:36:04,360 --> 00:36:07,990
that are output from a
particular subsystem.

727
00:36:07,990 --> 00:36:12,610
So let's zoom in now on
just one of the subsystems,

728
00:36:12,610 --> 00:36:17,000
let's say, aerodynamics,
and see how this works.

729
00:36:17,000 --> 00:36:20,800
So here we're zooming
in on one subsystem,

730
00:36:20,800 --> 00:36:22,750
and that's the aerodynamics.

731
00:36:22,750 --> 00:36:24,700
And the idea that
you can actually

732
00:36:24,700 --> 00:36:29,480
optimize just the aerodynamic
subsystem by itself.

733
00:36:29,480 --> 00:36:31,930
So what are the inputs into it?

734
00:36:31,930 --> 00:36:40,390
We have aspect ratio, wing
twist, and W. So what is W?

735
00:36:40,390 --> 00:36:42,480
And this is key to this
particular method--

736
00:36:42,480 --> 00:36:45,430
the W are the
weights, the weights

737
00:36:45,430 --> 00:36:47,750
that you give to the outputs.

738
00:36:47,750 --> 00:36:52,600
So what you're optimizing here
is some objective function, f,

739
00:36:52,600 --> 00:37:01,780
shown here below, which is W1,
Y1, plus W2, Y2 plus W3, Y3.

740
00:37:01,780 --> 00:37:04,460
So you're either maximizing
or minimizing that.

741
00:37:04,460 --> 00:37:05,410
And what is this here?

742
00:37:05,410 --> 00:37:09,460
It's lift, drag, and
the lift to drag ratio.

743
00:37:09,460 --> 00:37:12,100
So what's really cool
about this method is

744
00:37:12,100 --> 00:37:17,050
that what it essentially
is telling the subsystem

745
00:37:17,050 --> 00:37:21,340
is how to properly weight these
different outputs in order

746
00:37:21,340 --> 00:37:23,890
to optimize the
subsystem, taking into

747
00:37:23,890 --> 00:37:27,140
account the system
level objective.

748
00:37:27,140 --> 00:37:29,350
So the weights are
not pre-defined,

749
00:37:29,350 --> 00:37:33,310
but the weights themselves
are a design variable.

750
00:37:33,310 --> 00:37:35,950
And so you can see the
formulation on the right side.

751
00:37:35,950 --> 00:37:41,960
So given Q, which is
your shared variables,

752
00:37:41,960 --> 00:37:43,930
the Y stars, which
are the inputs

753
00:37:43,930 --> 00:37:46,780
from the other subsystems,
and the weights,

754
00:37:46,780 --> 00:37:51,820
minimize your objective function
by varying your local design

755
00:37:51,820 --> 00:37:56,660
variables subject to a
bunch of constraints, OK?

756
00:37:56,660 --> 00:38:00,490
And so what is one
of the consequences

757
00:38:00,490 --> 00:38:02,830
of being able to formulate
the problem this way?

758
00:38:02,830 --> 00:38:07,330
So this could be a very high
fidelity, aerodynamics model,

759
00:38:07,330 --> 00:38:11,800
computational fluid dynamics,
or it could be what?

760
00:38:15,410 --> 00:38:18,710
Or it could be a lower
fidelity code, right?

761
00:38:18,710 --> 00:38:22,100
As long as the inputs and
the outputs are the same,

762
00:38:22,100 --> 00:38:27,200
you can now plug in, you know,
different fidelity levels

763
00:38:27,200 --> 00:38:31,010
of analysis, you know
something simple or something

764
00:38:31,010 --> 00:38:33,470
based on experimental
results, you

765
00:38:33,470 --> 00:38:36,890
know, panel method,
very high fidelity CFD.

766
00:38:36,890 --> 00:38:38,960
This becomes very modular.

767
00:38:38,960 --> 00:38:40,550
The other thing
you can do is you

768
00:38:40,550 --> 00:38:44,700
can pre-compute a lot of
these responses, right?

769
00:38:44,700 --> 00:38:47,160
So for a given
set of inputs, you

770
00:38:47,160 --> 00:38:53,190
can pre-compute over a whole
range of input variables

771
00:38:53,190 --> 00:38:56,040
what the response of
that subsystem will be,

772
00:38:56,040 --> 00:38:58,290
and you can store
that information

773
00:38:58,290 --> 00:39:00,840
and then use what are
known approximation

774
00:39:00,840 --> 00:39:03,570
methods, like response
surface models,

775
00:39:03,570 --> 00:39:06,900
to save that information,
such that when

776
00:39:06,900 --> 00:39:08,980
you do the full
system optimization,

777
00:39:08,980 --> 00:39:12,090
you don't have to rerun,
you know, the detailed codes

778
00:39:12,090 --> 00:39:12,810
every time.

779
00:39:12,810 --> 00:39:16,040
You can just pull
from the approximation

780
00:39:16,040 --> 00:39:18,390
you've pre-computed already.

781
00:39:18,390 --> 00:39:20,520
So that's one of the
really powerful things

782
00:39:20,520 --> 00:39:23,120
about this method.

783
00:39:23,120 --> 00:39:26,880
And that's what's shown
here, is you substitute--

784
00:39:26,880 --> 00:39:29,700
you have a series of
approximation models, one

785
00:39:29,700 --> 00:39:31,710
for each of the
outputs of the system.

786
00:39:31,710 --> 00:39:32,580
You see that?

787
00:39:32,580 --> 00:39:34,860
So these little
wiggly things here

788
00:39:34,860 --> 00:39:37,200
are essentially an
approximation of

789
00:39:37,200 --> 00:39:41,100
that input/output relationship
at the subsystem level.

790
00:39:41,100 --> 00:39:44,730
Could be an empirical function,
a neural network, a response

791
00:39:44,730 --> 00:39:48,540
surface method, or you
know, simplified engineering

792
00:39:48,540 --> 00:39:49,860
analysis.

793
00:39:49,860 --> 00:39:53,400
So you do a lot of this
work offline ahead of time

794
00:39:53,400 --> 00:39:55,590
for each of the subsystems.

795
00:39:55,590 --> 00:39:57,060
Does that make sense?

796
00:39:57,060 --> 00:39:58,980
Any questions at EPFL?

797
00:39:58,980 --> 00:40:02,340
Is that clear how this
works, at least in theory?

798
00:40:05,960 --> 00:40:07,520
AUDIENCE: Everything's
fine, thanks.

799
00:40:07,520 --> 00:40:12,500
OLIVIER DE WECK: OK, so what
we have now is essentially

800
00:40:12,500 --> 00:40:17,570
these subsystem approximations,
and now we can use them

801
00:40:17,570 --> 00:40:20,470
for system level optimization.

802
00:40:20,470 --> 00:40:23,770
So now let's look at the
system level optimization.

803
00:40:23,770 --> 00:40:27,430
So that's essentially
given approximation

804
00:40:27,430 --> 00:40:30,970
models for optimized
subsystem outputs,

805
00:40:30,970 --> 00:40:34,070
we want to minimize
some objective function.

806
00:40:34,070 --> 00:40:38,920
So F, here in this case, which
is a function of our shared

807
00:40:38,920 --> 00:40:45,820
variables, our relevant outputs,
Y star, and the weighting

808
00:40:45,820 --> 00:40:48,370
factors, right, the omega--

809
00:40:48,370 --> 00:40:51,760
not the omegas, the W's, which
are the weighting factors given

810
00:40:51,760 --> 00:40:53,920
to each of the subsystems.

811
00:40:53,920 --> 00:40:57,670
And we of course, have
to satisfy constraints

812
00:40:57,670 --> 00:41:00,080
at the system level.

813
00:41:00,080 --> 00:41:04,780
So let me show you an example
then of what this looks like.

814
00:41:04,780 --> 00:41:09,250
So here's our supersonic jet.

815
00:41:09,250 --> 00:41:12,590
This is our cycle zero,
our initial condition.

816
00:41:12,590 --> 00:41:14,890
Here's the planned
form of the jet,

817
00:41:14,890 --> 00:41:16,900
and I'm going to show
you down here range

818
00:41:16,900 --> 00:41:21,160
and just one of the design
variables, or a set of design

819
00:41:21,160 --> 00:41:24,200
variables, which is our
wing box sheet thickness.

820
00:41:24,200 --> 00:41:29,320
So you go through this, and
it converges after 10 cycles

821
00:41:29,320 --> 00:41:30,340
already.

822
00:41:30,340 --> 00:41:34,660
So this is what our business
jet looks like after 10 cycles.

823
00:41:34,660 --> 00:41:37,050
And you know, you see there's
a pretty big difference.

824
00:41:37,050 --> 00:41:40,360
So the fuselage is much longer.

825
00:41:40,360 --> 00:41:42,810
This by the way, this is
the same in ships, right?

826
00:41:42,810 --> 00:41:46,900
Your speed that you can
reach, your max speed

827
00:41:46,900 --> 00:41:51,040
is driven by your fuselage
length, your water length.

828
00:41:51,040 --> 00:41:54,430
You can see the wings have
a pretty high sweep angle.

829
00:41:54,430 --> 00:41:58,060
This is a supersonic jet.

830
00:41:58,060 --> 00:42:00,370
And if we look at
the details, we

831
00:42:00,370 --> 00:42:03,520
see that the range
here, the range

832
00:42:03,520 --> 00:42:07,960
is about 3,000 nautical
miles, and what

833
00:42:07,960 --> 00:42:11,920
you see is that the exact
range that can be achieved

834
00:42:11,920 --> 00:42:14,210
and the approximate range merge.

835
00:42:14,210 --> 00:42:18,850
So the approximation models
that you have in this method

836
00:42:18,850 --> 00:42:19,900
are only approximate.

837
00:42:19,900 --> 00:42:21,430
There's some error.

838
00:42:21,430 --> 00:42:23,950
And you know, as you run
the model several times,

839
00:42:23,950 --> 00:42:26,050
several iterations,
the difference

840
00:42:26,050 --> 00:42:29,710
between the approximation and
the actual high fidelity result

841
00:42:29,710 --> 00:42:33,310
gets less and less, so that
it's actually a feasible design.

842
00:42:33,310 --> 00:42:36,340
And then over here, we see
the different thicknesses

843
00:42:36,340 --> 00:42:39,770
for the inner wing
box, the mid span wing

844
00:42:39,770 --> 00:42:41,900
box, and the outer wing box.

845
00:42:41,900 --> 00:42:44,320
And you can see that early
on in the first four or five

846
00:42:44,320 --> 00:42:47,840
six iterations, things
are kind of moving around,

847
00:42:47,840 --> 00:42:49,570
but then they're settling out.

848
00:42:49,570 --> 00:42:54,190
And so for this maximum
objective of range,

849
00:42:54,190 --> 00:42:59,290
you can see that the inner, mid,
and outer wing box thicknesses

850
00:42:59,290 --> 00:43:01,160
are quite different.

851
00:43:01,160 --> 00:43:04,930
So that's the general idea
here is that this result--

852
00:43:04,930 --> 00:43:10,630
this is a fairly detailed design
was produced based on a system

853
00:43:10,630 --> 00:43:13,450
level optimization that
included consideration

854
00:43:13,450 --> 00:43:18,370
of aerodynamic structures,
propulsion, and so forth.

855
00:43:18,370 --> 00:43:19,870
And there are several
other methods,

856
00:43:19,870 --> 00:43:23,170
but this is just one
of the methods that's

857
00:43:23,170 --> 00:43:27,010
been actually quite helpful.

858
00:43:27,010 --> 00:43:30,130
The challenges now in MDO,
in this multidisciplinary

859
00:43:30,130 --> 00:43:32,530
optimization are
different trade-offs.

860
00:43:32,530 --> 00:43:35,920
So one trade-off is between
fidelity versus how expensive

861
00:43:35,920 --> 00:43:38,290
is it computationally.

862
00:43:38,290 --> 00:43:39,760
So fidelity is over here.

863
00:43:39,760 --> 00:43:42,580
So high fidelity
means, you know,

864
00:43:42,580 --> 00:43:45,970
3D finite element model
analysis, computational fluid

865
00:43:45,970 --> 00:43:46,999
dynamics.

866
00:43:46,999 --> 00:43:48,790
Then you have sort of
intermediate fidelity

867
00:43:48,790 --> 00:43:51,610
or low fidelity
empirical models.

868
00:43:51,610 --> 00:43:55,060
And then on this axis, we
have just doing trade studies,

869
00:43:55,060 --> 00:43:57,070
meaning you have sort
of a baseline design

870
00:43:57,070 --> 00:43:59,960
and you just explore from
that baseline design.

871
00:43:59,960 --> 00:44:03,460
You do limited optimization
or iteration or full MDO,

872
00:44:03,460 --> 00:44:05,770
exploring the full space.

873
00:44:05,770 --> 00:44:08,230
And traditionally,
you had to choose,

874
00:44:08,230 --> 00:44:11,300
am I gonna use high
fidelity models?

875
00:44:11,300 --> 00:44:13,850
But then I can only
look at a few points.

876
00:44:13,850 --> 00:44:16,210
So then the question
is can you do better?

877
00:44:16,210 --> 00:44:18,730
Is there a better
solution out there?

878
00:44:18,730 --> 00:44:21,760
Or you can do lower
fidelity models

879
00:44:21,760 --> 00:44:24,610
and explore the design
space very widely,

880
00:44:24,610 --> 00:44:26,050
and then the
question is, can you

881
00:44:26,050 --> 00:44:28,510
believe the results
because you've done

882
00:44:28,510 --> 00:44:31,360
this with approximate models?

883
00:44:31,360 --> 00:44:34,840
So the real challenge here is we
would like to do high fidelity

884
00:44:34,840 --> 00:44:37,370
and explore the
full design space.

885
00:44:37,370 --> 00:44:40,750
So that's one of the big
research areas in MDO

886
00:44:40,750 --> 00:44:45,730
is high fidelity but also very
computationally efficient.

887
00:44:45,730 --> 00:44:49,160
And then the other
challenge is quite similar.

888
00:44:49,160 --> 00:44:52,150
So it's again the
fidelity here, but then

889
00:44:52,150 --> 00:44:54,730
should you just focus
on one subsystem?

890
00:44:54,730 --> 00:44:57,070
How many of the subsystems,
including, you know,

891
00:44:57,070 --> 00:45:01,150
the financial, operations,
manufacturing considerations,

892
00:45:01,150 --> 00:45:02,950
so how many of those
subsystems do you

893
00:45:02,950 --> 00:45:06,130
include in your optimization?

894
00:45:06,130 --> 00:45:10,930
And so you know, ideally, you'd
like to do the complete system,

895
00:45:10,930 --> 00:45:13,600
but then you may
have to sacrifice

896
00:45:13,600 --> 00:45:14,740
on some of the fidelity.

897
00:45:14,740 --> 00:45:19,440
So that's the breadth
versus depth trade-off.

898
00:45:19,440 --> 00:45:22,860
OK so any questions about--

899
00:45:22,860 --> 00:45:24,640
I know this was very fast.

900
00:45:24,640 --> 00:45:30,450
I do not expect you, by the
way, for assignment four or five

901
00:45:30,450 --> 00:45:33,840
to do MDO on your CanSat.

902
00:45:33,840 --> 00:45:37,950
This is really more to give
you a feeling for, you know,

903
00:45:37,950 --> 00:45:40,860
how far things
have come and some

904
00:45:40,860 --> 00:45:42,420
of the computational
frameworks that

905
00:45:42,420 --> 00:45:46,980
exist to get you to a really
fine tuned optimized design.

906
00:45:46,980 --> 00:45:49,050
If you are interested
in this, you know,

907
00:45:49,050 --> 00:45:50,910
we have the spring class.

908
00:45:50,910 --> 00:45:56,400
And you know, I hadn't offered
this class at EPFL in the past,

909
00:45:56,400 --> 00:45:59,100
but we actually do
offer a distance version

910
00:45:59,100 --> 00:46:00,810
with the SDM program.

911
00:46:00,810 --> 00:46:03,890
So there may be an option to
do this at a distance as well.

912
00:46:03,890 --> 00:46:05,190
We'll discuss it.

913
00:46:05,190 --> 00:46:07,860
But any questions about MDO?

914
00:46:07,860 --> 00:46:11,550
What it is, how it works,
at least in principle?

915
00:46:11,550 --> 00:46:14,160
Has anybody done
MDO kind of work?

916
00:46:14,160 --> 00:46:16,347
Sam, go ahead

917
00:46:16,347 --> 00:46:18,180
AUDIENCE: I just had
one question about if--

918
00:46:18,180 --> 00:46:21,920
so here, we're looking at
optimizing on one attribute, so

919
00:46:21,920 --> 00:46:25,850
range, but if we have
multiple attributes,

920
00:46:25,850 --> 00:46:28,280
it's still up to
someone to decide

921
00:46:28,280 --> 00:46:31,340
how important those
attributes are in the end.

922
00:46:31,340 --> 00:46:33,000
OLIVIER DE WECK: That's correct.

923
00:46:33,000 --> 00:46:35,630
That is exactly
right, unless you

924
00:46:35,630 --> 00:46:40,040
have a way to combine all these
into one objective function.

925
00:46:40,040 --> 00:46:43,180
And to be blatant
about it, to be-- you

926
00:46:43,180 --> 00:46:47,080
know, in the commercial
world, that's actually clear.

927
00:46:47,080 --> 00:46:50,550
It's profit or
net present value.

928
00:46:50,550 --> 00:46:53,740
So if you're designing
a commercial product

929
00:46:53,740 --> 00:46:58,220
in a competitive
industry, you can actually

930
00:46:58,220 --> 00:47:00,880
pretty easily combine
all this into a long--

931
00:47:00,880 --> 00:47:05,780
you know, as long as you include
sort of the key drivers of,

932
00:47:05,780 --> 00:47:07,880
you know, your
manufacturing costs.

933
00:47:07,880 --> 00:47:10,970
If the product is not as
optimal and not as performant,

934
00:47:10,970 --> 00:47:14,640
well, probably, you're not
gonna be as successful.

935
00:47:14,640 --> 00:47:16,304
Your market share
might be lower.

936
00:47:16,304 --> 00:47:17,720
So in the commercial
world, people

937
00:47:17,720 --> 00:47:21,800
are doing MDO with profitability
as the ultimate objective

938
00:47:21,800 --> 00:47:23,000
function.

939
00:47:23,000 --> 00:47:30,470
Now, for situations where it's
not about profit necessarily,

940
00:47:30,470 --> 00:47:32,626
but it's about a bunch
of other considerations,

941
00:47:32,626 --> 00:47:34,000
then you can't
easily combine it.

942
00:47:34,000 --> 00:47:35,960
You can still do utility.

943
00:47:35,960 --> 00:47:38,650
You can do a utility, you
know, combine it in utility.

944
00:47:38,650 --> 00:47:42,020
But I agree with you, I think
a good practice is then,

945
00:47:42,020 --> 00:47:44,930
in that case, to do
the multi-objective,

946
00:47:44,930 --> 00:47:47,630
and then you're gonna have
a bunch of solutions that

947
00:47:47,630 --> 00:47:50,750
are all non-dominated, right?

948
00:47:50,750 --> 00:47:53,150
And then it's a human
decision making process

949
00:47:53,150 --> 00:47:56,185
to pick among those.

950
00:47:56,185 --> 00:47:58,090
Any other questions or comments?

951
00:47:58,090 --> 00:47:59,500
What about at EPFL?

952
00:47:59,500 --> 00:48:03,360
MDO, anybody have
experience with this before,

953
00:48:03,360 --> 00:48:05,660
multidisciplinary
design optimization?

954
00:48:09,532 --> 00:48:11,020
AUDIENCE: OK, no one, again.

955
00:48:11,020 --> 00:48:13,900
OLIVIER DE WECK: OK,
that's all right.

956
00:48:13,900 --> 00:48:17,620
So just to tell you, this is
really a big deal in practice.

957
00:48:17,620 --> 00:48:20,500
So you know, the structural
optimization has been around,

958
00:48:20,500 --> 00:48:24,430
that's pretty standard
practice, and you know,

959
00:48:24,430 --> 00:48:26,800
optimizing control systems.

960
00:48:26,800 --> 00:48:31,150
So we've optimized individual
subsystems for a long time now,

961
00:48:31,150 --> 00:48:34,690
at least two or three decades,
but optimizing at the system

962
00:48:34,690 --> 00:48:36,620
level is relatively new.

963
00:48:36,620 --> 00:48:39,670
And I think the best companies,
the best organizations that

964
00:48:39,670 --> 00:48:43,150
are out there, they're really
investing in this technology,

965
00:48:43,150 --> 00:48:44,810
and they're quite good at it.

966
00:48:44,810 --> 00:48:47,110
And that's where you can
really distinguish yourself

967
00:48:47,110 --> 00:48:49,620
from the competition.

968
00:48:49,620 --> 00:48:53,200
So let's talk about
concurrent design facilities.

969
00:48:53,200 --> 00:48:57,850
And so I'm actually going to
use some material here from EPFL

970
00:48:57,850 --> 00:49:00,490
because they have
a very nice CDF,

971
00:49:00,490 --> 00:49:03,910
and I want to give
credit here to Dr. Anton

972
00:49:03,910 --> 00:49:09,670
Ivanov at the Swiss Space Center
for some of this material.

973
00:49:09,670 --> 00:49:13,380
And actually one of the papers
is also related to this.

974
00:49:13,380 --> 00:49:15,310
Anton used to be at JPL.

975
00:49:15,310 --> 00:49:21,400
He worked at JPL for close to
a decade before going to EPFL.

976
00:49:21,400 --> 00:49:24,940
So what is a CDF, a
concurrent design facility?

977
00:49:24,940 --> 00:49:27,010
It's essentially an
environment where

978
00:49:27,010 --> 00:49:28,600
engineers of
different disciplines

979
00:49:28,600 --> 00:49:30,730
come together to
perform, you know,

980
00:49:30,730 --> 00:49:33,130
a system engineering
study for a project.

981
00:49:33,130 --> 00:49:35,740
Key elements that you need
is the team, the process,

982
00:49:35,740 --> 00:49:37,510
the physical environment,
which includes

983
00:49:37,510 --> 00:49:40,690
the A/V and the software,
and then also some way

984
00:49:40,690 --> 00:49:42,970
to do knowledge
management, meaning

985
00:49:42,970 --> 00:49:48,730
capturing the results of prior
studies and ongoing studies.

986
00:49:48,730 --> 00:49:52,930
And I'll show you
some examples of CDFs.

987
00:49:52,930 --> 00:49:54,620
We also do this in academia.

988
00:49:54,620 --> 00:49:56,890
There's challenges for
doing this in academia,

989
00:49:56,890 --> 00:49:59,530
a short learning curve, you
know, synchronizing this

990
00:49:59,530 --> 00:50:01,390
with the academic
schedule, and then

991
00:50:01,390 --> 00:50:06,560
a lot of turnover, of course, as
students graduate and so forth.

992
00:50:06,560 --> 00:50:08,230
So there's challenges.

993
00:50:08,230 --> 00:50:12,790
And in practice, what we
essentially have is, in a CDF,

994
00:50:12,790 --> 00:50:18,050
we have a set of models that
are linked to each other.

995
00:50:18,050 --> 00:50:20,200
And these models can
be, for example, here's

996
00:50:20,200 --> 00:50:23,310
some Simulink models shown.

997
00:50:23,310 --> 00:50:27,520
Here, there is
some Excel models.

998
00:50:27,520 --> 00:50:29,140
You can have CAD.

999
00:50:29,140 --> 00:50:31,750
And one of the tricks is, you
know, to link this together

1000
00:50:31,750 --> 00:50:35,230
in a way that you can quickly
execute different designs

1001
00:50:35,230 --> 00:50:38,870
and look at the
results from that.

1002
00:50:38,870 --> 00:50:43,120
So in industrial settings,
probably the most famous CDF

1003
00:50:43,120 --> 00:50:46,540
is the one at JPL, at the
Jet Propulsion Laboratory,

1004
00:50:46,540 --> 00:50:52,600
called Team X. Team X has been
around for about 20 years,

1005
00:50:52,600 --> 00:50:58,900
and they just had an anniversary
where they celebrated

1006
00:50:58,900 --> 00:51:05,560
their 1000th study that was
done at Team X. A typical Team X

1007
00:51:05,560 --> 00:51:07,780
study is about a week.

1008
00:51:07,780 --> 00:51:10,570
So you do a little
bit of preparation,

1009
00:51:10,570 --> 00:51:14,110
but people work together in
the facility for about a week,

1010
00:51:14,110 --> 00:51:16,300
and at the end of
the week, you have

1011
00:51:16,300 --> 00:51:21,970
basically a conceptual design
pretty well worked out.

1012
00:51:21,970 --> 00:51:23,870
And the results have
been impressive.

1013
00:51:23,870 --> 00:51:26,470
So studies have shown
that the cost estimates

1014
00:51:26,470 --> 00:51:30,550
that Team X came up with were
within 10% of final mission

1015
00:51:30,550 --> 00:51:34,540
costs, for those missions that
were actually built and flown.

1016
00:51:34,540 --> 00:51:38,710
So of the 1,000 studies that
Team X did, a handful of them

1017
00:51:38,710 --> 00:51:40,820
were actual flight missions.

1018
00:51:40,820 --> 00:51:46,960
But every proposal
that is written

1019
00:51:46,960 --> 00:51:49,480
starts now as a Team X study.

1020
00:51:49,480 --> 00:51:52,120
So it's really a big deal.

1021
00:51:52,120 --> 00:51:57,310
The European Space Agency
ESTEC has a similar facility

1022
00:51:57,310 --> 00:52:00,360
in Noordwijk in the Netherlands.

1023
00:52:00,360 --> 00:52:04,020
As well, all the ESA projects
are going through their CDF.

1024
00:52:04,020 --> 00:52:06,100
So one of these is
[INAUDIBLE] which

1025
00:52:06,100 --> 00:52:09,430
is a space-based observatory.

1026
00:52:09,430 --> 00:52:11,350
EPFL is involved
in that project.

1027
00:52:11,350 --> 00:52:15,290
That went through
the CDF as well.

1028
00:52:15,290 --> 00:52:19,420
Most NASA centers and then
commercial applications

1029
00:52:19,420 --> 00:52:21,730
in medical devices,
painting, shipbuilding.

1030
00:52:21,730 --> 00:52:23,740
And like I mentioned
in my introduction,

1031
00:52:23,740 --> 00:52:26,350
even consumer packaged
good companies

1032
00:52:26,350 --> 00:52:29,920
are now building their
version of a CDF,

1033
00:52:29,920 --> 00:52:33,550
such that they can, you know,
very quickly, within a couple

1034
00:52:33,550 --> 00:52:38,590
weeks, come up with a new
proposal for a product and all

1035
00:52:38,590 --> 00:52:41,140
the major aspects, you
know, the underlying physics

1036
00:52:41,140 --> 00:52:45,650
and chemistry, the packaging,
the business case, everything

1037
00:52:45,650 --> 00:52:48,490
is considered at the same time.

1038
00:52:48,490 --> 00:52:49,630
Yes?

1039
00:52:49,630 --> 00:52:51,532
AUDIENCE: I was wondering why--

1040
00:52:51,532 --> 00:52:55,550
what makes the cost estimate
so accurate with this method?

1041
00:52:55,550 --> 00:52:58,160
OLIVIER DE WECK: So there
are three major ways

1042
00:52:58,160 --> 00:52:59,840
of doing-- oh,
Veronica should answer

1043
00:52:59,840 --> 00:53:03,500
that question, since
she's launching on cost

1044
00:53:03,500 --> 00:53:05,810
modeling for her research.

1045
00:53:05,810 --> 00:53:08,346
There are three major
cost estimation methods.

1046
00:53:08,346 --> 00:53:10,970
One is bottom up, you know, you
have the product decomposition,

1047
00:53:10,970 --> 00:53:13,100
and you add up all the costs.

1048
00:53:13,100 --> 00:53:16,100
One is called
costing by analogy,

1049
00:53:16,100 --> 00:53:18,980
where you look at a prior
mission or a prior product

1050
00:53:18,980 --> 00:53:20,930
and then you look at the deltas.

1051
00:53:20,930 --> 00:53:23,960
And then the third one is
based on so-called CERs,

1052
00:53:23,960 --> 00:53:26,000
cost estimation relationships.

1053
00:53:26,000 --> 00:53:27,650
It's a parametric method.

1054
00:53:27,650 --> 00:53:29,570
You can also blend these.

1055
00:53:29,570 --> 00:53:32,930
And you know, I think one of
the reasons why it's accurate

1056
00:53:32,930 --> 00:53:34,130
is because--

1057
00:53:34,130 --> 00:53:36,880
there are two reasons
why, I would say.

1058
00:53:36,880 --> 00:53:39,410
So you always have a cost share.

1059
00:53:39,410 --> 00:53:41,390
One of the subsystems
is costing.

1060
00:53:41,390 --> 00:53:44,440
So you have really people who
been doing this for a while.

1061
00:53:44,440 --> 00:53:47,660
They know these
models, and so there's

1062
00:53:47,660 --> 00:53:50,390
some legacy information.

1063
00:53:50,390 --> 00:53:55,070
And the more missions we fly,
the more products we make,

1064
00:53:55,070 --> 00:53:58,880
the more robust, essentially,
these cost estimates become.

1065
00:53:58,880 --> 00:54:03,650
And then the second, I think,
is because people that typically

1066
00:54:03,650 --> 00:54:07,430
work in a CDF,
they're non-advocates,

1067
00:54:07,430 --> 00:54:09,950
meaning it's kind
of neutral ground.

1068
00:54:09,950 --> 00:54:12,140
So rather than pushing
a particular idea

1069
00:54:12,140 --> 00:54:14,660
or trying to be
overly optimistic,

1070
00:54:14,660 --> 00:54:18,290
you know, usually in a
CDF, the calculations

1071
00:54:18,290 --> 00:54:21,470
that are done are
pretty non-biased,

1072
00:54:21,470 --> 00:54:23,760
and that's not always the case.

1073
00:54:23,760 --> 00:54:27,152
So that's one of the advantages
of the CDF is you're getting--

1074
00:54:27,152 --> 00:54:29,360
you know, people are trying
to say, but really, let's

1075
00:54:29,360 --> 00:54:30,590
make it look a little better.

1076
00:54:30,590 --> 00:54:34,170
No, no, we can't do this,
because the model says,

1077
00:54:34,170 --> 00:54:38,510
you know, it is gonna cost $280
million to do this satellite.

1078
00:54:38,510 --> 00:54:41,910
And that's based on, you
know, within prior models,

1079
00:54:41,910 --> 00:54:43,670
and if you overlay
different models,

1080
00:54:43,670 --> 00:54:46,550
you know, with some
error or confidence.

1081
00:54:46,550 --> 00:54:48,360
That's the right answer.

1082
00:54:48,360 --> 00:54:49,490
So it's these two reasons.

1083
00:54:49,490 --> 00:54:54,470
There's experience, there's
data, and non-advocacy, really

1084
00:54:54,470 --> 00:54:58,010
taking a kind of neutral,
unbiased position.

1085
00:54:58,010 --> 00:55:01,430
So benefits are, you know,
improvements of quality,

1086
00:55:01,430 --> 00:55:04,970
quick turnaround for ideas,
better cost estimates,

1087
00:55:04,970 --> 00:55:09,720
and then increased creativity
and productivity in a company.

1088
00:55:09,720 --> 00:55:11,730
So let me give you a
couple examples here.

1089
00:55:11,730 --> 00:55:14,000
This is a CubeSat.

1090
00:55:14,000 --> 00:55:18,530
So this is the ability to
design quickly new CubeSats.

1091
00:55:18,530 --> 00:55:20,030
This picture here
I'm showing you

1092
00:55:20,030 --> 00:55:23,510
is the Swiss Cube that was
launched in 2009, still

1093
00:55:23,510 --> 00:55:24,230
operating.

1094
00:55:24,230 --> 00:55:26,750
I think it's the longest
operating, continuously

1095
00:55:26,750 --> 00:55:28,850
operating CubeSat so far.

1096
00:55:28,850 --> 00:55:32,060
There's like over six years
of operational data now.

1097
00:55:32,060 --> 00:55:34,520
And so you know, for the
next generation, which

1098
00:55:34,520 --> 00:55:39,821
is shown here, how would you do
a Cubesat in such a facility?

1099
00:55:39,821 --> 00:55:40,820
And there's three steps.

1100
00:55:40,820 --> 00:55:42,890
You define your
decomposition level.

1101
00:55:42,890 --> 00:55:45,840
So you have your system
level, your subsystem level,

1102
00:55:45,840 --> 00:55:48,410
and then at the lowest
level it's called equipment.

1103
00:55:48,410 --> 00:55:50,960
These are the actual components.

1104
00:55:50,960 --> 00:55:53,660
And you have your product
decomposition here,

1105
00:55:53,660 --> 00:55:55,730
and then for each
of these components,

1106
00:55:55,730 --> 00:55:58,670
you have different sheets
that give you either component

1107
00:55:58,670 --> 00:56:04,220
choices or give you different
parameters like mass power

1108
00:56:04,220 --> 00:56:06,680
with margins that allow
you to then choose

1109
00:56:06,680 --> 00:56:11,490
from this like a menu and pull
together your overall design.

1110
00:56:11,490 --> 00:56:13,457
So that's a Cubesat example.

1111
00:56:13,457 --> 00:56:15,040
But I want to talk
to you a little bit

1112
00:56:15,040 --> 00:56:19,250
in more detail through a design
of a suborbital space plane

1113
00:56:19,250 --> 00:56:22,850
that was done in
that CDF, and this

1114
00:56:22,850 --> 00:56:26,360
was a project known as K1000.

1115
00:56:26,360 --> 00:56:30,240
So some of you may have heard
of the Hermes, the Hermes space

1116
00:56:30,240 --> 00:56:31,100
plane.

1117
00:56:31,100 --> 00:56:35,240
This was a project that the
European Space Agency had

1118
00:56:35,240 --> 00:56:39,650
starting even in the
'80s, and it wasn't built,

1119
00:56:39,650 --> 00:56:41,780
but there was a lot
of studies done.

1120
00:56:41,780 --> 00:56:44,310
And this is essentially
a Hermes-like vehicle

1121
00:56:44,310 --> 00:56:48,060
but for space tourism,
that's the basic application.

1122
00:56:48,060 --> 00:56:51,800
And so you know, how would
you design a space tourism

1123
00:56:51,800 --> 00:56:53,690
vehicle in a CDF?

1124
00:56:53,690 --> 00:56:56,150
Well, you have to start
with requirements, right?

1125
00:56:56,150 --> 00:56:59,581
Without requirements, you're
just-- it's like, you know,

1126
00:56:59,581 --> 00:57:00,080
setting--

1127
00:57:00,080 --> 00:57:02,420
my analogy for designing
without requirements

1128
00:57:02,420 --> 00:57:06,800
is like taking your ship and
going out to sea without a map,

1129
00:57:06,800 --> 00:57:10,310
without a chart, with no
destination port in mind.

1130
00:57:10,310 --> 00:57:12,990
You're just drifting
in the ocean.

1131
00:57:12,990 --> 00:57:14,810
You've got to have requirements.

1132
00:57:14,810 --> 00:57:17,150
So here are the high
level requirements

1133
00:57:17,150 --> 00:57:18,980
for this particular vehicle.

1134
00:57:18,980 --> 00:57:20,540
Level one
requirements-- you have

1135
00:57:20,540 --> 00:57:25,280
to reach an altitude of at least
100 kilometers over sea level.

1136
00:57:25,280 --> 00:57:26,930
Why is that?

1137
00:57:26,930 --> 00:57:29,270
Where does the 100
kilometers come from?

1138
00:57:29,270 --> 00:57:34,000
Why is it not 80 or 90 or 150?

1139
00:57:34,000 --> 00:57:36,050
AUDIENCE: That's where
space is defined.

1140
00:57:36,050 --> 00:57:38,600
OLIVIER DE WECK: Right,
and what happens to people

1141
00:57:38,600 --> 00:57:41,032
who go above 100 kilometers?

1142
00:57:41,032 --> 00:57:42,740
AUDIENCE: They get
their astronaut wings.

1143
00:57:42,740 --> 00:57:45,050
OLIVIER DE WECK:
They get their--

1144
00:57:45,050 --> 00:57:47,450
you have gone to
space, and like,

1145
00:57:47,450 --> 00:57:49,490
you know, you can
claim that you've gone

1146
00:57:49,490 --> 00:57:52,610
to space if you go above 100.

1147
00:57:52,610 --> 00:57:54,770
If you don't, if you are
at 80 or 90 you don't.

1148
00:57:54,770 --> 00:57:57,080
That's kind of
internationally accepted.

1149
00:57:57,080 --> 00:58:01,580
So that's really where that
requirement comes from.

1150
00:58:01,580 --> 00:58:04,640
The second one is
a zero g phase.

1151
00:58:04,640 --> 00:58:06,920
You should experience
weightlessness

1152
00:58:06,920 --> 00:58:09,480
for at least several minutes.

1153
00:58:09,480 --> 00:58:12,430
Who's done a parabolic
flight before here?

1154
00:58:12,430 --> 00:58:15,100
The zero g, you know,
the Vomit Comet?

1155
00:58:15,100 --> 00:58:16,680
Anybody?

1156
00:58:16,680 --> 00:58:19,100
Wow, we have to get
you guys out there.

1157
00:58:19,100 --> 00:58:21,630
At EPFL, are you still with us?

1158
00:58:24,920 --> 00:58:27,230
AUDIENCE: We are.

1159
00:58:27,230 --> 00:58:29,690
OLIVIER DE WECK: Has anybody
done parabolic flights?

1160
00:58:29,690 --> 00:58:32,960
I know there's a flight campaign
coming up in Switzerland

1161
00:58:32,960 --> 00:58:34,580
later this year?

1162
00:58:34,580 --> 00:58:36,780
AUDIENCE: Yeah, it will
be next year, I guess,

1163
00:58:36,780 --> 00:58:39,710
but unfortunately, no one
has done this right now here.

1164
00:58:39,710 --> 00:58:41,960
OLIVIER DE WECK: All right,
guys, we have to fix this.

1165
00:58:41,960 --> 00:58:43,070
This is unacceptable.

1166
00:58:43,070 --> 00:58:46,500
I mean, we have to work on this.

1167
00:58:46,500 --> 00:58:50,060
So it's very cool to do
these parabolic flights,

1168
00:58:50,060 --> 00:58:52,550
but you only get about
20 seconds, right?

1169
00:58:52,550 --> 00:58:54,590
20 seconds at the top.

1170
00:58:54,590 --> 00:58:57,560
It's like riding a big
roller coaster, basically.

1171
00:58:57,560 --> 00:59:00,320
And then you go down, and
then you suffer, you know,

1172
00:59:00,320 --> 00:59:02,880
1.7 g's or whatever.

1173
00:59:02,880 --> 00:59:04,220
Then you come back up.

1174
00:59:04,220 --> 00:59:05,540
But we want several minutes.

1175
00:59:08,720 --> 00:59:12,160
And then the next requirement
is that the passengers

1176
00:59:12,160 --> 00:59:15,710
should carry-- there should
be six passengers carried

1177
00:59:15,710 --> 00:59:16,940
on this vehicle.

1178
00:59:16,940 --> 00:59:19,520
Those are the top
level one requirements.

1179
00:59:19,520 --> 00:59:23,270
Lower level-- level two
requirements-- safety.

1180
00:59:23,270 --> 00:59:28,040
Don't exceed six g,
positive six g's.

1181
00:59:28,040 --> 00:59:31,140
The vehicle must be
controllable at all times.

1182
00:59:31,140 --> 00:59:33,180
So when you're out
of the atmosphere,

1183
00:59:33,180 --> 00:59:35,180
you have to be able to
control the vehicle, that

1184
00:59:35,180 --> 00:59:37,490
has certain implications.

1185
00:59:37,490 --> 00:59:39,140
The customer
experience-- you should

1186
00:59:39,140 --> 00:59:42,550
be able to see the Earth's
curvature and atmosphere,

1187
00:59:42,550 --> 00:59:46,010
that thin layer, at all times.

1188
00:59:46,010 --> 00:59:48,680
The spacecraft impact
on the environment

1189
00:59:48,680 --> 00:59:50,540
should be as small as possible.

1190
00:59:50,540 --> 00:59:53,210
And then the mass budget,
the spacecraft mass

1191
00:59:53,210 --> 00:59:57,050
should not exceed 11.6
tons with propellants.

1192
00:59:57,050 --> 00:59:59,060
Where do you think
that comes from?

1193
00:59:59,060 --> 01:00:01,250
What kind of a
requirement is that?

1194
01:00:01,250 --> 01:00:02,970
What flavor of
requirement is that?

1195
01:00:02,970 --> 01:00:03,470
Yeah?

1196
01:00:03,470 --> 01:00:05,554
AUDIENCE: Transporting it
for proper [INAUDIBLE]..

1197
01:00:05,554 --> 01:00:07,220
OLIVIER DE WECK:
Right, so in this case,

1198
01:00:07,220 --> 01:00:09,100
it's actually a
piggyback, right?

1199
01:00:09,100 --> 01:00:11,890
You either have
a vertical launch

1200
01:00:11,890 --> 01:00:15,270
or you have a first stage,
an airplane from which you

1201
01:00:15,270 --> 01:00:16,180
are dropped right?

1202
01:00:16,180 --> 01:00:18,980
So that's sort of an
interface requirement.

1203
01:00:18,980 --> 01:00:21,400
OK, so once you know
your requirements,

1204
01:00:21,400 --> 01:00:23,950
you can really start
developing your models.

1205
01:00:23,950 --> 01:00:31,880
So here's a model of essentially
the kinematics of the vehicle.

1206
01:00:31,880 --> 01:00:34,090
So you have your environment,
your atmospheric model,

1207
01:00:34,090 --> 01:00:38,850
atmospheric density, pressure,
mach number, speed of sound,

1208
01:00:38,850 --> 01:00:41,530
a gravity model.

1209
01:00:41,530 --> 01:00:45,580
Then computing the forces and
torques acting on the vehicle

1210
01:00:45,580 --> 01:00:50,170
itself, looking at the
derivatives, integrating that.

1211
01:00:50,170 --> 01:00:51,730
This is essentially
the equations

1212
01:00:51,730 --> 01:00:54,910
of motion of a vehicle
in its environment

1213
01:00:54,910 --> 01:00:57,160
that it's operating
in, and then updating

1214
01:00:57,160 --> 01:01:01,060
that with a certain
time step, delta T.

1215
01:01:01,060 --> 01:01:03,250
So you can zoom in
on one of these,

1216
01:01:03,250 --> 01:01:08,400
for example, this force model,
computation of the forces.

1217
01:01:08,400 --> 01:01:09,760
And there's two frames.

1218
01:01:09,760 --> 01:01:11,170
There's the Earth
reference frame

1219
01:01:11,170 --> 01:01:12,910
and, then there's the
body centric frame

1220
01:01:12,910 --> 01:01:14,680
of the vehicle itself.

1221
01:01:14,680 --> 01:01:19,330
So you have, in this case,
the aerodynamic coefficients.

1222
01:01:19,330 --> 01:01:22,690
You have the forces, lift,
drag, pitching moment,

1223
01:01:22,690 --> 01:01:25,270
and then you have the actuation.

1224
01:01:25,270 --> 01:01:28,540
This is assumed, you know, the
pilots are here in control.

1225
01:01:28,540 --> 01:01:31,690
The rocket engine itself,
secondary propulsion,

1226
01:01:31,690 --> 01:01:34,810
the rudders, and then
attitude thrusters.

1227
01:01:34,810 --> 01:01:37,675
You need those especially--

1228
01:01:37,675 --> 01:01:40,480
you know, once you're out of
the atmosphere, the rudders--

1229
01:01:40,480 --> 01:01:43,750
the aerodynamic surfaces
are no longer effective.

1230
01:01:43,750 --> 01:01:45,550
And we knew this in the '60s.

1231
01:01:45,550 --> 01:01:48,670
Vehicles-- like the
X15 is very famous.

1232
01:01:48,670 --> 01:01:50,500
Well, it had thrusters
that you could

1233
01:01:50,500 --> 01:01:54,620
you could actuate once
you're out of the atmosphere.

1234
01:01:54,620 --> 01:01:56,620
So you know, you can go
pretty detailed in this.

1235
01:01:59,200 --> 01:02:02,920
You would then design the
vehicle in a particular way,

1236
01:02:02,920 --> 01:02:08,560
you know, particular engine,
wing dimensions, aspect ratio,

1237
01:02:08,560 --> 01:02:11,920
a particular flight
profile, and then

1238
01:02:11,920 --> 01:02:15,130
that would be a typical result
that you would get in a CDF.

1239
01:02:15,130 --> 01:02:16,990
There's two curves shown here.

1240
01:02:16,990 --> 01:02:21,050
The blue curve is essentially
altitude and kilometers.

1241
01:02:21,050 --> 01:02:23,320
So you see there's
the 100 kilometers.

1242
01:02:23,320 --> 01:02:24,760
We need to get
above this, right?

1243
01:02:24,760 --> 01:02:26,540
This was our requirement.

1244
01:02:26,540 --> 01:02:27,820
So we go above 100.

1245
01:02:27,820 --> 01:02:28,390
That's good.

1246
01:02:28,390 --> 01:02:29,750
We satisfy that requirement.

1247
01:02:29,750 --> 01:02:32,560
So looks like a
parabola, and then we

1248
01:02:32,560 --> 01:02:35,860
have a-- we glide here,
essentially, to the ground.

1249
01:02:35,860 --> 01:02:40,810
The whole mission takes about
900 seconds, 15 minutes.

1250
01:02:40,810 --> 01:02:44,800
And then the yellow curve here
is your load factor, your NZ,

1251
01:02:44,800 --> 01:02:47,030
your vertical load factor.

1252
01:02:47,030 --> 01:02:50,500
And you can see that's a
little noisier as a curve.

1253
01:02:50,500 --> 01:02:55,990
So we have higher acceleration
as we accelerate to altitude.

1254
01:02:55,990 --> 01:02:59,650
And you can see that
roughly here, roughly

1255
01:02:59,650 --> 01:03:04,770
in the middle of the parabola,
the load factor starts to--

1256
01:03:04,770 --> 01:03:06,160
you're basically in free fall.

1257
01:03:06,160 --> 01:03:10,840
You're basically ballistic,
and the peak acceleration

1258
01:03:10,840 --> 01:03:16,210
is about 4.3 g's here, so
we satisfy that requirement.

1259
01:03:16,210 --> 01:03:20,050
And the whole reason we're doing
it is for this period here,

1260
01:03:20,050 --> 01:03:22,360
that zero g experience.

1261
01:03:22,360 --> 01:03:26,410
So when the load factor
goes to zero, or near zero,

1262
01:03:26,410 --> 01:03:28,130
that's when you
get weightlessness.

1263
01:03:28,130 --> 01:03:31,240
And it turns out for this
particular flight profile,

1264
01:03:31,240 --> 01:03:36,910
we would experience 184
seconds, about three minutes

1265
01:03:36,910 --> 01:03:40,380
of weightlessness.

1266
01:03:40,380 --> 01:03:43,110
This is pretty similar
to the flight profile

1267
01:03:43,110 --> 01:03:47,440
that Virgin Galactic is
pursuing, by the way.

1268
01:03:47,440 --> 01:03:54,100
So you know, looking at
this result, you would say,

1269
01:03:54,100 --> 01:03:57,480
well, I don't know about
the mass here, but from,

1270
01:03:57,480 --> 01:04:00,750
you know, reaching
100 kilometers keeping

1271
01:04:00,750 --> 01:04:07,180
accelerations below six g, this
looks like a feasible profile.

1272
01:04:07,180 --> 01:04:10,090
Of course, the other
thing that we care about

1273
01:04:10,090 --> 01:04:13,600
is the view, since this
is for space tourism.

1274
01:04:13,600 --> 01:04:17,920
We want to make
sure that, you know,

1275
01:04:17,920 --> 01:04:20,590
that the experience
is great, not just

1276
01:04:20,590 --> 01:04:23,230
from an experiencing
weightlessness, which

1277
01:04:23,230 --> 01:04:26,950
you can do in a closed can but
you actually see something.

1278
01:04:26,950 --> 01:04:31,480
So this was essentially
coupled with a virtual reality

1279
01:04:31,480 --> 01:04:32,980
visualization.

1280
01:04:32,980 --> 01:04:35,170
So you can see, you
know, the landing here.

1281
01:04:38,200 --> 01:04:40,000
You can see these
three views here.

1282
01:04:40,000 --> 01:04:42,430
And you can see
the vehicle itself.

1283
01:04:42,430 --> 01:04:44,710
So the pilots are up
front, and then here

1284
01:04:44,710 --> 01:04:46,660
is our six passengers.

1285
01:04:46,660 --> 01:04:51,034
Every one of the passengers
gets their own window.

1286
01:04:51,034 --> 01:04:52,450
So this is the
view that you would

1287
01:04:52,450 --> 01:04:55,910
get from the front window.

1288
01:04:55,910 --> 01:04:58,150
This is the view from
the middle window.

1289
01:04:58,150 --> 01:05:00,970
And then this is the view
from the back window,

1290
01:05:00,970 --> 01:05:04,030
from that last window here,
which is closer to the back,

1291
01:05:04,030 --> 01:05:05,570
to the engines.

1292
01:05:05,570 --> 01:05:09,330
So what would you, looking
at these pictures-- you know,

1293
01:05:09,330 --> 01:05:13,740
this comes directly out of
the simulation and the CDF--

1294
01:05:13,740 --> 01:05:15,630
what would you say
looking at this?

1295
01:05:18,190 --> 01:05:19,240
Anything you notice?

1296
01:05:19,240 --> 01:05:21,580
EPFL anything you guys notice?

1297
01:05:21,580 --> 01:05:24,460
Look at the view from
the front, the middle,

1298
01:05:24,460 --> 01:05:25,570
and the back window.

1299
01:05:28,755 --> 01:05:30,380
AUDIENCE: From the
front, looks better.

1300
01:05:30,380 --> 01:05:32,650
I guess it will be
more expensive, again,

1301
01:05:32,650 --> 01:05:34,360
for the customer.

1302
01:05:34,360 --> 01:05:37,920
OLIVIER DE WECK: Yeah, so that's
exactly what I was asking.

1303
01:05:37,920 --> 01:05:42,790
So it looks like either you
need to redesign the vehicle so

1304
01:05:42,790 --> 01:05:46,330
that the view is the same, more
or less, from every window,

1305
01:05:46,330 --> 01:05:49,450
or you need to start
charging differential prices.

1306
01:05:49,450 --> 01:05:54,310
Because clearly,
from the back window,

1307
01:05:54,310 --> 01:05:59,167
half your view is obstructed
by the wing structure, right?

1308
01:05:59,167 --> 01:06:00,250
Would you agree with that?

1309
01:06:03,610 --> 01:06:06,430
Would you charge different
ticket prices for this ride,

1310
01:06:06,430 --> 01:06:08,170
depending on which seat you get?

1311
01:06:11,770 --> 01:06:14,110
AUDIENCE: I guess
that's offer and demand?

1312
01:06:14,110 --> 01:06:16,030
OLIVIER DE WECK:
Supply and demand, yes.

1313
01:06:16,030 --> 01:06:17,260
That's right.

1314
01:06:17,260 --> 01:06:18,150
Go ahead.

1315
01:06:18,150 --> 01:06:22,240
AUDIENCE: It could also merit
a change in flight plan.

1316
01:06:22,240 --> 01:06:24,820
If your structure blocks
the downward view,

1317
01:06:24,820 --> 01:06:27,160
you might want to
consider flipping over

1318
01:06:27,160 --> 01:06:31,030
if like the Earth is what
you're trying to see from above.

1319
01:06:31,030 --> 01:06:34,390
So it may have broader
implications for the mission.

1320
01:06:34,390 --> 01:06:37,150
OLIVIER DE WECK: OK,
so the point here,

1321
01:06:37,150 --> 01:06:41,830
is there a way to operationally
improve the experience,

1322
01:06:41,830 --> 01:06:43,670
keeping the vehicle
design the same?

1323
01:06:43,670 --> 01:06:45,670
Absolutely, but
what that means is,

1324
01:06:45,670 --> 01:06:47,650
you would have to
put the operation--

1325
01:06:47,650 --> 01:06:50,440
so the orientation
of the vehicle,

1326
01:06:50,440 --> 01:06:53,620
you'd have to put that
into your design vector

1327
01:06:53,620 --> 01:06:55,610
if you want to change it.

1328
01:06:55,610 --> 01:06:57,310
OK, so I don't know.

1329
01:06:57,310 --> 01:06:59,000
I think this is pretty cool.

1330
01:06:59,000 --> 01:07:03,610
So you can actually do a lot in
this kind of CDF type setting.

1331
01:07:03,610 --> 01:07:08,620
And by the way, of the things
that spun out of this project

1332
01:07:08,620 --> 01:07:11,860
is this company called S3.

1333
01:07:11,860 --> 01:07:15,070
Their headquarters in
Payerne in Switzerland.

1334
01:07:15,070 --> 01:07:17,350
This is actually-- I did a
lot of my military service

1335
01:07:17,350 --> 01:07:18,730
at that airport.

1336
01:07:18,730 --> 01:07:20,680
Their goal is not
space tourism, you

1337
01:07:20,680 --> 01:07:22,630
see there's no windows
on this vehicle,

1338
01:07:22,630 --> 01:07:25,720
but deployment of small
satellites, custom

1339
01:07:25,720 --> 01:07:29,280
deployment for small satellites
with a reusable vehicle.

1340
01:07:29,280 --> 01:07:34,900
So you know, I think it's
a very interesting concept,

1341
01:07:34,900 --> 01:07:37,330
but the question is,
is it feasible, right?

1342
01:07:37,330 --> 01:07:42,280
Is it feasible-- the
aerodynamics, the economics,

1343
01:07:42,280 --> 01:07:44,100
structurally, engine reuse?

1344
01:07:44,100 --> 01:07:45,030
There's a lot of--

1345
01:07:45,030 --> 01:07:48,100
this is a great
picture, but you know,

1346
01:07:48,100 --> 01:07:49,490
the devil is in the details.

1347
01:07:49,490 --> 01:07:53,230
So really working out the
details of such a design,

1348
01:07:53,230 --> 01:07:58,720
that's what a CDF is for,
and you can really go deep.

1349
01:07:58,720 --> 01:08:01,260
OK so any questions about this?

1350
01:08:01,260 --> 01:08:05,380
I'd like to do a little
partner exercise here.

1351
01:08:05,380 --> 01:08:06,700
Any questions?

1352
01:08:06,700 --> 01:08:12,910
So let's do a partner
exercise, and the question is,

1353
01:08:12,910 --> 01:08:16,090
what are your experiences with
concurrent design facilities

1354
01:08:16,090 --> 01:08:17,750
so far?

1355
01:08:17,750 --> 01:08:21,279
Whether you've done an
internship at ESA or JPL,

1356
01:08:21,279 --> 01:08:24,729
or some other place that
has a facility like this,

1357
01:08:24,729 --> 01:08:28,840
or even if you
haven't been there,

1358
01:08:28,840 --> 01:08:32,560
kind of speculate and discuss
what it would be like.

1359
01:08:32,560 --> 01:08:35,920
Maybe, you know, what would
be your favorite chair?

1360
01:08:35,920 --> 01:08:42,609
Would you pick propulsion or
costing or flight trajectory?

1361
01:08:42,609 --> 01:08:46,250
You know, just sort of speculate
and discuss with your partner,

1362
01:08:46,250 --> 01:08:49,210
so for which project or
application did you use it?

1363
01:08:49,210 --> 01:08:49,960
What went well?

1364
01:08:49,960 --> 01:08:51,010
What did not?

1365
01:08:51,010 --> 01:08:52,120
What could be improved?

1366
01:08:52,120 --> 01:08:54,700
And again, if you don't
have much experiences,

1367
01:08:54,700 --> 01:08:57,799
just think ahead what
it would be like.

1368
01:08:57,799 --> 01:08:59,590
What would you like to
get out of it if you

1369
01:08:59,590 --> 01:09:02,460
have no experience in it?

1370
01:09:02,460 --> 01:09:06,100
AUDIENCE: So there is at ESA,
a concurrent design facility,

1371
01:09:06,100 --> 01:09:12,250
and I was talking with Katya
that maybe also United Nations

1372
01:09:12,250 --> 01:09:16,250
can be considered, like,
the facilities there

1373
01:09:16,250 --> 01:09:20,120
can be considered a
concurrent design facility.

1374
01:09:20,120 --> 01:09:22,510
I don't know if it
worked, this approach.

1375
01:09:22,510 --> 01:09:24,490
OLIVIER DE WECK: Do
you mean in Geneva,

1376
01:09:24,490 --> 01:09:28,066
United Nations in Geneva
or London or New York,

1377
01:09:28,066 --> 01:09:29,149
or do you mean in general?

1378
01:09:29,149 --> 01:09:29,950
AUDIENCE: Exactly.

1379
01:09:29,950 --> 01:09:31,569
OLIVIER DE WECK:
But do they actually

1380
01:09:31,569 --> 01:09:35,859
have a physical facility
with models and screens,

1381
01:09:35,859 --> 01:09:39,590
or are you saying this
more metaphorically?

1382
01:09:39,590 --> 01:09:40,840
AUDIENCE: Metaphorically more.

1383
01:09:40,840 --> 01:09:43,130
OLIVIER DE WECK:
OK, but you know I--

1384
01:09:43,130 --> 01:09:43,999
go ahead.

1385
01:09:43,999 --> 01:09:44,630
Go ahead.

1386
01:09:44,630 --> 01:09:48,220
AUDIENCE: Yeah, we're referring
to the layout, to the site,

1387
01:09:48,220 --> 01:09:49,870
we're looking at the
ESA layout, where

1388
01:09:49,870 --> 01:09:52,665
you have one big room, where
you have all the people coming

1389
01:09:52,665 --> 01:09:54,790
together, all the different
models coming together.

1390
01:09:54,790 --> 01:09:56,050
So that's like all
the different experts.

1391
01:09:56,050 --> 01:09:58,530
Then you have subrooms for
different project rooms,

1392
01:09:58,530 --> 01:10:00,680
and then you have, like,
a big conference room.

1393
01:10:00,680 --> 01:10:05,470
That's kind of the structure of
the layout of the facilities.

1394
01:10:05,470 --> 01:10:08,320
OLIVIER DE WECK: But you know,
maybe a place like the UN

1395
01:10:08,320 --> 01:10:12,430
could really benefit when they
deal with a refugee crisis

1396
01:10:12,430 --> 01:10:15,850
or with, you know, climate
change negotiations.

1397
01:10:15,850 --> 01:10:17,410
I like this idea.

1398
01:10:17,410 --> 01:10:18,610
Great suggestion.

1399
01:10:18,610 --> 01:10:19,960
What about here at MIT?

1400
01:10:19,960 --> 01:10:21,370
Any experiences, thoughts?

1401
01:10:21,370 --> 01:10:22,730
Yeah, please go ahead.

1402
01:10:22,730 --> 01:10:26,730
AUDIENCE: Yeah, AFRL, we have
a rapid innovation officer.

1403
01:10:26,730 --> 01:10:30,880
He's like a general equivalent
as a civilian, a LOC BOS,

1404
01:10:30,880 --> 01:10:36,070
and he basically is tasked with
breaking all the rules in AFRL.

1405
01:10:36,070 --> 01:10:39,460
Whereas most projects
take years to complete,

1406
01:10:39,460 --> 01:10:42,880
he fields questions from
the field, from the theater,

1407
01:10:42,880 --> 01:10:46,010
from war fighters, and
they have an urgent need.

1408
01:10:46,010 --> 01:10:49,550
And he gets them a solution
in under 12 months.

1409
01:10:49,550 --> 01:10:54,610
So he pulls in all these
experts from industry, academia,

1410
01:10:54,610 --> 01:10:55,750
wherever.

1411
01:10:55,750 --> 01:10:58,150
He'll lock them in
a room, and they

1412
01:10:58,150 --> 01:11:00,370
don't leave until they
come up with a solution.

1413
01:11:00,370 --> 01:11:03,460
And he has the money
and the influence,

1414
01:11:03,460 --> 01:11:05,699
the backing by the AFRL
Commander to get it done.

1415
01:11:05,699 --> 01:11:07,240
OLIVIER DE WECK:
And what does that--

1416
01:11:07,240 --> 01:11:09,040
I mean, if you
can share, roughly

1417
01:11:09,040 --> 01:11:11,470
what does that facility
look like roughly?

1418
01:11:11,470 --> 01:11:15,520
Does it have, like, models
and computers and screens?

1419
01:11:15,520 --> 01:11:19,050
AUDIENCE: Just white walls,
like, whiteboard walls,

1420
01:11:19,050 --> 01:11:22,112
and they will just be covered
with everybody's ideas

1421
01:11:22,112 --> 01:11:23,320
by the time they leave there.

1422
01:11:23,320 --> 01:11:24,445
OLIVIER DE WECK: Oh, I see.

1423
01:11:24,445 --> 01:11:27,170
So it's not as much
computationally supported.

1424
01:11:27,170 --> 01:11:28,405
It's more of a human process.

1425
01:11:28,405 --> 01:11:29,220
AUDIENCE: That's right.

1426
01:11:29,220 --> 01:11:29,860
OLIVIER DE WECK: Cool.

1427
01:11:29,860 --> 01:11:30,310
OK.

1428
01:11:30,310 --> 01:11:30,810
Great.

1429
01:11:30,810 --> 01:11:34,150
Interesting.

1430
01:11:34,150 --> 01:11:35,980
Any other experiences?

1431
01:11:35,980 --> 01:11:39,250
Anybody been at JPL or
seen a commercial facility

1432
01:11:39,250 --> 01:11:42,340
in a commercial company
that does this kind of work?

1433
01:11:42,340 --> 01:11:43,782
Please, go ahead, Justice.

1434
01:11:46,490 --> 01:11:48,327
AUDIENCE: Well it's
less of a commercial--

1435
01:11:48,327 --> 01:11:50,910
I was interning at the Aerospace
Corporation this past summer,

1436
01:11:50,910 --> 01:11:53,240
and I was working in their
concurrent design center.

1437
01:11:53,240 --> 01:11:58,970
And it uses a similar layout
as JPL's Team X's design, where

1438
01:11:58,970 --> 01:12:01,010
it's basically using models.

1439
01:12:01,010 --> 01:12:04,205
That they have, like, a group
of engineers who sit at desks,

1440
01:12:04,205 --> 01:12:06,080
and there's multiple
engineers per subsystem,

1441
01:12:06,080 --> 01:12:07,430
and they just discuss.

1442
01:12:07,430 --> 01:12:09,620
They're given a
mission objective,

1443
01:12:09,620 --> 01:12:12,990
and they discuss how they can
best design the system based

1444
01:12:12,990 --> 01:12:13,490
on that.

1445
01:12:13,490 --> 01:12:15,870
And they model it and
optimize it in the subsystem,

1446
01:12:15,870 --> 01:12:18,635
and then they feed it back
to a systems level design.

1447
01:12:18,635 --> 01:12:20,510
OLIVIER DE WECK: Did
you participate in that?

1448
01:12:20,510 --> 01:12:21,230
AUDIENCE: Yes.

1449
01:12:21,230 --> 01:12:24,410
OLIVIER DE WECK: So what
did you think of it?

1450
01:12:24,410 --> 01:12:26,000
AUDIENCE: A lot--
it's a lot to take

1451
01:12:26,000 --> 01:12:28,100
in at first just
trying to understand

1452
01:12:28,100 --> 01:12:29,480
how each of the models work.

1453
01:12:29,480 --> 01:12:33,280
You're not necessarily
supposed to work on each model,

1454
01:12:33,280 --> 01:12:35,330
but as an intern, they
wanted me to kind of see

1455
01:12:35,330 --> 01:12:39,602
what each of the models would
do and not necessarily exactly

1456
01:12:39,602 --> 01:12:41,810
how the optimization techniques
work because I didn't

1457
01:12:41,810 --> 01:12:44,030
necessarily understand
it, but just more

1458
01:12:44,030 --> 01:12:47,960
of have an idea of how it
actually works as a whole.

1459
01:12:47,960 --> 01:12:49,490
OLIVIER DE WECK: OK, good.

1460
01:12:49,490 --> 01:12:54,140
Anybody else at EPFL want
to share any experiences?

1461
01:12:58,410 --> 01:13:01,960
AUDIENCE: Apparently not, but
just a problem with the sound.

1462
01:13:01,960 --> 01:13:05,640
We didn't hear anything on what
the guy before just told us.

1463
01:13:05,640 --> 01:13:09,000
OLIVIER DE WECK: OK, Justice,
did you push the mic button?

1464
01:13:09,000 --> 01:13:11,260
OK, yeah, so
basically, he worked--

1465
01:13:11,260 --> 01:13:12,760
let me try to summarize.

1466
01:13:12,760 --> 01:13:14,710
He worked at the
Aerospace Corporation.

1467
01:13:14,710 --> 01:13:16,900
He interned at the
Airspace Corporation.

1468
01:13:16,900 --> 01:13:19,240
Was it in Virginia
or California?

1469
01:13:19,240 --> 01:13:21,280
California, in California.

1470
01:13:21,280 --> 01:13:24,580
And Aerospace Corporation
is like a think tank,

1471
01:13:24,580 --> 01:13:28,900
and it was a very similar set
up with different subsystem

1472
01:13:28,900 --> 01:13:30,910
experts, different models.

1473
01:13:30,910 --> 01:13:33,670
And Justice was saying
one of the big challenges

1474
01:13:33,670 --> 01:13:36,610
is so much information,
taking it all in

1475
01:13:36,610 --> 01:13:40,280
and understanding all the models
and coupling, and so forth.

1476
01:13:40,280 --> 01:13:41,860
So and I agree.

1477
01:13:41,860 --> 01:13:42,700
It takes a while.

1478
01:13:42,700 --> 01:13:44,770
If you haven't been working
in this environment,

1479
01:13:44,770 --> 01:13:47,540
it takes a while to
really get used to it.

1480
01:13:47,540 --> 01:13:52,180
So the last thing I'll
say here is that--

1481
01:13:52,180 --> 01:13:53,090
well, two things.

1482
01:13:53,090 --> 01:13:55,390
One is-- and they're
linked-- one is

1483
01:13:55,390 --> 01:13:59,230
that these CDFs are not cheap.

1484
01:13:59,230 --> 01:14:02,380
If you really want a
high-functioning CDF,

1485
01:14:02,380 --> 01:14:05,260
you have to have people
that are dedicated to this.

1486
01:14:05,260 --> 01:14:08,230
You know, the software
licenses expire.

1487
01:14:08,230 --> 01:14:10,420
The technology gets old.

1488
01:14:10,420 --> 01:14:12,250
You need to not
just use it, but you

1489
01:14:12,250 --> 01:14:16,480
need to constantly refresh
a CDF, keep it up to date,

1490
01:14:16,480 --> 01:14:19,250
and it just doesn't
happen automatically.

1491
01:14:19,250 --> 01:14:21,760
You've got to have
budget for that.

1492
01:14:21,760 --> 01:14:24,310
You have to have people who
are dedicated to updating

1493
01:14:24,310 --> 01:14:28,280
the models, updating the
facility, and so forth.

1494
01:14:28,280 --> 01:14:31,120
So it's definitely
something that

1495
01:14:31,120 --> 01:14:34,900
needs to be invested
in on a regular basis.

1496
01:14:34,900 --> 01:14:39,310
So what happened to us
here at MIT in air astro,

1497
01:14:39,310 --> 01:14:41,650
actually, when I
was sitting where

1498
01:14:41,650 --> 01:14:44,170
you are sitting
as a grad student,

1499
01:14:44,170 --> 01:14:47,320
we created a CDF
in the mid '90s.

1500
01:14:47,320 --> 01:14:50,770
And you know, it was
just like the one at JPL.

1501
01:14:50,770 --> 01:14:55,840
We had desktops, you know, with
labels-- thermal, propulsion.

1502
01:14:55,840 --> 01:14:59,870
And we used it quite
actively for maybe five, six,

1503
01:14:59,870 --> 01:15:04,030
seven years, but we didn't put
the budget into it every year.

1504
01:15:04,030 --> 01:15:06,610
We didn't have a dedicated
person who was really.

1505
01:15:06,610 --> 01:15:07,840
looking after it.

1506
01:15:07,840 --> 01:15:13,240
So what happened to our CDF
is that entropy took over,

1507
01:15:13,240 --> 01:15:15,490
and it sort of got
more and more dated.

1508
01:15:15,490 --> 01:15:17,530
And when you go and
look in this room now,

1509
01:15:17,530 --> 01:15:20,330
it kind of looks like a
regular conference room.

1510
01:15:20,330 --> 01:15:22,810
So all the desktops
that we had are gone.

1511
01:15:22,810 --> 01:15:25,180
I guess they'd be
pretty old by now.

1512
01:15:25,180 --> 01:15:31,360
And there has been a pretty
strong debate among the faculty

1513
01:15:31,360 --> 01:15:32,960
whether actually we need this.

1514
01:15:32,960 --> 01:15:34,690
Should we recreate one?

1515
01:15:34,690 --> 01:15:36,880
And the argument
goes as follows,

1516
01:15:36,880 --> 01:15:43,010
the proponents say that a CDF
is really something special.

1517
01:15:43,010 --> 01:15:47,510
When you enter in the CDF, you
know this is a special place.

1518
01:15:47,510 --> 01:15:49,960
This is a facility that's
designed specifically

1519
01:15:49,960 --> 01:15:52,630
for doing rapid studies.

1520
01:15:52,630 --> 01:15:55,000
All the information
you need is there.

1521
01:15:55,000 --> 01:15:58,630
The walls are plastered
with reference information.

1522
01:15:58,630 --> 01:16:02,170
You know, really,
that's what it's for.

1523
01:16:02,170 --> 01:16:08,030
The counter argument is that
this is no longer the mid-'90s.

1524
01:16:08,030 --> 01:16:11,020
Now, everybody has laptops.

1525
01:16:11,020 --> 01:16:13,330
Everybody does mobile computing.

1526
01:16:13,330 --> 01:16:16,840
All the models and data
should be on the cloud,

1527
01:16:16,840 --> 01:16:20,170
not sitting anymore in a
dedicated server facility,

1528
01:16:20,170 --> 01:16:21,580
but it's sitting on the cloud.

1529
01:16:21,580 --> 01:16:24,760
And therefore, you
know, physically,

1530
01:16:24,760 --> 01:16:27,820
a facility that's dedicated
is no longer needed.

1531
01:16:27,820 --> 01:16:30,670
What you do is you
get together in a more

1532
01:16:30,670 --> 01:16:33,280
kind of flexible ad hoc way.

1533
01:16:33,280 --> 01:16:35,080
Everybody brings their laptops.

1534
01:16:35,080 --> 01:16:37,600
You have the big white
walls, like you said.

1535
01:16:37,600 --> 01:16:41,740
And you can just create
a CDF on the fly, right?

1536
01:16:41,740 --> 01:16:43,900
Everybody brings their
computing resources.

1537
01:16:43,900 --> 01:16:47,200
Stuff is on the cloud.

1538
01:16:47,200 --> 01:16:50,380
And then you get a lot done,
and then you disband again.

1539
01:16:50,380 --> 01:16:53,020
You clean the white
boards and you're done.

1540
01:16:53,020 --> 01:16:56,740
Therefore, no longer
needed a dedicated CDF.

1541
01:16:56,740 --> 01:17:04,410
Those are the two countervailing
opinions right now.

1542
01:17:04,410 --> 01:17:09,570
My view is, I think a CDF
still has value today.

1543
01:17:09,570 --> 01:17:14,650
I think it is valuable
as a dedicated facility,

1544
01:17:14,650 --> 01:17:16,030
but it needs to be updated.

1545
01:17:16,030 --> 01:17:18,570
So yes, the models
should be on the cloud.

1546
01:17:18,570 --> 01:17:20,760
Yes, you can bring your
laptops, but it is still

1547
01:17:20,760 --> 01:17:23,940
valuable to have dedicated
computing resources, again.

1548
01:17:23,940 --> 01:17:30,250
So I now think a hybrid model
is actually the right way to go.

1549
01:17:30,250 --> 01:17:33,940
So we'll see where it goes, but
it's a pretty active debate.

1550
01:17:33,940 --> 01:17:35,820
There have also been,
you know, theses

1551
01:17:35,820 --> 01:17:39,880
written on CDFs, comparing
different CDFs and so forth.

1552
01:17:39,880 --> 01:17:45,840
So take a look at the Anton
Ivanov paper in the reading

1553
01:17:45,840 --> 01:17:47,730
and just think about this.

1554
01:17:47,730 --> 01:17:51,210
I think this is a
big deal, and it's

1555
01:17:51,210 --> 01:17:53,910
going to be even more important
than the future, whether you're

1556
01:17:53,910 --> 01:17:57,390
doing space missions,
aeronautics,

1557
01:17:57,390 --> 01:17:59,550
whether you're designing
commercial products,

1558
01:17:59,550 --> 01:18:03,420
you know, or in the
military, these facilities

1559
01:18:03,420 --> 01:18:06,030
that allow you to make these
design decisions in a very

1560
01:18:06,030 --> 01:18:09,420
interactive, very
quick way are gonna

1561
01:18:09,420 --> 01:18:11,130
be more and more
important in the future,

1562
01:18:11,130 --> 01:18:15,205
and this is a big part
of system engineering.

1563
01:18:15,205 --> 01:18:19,700
Any final thoughts on
this before we move on?

1564
01:18:19,700 --> 01:18:26,400
OK, so here some lessons
learned from the EPFL CDF.

1565
01:18:26,400 --> 01:18:29,970
So this is particularly in
the university environment,

1566
01:18:29,970 --> 01:18:33,960
so you know, giving access
to a wide body of students.

1567
01:18:33,960 --> 01:18:36,810
You need to adapt this to
the university's schedule,

1568
01:18:36,810 --> 01:18:43,410
the learning curves, the analogy
of a smoke-filled room or war

1569
01:18:43,410 --> 01:18:46,080
room, optimal team size.

1570
01:18:46,080 --> 01:18:49,680
You know a CDF with 50 people
on it is not as effective as one

1571
01:18:49,680 --> 01:18:52,800
with, you know, seven
plus or minus two.

1572
01:18:52,800 --> 01:18:54,660
And then distributed CDFs--

1573
01:18:54,660 --> 01:18:56,860
they don't typically
work as well.

1574
01:18:56,860 --> 01:19:00,780
So I almost wore my--

1575
01:19:00,780 --> 01:19:03,720
I have a shirt that's
almost the same.

1576
01:19:03,720 --> 01:19:07,390
Anybody know who this is
or where this is from?

1577
01:19:07,390 --> 01:19:10,090
This is Sheldon Cooper
in the Big Bang Theory,

1578
01:19:10,090 --> 01:19:12,310
his virtual reality presence.

1579
01:19:12,310 --> 01:19:18,100
So the point here is that
virtual reality is OK.

1580
01:19:18,100 --> 01:19:23,380
So you can imagine a CDF where
half the people are in the room

1581
01:19:23,380 --> 01:19:26,290
and half the people are
dialing in remotely,

1582
01:19:26,290 --> 01:19:30,720
but it's never quite as good as
everybody is there physically.

1583
01:19:30,720 --> 01:19:34,140
So human interaction
is critical.

1584
01:19:34,140 --> 01:19:39,380
OK, the last thing I want to
talk about today is the CDR.

1585
01:19:39,380 --> 01:19:41,240
This is a big, big milestone.

1586
01:19:41,240 --> 01:19:42,860
This is a very big deal.

1587
01:19:42,860 --> 01:19:46,520
The critical design review,
the main purpose of it

1588
01:19:46,520 --> 01:19:50,720
is to approve the final
design in all its glory,

1589
01:19:50,720 --> 01:19:52,530
in all its details.

1590
01:19:52,530 --> 01:19:56,240
So generally, the
way we talk about it

1591
01:19:56,240 --> 01:20:00,470
is, if you pass the CDR, you've
got the green light to quote,

1592
01:20:00,470 --> 01:20:02,210
end quote, cut metal.

1593
01:20:02,210 --> 01:20:06,750
Of course, you know, our systems
have software and hardware.

1594
01:20:06,750 --> 01:20:10,520
So the cut metal is
essentially a euphemism

1595
01:20:10,520 --> 01:20:14,420
for you can go ahead and
manufacture the system now.

1596
01:20:14,420 --> 01:20:17,450
CDRs are typically
the biggest review.

1597
01:20:17,450 --> 01:20:22,400
The largest number of people
you will have at the CDR.

1598
01:20:22,400 --> 01:20:26,260
You know, if you had
10 people at the SRR

1599
01:20:26,260 --> 01:20:28,850
and 20 people at the
PDR, you're gonna

1600
01:20:28,850 --> 01:20:31,790
have 50 or 100
people at the CDR.

1601
01:20:31,790 --> 01:20:35,480
Because any question that
comes up on any of the parts--

1602
01:20:35,480 --> 01:20:37,460
say, is this really right?

1603
01:20:37,460 --> 01:20:40,490
Well, what could be the
impact of this detail

1604
01:20:40,490 --> 01:20:41,990
on the rest of the systems?

1605
01:20:41,990 --> 01:20:45,070
Somebody should be there
who designed that part

1606
01:20:45,070 --> 01:20:48,290
or who can answer the
question at least.

1607
01:20:48,290 --> 01:20:51,620
And CDRs are typically
longer than one day.

1608
01:20:51,620 --> 01:20:53,810
In fact, for a
big project, a CDR

1609
01:20:53,810 --> 01:20:56,300
can be a whole week of reviews.

1610
01:20:56,300 --> 01:20:58,850
So people get really tired.

1611
01:20:58,850 --> 01:21:01,880
Big CDRs are very,
very detailed.

1612
01:21:01,880 --> 01:21:05,660
You really have to drink lots
of coffee and stay awake.

1613
01:21:05,660 --> 01:21:09,140
So here's the description
of what a CDR is.

1614
01:21:09,140 --> 01:21:13,670
One of the guidelines is
that approximately 90%

1615
01:21:13,670 --> 01:21:16,640
of the engineering drawings
are approved and released.

1616
01:21:16,640 --> 01:21:21,860
So it's OK to have a CDR if
you have not 100% of everything

1617
01:21:21,860 --> 01:21:25,880
done, but you need to have
at least 90% complete.

1618
01:21:25,880 --> 01:21:29,870
You can't do a CDR, with, you
know, half the drawings done.

1619
01:21:29,870 --> 01:21:34,070
And it happens at the end
of phase C, the final design

1620
01:21:34,070 --> 01:21:35,840
phase C.

1621
01:21:35,840 --> 01:21:41,450
For very large projects,
you can have sub CDRs.

1622
01:21:41,450 --> 01:21:44,450
You'll have a system
level CDR, and then you

1623
01:21:44,450 --> 01:21:47,810
can have sub CDRs for
every major subsystem

1624
01:21:47,810 --> 01:21:49,760
in your project.

1625
01:21:49,760 --> 01:21:52,100
So an example, I
put a link here.

1626
01:21:52,100 --> 01:21:58,640
This is from late 2014, the
CDR for the James Webb Space

1627
01:21:58,640 --> 01:22:00,710
Telescope that I
mentioned earlier.

1628
01:22:00,710 --> 01:22:03,440
They passed the last
major element level

1629
01:22:03,440 --> 01:22:08,220
critical design review,
which was a big deal.

1630
01:22:08,220 --> 01:22:11,090
So launch is planned for 2018.

1631
01:22:11,090 --> 01:22:14,420
But for big projects,
what I'm trying to say is,

1632
01:22:14,420 --> 01:22:16,350
there may be more
than one CDR. There

1633
01:22:16,350 --> 01:22:19,520
may be the master CDR
for the whole thing

1634
01:22:19,520 --> 01:22:22,670
and then sub CDRs for
different subsystems.

1635
01:22:22,670 --> 01:22:25,860
Now, if you don't-- well, let me
get to this on the next chart.

1636
01:22:25,860 --> 01:22:27,450
Yeah, go ahead.

1637
01:22:27,450 --> 01:22:30,900
AUDIENCE: Are the, like, CDR
slides for NASA projects,

1638
01:22:30,900 --> 01:22:34,950
are those available
anywhere as a public entity

1639
01:22:34,950 --> 01:22:38,199
and can you get access
to any of that stuff?

1640
01:22:38,199 --> 01:22:39,365
OLIVIER DE WECK: It depends.

1641
01:22:42,870 --> 01:22:46,440
Not always, because especially
if commercial companies,

1642
01:22:46,440 --> 01:22:50,550
like, you know, Lockheed-Martin
or somebody is doing, like,

1643
01:22:50,550 --> 01:22:53,550
the spacecraft bus, they
would consider those details

1644
01:22:53,550 --> 01:22:55,330
proprietary.

1645
01:22:55,330 --> 01:22:58,770
So in general, I would say no.

1646
01:22:58,770 --> 01:23:04,080
But for CanSat, actually, you've
probably already found them.

1647
01:23:04,080 --> 01:23:07,080
For CanSat, you can
actually find a lot of this.

1648
01:23:07,080 --> 01:23:10,870
So prior PDRs and
CDRs, you can see that.

1649
01:23:10,870 --> 01:23:12,820
I encourage you not
to look too much.

1650
01:23:12,820 --> 01:23:15,236
You know, you kind of want to
come up with your own thing.

1651
01:23:15,236 --> 01:23:18,030
But you know, for
projects that don't

1652
01:23:18,030 --> 01:23:21,090
have a commercial interest
or proprietary technology,

1653
01:23:21,090 --> 01:23:23,610
it's often available.

1654
01:23:23,610 --> 01:23:28,610
So to show you just-- this is
a table from the Handbook, page

1655
01:23:28,610 --> 01:23:30,710
178.

1656
01:23:30,710 --> 01:23:35,480
This is the entrance criteria
and the exit criteria for CDR.

1657
01:23:35,480 --> 01:23:39,560
So entrance criteria means what
do you have to have in order

1658
01:23:39,560 --> 01:23:43,610
to be allowed to go
forward with a CDR?

1659
01:23:43,610 --> 01:23:45,380
And it's mainly three things--

1660
01:23:45,380 --> 01:23:49,100
successful completion of
the PDR, and responses

1661
01:23:49,100 --> 01:23:52,550
made to all the
PDR RFAs and RIDs.

1662
01:23:52,550 --> 01:23:56,960
RFAs and RIDs are essentially
deficiencies, open items

1663
01:23:56,960 --> 01:23:58,250
that came up in the PDR.

1664
01:23:58,250 --> 01:24:00,170
So you have to close them out.

1665
01:24:00,170 --> 01:24:02,570
A CDR agenda, and
then number three

1666
01:24:02,570 --> 01:24:05,450
is all the technical
work packages.

1667
01:24:05,450 --> 01:24:07,970
And so you see, it's
a huge list here--

1668
01:24:07,970 --> 01:24:11,010
baseline documents,
specification, fabrication,

1669
01:24:11,010 --> 01:24:14,450
assembly and integration,
technical data package,

1670
01:24:14,450 --> 01:24:18,020
acceptance criteria,
verification and validation

1671
01:24:18,020 --> 01:24:21,740
plan, launch site selection,
and operations plan--

1672
01:24:21,740 --> 01:24:24,140
super, super, super detailed.

1673
01:24:24,140 --> 01:24:26,240
This is a lot of information.

1674
01:24:26,240 --> 01:24:30,320
And at the CDR, you're
gonna go through all this.

1675
01:24:30,320 --> 01:24:34,650
So you can imagine
what that means.

1676
01:24:34,650 --> 01:24:36,890
And then success
criteria, on the right,

1677
01:24:36,890 --> 01:24:42,380
is, you know, in order to
successfully pass the CDR,

1678
01:24:42,380 --> 01:24:47,270
these are all the things that
you have to basically have.

1679
01:24:47,270 --> 01:24:49,970
You have to have approval
for all these things.

1680
01:24:49,970 --> 01:24:54,740
Now in reality, just
like a PDR, at a CDR,

1681
01:24:54,740 --> 01:24:56,480
things will come
up that you missed

1682
01:24:56,480 --> 01:24:58,640
or that people have
concerns about.

1683
01:24:58,640 --> 01:25:02,270
And so you will make
a list of those,

1684
01:25:02,270 --> 01:25:04,790
and then you have to address
these action items coming out

1685
01:25:04,790 --> 01:25:08,570
of a CDR. If those action
items are serious--

1686
01:25:08,570 --> 01:25:13,250
if they're really serious, you
could fail the CDR and say,

1687
01:25:13,250 --> 01:25:15,200
you know, this is
not gonna work.

1688
01:25:17,870 --> 01:25:19,610
We should not proceed.

1689
01:25:19,610 --> 01:25:20,720
And they send you back.

1690
01:25:20,720 --> 01:25:22,490
Or they cancel the project.

1691
01:25:22,490 --> 01:25:24,620
That's happened also.

1692
01:25:24,620 --> 01:25:28,370
More likely, is that they'll
ask you to do a so-called delta

1693
01:25:28,370 --> 01:25:31,250
CDR. So they'll
have you, you know,

1694
01:25:31,250 --> 01:25:33,290
it's like retaking
an exam or something.

1695
01:25:33,290 --> 01:25:35,840
They'll have you
come back, and you'll

1696
01:25:35,840 --> 01:25:41,150
have to do a CDR again but
only on the delta, the portion

1697
01:25:41,150 --> 01:25:43,730
that raised concerns
at the main CDR.

1698
01:25:43,730 --> 01:25:45,800
You don't have to
do everything again.

1699
01:25:45,800 --> 01:25:47,570
That's called a delta CDR.

1700
01:25:47,570 --> 01:25:50,990
OK, any questions about CDR?

1701
01:25:50,990 --> 01:25:53,400
This is really the--
this is the big one.

1702
01:25:53,400 --> 01:25:54,980
This is the big milestone.

1703
01:25:54,980 --> 01:25:58,490
And obviously, you know, the
concept or the architecture

1704
01:25:58,490 --> 01:26:02,640
you chose at the PDR turns
out not to be a good one,

1705
01:26:02,640 --> 01:26:06,800
it's very difficult to make
changes to that here, right?

1706
01:26:06,800 --> 01:26:10,040
You're kind of locked in to
whatever design or architecture

1707
01:26:10,040 --> 01:26:11,180
you chose at PDR.

1708
01:26:11,180 --> 01:26:14,004
That's why PDR is also
so important, even

1709
01:26:14,004 --> 01:26:15,170
though you have less detail.

1710
01:26:15,170 --> 01:26:17,071
Yeah, Sam, go ahead.

1711
01:26:17,071 --> 01:26:19,070
AUDIENCE: Do you have to
have contracts in place

1712
01:26:19,070 --> 01:26:22,267
already for the, like,
manufacturing and operations?

1713
01:26:22,267 --> 01:26:23,600
OLIVIER DE WECK: Oh, absolutely.

1714
01:26:23,600 --> 01:26:27,500
I mean, basically green
light CDR, the next day,

1715
01:26:27,500 --> 01:26:28,760
you start manufacturing.

1716
01:26:28,760 --> 01:26:31,775
So you know, every portion,
all the components,

1717
01:26:31,775 --> 01:26:34,040
all the manufacturing,
all of that

1718
01:26:34,040 --> 01:26:38,750
has been bid,
selected, contracted.

1719
01:26:38,750 --> 01:26:43,590
Everything's in place
contractually, absolutely.

1720
01:26:43,590 --> 01:26:46,830
OK, any questions
at EPFL about CDR?

1721
01:26:46,830 --> 01:26:49,710
Who has actually gone
through a CDR on your end?

1722
01:26:49,710 --> 01:26:53,460
Has anybody done a CDR
for any of the projects?

1723
01:26:53,460 --> 01:26:55,970
OK, go ahead.

1724
01:26:55,970 --> 01:26:57,330
What was it like?

1725
01:26:57,330 --> 01:27:01,950
AUDIENCE: I did also a CDR
because I had the PDR last time

1726
01:27:01,950 --> 01:27:04,290
at the [INAUDIBLE] program.

1727
01:27:04,290 --> 01:27:06,030
And one comment I want to make.

1728
01:27:06,030 --> 01:27:11,670
So the teams were allowed to
fail the PDR and still fly.

1729
01:27:11,670 --> 01:27:16,510
If they failed the CDR,
then they didn't fly at all.

1730
01:27:16,510 --> 01:27:19,452
So it was very
strict and demanding

1731
01:27:19,452 --> 01:27:20,910
OLIVIER DE WECK:
Yeah, so would you

1732
01:27:20,910 --> 01:27:25,050
say that the CDR is more
strict than the PDR?

1733
01:27:25,050 --> 01:27:27,090
AUDIENCE: The CDR
is more strict.

1734
01:27:27,090 --> 01:27:28,740
OLIVIER DE WECK: Yes, I agree.

1735
01:27:28,740 --> 01:27:29,430
I agree.

1736
01:27:29,430 --> 01:27:30,740
Definitely.

1737
01:27:30,740 --> 01:27:32,810
OK, anybody else?

1738
01:27:35,640 --> 01:27:40,902
Are any of you working in
systems other than aerospace,

1739
01:27:40,902 --> 01:27:44,170
like, I don't know, medical
devices or you know,

1740
01:27:44,170 --> 01:27:45,670
scientific instruments?

1741
01:27:50,580 --> 01:27:53,160
So the reason I bring
this up is because you

1742
01:27:53,160 --> 01:28:00,600
know, medical devices have to
be approved by, here in the US,

1743
01:28:00,600 --> 01:28:03,542
it's the FDA, the Food
and Drug Administration,

1744
01:28:03,542 --> 01:28:05,625
and then in Europe, there
is equivalent, you know,

1745
01:28:05,625 --> 01:28:07,170
European level thing.

1746
01:28:07,170 --> 01:28:08,850
It's very, very detailed.

1747
01:28:08,850 --> 01:28:10,330
It's really not that different.

1748
01:28:10,330 --> 01:28:13,020
So presenting your
medical device design

1749
01:28:13,020 --> 01:28:14,580
to a panel of experts--

1750
01:28:14,580 --> 01:28:23,410
doctors, you know, people in
medical safety, regulators--

1751
01:28:23,410 --> 01:28:26,921
it is like a CDR. You basically
are going through a CDR

1752
01:28:26,921 --> 01:28:29,170
when you're trying to get
your medical device approved

1753
01:28:29,170 --> 01:28:30,430
for sale and use.

1754
01:28:30,430 --> 01:28:35,350
So it's not often called a CDR,
but it effectively is a CDR.

1755
01:28:35,350 --> 01:28:39,280
So this is, especially
in industries

1756
01:28:39,280 --> 01:28:43,600
where there's a lot of
money at stake and people,

1757
01:28:43,600 --> 01:28:47,830
you know, safety, people's
health is at stake,

1758
01:28:47,830 --> 01:28:50,560
this is common practice.

1759
01:28:50,560 --> 01:28:52,570
OK, so let me summarize here.

1760
01:28:52,570 --> 01:28:55,770
So this is the detailed
design phase we discussed.

1761
01:28:55,770 --> 01:28:57,790
It's very important.

1762
01:28:57,790 --> 01:29:00,250
Basically, you take
your PDR level design

1763
01:29:00,250 --> 01:29:04,180
and you go define all the
details in full maturity.

1764
01:29:04,180 --> 01:29:06,760
You create design
documents and models.

1765
01:29:06,760 --> 01:29:10,510
And you know, as we
talked about last time,

1766
01:29:10,510 --> 01:29:12,250
we're sort of in
this transition phase

1767
01:29:12,250 --> 01:29:15,130
of producing documents
versus producing models.

1768
01:29:15,130 --> 01:29:18,220
Typically, we have both now.

1769
01:29:18,220 --> 01:29:21,520
Examples would be detailed
bill of materials,

1770
01:29:21,520 --> 01:29:25,120
computer-aided design files--
we'll talk about CAD next week,

1771
01:29:25,120 --> 01:29:26,590
computer-aided design--

1772
01:29:26,590 --> 01:29:28,660
software and control
system definition,

1773
01:29:28,660 --> 01:29:32,420
user interfaces, et cetera,
et cetera, et cetera.

1774
01:29:32,420 --> 01:29:36,326
Multidisciplinary
design optimization

1775
01:29:36,326 --> 01:29:40,360
is a very powerful
concept to optimize

1776
01:29:40,360 --> 01:29:42,680
at the system and
the subsystem level

1777
01:29:42,680 --> 01:29:45,730
and trade off between
disciplines and objectives.

1778
01:29:45,730 --> 01:29:49,040
The CDFs are the concurrent
design facilities,

1779
01:29:49,040 --> 01:29:52,880
and they've now become standard
practice both in aerospace

1780
01:29:52,880 --> 01:29:55,880
and product design companies.

1781
01:29:55,880 --> 01:29:59,720
If you have not had the
chance to experience and work

1782
01:29:59,720 --> 01:30:03,200
in a CDF, I really
encourage you to do this.

1783
01:30:03,200 --> 01:30:05,750
I think it's a
life-changing experience.

1784
01:30:05,750 --> 01:30:10,100
It really-- it's not easy,
because like, you know, Justice

1785
01:30:10,100 --> 01:30:12,740
said, there's a lot of
information to absorb.

1786
01:30:12,740 --> 01:30:16,040
But once you've done
this and experienced it,

1787
01:30:16,040 --> 01:30:19,260
you realize if you
don't have a CDF,

1788
01:30:19,260 --> 01:30:23,750
if you don't design products
this way, you really should.

1789
01:30:23,750 --> 01:30:26,150
And then CDR is
the big milestone,

1790
01:30:26,150 --> 01:30:30,520
the last gate before, quote,
end quote, cutting metal.