WEBVTT
Language: en-US

1
00:00:01.389 --> 00:00:05.070
<v Speaker 1>Welcome to the trend detection podcast
powered by Siemens.

2
00:00:08.329 --> 00:00:11.109
<v Speaker 2>Thank you very much and hello to everyone
online.

3
00:00:12.829 --> 00:00:16.049
<v Speaker 2>So, let's start off with what is
predictive maintenance.

4
00:00:16.780 --> 00:00:21.340
<v Speaker 2>And I always do like to start off at this
point because whenever I engage with new

5
00:00:21.340 --> 00:00:25.299
<v Speaker 2>people or a new audience, some of you
would have heard about predictive maintenance,

6
00:00:25.540 --> 00:00:28.120
<v Speaker 2>potentially even investigated predictive
maintenance.

7
00:00:28.770 --> 00:00:33.159
<v Speaker 2>But quite often, the understanding of what
predictive maintenance is falls in line

8
00:00:33.159 --> 00:00:36.150
<v Speaker 2>with machine magic rather than machine
learning.

9
00:00:36.650 --> 00:00:41.069
<v Speaker 2>So predictive maintenance is an evolution
from traditional condition monitoring.

10
00:00:41.610 --> 00:00:45.290
<v Speaker 2>But instead of having a specialist go
around the facility, taking manual

11
00:00:45.290 --> 00:00:49.689
<v Speaker 2>measurements, then uploading them to the
computer, analyzing them and reporting back

12
00:00:49.689 --> 00:00:53.610
<v Speaker 2>on them, this function is now automated
and online, so continuous.

13
00:00:55.139 --> 00:00:59.119
<v Speaker 2>Also, with traditional condition
monitoring, it was a snapshot in time.

14
00:00:59.930 --> 00:01:04.660
<v Speaker 2>When I used to run uh condition monitoring
programs, it would be once a month, you'd

15
00:01:04.660 --> 00:01:09.099
<v Speaker 2>manage to get to that machine again on
your route-based, time-based schedule, and

16
00:01:09.099 --> 00:01:13.209
<v Speaker 2>you would take a snapshot to identify the
condition of that asset at that time.

17
00:01:13.959 --> 00:01:19.360
<v Speaker 2>However, with online continuous monitoring
and an understanding of an uh an

18
00:01:19.360 --> 00:01:24.010
<v Speaker 2>operation, a machine's behavior, you can
really start to get amazing insight into

19
00:01:24.010 --> 00:01:27.970
<v Speaker 2>the condition of that asset, as well as
highlighting signs of degradation.

20
00:01:29.199 --> 00:01:32.910
<v Speaker 2>But like with any good condition
monitoring engineer, a predictive maintenance

21
00:01:32.910 --> 00:01:34.830
<v Speaker 2>solution should evolve and learn.

22
00:01:35.360 --> 00:01:39.400
<v Speaker 2>Learn about failure modes, failure
patterns, maintenance events, and give you

23
00:01:39.400 --> 00:01:41.120
<v Speaker 2>recommended actions based off of that.

24
00:01:41.989 --> 00:01:45.830
<v Speaker 2>It's also important to understand that
predictive maintenance is not a technology,

25
00:01:45.949 --> 00:01:47.269
<v Speaker 2>it's a maintenance strategy.

26
00:01:48.059 --> 00:01:50.680
<v Speaker 2>Predictive maintenance cannot be installed
in a site.

27
00:01:51.059 --> 00:01:56.610
<v Speaker 2>Um, really, it's the usage of the
information provided by a system that enables a

28
00:01:56.610 --> 00:01:59.040
<v Speaker 2>predictive maintenance strategy in your
environment.

29
00:01:59.779 --> 00:02:02.669
<v Speaker 2>It's also not about predicting when a
machine will fail.

30
00:02:03.099 --> 00:02:08.500
<v Speaker 2>Quite often I get asked, um, as a
requirement, you know, we want to know a year in

31
00:02:08.500 --> 00:02:12.289
<v Speaker 2>advance of exactly what the failure is and
exactly the day it will fail.

32
00:02:12.860 --> 00:02:17.380
<v Speaker 2>And the further out you're trying to
predict a negative event like that, the less

33
00:02:17.380 --> 00:02:18.500
<v Speaker 2>specific you can be.

34
00:02:19.309 --> 00:02:24.500
<v Speaker 2>Really what is of use to industries,
maintenance teams, to project teams, is an

35
00:02:24.500 --> 00:02:28.740
<v Speaker 2>understanding of the early signs of
degradation with enough information that they

36
00:02:28.740 --> 00:02:33.100
<v Speaker 2>can go and perform the corrective action
when it best suits their, their teams,

37
00:02:33.179 --> 00:02:36.479
<v Speaker 2>their production requirements and their
maintenance availability.

38
00:02:37.809 --> 00:02:43.279
<v Speaker 2>It's also not a big data problem, uh,
especially when you're looking at cloud

39
00:02:43.279 --> 00:02:47.229
<v Speaker 2>solutions and having to send large volumes
of data to the cloud continuously.

40
00:02:47.570 --> 00:02:50.800
<v Speaker 2>That can put a huge strain on your network
requirements.

41
00:02:51.210 --> 00:02:57.160
<v Speaker 2>So it's not about big data, really looking
at the, the fine details inside of it and

42
00:02:57.160 --> 00:03:01.110
<v Speaker 2>using machine learning to understand
inherently noisy data.

43
00:03:01.729 --> 00:03:05.589
<v Speaker 2>It's actually about providing the system
with the good quality information.

44
00:03:06.270 --> 00:03:11.369
<v Speaker 2>Um, making it, uh, a solid baseline to
learn from and highlight results.

45
00:03:12.220 --> 00:03:17.470
<v Speaker 2>Predictive maintenance should also not be
restricted to your high criticality assets

46
00:03:17.470 --> 00:03:19.649
<v Speaker 2>or only your high criticality assets.

47
00:03:20.259 --> 00:03:24.479
<v Speaker 2>The real value from a predictive
maintenance program in your environment is to cover

48
00:03:24.479 --> 00:03:26.059
<v Speaker 2>the entirety of its plant.

49
00:03:26.820 --> 00:03:30.979
<v Speaker 2>And probably the most important factor, at
least for myself, is that this isn't

50
00:03:30.979 --> 00:03:33.910
<v Speaker 2>about replacing expert knowledge or skills
you have.

51
00:03:34.220 --> 00:03:36.419
<v Speaker 2>It's about making them more efficient.

52
00:03:36.699 --> 00:03:41.059
<v Speaker 2>It's about scaling their efforts and
availability in your environment.

53
00:03:42.850 --> 00:03:47.050
<v Speaker 2>And to look at predictive maintenance, at
the core of it is condition monitoring

54
00:03:47.050 --> 00:03:47.389
<v Speaker 2>data.

55
00:03:47.970 --> 00:03:53.740
<v Speaker 2>And this is data which can come from PLCs,
drives, inverters, um, even your process

56
00:03:53.740 --> 00:03:56.089
<v Speaker 2>data is used for predictive maintenance.

57
00:03:56.449 --> 00:04:01.369
<v Speaker 2>Things like flows, pressures,
temperatures, um, understanding of batch requirements

58
00:04:01.369 --> 00:04:02.050
<v Speaker 2>and so on.

59
00:04:03.399 --> 00:04:07.720
<v Speaker 2>You know, condition-based maintenance is
in maintenance, not done on a reactive or

60
00:04:07.720 --> 00:04:08.960
<v Speaker 2>planned approach.

61
00:04:09.440 --> 00:04:14.279
<v Speaker 2>However, it is done on condition when the
machine or asset requires the maintenance

62
00:04:14.279 --> 00:04:14.839
<v Speaker 2>to be done.

63
00:04:16.179 --> 00:04:22.260
<v Speaker 2>Predictive maintenance sits around all of
those and really enables a best practice

64
00:04:22.260 --> 00:04:24.559
<v Speaker 2>approach to maintenance in an environment.

65
00:04:25.309 --> 00:04:28.209
<v Speaker 2>And with technology being what it is these
days.

66
00:04:29.070 --> 00:04:33.350
<v Speaker 2>There is now the advent of uh
conversational predictive maintenance.

67
00:04:33.869 --> 00:04:38.829
<v Speaker 2>So this is now making the information
accessible to everyone, and we will touch on

68
00:04:38.829 --> 00:04:40.489
<v Speaker 2>that a bit more in a little while.

69
00:04:42.350 --> 00:04:44.320
<v Speaker 2>But what does predictive maintenance give?

70
00:04:44.399 --> 00:04:45.179
<v Speaker 2>What is the point?

71
00:04:45.660 --> 00:04:49.140
<v Speaker 2>So I'm sure you would have all seen a PF
curve with within your time.

72
00:04:49.940 --> 00:04:55.299
<v Speaker 2>So, what predictive maintenance enables
from the very early stages is when you start

73
00:04:55.299 --> 00:04:57.399
<v Speaker 2>to see the signs of degradation.

74
00:04:58.160 --> 00:05:00.390
<v Speaker 2>It will give multiple sets of early
warning.

75
00:05:00.679 --> 00:05:05.000
<v Speaker 2>So an indication that something is
degrading with enough information that you can

76
00:05:05.000 --> 00:05:09.989
<v Speaker 2>perform maintenance before your existing
systems will set off an alarm before the

77
00:05:10.010 --> 00:05:13.179
<v Speaker 2>the process production machine has failed.

78
00:05:14.209 --> 00:05:16.019
<v Speaker 2>But as I said, it continues to learn.

79
00:05:16.649 --> 00:05:21.529
<v Speaker 2>So once it does learn and starts to
understand what these signs of degradation mean,

80
00:05:21.790 --> 00:05:23.549
<v Speaker 2>it can start to be more specific.

81
00:05:24.089 --> 00:05:28.730
<v Speaker 2>So from the, the stage of early
degradation, from those early signs, you may get

82
00:05:28.730 --> 00:05:30.519
<v Speaker 2>some anomalies or trends detected.

83
00:05:31.010 --> 00:05:36.649
<v Speaker 2>Then insight descriptions, letting the
user know what is happening, um, interpreting

84
00:05:36.649 --> 00:05:38.149
<v Speaker 2>the, the data.

85
00:05:38.570 --> 00:05:42.730
<v Speaker 2>You may even get some prognostic, uh,
forecasting and a lead time to repair

86
00:05:42.730 --> 00:05:43.709
<v Speaker 2>indications.

87
00:05:43.970 --> 00:05:48.920
<v Speaker 2>Again, long before any conditions or
thresholds have been violated on your

88
00:05:48.920 --> 00:05:52.890
<v Speaker 2>environment and your existing systems
would have notified you that something has

89
00:05:52.890 --> 00:05:53.609
<v Speaker 2>gone bang.

90
00:05:53.820 --> 00:05:55.350
<v Speaker 2>Whoops, it's all gone wrong.

91
00:05:55.769 --> 00:06:01.570
<v Speaker 2>So it's about giving insight early on that
is actionable with enough time for users

92
00:06:01.570 --> 00:06:06.089
<v Speaker 2>to, to make use of this and to plan in
maintenance when it best suits the company

93
00:06:06.089 --> 00:06:07.609
<v Speaker 2>and the environment they're working in.

94
00:06:10.339 --> 00:06:14.760
<v Speaker 2>It's also important to note that
predictive maintenance is not about thresholds.

95
00:06:15.019 --> 00:06:19.160
<v Speaker 2>And this is something I've personally have
had to try and combat quite a lot in the

96
00:06:19.339 --> 00:06:25.380
<v Speaker 2>eastern part of the world, there are
hundreds if not thousands of companies with

97
00:06:25.380 --> 00:06:26.769
<v Speaker 2>hardware, sensors.

98
00:06:27.100 --> 00:06:31.179
<v Speaker 2>They have potentially developed their own
software solution, which is rule or

99
00:06:31.179 --> 00:06:32.179
<v Speaker 2>threshold-based.

100
00:06:32.779 --> 00:06:36.059
<v Speaker 2>Um, and they're selling that as predictive
maintenance, and it just isn't.

101
00:06:37.089 --> 00:06:41.559
<v Speaker 2>So if we look at the graph on the
right-hand side, let's say that is uh the current

102
00:06:41.559 --> 00:06:45.420
<v Speaker 2>profile being amps for an electric motor
over a week period.

103
00:06:46.079 --> 00:06:51.019
<v Speaker 2>So considering your, your production
cycles, Monday morning, there's an idle period

104
00:06:51.359 --> 00:06:54.779
<v Speaker 2>and then you start producing production uh
Y or X.

105
00:06:55.869 --> 00:06:59.640
<v Speaker 2>Then you may go down for commissioning,
uh, pre-setup and an idle phase, and then

106
00:06:59.640 --> 00:07:02.239
<v Speaker 2>you start in production for product Y.

107
00:07:03.380 --> 00:07:07.179
<v Speaker 2>You may go down for a little bit of
maintenance midweek, um, something that which

108
00:07:07.179 --> 00:07:07.959
<v Speaker 2>was planned in.

109
00:07:08.140 --> 00:07:13.339
<v Speaker 2>You carry on with production of P Y or
product Y, and then you tail off for the rest

110
00:07:13.339 --> 00:07:13.920
<v Speaker 2>of the week.

111
00:07:14.640 --> 00:07:20.230
<v Speaker 2>So that is a, an example of a typical
production requirements and profile of an

112
00:07:20.230 --> 00:07:21.730
<v Speaker 2>electric motor for a week's period.

113
00:07:22.670 --> 00:07:27.730
<v Speaker 2>So with the, the system in question, the
threshold based, the company will come in,

114
00:07:28.010 --> 00:07:32.619
<v Speaker 2>install sensors, and possibly even set up
some thresholds or limits for you.

115
00:07:33.029 --> 00:07:37.510
<v Speaker 2>These could be based off of ISO standards,
the size of the motor, the bearing type,

116
00:07:37.589 --> 00:07:41.209
<v Speaker 2>the uh rigidity of the, the base and
mounting.

117
00:07:41.829 --> 00:07:44.029
<v Speaker 2>But then you have a simple threshold
limit.

118
00:07:45.059 --> 00:07:50.380
<v Speaker 2>Well, it is missing all of the important
detail, especially around the operational

119
00:07:50.380 --> 00:07:51.500
<v Speaker 2>complexity that you have.

120
00:07:52.869 --> 00:07:56.130
<v Speaker 2>So, maybe someone in your team needs to go
and reset the threshold.

121
00:07:58.040 --> 00:08:02.000
<v Speaker 2>Well, maybe the 2nd time, they have gotten
it a little bit more rights, a little bit

122
00:08:02.000 --> 00:08:07.279
<v Speaker 2>more fit for purpose for um an
understanding of the production requirements for

123
00:08:07.279 --> 00:08:07.959
<v Speaker 2>product X.

124
00:08:08.709 --> 00:08:11.760
<v Speaker 2>However, it's missed the detail for
product Y.

125
00:08:12.730 --> 00:08:15.429
<v Speaker 2>So, then the threshold needs to be reset
again.

126
00:08:16.720 --> 00:08:21.480
<v Speaker 2>And potentially this time, they've
captured or managed to get the right limit to

127
00:08:21.480 --> 00:08:23.279
<v Speaker 2>capture the event that you're looking for.

128
00:08:24.059 --> 00:08:29.739
<v Speaker 2>However, by this time, the teams uh on
site that need to use this have lost trust in

129
00:08:29.739 --> 00:08:29.940
<v Speaker 2>it.

130
00:08:29.980 --> 00:08:31.140
<v Speaker 2>They've been overalerted.

131
00:08:31.220 --> 00:08:36.619
<v Speaker 2>They've uh performed maintenance and
essentially wasted money because there is

132
00:08:36.619 --> 00:08:42.539
<v Speaker 2>limited understanding of the complexity of
the asset, um, and the production

133
00:08:42.539 --> 00:08:43.169
<v Speaker 2>requirements.

134
00:08:43.580 --> 00:08:45.840
<v Speaker 2>So really, it's not fit for purpose.

135
00:08:46.489 --> 00:08:51.159
<v Speaker 2>What is needed, and especially within the
pharmaceutical industries, is an

136
00:08:51.159 --> 00:08:56.239
<v Speaker 2>understanding of normal production
complexity, an understanding of your machine

137
00:08:56.239 --> 00:08:56.479
<v Speaker 2>state.

138
00:08:56.590 --> 00:08:57.710
<v Speaker 2>What is it trying to do?

139
00:08:57.799 --> 00:08:59.159
<v Speaker 2>What speed is it running at?

140
00:08:59.229 --> 00:09:03.440
<v Speaker 2>What product does it have or batch is it
running through, um, what are the, the

141
00:09:03.440 --> 00:09:05.500
<v Speaker 2>output capacity requirements?

142
00:09:06.119 --> 00:09:10.719
<v Speaker 2>And with that understanding, a system
which can highlight the specific signs of

143
00:09:10.719 --> 00:09:14.320
<v Speaker 2>degradation with an understanding of
machine state.

144
00:09:18.280 --> 00:09:23.400
<v Speaker 2>And really, I think this is gonna be a
universal requirement, but also very

145
00:09:23.400 --> 00:09:27.679
<v Speaker 2>specifically within the phar
pharmaceutical industry, is a system which is

146
00:09:27.679 --> 00:09:31.409
<v Speaker 2>applicable to every single type of machine
that you have.

147
00:09:31.760 --> 00:09:36.479
<v Speaker 2>You know, from your process environments,
having your fermenters, uh, chromatography

148
00:09:36.479 --> 00:09:40.659
<v Speaker 2>columns, uh, Lilysers, all the way through
to your warehousing and distribution.

149
00:09:41.919 --> 00:09:47.760
<v Speaker 2>If you were to implement a solution or a
system that is specialist in one of these

150
00:09:47.760 --> 00:09:49.469
<v Speaker 2>areas, fantastic.

151
00:09:49.520 --> 00:09:52.179
<v Speaker 2>I'm sure it will be very good at that one
area.

152
00:09:52.640 --> 00:09:57.789
<v Speaker 2>But when you look at the full range of
production um and environments that you have,

153
00:09:58.159 --> 00:10:02.299
<v Speaker 2>you could now end up with 5, 10, even 20
individual solutions.

154
00:10:02.919 --> 00:10:07.880
<v Speaker 2>Therefore, that is problematic because you
now have multiple logins, people that

155
00:10:07.880 --> 00:10:09.469
<v Speaker 2>need multiple sets of training.

156
00:10:09.880 --> 00:10:16.150
<v Speaker 2>It's really hard to I truly understand the
differences in capability, layouts, user

157
00:10:16.150 --> 00:10:18.380
<v Speaker 2>interfaces between these different
systems.

158
00:10:18.909 --> 00:10:24.109
<v Speaker 2>What is needed uh in industry and in the
and in the pharmaceutical industry is one

159
00:10:24.109 --> 00:10:26.690
<v Speaker 2>solution which is fit for purpose for
everything.

160
00:10:28.909 --> 00:10:34.229
<v Speaker 2>Predictive maintenance is evolving and
these uh requirements are very much uh in

161
00:10:34.229 --> 00:10:34.969
<v Speaker 2>your grasp.

162
00:10:35.950 --> 00:10:40.250
<v Speaker 2>So predictive maintenance, really, if we
think about it as uh online condition

163
00:10:40.250 --> 00:10:44.590
<v Speaker 2>monitoring with uh automated knowledge and
understanding, that could fall under

164
00:10:44.590 --> 00:10:46.969
<v Speaker 2>diagnostics and statistical forecasting.

165
00:10:47.380 --> 00:10:52.429
<v Speaker 2>Essentially, how is my machine, my
process, my line behaving now, and how is it

166
00:10:52.429 --> 00:10:53.989
<v Speaker 2>going to be behaving in the future?

167
00:10:54.070 --> 00:10:56.229
<v Speaker 2>In a few days, in a few weeks, in a few
months.

168
00:10:58.130 --> 00:11:03.049
<v Speaker 2>Uh, the solutions that you are assessing
should be able to prioritize maintenance.

169
00:11:03.510 --> 00:11:07.349
<v Speaker 2>And again, this is not based off of
machine or asset criticality.

170
00:11:07.729 --> 00:11:11.929
<v Speaker 2>It shouldn't just be, oh, my very
critical, uh, machine is degrading.

171
00:11:11.969 --> 00:11:16.650
<v Speaker 2>I need to run off to that immediately at
the first sign of, uh, an increase in

172
00:11:16.650 --> 00:11:17.969
<v Speaker 2>current or vibration.

173
00:11:18.570 --> 00:11:25.080
<v Speaker 2>No, it should be about understanding how
Concerned the user should be about an

174
00:11:25.080 --> 00:11:29.919
<v Speaker 2>event, about an insight, and to be able to
prioritize this so that your maintenance

175
00:11:29.919 --> 00:11:35.960
<v Speaker 2>teams and uh project teams can decide
which assets they want to work on, which is

176
00:11:35.960 --> 00:11:39.440
<v Speaker 2>going to have the highest impact or which
is at risk of failing first.

177
00:11:40.349 --> 00:11:47.109
<v Speaker 2>And we, with technology being where it is
and the um the huge boom in generative

178
00:11:47.109 --> 00:11:51.900
<v Speaker 2>AI functionalities, generative AI and
large language models have really

179
00:11:51.900 --> 00:11:56.950
<v Speaker 2>revolutionized the predictive maintenance
space, really making that information

180
00:11:56.950 --> 00:12:02.820
<v Speaker 2>accessible to everyone and also giving The
ability to ask questions to the system

181
00:12:02.820 --> 00:12:04.679
<v Speaker 2>which is doing the monitoring for you.

182
00:12:06.659 --> 00:12:09.739
<v Speaker 2>Oh, why would geneative AI even be a
thing?

183
00:12:09.849 --> 00:12:11.809
<v Speaker 2>Why would it be used for predictive
maintenance?

184
00:12:12.250 --> 00:12:16.190
<v Speaker 2>Well, what we've found is that there is an
inherent lack of data literacy.

185
00:12:17.119 --> 00:12:19.640
<v Speaker 2>And this is at no fault, uh, for the
maintenance teams.

186
00:12:19.799 --> 00:12:25.020
<v Speaker 2>They are extremely experienced, extremely
knowledgeable in maintenance, in

187
00:12:25.020 --> 00:12:26.869
<v Speaker 2>engineering, in their field.

188
00:12:27.359 --> 00:12:33.140
<v Speaker 2>However, Expecting them to have any sort
of data analytics skills is in many cases

189
00:12:33.140 --> 00:12:35.739
<v Speaker 2>unreasonable, and what we found is they
just don't.

190
00:12:36.760 --> 00:12:38.830
<v Speaker 2>They also don't have time to interpret
graphs.

191
00:12:39.669 --> 00:12:44.109
<v Speaker 2>Most of the, the customers that we, we
assist, their maintenance teams are stretched

192
00:12:44.109 --> 00:12:44.989
<v Speaker 2>thin already.

193
00:12:45.190 --> 00:12:49.390
<v Speaker 2>There have been a huge amount of budget
constraints limiting their, their team's

194
00:12:49.390 --> 00:12:54.390
<v Speaker 2>size, the, the team's availability, and
really having to make do with what they have

195
00:12:54.390 --> 00:12:54.989
<v Speaker 2>on site.

196
00:12:55.229 --> 00:12:56.590
<v Speaker 2>So time is a huge thing.

197
00:12:57.299 --> 00:13:00.809
<v Speaker 2>Also, whenever using a new system, who do
you ask questions to?

198
00:13:01.260 --> 00:13:06.619
<v Speaker 2>If it's just been installed and
preconfigured and then if you have to potentially

199
00:13:06.619 --> 00:13:11.619
<v Speaker 2>pay the supplier for any bits of uh
consultancy afterwards, you are very limited

200
00:13:11.619 --> 00:13:15.940
<v Speaker 2>with the ability to ask questions and
really get use from a, from a system.

201
00:13:17.119 --> 00:13:22.929
<v Speaker 2>Also, language barriers for global
organizations, you know, you will have a wide

202
00:13:22.929 --> 00:13:26.270
<v Speaker 2>range of languages spoken from your
skilled expertise.

203
00:13:26.820 --> 00:13:32.330
<v Speaker 2>So being able to speak to someone or
experts in your native language so you get a

204
00:13:32.330 --> 00:13:34.090
<v Speaker 2>true understanding is difficult.

205
00:13:35.590 --> 00:13:37.739
<v Speaker 2>There's also the aspect of lost knowledge.

206
00:13:38.229 --> 00:13:42.010
<v Speaker 2>I'm sure we are all facing and
experiencing these challenges.

207
00:13:42.429 --> 00:13:46.309
<v Speaker 2>In your environment, you'll have
maintenance engineers who have been working with

208
00:13:46.309 --> 00:13:51.150
<v Speaker 2>you or in your environment or in your
field for decades, and now the up and comers,

209
00:13:51.190 --> 00:13:52.770
<v Speaker 2>the apprentices coming through.

210
00:13:53.349 --> 00:13:58.070
<v Speaker 2>So anything that your experienced
engineers cannot impart onto the, the new up and

211
00:13:58.070 --> 00:13:59.770
<v Speaker 2>comers, that knowledge is lost.

212
00:14:00.400 --> 00:14:05.299
<v Speaker 2>And we all know of uh, that person on site
who can walk around and he knows when,

213
00:14:05.640 --> 00:14:07.799
<v Speaker 2>uh, when the homogenizer isn't sounding
right.

214
00:14:07.880 --> 00:14:11.840
<v Speaker 2>He knows that there may be something going
wrong and can go and fix it just based

215
00:14:11.840 --> 00:14:12.500
<v Speaker 2>off of that knowledge.

216
00:14:13.239 --> 00:14:18.659
<v Speaker 2>That is experience which, yes, can be
taught, but isn't always imparted on.

217
00:14:19.039 --> 00:14:22.989
<v Speaker 2>So that, that type of knowledge and
understanding of an environment, of a process,

218
00:14:23.039 --> 00:14:25.840
<v Speaker 2>of a, a problem, more importantly, does
get lost.

219
00:14:27.349 --> 00:14:31.770
<v Speaker 2>It's also it's also difficult to share
knowledge and collaborate with global

220
00:14:31.770 --> 00:14:32.289
<v Speaker 2>organizations.

221
00:14:32.820 --> 00:14:36.549
<v Speaker 2>Quite often you'll find a company doing
exactly the same thing.

222
00:14:36.669 --> 00:14:41.140
<v Speaker 2>Maybe they have a facility in India and
one in the UK as an example.

223
00:14:41.669 --> 00:14:44.309
<v Speaker 2>Their processes are pretty much the same.

224
00:14:44.590 --> 00:14:46.270
<v Speaker 2>Um, the products are identical.

225
00:14:46.349 --> 00:14:49.849
<v Speaker 2>The machinery, may be some variation on
OEMs.

226
00:14:50.429 --> 00:14:53.669
<v Speaker 2>However, they are like for like and
extremely comparable.

227
00:14:54.219 --> 00:14:55.520
<v Speaker 2>But how do they collaborate?

228
00:14:55.700 --> 00:14:57.520
<v Speaker 2>How do they share knowledge between each
other?

229
00:14:57.979 --> 00:15:03.340
<v Speaker 2>Often we find these organizational silos
in place where there isn't collaboration,

230
00:15:03.419 --> 00:15:06.109
<v Speaker 2>there isn't best practice sharing between
them.

231
00:15:06.539 --> 00:15:08.539
<v Speaker 2>And this is where generative AI comes in.

232
00:15:10.799 --> 00:15:13.299
<v Speaker 2>First of all, it's about getting instant
answers to your questions.

233
00:15:13.719 --> 00:15:20.630
<v Speaker 2>So generative AI is able to interpret the
data, the graphs, the, the data analytics

234
00:15:20.630 --> 00:15:24.119
<v Speaker 2>for you and present it to the user in the
natural language.

235
00:15:24.719 --> 00:15:30.020
<v Speaker 2>It also provides the ability to ask
questions to, to the, the system.

236
00:15:31.150 --> 00:15:35.869
<v Speaker 2>Um, the answers you'll get will be from an
inherent knowledge, you know, trained off

237
00:15:35.869 --> 00:15:37.250
<v Speaker 2>of the knowledge on the internet.

238
00:15:37.750 --> 00:15:42.270
<v Speaker 2>But for companies that are, are feeding
their large language models with expert

239
00:15:42.270 --> 00:15:44.770
<v Speaker 2>knowledge, it starts to become extremely
powerful.

240
00:15:45.190 --> 00:15:49.869
<v Speaker 2>So this expert knowledge could be in the
form of uh condition monitoring techniques,

241
00:15:49.989 --> 00:15:55.150
<v Speaker 2>maintenance and reliability, uh,
practices, as well as user manuals, maintenance

242
00:15:55.150 --> 00:15:57.429
<v Speaker 2>manuals, decision trees.

243
00:15:57.890 --> 00:16:02.809
<v Speaker 2>Um, and having instant answers to that
expert knowledge and having instant reference

244
00:16:02.809 --> 00:16:04.710
<v Speaker 2>to that knowledge is extremely powerful.

245
00:16:05.859 --> 00:16:11.650
<v Speaker 2>Generative AI retains the knowledge
forever, and I know forever is a definite term,

246
00:16:11.809 --> 00:16:14.309
<v Speaker 2>but for the foreseeable future we could
say.

247
00:16:15.299 --> 00:16:18.539
<v Speaker 2>Because it is going to continuously learn.

248
00:16:18.859 --> 00:16:24.900
<v Speaker 2>It's going to have the storm of Uh,
failure patterns of maintenance events.

249
00:16:25.119 --> 00:16:29.179
<v Speaker 2>It has reference to the expert knowledge,
your user manuals, maintenance manuals,

250
00:16:29.479 --> 00:16:34.359
<v Speaker 2>and it can really start to build a deep,
deep understanding of your environment and

251
00:16:34.359 --> 00:16:39.119
<v Speaker 2>what is important to you and make this
knowledge accessible to all of you new up and

252
00:16:39.119 --> 00:16:44.609
<v Speaker 2>comers, the, the apprentices who don't
necessarily know what this components for or

253
00:16:44.609 --> 00:16:47.440
<v Speaker 2>uh what the part numbers of that um asset
are.

254
00:16:48.570 --> 00:16:54.169
<v Speaker 2>And it makes it extremely easy to share
knowledge across all locations because as

255
00:16:54.169 --> 00:16:58.429
<v Speaker 2>you would expect, large language models
work in pretty much every language available

256
00:16:58.849 --> 00:17:04.540
<v Speaker 2>and because the users can interact with it
in their native language and get results

257
00:17:04.540 --> 00:17:10.250
<v Speaker 2>back in their native language, it means
that your, your cross-country and uh cross

258
00:17:10.250 --> 00:17:13.229
<v Speaker 2>geography collaboration is heavily
enhanced.

259
00:17:13.719 --> 00:17:19.260
<v Speaker 2>Really tying in all of this human
knowledge into one space and making it accessible

260
00:17:19.260 --> 00:17:22.000
<v Speaker 2>to every single one that uses the system.

261
00:17:24.150 --> 00:17:29.250
<v Speaker 2>Really, predictive maintenance should be
the convergence of the human-machine

262
00:17:29.250 --> 00:17:30.020
<v Speaker 2>interface.

263
00:17:30.310 --> 00:17:35.510
<v Speaker 2>It should be having a true understanding
of your machines, of your processes, using

264
00:17:35.510 --> 00:17:41.109
<v Speaker 2>your OT data, IT data, but also to learn
from the human element, the expertise that

265
00:17:41.109 --> 00:17:41.910
<v Speaker 2>you have on site.

266
00:17:43.069 --> 00:17:48.140
<v Speaker 2>But predictive maintenance or any
digitalization program should be focused on

267
00:17:48.140 --> 00:17:49.069
<v Speaker 2>business outcomes.

268
00:17:50.430 --> 00:17:52.930
<v Speaker 2>And we'll discuss some of those in a
while.

269
00:17:54.650 --> 00:17:58.869
<v Speaker 2>An example of deployments within the, the,
the farmer industry.

270
00:17:59.810 --> 00:18:05.250
<v Speaker 2>Um, this this specific customer, uh, they
make insulin, um, and the machines that

271
00:18:05.250 --> 00:18:10.329
<v Speaker 2>we're monitoring for them specifically is
the vial filling line, so little vials.

272
00:18:10.810 --> 00:18:14.810
<v Speaker 2>One of the key factors though is that we
had to use existing data from their

273
00:18:14.810 --> 00:18:16.050
<v Speaker 2>existing infrastructure.

274
00:18:16.369 --> 00:18:20.880
<v Speaker 2>This wasn't a a project to come in and
install huge amounts of hardware, sensors,

275
00:18:21.089 --> 00:18:22.520
<v Speaker 2>data collection devices.

276
00:18:22.880 --> 00:18:27.609
<v Speaker 2>It was about can we derive value from
their existing infrastructure, their existing

277
00:18:27.609 --> 00:18:28.050
<v Speaker 2>investment.

278
00:18:28.630 --> 00:18:29.880
<v Speaker 2>And the answer is yes.

279
00:18:30.770 --> 00:18:36.579
<v Speaker 2>Um, some of their success criteria and the
pain points was on the silicone valves

280
00:18:36.579 --> 00:18:39.739
<v Speaker 2>specifically and the degradation that
these caused.

281
00:18:40.180 --> 00:18:44.150
<v Speaker 2>The fact that the, this company didn't
have a centralized visibility, these were

282
00:18:44.150 --> 00:18:49.790
<v Speaker 2>happening unexpectedly and halting
production until they were able to inspect 20 or

283
00:18:49.790 --> 00:18:53.420
<v Speaker 2>30 silicone valves to find which one had
actually failed.

284
00:18:55.609 --> 00:18:59.449
<v Speaker 2>So we started working with this company in
January 2023.

285
00:18:59.969 --> 00:19:04.329
<v Speaker 2>Thankfully for us and I would say for them
as well, they actually had tried to

286
00:19:04.329 --> 00:19:06.040
<v Speaker 2>develop their own system in-house.

287
00:19:06.609 --> 00:19:08.609
<v Speaker 2>Unfortunately, that did fail.

288
00:19:09.050 --> 00:19:13.109
<v Speaker 2>However, fortunately for us, it meant that
they had done a huge amount of

289
00:19:13.160 --> 00:19:13.290
<v Speaker 2>investigation.

290
00:19:13.369 --> 00:19:15.819
<v Speaker 2>They weren't starting at a zero knowledge
base.

291
00:19:16.319 --> 00:19:20.500
<v Speaker 2>They knew what the complexity of
predictive maintenance and had done a lot of the

292
00:19:20.500 --> 00:19:22.170
<v Speaker 2>data work beforehand.

293
00:19:22.939 --> 00:19:27.579
<v Speaker 2>So, quite rapidly, um, you know, a month
and a half later, we had the system

294
00:19:27.579 --> 00:19:30.900
<v Speaker 2>integration with data being pushed into
the application.

295
00:19:31.979 --> 00:19:35.300
<v Speaker 2>Their first success was roughly 6 months
later, from there.

296
00:19:36.050 --> 00:19:41.770
<v Speaker 2>And This first success was just the first
interesting event that was detected.

297
00:19:42.150 --> 00:19:46.349
<v Speaker 2>And this also goes to show why going for
your most high criticality assets isn't

298
00:19:46.349 --> 00:19:50.790
<v Speaker 2>always the best approach, because high
criticality assets are typically

299
00:19:50.790 --> 00:19:51.750
<v Speaker 2>over-maintained.

300
00:19:52.400 --> 00:19:55.969
<v Speaker 2>Um, a huge amount of maintenance, effort,
time, and money is spent on them.

301
00:19:56.469 --> 00:20:01.150
<v Speaker 2>So when you're looking at trying to
achieve value and a return on investment, in a

302
00:20:01.150 --> 00:20:05.550
<v Speaker 2>twelve-month period, what are, what is the
likelihood that something interesting is

303
00:20:05.550 --> 00:20:09.630
<v Speaker 2>gonna happen, that something is actually
going to degrade, that you can show results

304
00:20:09.630 --> 00:20:11.569
<v Speaker 2>and show something working?

305
00:20:12.619 --> 00:20:15.849
<v Speaker 2>So that is why we go for value.

306
00:20:15.939 --> 00:20:19.780
<v Speaker 2>And again, I will discuss value a little
bit later on in this presentation and how

307
00:20:19.780 --> 00:20:21.680
<v Speaker 2>we target where the value is.

308
00:20:22.219 --> 00:20:27.140
<v Speaker 2>But for this customer, even though the
silicone valves were the main cause of

309
00:20:27.140 --> 00:20:32.140
<v Speaker 2>frustration and loss in production, it
still took over 6 months before anything

310
00:20:32.140 --> 00:20:33.890
<v Speaker 2>happened for us to show results.

311
00:20:34.380 --> 00:20:39.819
<v Speaker 2>However, with one notification, they were
able to achieve a 3 times return on

312
00:20:39.819 --> 00:20:40.339
<v Speaker 2>investment.

313
00:20:41.300 --> 00:20:47.459
<v Speaker 2>By the end of that year, over 300,000
units or additional units were able to be

314
00:20:47.459 --> 00:20:50.479
<v Speaker 2>produced because of reduction in
breakdowns.

315
00:20:50.939 --> 00:20:55.219
<v Speaker 2>So, early indication that something was
degrading with enough time for the

316
00:20:55.219 --> 00:21:00.050
<v Speaker 2>maintenance and project teams to assess
when they wanted to do the production, to

317
00:21:00.050 --> 00:21:05.099
<v Speaker 2>fit it in with their planned stops, and to
still maintain production through all of

318
00:21:05.099 --> 00:21:06.849
<v Speaker 2>those um additional units.

319
00:21:07.260 --> 00:21:10.959
<v Speaker 2>Huge amounts of value in a relatively
short space of time.

320
00:21:14.369 --> 00:21:16.729
<v Speaker 2>Let's talk about some of the challenges
faced.

321
00:21:18.780 --> 00:21:22.619
<v Speaker 2>Again, I would say these are quite
universal, the maintenance of uptime.

322
00:21:23.530 --> 00:21:28.109
<v Speaker 2>For maintenance teams, again, they are
very good, very specialized, very uh

323
00:21:28.810 --> 00:21:31.170
<v Speaker 2>experienced in performing maintenance.

324
00:21:31.689 --> 00:21:34.010
<v Speaker 2>However, when do you perform maintenance?

325
00:21:34.410 --> 00:21:35.969
<v Speaker 2>Well, when something fails.

326
00:21:36.589 --> 00:21:41.469
<v Speaker 2>You know, the process is something fails,
potentially an operator, uh, pick up the

327
00:21:41.469 --> 00:21:45.989
<v Speaker 2>phone, call the maintenance department and
say, hey, my, uh, chromatography column

328
00:21:45.989 --> 00:21:48.020
<v Speaker 2>has stopped, uh, something's wrong.

329
00:21:48.630 --> 00:21:52.949
<v Speaker 2>They would go out, then investigate, go
back, planning, make sure they have the

330
00:21:52.949 --> 00:21:57.469
<v Speaker 2>spares, then go back through, perform the
repair, make sure everything's tight and

331
00:21:57.469 --> 00:22:02.349
<v Speaker 2>fitted, start it up, oh, didn't quite get
it right, and then they have to start the

332
00:22:02.349 --> 00:22:04.469
<v Speaker 2>process of investigating, trying something
new.

333
00:22:05.099 --> 00:22:08.550
<v Speaker 2>And repairing it, uh, iteratively, let's
say.

334
00:22:09.060 --> 00:22:12.640
<v Speaker 2>That's potentially the worst-case
scenario, but still quite common.

335
00:22:13.140 --> 00:22:18.819
<v Speaker 2>So without the visibility of the signs of
degradation, the maintenance of uptime can

336
00:22:18.819 --> 00:22:20.479
<v Speaker 2>be extremely difficult.

337
00:22:21.060 --> 00:22:25.380
<v Speaker 2>Likewise, precise and correct maintenance
because in any environment, from my

338
00:22:25.380 --> 00:22:32.010
<v Speaker 2>experience, About 5 to 6% of your assets
or machines will have signs of

339
00:22:32.010 --> 00:22:32.689
<v Speaker 2>degradation.

340
00:22:33.349 --> 00:22:35.430
<v Speaker 2>But what about the other 95%?

341
00:22:36.130 --> 00:22:40.369
<v Speaker 2>You know, precise and correct maintenance
is, is about having an understanding that

342
00:22:40.369 --> 00:22:44.569
<v Speaker 2>for the 95%, which are in perfectly good
condition, that they're fine.

343
00:22:44.650 --> 00:22:49.349
<v Speaker 2>Don't touch them, don't induce maintenance
uh induced failure to them.

344
00:22:49.849 --> 00:22:51.250
<v Speaker 2>They're in perfectly good condition.

345
00:22:51.329 --> 00:22:55.849
<v Speaker 2>You can extend your planned maintenance
intervals out a little bit until you have

346
00:22:55.849 --> 00:23:00.689
<v Speaker 2>true trust and visibility of the signs of
degradation, and you can fit that in with

347
00:23:00.689 --> 00:23:05.189
<v Speaker 2>your production cycles and really benefit
from precise and correct maintenance.

348
00:23:06.609 --> 00:23:10.689
<v Speaker 2>In any environment, we all want to reduce
the risk and operating costs.

349
00:23:11.400 --> 00:23:14.199
<v Speaker 2>And a huge factor of risk is maintenance.

350
00:23:14.869 --> 00:23:19.229
<v Speaker 2>The more someone has to touch a machine,
the more risk and the more cost there is

351
00:23:19.229 --> 00:23:19.859
<v Speaker 2>involved with that.

352
00:23:21.359 --> 00:23:23.339
<v Speaker 2>I think what we're all finding is uh.

353
00:23:24.380 --> 00:23:30.300
<v Speaker 2>Having separate systems which together
make your entire production, but individually

354
00:23:30.300 --> 00:23:33.270
<v Speaker 2>they are completely separate and really
hard to talk to.

355
00:23:33.699 --> 00:23:36.040
<v Speaker 2>So this is the OTIT convergence.

356
00:23:36.380 --> 00:23:42.619
<v Speaker 2>You know, your OT environments, PLCs,
drives, inverters, automating a lot of your

357
00:23:42.619 --> 00:23:48.180
<v Speaker 2>process, along with IT systems such as
your maintenance management system or even

358
00:23:48.180 --> 00:23:50.479
<v Speaker 2>your EAM ERP systems.

359
00:23:51.010 --> 00:23:55.260
<v Speaker 2>Which together give you a nice big
picture, but they are still separated systems.

360
00:23:55.339 --> 00:23:59.819
<v Speaker 2>They don't talk to each other and they
make a true understanding and true value

361
00:23:59.819 --> 00:24:00.420
<v Speaker 2>extremely hard.

362
00:24:01.390 --> 00:24:06.430
<v Speaker 2>And sustainability, you know, this day and
age, we do all need to have an eye on our

363
00:24:06.430 --> 00:24:09.750
<v Speaker 2>sustainability and carbon footprints,
let's say.

364
00:24:11.170 --> 00:24:13.479
<v Speaker 2>This is also compounded by scalability.

365
00:24:13.859 --> 00:24:17.119
<v Speaker 2>Again, I spoke about a, a single solution
to monitor everything.

366
00:24:18.069 --> 00:24:21.089
<v Speaker 2>Um, and also not being specialist in any
one area.

367
00:24:21.430 --> 00:24:25.349
<v Speaker 2>In an environment, uh, we need to have the
effects of scale.

368
00:24:25.430 --> 00:24:26.729
<v Speaker 2>It needs to be scalable.

369
00:24:27.150 --> 00:24:31.739
<v Speaker 2>And not scalable in the sense of throwing
more humans at it, more resources at it,

370
00:24:31.829 --> 00:24:34.430
<v Speaker 2>more money at it, but truly scalable.

371
00:24:35.540 --> 00:24:37.979
<v Speaker 2>Something which can target your data
silos.

372
00:24:38.579 --> 00:24:43.260
<v Speaker 2>You know, I spoke about the difference
between IT and OT, but what about within

373
00:24:43.260 --> 00:24:44.180
<v Speaker 2>those systems?

374
00:24:44.290 --> 00:24:47.000
<v Speaker 2>Uh, you know, the individual OT.

375
00:24:47.880 --> 00:24:54.020
<v Speaker 2>Uh, layers of infrastructure talking to
each other, being able to be interoperable

376
00:24:54.280 --> 00:24:56.599
<v Speaker 2>and really get the value together.

377
00:24:57.400 --> 00:24:59.260
<v Speaker 2>There is a skills shortage going on.

378
00:24:59.560 --> 00:25:04.959
<v Speaker 2>Um, skills in the sense of, uh, new
maintenance engineers, new production teams, um,

379
00:25:05.040 --> 00:25:10.680
<v Speaker 2>people not wanting to be influencers or
social media personalities, but going into

380
00:25:10.680 --> 00:25:13.020
<v Speaker 2>the STEM, uh, world, really.

381
00:25:14.530 --> 00:25:20.530
<v Speaker 2>Adoption is always going to be a massive
thing in any digitization project, um, any

382
00:25:20.530 --> 00:25:21.949
<v Speaker 2>kind of predictive maintenance.

383
00:25:22.729 --> 00:25:27.949
<v Speaker 2>Uh, project that you would run, getting
the human element to start using a new

384
00:25:27.949 --> 00:25:32.050
<v Speaker 2>technology, a new system, and being happy
with it, as well as collaboration and

385
00:25:32.050 --> 00:25:33.869
<v Speaker 2>engagements between your global teams.

386
00:25:34.729 --> 00:25:39.969
<v Speaker 2>And then if we look at something more
pharmaceutical specific, you know, regulatory

387
00:25:39.969 --> 00:25:41.640
<v Speaker 2>compliance is massive.

388
00:25:42.089 --> 00:25:45.640
<v Speaker 2>Your GXP uh processes are insane.

389
00:25:45.930 --> 00:25:49.520
<v Speaker 2>I'm sure a lot of you will know the, the
trouble you have to go through, even just

390
00:25:49.520 --> 00:25:53.160
<v Speaker 2>to drill a single hole within your process
environment.

391
00:25:53.329 --> 00:25:57.589
<v Speaker 2>The, the amount of paperwork required to
just to drill a hole is huge.

392
00:25:58.170 --> 00:26:01.209
<v Speaker 2>Within pharmaceuticals, however, my lights
have gone down.

393
00:26:01.369 --> 00:26:02.550
<v Speaker 2>Give me one second.

394
00:26:05.349 --> 00:26:05.989
<v Speaker 2>There you go.

395
00:26:07.709 --> 00:26:08.540
<v Speaker 2>Apologies for that.

396
00:26:09.540 --> 00:26:14.140
<v Speaker 2>Within the pharmaceutical environment, the
processes are complex, the machinery is

397
00:26:14.140 --> 00:26:15.040
<v Speaker 2>quite complex.

398
00:26:15.339 --> 00:26:22.300
<v Speaker 2>You know, a lot can happen with your loss
in production, um, based off of very

399
00:26:22.300 --> 00:26:25.020
<v Speaker 2>small things happening inside of the
internal workings.

400
00:26:25.420 --> 00:26:30.420
<v Speaker 2>Um, so a true understanding of the
complexity of the machines and production is, is

401
00:26:30.420 --> 00:26:32.400
<v Speaker 2>a problem or a challenge.

402
00:26:32.979 --> 00:26:37.839
<v Speaker 2>The high cost and loss production for
pharmaceuticals, as you all know, is huge.

403
00:26:38.369 --> 00:26:39.890
<v Speaker 2>Especially when you start looking at.

404
00:26:40.530 --> 00:26:41.819
<v Speaker 2>In a sense, batch production.

405
00:26:41.939 --> 00:26:48.050
<v Speaker 2>So, if we look at fermentation, starting
off at 2 L, then going to 20, then 200,

406
00:26:48.250 --> 00:26:53.680
<v Speaker 2>2000, 20,000, if any one of those, uh,

407
00:26:54.050 --> 00:26:59.670
<v Speaker 2>fermentation processes were to fail, you
are losing that 20,000 L capability.

408
00:27:00.589 --> 00:27:03.140
<v Speaker 2>Even if it starts off at the 2 L point.

409
00:27:03.589 --> 00:27:07.969
<v Speaker 2>So it has a huge cost implications, along
with product quality.

410
00:27:08.550 --> 00:27:14.030
<v Speaker 2>If contamination gets in at any one of
those points, you've also lost the 20,000 L

411
00:27:14.030 --> 00:27:15.589
<v Speaker 2>capability in your production.

412
00:27:16.680 --> 00:27:18.699
<v Speaker 2>And of course, cybersecurity.

413
00:27:20.000 --> 00:27:25.239
<v Speaker 2>I don't think any one of you would want
your IP getting leaked or even being held

414
00:27:25.239 --> 00:27:27.550
<v Speaker 2>under a cyberattack and being held for
ransom.

415
00:27:27.920 --> 00:27:34.719
<v Speaker 2>These are extremely, extremely costly
events that could cripple, um, or have huge,

416
00:27:34.880 --> 00:27:37.219
<v Speaker 2>huge dire impact, uh, in your environment.

417
00:27:40.000 --> 00:27:46.650
<v Speaker 2>Now, Predictive maintenance often gets
confused with predictive analytics,

418
00:27:47.050 --> 00:27:52.290
<v Speaker 2>or at least there are approaches which
rely on pre-built machine learning models or

419
00:27:52.290 --> 00:27:53.890
<v Speaker 2>customized model creation.

420
00:27:54.609 --> 00:27:59.130
<v Speaker 2>And essentially, I'm not gonna go through
all of these in great detail, but you can

421
00:27:59.130 --> 00:28:02.589
<v Speaker 2>see there, there are multiple steps which
take a huge amount of time.

422
00:28:02.969 --> 00:28:05.479
<v Speaker 2>So your problem definition, roughly 3
weeks.

423
00:28:05.819 --> 00:28:11.650
<v Speaker 2>Data collection, another 3 months, getting
data at a huge high frequency rate.

424
00:28:12.189 --> 00:28:17.349
<v Speaker 2>There's an exploratory data analysis,
which takes a team of data scientists weeks to

425
00:28:17.349 --> 00:28:17.650
<v Speaker 2>perform.

426
00:28:18.310 --> 00:28:22.469
<v Speaker 2>Then algorithm selection and model
development, making it fit for purpose to take

427
00:28:22.469 --> 00:28:26.050
<v Speaker 2>into account all of this complexity and
huge volumes of data.

428
00:28:26.790 --> 00:28:31.550
<v Speaker 2>There's the evaluation in a real, uh, real
world environment, making sure that these

429
00:28:31.550 --> 00:28:33.180
<v Speaker 2>are fit for purpose.

430
00:28:33.709 --> 00:28:38.109
<v Speaker 2>The rollout again could take a number of
months, and along with post-deployment

431
00:28:38.109 --> 00:28:39.410
<v Speaker 2>monitoring, which is ongoing.

432
00:28:41.479 --> 00:28:45.579
<v Speaker 2>Time to value for these approaches is 14
to 24 months.

433
00:28:45.859 --> 00:28:50.390
<v Speaker 2>It's a huge amount of time and this
process that I've just described has to be

434
00:28:50.390 --> 00:28:53.829
<v Speaker 2>iterated for every single machine type you
have.

435
00:28:55.319 --> 00:29:01.119
<v Speaker 2>So that really means that to get value out
of these takes a huge amount of time, a

436
00:29:01.119 --> 00:29:07.189
<v Speaker 2>huge amount of resource, and a lot of your
efforts on site will need to be put into

437
00:29:07.189 --> 00:29:09.859
<v Speaker 2>helping create these fit for purpose
models.

438
00:29:11.020 --> 00:29:16.400
<v Speaker 2>Really, what the industry needs is
something that can be deployed once and scale.

439
00:29:16.780 --> 00:29:18.119
<v Speaker 2>You're not looking for something.

440
00:29:19.079 --> 00:29:23.130
<v Speaker 2>Uh, which is specialized and runs through
this iterative process and takes huge

441
00:29:23.130 --> 00:29:24.229
<v Speaker 2>amounts of time, as I've said.

442
00:29:24.910 --> 00:29:28.849
<v Speaker 2>You need something which is fit for
purpose and can be deployed rapidly in scale

443
00:29:29.150 --> 00:29:33.150
<v Speaker 2>from your site in the northern hemisphere,
southern hemisphere, and all over the

444
00:29:33.150 --> 00:29:33.609
<v Speaker 2>world.

445
00:29:34.569 --> 00:29:36.969
<v Speaker 2>So how does Siemens do this?

446
00:29:38.349 --> 00:29:43.400
<v Speaker 2>For us, it's important really to identify
and define condition indicators.

447
00:29:44.189 --> 00:29:47.469
<v Speaker 2>So condition indicators are derived from
time series data.

448
00:29:47.660 --> 00:29:52.420
<v Speaker 2>They come from condition monitoring data
um in the pharmaceutical industry as well

449
00:29:52.420 --> 00:29:59.119
<v Speaker 2>uh specifically also processed data, but
it's Derived from uh your process and

450
00:29:59.119 --> 00:30:03.719
<v Speaker 2>condition monitoring data to expose signs
of degradation, and it is to give you

451
00:30:03.719 --> 00:30:05.219
<v Speaker 2>leading indicators of failure.

452
00:30:06.130 --> 00:30:11.609
<v Speaker 2>It also helps to reduce the amount of
noise within the data, because a current

453
00:30:11.609 --> 00:30:16.849
<v Speaker 2>profile for an electric motor again, can
have a huge variation in normal operating

454
00:30:16.849 --> 00:30:17.069
<v Speaker 2>condition.

455
00:30:17.329 --> 00:30:23.449
<v Speaker 2>So when we start to expose condition
indicators, you provide a more stable baseline

456
00:30:23.449 --> 00:30:23.890
<v Speaker 2>understanding.

457
00:30:24.469 --> 00:30:28.589
<v Speaker 2>And condition indicators can be created in
a range of places within your

458
00:30:28.589 --> 00:30:33.900
<v Speaker 2>infrastructure derived in the or on the
edge, and within factory historians, SQL

459
00:30:33.900 --> 00:30:36.390
<v Speaker 2>databases or even cloud-based systems.

460
00:30:37.439 --> 00:30:40.910
<v Speaker 2>And I'll give you some examples of these
condi condition indicators.

461
00:30:41.359 --> 00:30:43.640
<v Speaker 2>So we've got 3 assets as the examples.

462
00:30:43.680 --> 00:30:47.619
<v Speaker 2>There's a, a robot, a pump, and a
conveyor, quite universal.

463
00:30:48.280 --> 00:30:52.380
<v Speaker 2>So if we will look at the robot, we may be
looking or the condition monitoring data

464
00:30:52.380 --> 00:30:54.560
<v Speaker 2>is potentially torque per axis.

465
00:30:55.849 --> 00:31:02.530
<v Speaker 2>However, we want to capture this data or
process data to be consistent because if

466
00:31:02.530 --> 00:31:07.369
<v Speaker 2>this is a palletizing robot, well, it
could be picking up different weights and

467
00:31:07.369 --> 00:31:12.050
<v Speaker 2>configuring them, or it could be running
at different configurations of um product

468
00:31:12.050 --> 00:31:13.329
<v Speaker 2>placements on your pallets.

469
00:31:14.209 --> 00:31:19.569
<v Speaker 2>Therefore, a huge variability in the data,
the torque requirements in any given

470
00:31:19.569 --> 00:31:20.250
<v Speaker 2>period of time.

471
00:31:21.040 --> 00:31:25.040
<v Speaker 2>So a condition indicator for that or an
example of a condition indicator could be

472
00:31:25.040 --> 00:31:29.619
<v Speaker 2>the max torque per axis after the last
movement and its return to home.

473
00:31:30.239 --> 00:31:35.520
<v Speaker 2>This is an unloaded movement and it brings
consistency to the the monitoring of

474
00:31:35.520 --> 00:31:36.239
<v Speaker 2>these axes.

475
00:31:37.449 --> 00:31:43.349
<v Speaker 2>For a pump, the data could be input and
output pressure, but we may look at a min

476
00:31:43.349 --> 00:31:46.449
<v Speaker 2>maximum average differential pressure
across the pump.

477
00:31:47.050 --> 00:31:51.329
<v Speaker 2>Again, by looking at your min maximum
average, you're still understanding the the

478
00:31:51.329 --> 00:31:56.410
<v Speaker 2>profile of those measures while heavily
reducing the amount of data you send to the

479
00:31:56.410 --> 00:31:58.689
<v Speaker 2>cloud and bringing stability to those.

480
00:31:59.130 --> 00:32:01.390
<v Speaker 2>Also highlighting when one of them
changes.

481
00:32:02.010 --> 00:32:06.089
<v Speaker 2>Because your, your minimum and average
could be relatively stable, but if your max

482
00:32:06.089 --> 00:32:09.989
<v Speaker 2>differential pressure is increasing, that
could give you an indication for

483
00:32:09.989 --> 00:32:12.729
<v Speaker 2>individual events happening which are
affecting the pressure.

484
00:32:14.000 --> 00:32:18.180
<v Speaker 2>Motor currents, we could look for motor
current for the driver of the conveyor.

485
00:32:18.560 --> 00:32:22.280
<v Speaker 2>We could look to extract statistical
characteristics like your peak start current,

486
00:32:22.400 --> 00:32:24.869
<v Speaker 2>your mean running current, and your peak
running current.

487
00:32:25.319 --> 00:32:30.560
<v Speaker 2>Again, still giving profiles of the
behavior while bringing stability into the

488
00:32:30.560 --> 00:32:31.180
<v Speaker 2>monitoring.

489
00:32:33.829 --> 00:32:37.530
<v Speaker 2>How do you deploy scalable predictive
maintenance in your environment?

490
00:32:39.569 --> 00:32:46.310
<v Speaker 2>At Demons, we have come up with what we
believe is a best practice approach, um, To

491
00:32:46.310 --> 00:32:50.930
<v Speaker 2>implementing, uh, investigating,
implementing, and scaling predictive maintenance,

492
00:32:51.229 --> 00:32:54.180
<v Speaker 2>and it's designed to be customized to your
environment.

493
00:32:54.790 --> 00:33:00.150
<v Speaker 2>So this methodology comes in 5 phases, and
we'll go into the detail of these, but

494
00:33:00.150 --> 00:33:02.979
<v Speaker 2>this is scope, design, deploy, operate,
and refine.

495
00:33:03.270 --> 00:33:08.369
<v Speaker 2>And it is the best practice approach that
we have created fit for purpose for us.

496
00:33:09.729 --> 00:33:15.130
<v Speaker 2>If we look at these a bit further, scope,
this is typically pre-contract

497
00:33:15.130 --> 00:33:15.479
<v Speaker 2>investigation.

498
00:33:15.810 --> 00:33:19.449
<v Speaker 2>Getting a true understanding of your
environment, the complexities that you have,

499
00:33:19.510 --> 00:33:23.969
<v Speaker 2>you know, um, what is specific to you that
we need to take into account before we

500
00:33:23.969 --> 00:33:26.150
<v Speaker 2>can say, yes, we will have a successful
project.

501
00:33:26.880 --> 00:33:29.420
<v Speaker 2>So, understanding the value and benefits.

502
00:33:30.040 --> 00:33:33.390
<v Speaker 2>For predictive maintenance, how do you
define what is valuable and where the

503
00:33:33.390 --> 00:33:34.339
<v Speaker 2>benefits lie?

504
00:33:34.920 --> 00:33:36.459
<v Speaker 2>Starts off with maintenance time.

505
00:33:36.959 --> 00:33:41.910
<v Speaker 2>So for your global operations, let's say
you have 4 or 5 factories around the world.

506
00:33:42.359 --> 00:33:47.680
<v Speaker 2>Which one of them, if we were to list the
reactive and planned maintenance for each

507
00:33:47.680 --> 00:33:50.069
<v Speaker 2>of them, we would then get a total time.

508
00:33:50.420 --> 00:33:54.579
<v Speaker 2>The, the factory that has the, the highest
total maintenance time.

509
00:33:55.010 --> 00:33:56.880
<v Speaker 2>That's where we now bring into focus.

510
00:33:57.290 --> 00:33:57.800
<v Speaker 2>Perfect.

511
00:33:58.030 --> 00:34:05.020
<v Speaker 2>So, we then split that up into So we then
split that up into the

512
00:34:05.020 --> 00:34:08.860
<v Speaker 2>various areas, your upstream, uh,
downstream, and services.

513
00:34:09.139 --> 00:34:11.820
<v Speaker 2>So again, which one of those have got the
highest maintenance type?

514
00:34:12.600 --> 00:34:13.000
<v Speaker 2>Perfect.

515
00:34:13.080 --> 00:34:16.919
<v Speaker 2>Well, it's actually your utilities and
services, and then we can start to focus on

516
00:34:16.919 --> 00:34:19.239
<v Speaker 2>machine types and so on.

517
00:34:20.580 --> 00:34:26.300
<v Speaker 2>Uh, we also then look at the specifics of
your environment, the infrastructure, the

518
00:34:26.300 --> 00:34:30.020
<v Speaker 2>data availability, the people involved,
and their time.

519
00:34:30.419 --> 00:34:36.580
<v Speaker 2>We try and develop a project team with the
right characters involved and then create

520
00:34:36.580 --> 00:34:40.340
<v Speaker 2>a project plan and a proposal with all of
these factors considered.

521
00:34:41.330 --> 00:34:45.629
<v Speaker 2>Design is when the project starts and it
goes even deeper into the investigation.

522
00:34:45.969 --> 00:34:50.159
<v Speaker 2>So there's a look at the assets, you know,
the assets that have been identified and

523
00:34:50.159 --> 00:34:55.090
<v Speaker 2>selected, but you look at the operational
complexity and operational interdependency

524
00:34:55.090 --> 00:34:55.659
<v Speaker 2>of these.

525
00:34:56.959 --> 00:35:01.479
<v Speaker 2>Um, we look at how the, these machines and
assets are failing and what data is

526
00:35:01.479 --> 00:35:02.899
<v Speaker 2>available to cover those.

527
00:35:03.280 --> 00:35:08.320
<v Speaker 2>We can also do a gap analysis to see is
there a need for additional sensing, um, or

528
00:35:08.320 --> 00:35:09.060
<v Speaker 2>additional data.

529
00:35:09.830 --> 00:35:14.850
<v Speaker 2>We make sure that we define the goals and
success criteria with the project team,

530
00:35:15.350 --> 00:35:19.179
<v Speaker 2>and we look at the data architecture and
data transformation capabilities.

531
00:35:20.750 --> 00:35:24.219
<v Speaker 2>From there we get into the deploy phase,
which is about making the data move.

532
00:35:24.860 --> 00:35:28.899
<v Speaker 2>So if there is need for additional
hardware installation, sensor, data collection

533
00:35:28.899 --> 00:35:33.899
<v Speaker 2>and so on, that that gets done, that the
data is then processed and streaming

534
00:35:33.899 --> 00:35:34.739
<v Speaker 2>through to the cloud.

535
00:35:35.810 --> 00:35:39.969
<v Speaker 2>We make sure that the assets are on
boarded, the users are on boarded and trained,

536
00:35:40.050 --> 00:35:44.469
<v Speaker 2>and any contextual data is also fed in to
the application.

537
00:35:45.500 --> 00:35:49.860
<v Speaker 2>We then go to operate, which runs through
best practice approach on using the

538
00:35:49.860 --> 00:35:50.139
<v Speaker 2>application.

539
00:35:50.659 --> 00:35:54.979
<v Speaker 2>So how to monitor notifications, how best
to add in maintenance events, how to do

540
00:35:54.979 --> 00:35:59.320
<v Speaker 2>daily, weekly or monthly case reviews, as
well as periodic reviews of the system.

541
00:36:00.110 --> 00:36:03.889
<v Speaker 2>And then in refined, it's about making it
as good as it can possibly be.

542
00:36:04.350 --> 00:36:08.889
<v Speaker 2>So we look at the need for additional data
or additional processing of the data.

543
00:36:09.780 --> 00:36:11.770
<v Speaker 2>We look at who is using the application.

544
00:36:11.870 --> 00:36:14.169
<v Speaker 2>Is the technology being adopted on site?

545
00:36:14.429 --> 00:36:16.610
<v Speaker 2>Because without people using it, it is
useless.

546
00:36:17.229 --> 00:36:22.239
<v Speaker 2>So, our customers have got many different
approaches on the adoption factor.

547
00:36:22.790 --> 00:36:27.949
<v Speaker 2>Some of them will define a single or a
small group of PDM champions.

548
00:36:28.550 --> 00:36:33.310
<v Speaker 2>One, people whose responsibility it is to
log in daily and make sure that the

549
00:36:33.310 --> 00:36:36.090
<v Speaker 2>information is then distributed to the
wider maintenance teams.

550
00:36:36.979 --> 00:36:41.389
<v Speaker 2>For other customers, they make sure the
entire maintenance team has a responsibility

551
00:36:41.929 --> 00:36:45.659
<v Speaker 2>uh to log in and incorporate the, the
knowledge into their day to day.

552
00:36:46.050 --> 00:36:50.810
<v Speaker 2>So for your area that you are responsible
for, you need to make sure you log in, are

553
00:36:50.810 --> 00:36:54.889
<v Speaker 2>looking at the new notifications and bring
that to the morning standard so that you

554
00:36:54.889 --> 00:36:59.909
<v Speaker 2>are on top of what has happened overnight
and what your plan is for the day.

555
00:37:00.860 --> 00:37:04.959
<v Speaker 2>Also look at what value has been achieved
and how do you scale that outwards.

556
00:37:05.550 --> 00:37:08.389
<v Speaker 2>And value for predictive maintenance can
come in many forms.

557
00:37:08.679 --> 00:37:12.689
<v Speaker 2>Um, downtime avoidance, spares, uh,
reductions.

558
00:37:13.120 --> 00:37:16.919
<v Speaker 2>You know, you can have, uh, planned
maintenance efficiencies and even pushing those

559
00:37:16.919 --> 00:37:18.300
<v Speaker 2>planned maintenance events out.

560
00:37:18.669 --> 00:37:24.659
<v Speaker 2>Um, less consumables and even, uh, uh less
of a carbon impact because, uh,

561
00:37:24.679 --> 00:37:26.699
<v Speaker 2>environmental assets have not gone down.

562
00:37:27.699 --> 00:37:33.149
<v Speaker 2>Um, you had less spills of chemicals or
even oils, let's say.

563
00:37:34.010 --> 00:37:36.800
<v Speaker 2>And then finally we look to integrate into
daily workflows.

564
00:37:37.050 --> 00:37:41.530
<v Speaker 2>So this could be a an integration with
your maintenance management system or even

565
00:37:41.530 --> 00:37:46.229
<v Speaker 2>providing the information into whatever
systems best suit your day to day.

566
00:37:48.409 --> 00:37:54.209
<v Speaker 2>And when we, and if we look at how this
project plan can roll out, um, it can follow

567
00:37:54.209 --> 00:37:55.199
<v Speaker 2>something similar to this.

568
00:37:55.370 --> 00:37:59.290
<v Speaker 2>So for the asset selection process, this
could take about 2 weeks.

569
00:37:59.889 --> 00:38:04.169
<v Speaker 2>At this, also starting in week two, we
would look at the data architecture and the

570
00:38:04.169 --> 00:38:05.350
<v Speaker 2>data transformation.

571
00:38:06.239 --> 00:38:10.760
<v Speaker 2>At week 5, we could be looking at the
pre-processing and data streaming through to

572
00:38:10.760 --> 00:38:15.020
<v Speaker 2>the application, and in week 6, the start
of the onboarding and configuration.

573
00:38:16.090 --> 00:38:20.459
<v Speaker 2>Which also means at the start of week 6,
there is now data and valuable insights

574
00:38:20.459 --> 00:38:21.570
<v Speaker 2>within the application.

575
00:38:22.020 --> 00:38:28.250
<v Speaker 2>So users can really start to rapidly gain
value from a solution if implemented in

576
00:38:28.250 --> 00:38:30.719
<v Speaker 2>the right way with the right project plan
behind it.

577
00:38:32.409 --> 00:38:36.379
<v Speaker 2>But for the human element, uh, the people
onsite, you must be thinking, wow, this is

578
00:38:36.379 --> 00:38:41.500
<v Speaker 2>gonna take a full team of us, a huge
amount of time and effort, and really we are

579
00:38:41.500 --> 00:38:42.659
<v Speaker 2>already stretched thin.

580
00:38:43.669 --> 00:38:46.169
<v Speaker 2>And the fact is, yes, it does require
time.

581
00:38:46.239 --> 00:38:51.459
<v Speaker 2>If you're going to undertake this process,
be prepared to invest in time.

582
00:38:52.040 --> 00:38:58.159
<v Speaker 2>However, we have found that the initial
stages are resource intensive to a certain

583
00:38:58.159 --> 00:38:58.719
<v Speaker 2>extent.

584
00:38:59.199 --> 00:39:04.959
<v Speaker 2>But however, scaled or steady-state usage
is quite easy, um, and doesn't require a

585
00:39:04.959 --> 00:39:06.860
<v Speaker 2>lot of, uh, huge amounts of time.

586
00:39:07.280 --> 00:39:11.120
<v Speaker 2>And if we were to look at the, the person
or the character within the project from

587
00:39:11.120 --> 00:39:16.110
<v Speaker 2>your teams, who would be The most
prominent, it could be the maintenance engineer at

588
00:39:16.110 --> 00:39:19.110
<v Speaker 2>roughly 2 hours per day in the initial
stages.

589
00:39:19.530 --> 00:39:22.840
<v Speaker 2>That is during your commissioning and
asset configuration.

590
00:39:23.370 --> 00:39:29.149
<v Speaker 2>But once the onboarding has taken place,
once the, The training and the adoption and

591
00:39:29.149 --> 00:39:34.469
<v Speaker 2>confidence in the system is there, that
then goes down to half an hour a day.

592
00:39:34.949 --> 00:39:39.750
<v Speaker 2>It's not about giving your teams an
additional thing to use, an additional

593
00:39:39.750 --> 00:39:40.929
<v Speaker 2>responsibility.

594
00:39:41.350 --> 00:39:45.500
<v Speaker 2>It's about giving them information which
is easy to access, easy to understand, and

595
00:39:45.500 --> 00:39:48.750
<v Speaker 2>is going to make their efforts in any day
more efficient.

596
00:39:51.870 --> 00:39:54.350
<v Speaker 2>So, we can run through a bit of a process
together.

597
00:39:54.550 --> 00:39:58.550
<v Speaker 2>Let's assess your data and cultural
readiness, because these are two aspects which

598
00:39:58.550 --> 00:40:01.110
<v Speaker 2>are extremely important in any project.

599
00:40:01.669 --> 00:40:03.929
<v Speaker 2>So we start off with the data evaluation.

600
00:40:04.469 --> 00:40:10.189
<v Speaker 2>We've got 5 categories here, 1 to 51 being
no sensors installed, poor access to

601
00:40:10.189 --> 00:40:14.270
<v Speaker 2>control systems, and no maintenance
records or no maintenance recorded.

602
00:40:15.229 --> 00:40:20.310
<v Speaker 2>Your ranking of 3 would be an online, raw
or simple data captured potentially from

603
00:40:20.310 --> 00:40:22.209
<v Speaker 2>systems, or do you have some access?

604
00:40:22.560 --> 00:40:22.830
<v Speaker 2>Boom.

605
00:40:25.510 --> 00:40:26.989
<v Speaker 2>Sorry, fire alarm.

606
00:40:28.469 --> 00:40:29.199
<v Speaker 2>There we go.

607
00:40:29.590 --> 00:40:35.389
<v Speaker 2>So 3 is online or raw simple data captured
potentially from sensors or you have

608
00:40:35.389 --> 00:40:39.919
<v Speaker 2>access to some of your control systems,
potentially even a historian, and you will

609
00:40:39.919 --> 00:40:43.149
<v Speaker 2>have access to some maintenance data
recorded electronically.

610
00:40:44.010 --> 00:40:48.290
<v Speaker 2>And in 5 being online condition monitoring
data and advanced condition indicators

611
00:40:48.290 --> 00:40:52.310
<v Speaker 2>coming from sensors and control systems
with the ability to stream this already

612
00:40:52.860 --> 00:40:57.449
<v Speaker 2>through to the cloud, and your maintenance
data is available via integration or

613
00:40:57.449 --> 00:40:59.489
<v Speaker 2>exports through a CMMS.

614
00:41:00.919 --> 00:41:02.669
<v Speaker 2>Then there is the cultural evaluation.

615
00:41:03.239 --> 00:41:08.179
<v Speaker 2>So again, 1 to 51 being the maintenance
strategy is mainly reactive.

616
00:41:08.270 --> 00:41:11.199
<v Speaker 2>Um, you'll have some planned work done.

617
00:41:11.320 --> 00:41:15.939
<v Speaker 2>There's no PDM in place or no, not even
traditional condition monitoring in place

618
00:41:16.320 --> 00:41:20.340
<v Speaker 2>and no real KPIs or targets to increase
capacity.

619
00:41:20.889 --> 00:41:25.469
<v Speaker 2>And this is more in line with a, a mom and
pop operation, let's say.

620
00:41:25.889 --> 00:41:30.070
<v Speaker 2>Your ranking of 3 is, uh, your maintenance
strategy will be extensively preventative

621
00:41:30.330 --> 00:41:34.929
<v Speaker 2>with the use of condition monitoring and
potentially even a PDM team being

622
00:41:34.929 --> 00:41:39.850
<v Speaker 2>established, and you will have as a
company targets to drive up efficiency and

623
00:41:39.850 --> 00:41:40.550
<v Speaker 2>productivity.

624
00:41:41.419 --> 00:41:46.929
<v Speaker 2>And then 5 being an extensive preventative
maintenance project or plan strategy

625
00:41:46.929 --> 00:41:51.939
<v Speaker 2>underway with PDM being investigated or
maybe you've even done predictive

626
00:41:51.939 --> 00:41:56.820
<v Speaker 2>maintenance in the past, and your company
has clear and ambitious KPIs and targets.

627
00:41:57.780 --> 00:42:01.739
<v Speaker 2>So from there, we can then give a
functional maturity, which is essentially an

628
00:42:01.739 --> 00:42:03.840
<v Speaker 2>aggregation of these two elements.

629
00:42:04.489 --> 00:42:09.090
<v Speaker 2>You know, basic group-based condition
monitoring, um, through to automated condition

630
00:42:09.090 --> 00:42:11.959
<v Speaker 2>monitoring, which is where we would really
start at.

631
00:42:12.370 --> 00:42:17.239
<v Speaker 2>And then, um, Four prognostic capabilities
being a 5.

632
00:42:17.780 --> 00:42:22.939
<v Speaker 2>But using this ranking, we can get a
project or we can predict a project's success

633
00:42:22.939 --> 00:42:24.010
<v Speaker 2>using these elements.

634
00:42:24.379 --> 00:42:29.820
<v Speaker 2>So how likely are you to succeed based off
of a very high level understanding of

635
00:42:29.820 --> 00:42:32.780
<v Speaker 2>your data availability and your cultural
availability?

636
00:42:34.149 --> 00:42:37.600
<v Speaker 2>But if you find yourself on the lower end
of the scale, don't worry.

637
00:42:37.800 --> 00:42:42.149
<v Speaker 2>There are definitely ways to develop both
data and cultural maturity.

638
00:42:42.719 --> 00:42:47.159
<v Speaker 2>So, from a data perspective, this could be
starting off with additional hardware and

639
00:42:47.159 --> 00:42:47.919
<v Speaker 2>sensors.

640
00:42:48.320 --> 00:42:54.000
<v Speaker 2>You know, let's get some vibration
currents, uh, temperature sensors in place.

641
00:42:54.790 --> 00:42:56.459
<v Speaker 2>So that you can start monitoring.

642
00:42:56.909 --> 00:43:01.070
<v Speaker 2>Then we can start, you can start looking
at your IT environments and your existing

643
00:43:01.070 --> 00:43:06.250
<v Speaker 2>data availability and how would you make
that data accessible all the way through to

644
00:43:06.550 --> 00:43:12.110
<v Speaker 2>having specialists configure the
algorithms and machine learning to be very specific

645
00:43:12.110 --> 00:43:15.949
<v Speaker 2>in understanding your assets and machines
and their requirements.

646
00:43:17.629 --> 00:43:23.139
<v Speaker 2>Agricultural maturity, you know, for
companies that don't have the people or culture

647
00:43:23.139 --> 00:43:29.340
<v Speaker 2>in place, we can increase all of this by
awareness, market studies, educations,

648
00:43:29.500 --> 00:43:35.169
<v Speaker 2>workshops, um, full in-depth training and
looking at specifics within your

649
00:43:35.169 --> 00:43:41.060
<v Speaker 2>environment and even For those who are
already starting at a mature state, you could

650
00:43:41.060 --> 00:43:45.100
<v Speaker 2>look at getting people within the project
team involved in wider communities and

651
00:43:45.100 --> 00:43:47.120
<v Speaker 2>start to become real champions

652
00:43:47.459 --> 00:43:49.840
<v Speaker 2>of predictive maintenance in your
environment.

653
00:43:51.959 --> 00:43:57.040
<v Speaker 2>And then calculating a return on
investment because essentially every project needs

654
00:43:57.040 --> 00:43:58.750
<v Speaker 2>to be of value.

655
00:44:00.169 --> 00:44:04.050
<v Speaker 2>It's important to understand though, when
you start a project, there is a difference

656
00:44:04.050 --> 00:44:06.449
<v Speaker 2>between a proof of concept and a proof of
value.

657
00:44:07.139 --> 00:44:10.060
<v Speaker 2>A proof of concept is about testing the
technology.

658
00:44:10.260 --> 00:44:12.770
<v Speaker 2>Is this going to work in my environment?

659
00:44:13.219 --> 00:44:18.179
<v Speaker 2>Um, is the technology going to give
results based off of my, the complexity of my

660
00:44:18.179 --> 00:44:19.050
<v Speaker 2>machines and areas?

661
00:44:20.100 --> 00:44:26.639
<v Speaker 2>This is really about Small number of
assets, probably for a short amount of time,

662
00:44:26.729 --> 00:44:32.030
<v Speaker 2>less than 6 months, to see, is this going
to work with my infrastructure, my data,

663
00:44:32.489 --> 00:44:33.209
<v Speaker 2>my machines.

664
00:44:33.290 --> 00:44:34.590
<v Speaker 2>That's a proof of concept.

665
00:44:35.010 --> 00:44:39.090
<v Speaker 2>But at this point, you cannot expect
realistically to get an ROI.

666
00:44:39.449 --> 00:44:41.719
<v Speaker 2>Maybe you get lucky, probably not.

667
00:44:41.770 --> 00:44:43.149
<v Speaker 2>It's about testing the technology.

668
00:44:44.030 --> 00:44:49.419
<v Speaker 2>A proof of value, which is where we drive
our projects, is about getting an ROI.

669
00:44:49.469 --> 00:44:51.850
<v Speaker 2>It's about really testing the methodology.

670
00:44:52.429 --> 00:44:56.989
<v Speaker 2>So still seeing, do I have the right data,
infrastructure, machines, and is this

671
00:44:56.989 --> 00:45:01.659
<v Speaker 2>system Suitable for my data infrastructure
machines.

672
00:45:02.370 --> 00:45:05.100
<v Speaker 2>But also, am I gonna get the adoption by
the users?

673
00:45:05.340 --> 00:45:08.770
<v Speaker 2>Is it easy to interpret and get value
from?

674
00:45:08.969 --> 00:45:13.530
<v Speaker 2>And am I going to essentially save money
by implementing this solution?

675
00:45:14.939 --> 00:45:18.020
<v Speaker 2>But there's a full range of aspects for
success.

676
00:45:18.770 --> 00:45:21.129
<v Speaker 2>And it really does depend on your project
size.

677
00:45:21.649 --> 00:45:25.530
<v Speaker 2>So if we look at the the table on the top
left-hand side, there's a proof of value.

678
00:45:25.929 --> 00:45:30.489
<v Speaker 2>This is typically a single area for less
than 12 months or maybe even up to 12

679
00:45:30.489 --> 00:45:35.129
<v Speaker 2>months, but still on a limited number of
machines or assets, between 50.

680
00:45:35.959 --> 00:45:37.479
<v Speaker 2>50 to 100, let's say.

681
00:45:37.919 --> 00:45:41.790
<v Speaker 2>And what can be achieved with that is
avoiding potential failure on those machines.

682
00:45:42.360 --> 00:45:46.679
<v Speaker 2>There is capturing the maintenance uh
activity and then knowledge capture with the

683
00:45:46.679 --> 00:45:47.639
<v Speaker 2>generative AI.

684
00:45:48.520 --> 00:45:54.050
<v Speaker 2>When you start to move to a larger phase
one project, being a single area for 12

685
00:45:54.050 --> 00:45:58.689
<v Speaker 2>months, maybe potentially more, that's
when companies can benefit from uptime

686
00:45:58.689 --> 00:45:59.389
<v Speaker 2>improvements.

687
00:46:00.050 --> 00:46:03.889
<v Speaker 2>However, when you go to a phase 2, which
is looking at hundreds of assets in an

688
00:46:03.889 --> 00:46:08.489
<v Speaker 2>environment for at least 12 months, that's
when you would benefit from efficient

689
00:46:08.489 --> 00:46:13.290
<v Speaker 2>maintenance, better maintenance planning,
asset lifetime extensions, impact on your

690
00:46:13.290 --> 00:46:17.850
<v Speaker 2>safety KPIs, planned maintenance
reductions, and avoided overtime.

691
00:46:18.770 --> 00:46:20.850
<v Speaker 2>When you're looking at multiple sites.

692
00:46:21.469 --> 00:46:25.229
<v Speaker 2>Um, that is when you can benefit from all
of those good things along with energy

693
00:46:25.229 --> 00:46:29.610
<v Speaker 2>efficiency, reduced labor, and OEE
improvements and on a large scale.

694
00:46:30.070 --> 00:46:34.080
<v Speaker 2>And when you go for a global expansion,
that is when a predictive maintenance, uh,

695
00:46:34.110 --> 00:46:39.050
<v Speaker 2>program can really, really impact your
sustainability goals.

696
00:46:39.790 --> 00:46:44.669
<v Speaker 1>Please like and subscribe on all major
podcast channels and find out more about

697
00:46:44.669 --> 00:46:48.050
<v Speaker 1>Sensei predictive maintenance at
Siemens.com.
