1
00:00:00,850 --> 00:00:03,638
In a cement plant, data never stops.

2
00:00:03,838 --> 00:00:07,679
The kiln, the mills, the energy meters, the laboratory,

3
00:00:07,679 --> 00:00:09,355
the maintenance records…

4
00:00:09,355 --> 00:00:13,546
Each one sits in its own system and speaks its own language.

5
00:00:13,746 --> 00:00:17,350
So who will deliver this data to the right person,

6
00:00:17,350 --> 00:00:20,234
reliably and in a form that makes sense?

7
00:00:20,634 --> 00:00:23,543
The answer is HighByte Intelligence Hub:

8
00:00:23,543 --> 00:00:25,797
an industrial DataOps platform.

9
00:00:27,597 --> 00:00:31,323
In most plants, systems are wired to each other one by one,

10
00:00:31,323 --> 00:00:32,270
point to point.

11
00:00:32,270 --> 00:00:37,322
Every new connection means another script that has to be written and maintained.

12
00:00:37,522 --> 00:00:40,348
When a source changes, the chain breaks.

13
00:00:40,348 --> 00:00:46,565
And the same temperature reading goes by five different names in five different systems.

14
00:00:46,765 --> 00:00:47,593
In the end,

15
00:00:47,593 --> 00:00:53,016
teams lose their time trying to understand the data instead of using it.

16
00:00:54,516 --> 00:01:01,232
HighByte is a data infrastructure that sits between the machines on the shop floor and the systems across the business.

17
00:01:01,532 --> 00:01:05,463
It runs inside the plant, at the edge. It connects the data,

18
00:01:05,463 --> 00:01:07,297
models it, gives it meaning,

19
00:01:07,297 --> 00:01:11,097
and delivers it in the same form to everyone who needs it.

20
00:01:11,647 --> 00:01:17,980
First, it connects: OPC UA, MQTT, REST, databases and files.

21
00:01:17,980 --> 00:01:21,780
And it does so without writing code.

22
00:01:22,730 --> 00:01:25,743
Then it models. For each type of equipment,

23
00:01:25,743 --> 00:01:28,685
a standard structure is defined just once.

24
00:01:29,635 --> 00:01:34,725
Next, pipelines process the data. They filter, calculate, enrich,

25
00:01:34,725 --> 00:01:37,779
and make decisions based on conditions.

26
00:01:38,729 --> 00:01:42,279
Finally, it publishes: to the Unified Namespace,

27
00:01:42,279 --> 00:01:45,532
to the cloud and to databases. When we want,

28
00:01:45,532 --> 00:01:48,934
we can even write data back to the shop floor.

29
00:01:49,884 --> 00:01:52,924
So the data becomes standardized in one place.

30
00:01:52,924 --> 00:01:56,427
The same model works unchanged on a hundred machines.

31
00:01:56,527 --> 00:01:59,158
And it is built not by software developers,

32
00:01:59,158 --> 00:02:01,728
but by the engineers who know the process.

33
00:02:03,228 --> 00:02:08,236
Now let's see this in a realistic example: Kızılırmak Çimento,

34
00:02:08,236 --> 00:02:09,529
Kırıkkale plant.

35
00:02:09,729 --> 00:02:13,783
The whole line, from raw material to packaging,

36
00:02:13,783 --> 00:02:19,734
is brought together in a single HighByte project: two OPC UA servers,

37
00:02:19,734 --> 00:02:22,925
MQTT, REST, four databases and files.

38
00:02:24,425 --> 00:02:27,124
This is HighByte's browser-based interface.

39
00:02:27,124 --> 00:02:29,383
The project has sixteen connections,

40
00:02:29,383 --> 00:02:32,270
twenty-eight models and twenty-nine pipelines,

41
00:02:32,270 --> 00:02:34,529
and all of them are running healthy.

42
00:02:37,079 --> 00:02:43,119
The Connections section shows two OPC UA PLCs, a REST service,

43
00:02:43,119 --> 00:02:46,919
four SQLite databases and file sources.

44
00:02:48,769 --> 00:02:54,524
We open the burning PLC. All of the kiln's tags are defined as inputs.

45
00:02:55,424 --> 00:02:59,825
When we hit Test, we immediately see the kiln's live values.

46
00:03:01,875 --> 00:03:06,725
In the Models section, we define the structure of the equipment once.

47
00:03:06,725 --> 00:03:10,661
The kiln model is made up of groups such as feed, drive,

48
00:03:10,661 --> 00:03:12,489
burning zone and flue gas.

49
00:03:12,639 --> 00:03:15,485
Models can also be derived from one another:

50
00:03:15,485 --> 00:03:18,137
shared attributes stay in the base model,

51
00:03:18,137 --> 00:03:22,083
and each piece of equipment adds only what is specific to it.

52
00:03:24,333 --> 00:03:28,453
An instance, in turn, binds the model to real data.

53
00:03:28,453 --> 00:03:32,976
Each attribute is mapped to the matching tag in the PLC.

54
00:03:33,876 --> 00:03:36,135
When we test it, we get an orderly,

55
00:03:36,135 --> 00:03:39,426
structured output that carries quality information.

56
00:03:40,176 --> 00:03:43,658
Thanks to parameters, a model works like a template.

57
00:03:43,658 --> 00:03:45,600
In the Energy Panel instance,

58
00:03:45,600 --> 00:03:48,547
you only need to change the panel parameter:

59
00:03:48,547 --> 00:03:53,168
the same definition attaches to the main incomer, the kiln or a mill.

60
00:03:53,368 --> 00:03:56,874
Inputs take parameters too. A REST request,

61
00:03:56,874 --> 00:04:04,131
a SQL query or an MQTT topic can be applied to dozens of assets with a single definition.

62
00:04:06,581 --> 00:04:11,230
Instances are published in the Unified Namespace, in order:

63
00:04:11,230 --> 00:04:14,224
enterprise, plant, area and equipment.

64
00:04:15,124 --> 00:04:19,637
In the UNS client, we watch the data coming from the kiln live,

65
00:04:19,637 --> 00:04:21,213
inside this hierarchy.

66
00:04:23,463 --> 00:04:26,355
Pipelines define the path the data takes.

67
00:04:26,355 --> 00:04:30,657
This one extracts the vibration data of the kiln drive motor,

68
00:04:30,657 --> 00:04:33,408
splits it into windows and analyzes it.

69
00:04:34,308 --> 00:04:37,936
It writes the result to the namespace, to a database,

70
00:04:37,936 --> 00:04:42,658
and to maintenance and enterprise systems. How many times it has run,

71
00:04:42,658 --> 00:04:45,875
and whether it has failed, is visible here too.

72
00:04:46,625 --> 00:04:51,172
When needed, you can write your own JavaScript in between. Here,

73
00:04:51,172 --> 00:04:52,878
a moving average, slope,

74
00:04:52,878 --> 00:04:57,425
ISO zone and health score are calculated for the vibration data.

75
00:04:57,625 --> 00:05:01,744
But most jobs don't need code. The stage library has triggers,

76
00:05:01,744 --> 00:05:04,866
filters, buffers, conditional branching, loops,

77
00:05:04,866 --> 00:05:07,789
and read and write stages ready and waiting.

78
00:05:10,239 --> 00:05:12,418
Pipelines can also call each other.

79
00:05:12,418 --> 00:05:15,966
Thanks to the Callable trigger and the Subpipeline stage,

80
00:05:15,966 --> 00:05:20,136
one pipeline uses another like a function and gets its result back.

81
00:05:20,336 --> 00:05:22,953
Even lookup tables can be fed by a pipeline.

82
00:05:22,953 --> 00:05:27,889
The equipment fault table takes its data from a pipeline running in the background,

83
00:05:27,889 --> 00:05:29,970
and keeps it cached for one minute.

84
00:05:32,420 --> 00:05:34,861
The Usage tab shows every connection:

85
00:05:34,861 --> 00:05:39,479
which inputs this pipeline reads from, and which outputs it writes to.

86
00:05:39,479 --> 00:05:43,041
You can see the impact of a change before you make it.

87
00:05:43,241 --> 00:05:44,384
With the Test tab,

88
00:05:44,384 --> 00:05:49,148
you can send sample data into the pipeline and watch the result right away.

89
00:05:51,598 --> 00:05:56,126
The closed-loop pipeline collects data, produces a decision,

90
00:05:56,126 --> 00:05:59,900
and writes the result back to the PLC over OPC UA.

91
00:06:00,800 --> 00:06:02,763
So data doesn't only flow upward.

92
00:06:02,763 --> 00:06:05,143
We can write back to the shop floor too.

93
00:06:07,793 --> 00:06:12,243
Conditions are the foundation of event-based operation. For example,

94
00:06:12,243 --> 00:06:18,786
here is a condition that kicks in when the burning-zone temperature exceeds fifteen hundred degrees.

95
00:06:19,686 --> 00:06:21,388
Lookup tables, in turn,

96
00:06:21,388 --> 00:06:26,864
translate an alarm code into the matching action and notification details.

97
00:06:29,514 --> 00:06:32,606
The Modeling Agent speeds up creating models,

98
00:06:32,606 --> 00:06:35,354
inputs and instances with AI assistance.

99
00:06:35,754 --> 00:06:40,393
If you like, you can also expose pipelines to AI agents as tools.

100
00:06:40,393 --> 00:06:43,177
HighByte offers an MCP server for this;

101
00:06:43,177 --> 00:06:46,032
we keep it switched off in this project.

102
00:06:46,882 --> 00:06:48,609
On the administration side,

103
00:06:48,609 --> 00:06:52,767
the Project section exports and imports the entire configuration,

104
00:06:52,767 --> 00:06:56,541
and can even back it up to a Git repository and version it.

105
00:06:56,741 --> 00:07:00,810
In Settings, enterprise features such as the central hub,

106
00:07:00,810 --> 00:07:02,666
automatic backup, secrets,

107
00:07:02,666 --> 00:07:06,450
variables and certificates are gathered in one place.

108
00:07:08,650 --> 00:07:12,669
So what can you do with this? Let's look at a few example scenarios.

109
00:07:13,319 --> 00:07:15,390
First, predictive maintenance.

110
00:07:15,390 --> 00:07:18,289
Motor vibration is monitored continuously;

111
00:07:18,289 --> 00:07:20,291
when a threshold is exceeded,

112
00:07:20,291 --> 00:07:24,364
a record is opened automatically in the maintenance system.

113
00:07:25,814 --> 00:07:30,585
Second, energy cost. Meter data and the market price are combined,

114
00:07:30,585 --> 00:07:34,560
and the energy cost per ton is calculated in real time.

115
00:07:36,010 --> 00:07:39,485
Third, quality. When a laboratory result arrives,

116
00:07:39,485 --> 00:07:45,725
a fine-tuning recommendation is calculated and the value is written straight to the PLC.

117
00:07:47,175 --> 00:07:50,443
Fourth, alarm management. A condition triggers,

118
00:07:50,443 --> 00:07:53,642
a lookup table turns the alarm into an action,

119
00:07:53,642 --> 00:07:56,701
and a notification goes to the right person.

120
00:07:58,151 --> 00:08:02,378
Fifth, open data. From the same model, cloud platforms,

121
00:08:02,378 --> 00:08:07,528
analytics tools and AI applications receive clean, consistent data.

122
00:08:08,978 --> 00:08:12,726
The same approach scales from a single plant to many sites.

123
00:08:12,726 --> 00:08:15,585
Adding a new source doesn't break the others.

124
00:08:17,785 --> 00:08:23,082
So what changes at enterprise scale? The architecture stays the same;

125
00:08:23,082 --> 00:08:25,001
only the layers multiply.

126
00:08:25,651 --> 00:08:30,409
You define the model just once. Thanks to parameterized instances,

127
00:08:30,409 --> 00:08:32,644
the configuration doesn't grow,

128
00:08:32,644 --> 00:08:36,105
even as the number of panels or silos increases.

129
00:08:37,355 --> 00:08:42,431
You keep pipelines small and single-purpose: one collects the data,

130
00:08:42,431 --> 00:08:44,856
one calculates, one distributes.

131
00:08:44,856 --> 00:08:47,962
Each one is tested separately and reused.

132
00:08:49,212 --> 00:08:51,798
You start with the ready-made stages.

133
00:08:51,798 --> 00:08:56,691
When a custom calculation is needed, JavaScript or JSONata takes over;

134
00:08:56,691 --> 00:09:01,095
global functions are written once and used in every expression.

135
00:09:02,345 --> 00:09:05,807
Hubs at remote sites connect to a central hub.

136
00:09:05,807 --> 00:09:11,000
Configuration is monitored from one place, and namespaces are merged.

137
00:09:11,200 --> 00:09:15,251
High availability, version control with Git, an audit log,

138
00:09:15,251 --> 00:09:16,858
role-based permissions,

139
00:09:16,858 --> 00:09:20,420
and SAML or LDAP authentication come along with it.

140
00:09:21,520 --> 00:09:23,635
If the target connection drops,

141
00:09:23,635 --> 00:09:28,205
data is buffered to disk and delivered when the connection returns.

142
00:09:29,805 --> 00:09:33,468
In short: data is connected once, modeled once,

143
00:09:33,468 --> 00:09:36,975
and reaches everywhere with the same meaning.

144
00:09:37,175 --> 00:09:40,575
For engineers, this means parameterized templates,

145
00:09:40,575 --> 00:09:43,362
the flexibility of JavaScript, debugging,

146
00:09:43,362 --> 00:09:45,334
and version control with Git.

147
00:09:45,534 --> 00:09:46,487
For managers,

148
00:09:46,487 --> 00:09:51,621
it means a data infrastructure that is monitored from a single screen,

149
00:09:51,621 --> 00:09:56,388
works the same way at every site, and is AI-ready and manageable.

150
00:09:56,588 --> 00:10:01,117
Teams spend their time designing data instead of writing integration code.

151
00:10:02,717 --> 00:10:07,890
ASP Dijital, as HighByte's regional distributor, is by your side,

152
00:10:07,890 --> 00:10:12,346
from installation to modeling, from training to support.

153
00:10:12,446 --> 00:10:15,370
Let's make your plant's data useful, together.