Repository navigation
Expand file tree
/
Copy pathslides.html
More file actions
1446 lines (1378 loc) · 85.2 KB
/
Copy pathslides.html
File metadata and controls
1446 lines (1378 loc) · 85.2 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>Modeling the Factory — Why Processes Trump Assets</title>
<style>
:root{
--bg:#0b1020; --bg2:#111830; --ink:#e8edf7; --dim:#94a3c4;
--line:#26314f; --accent:#f5b544; --accent2:#4fd1c5; --bad:#ff6b6b;
--good:#4ade80; --panel:#151d38;
--mono:"SF Mono",ui-monospace,Menlo,Consolas,monospace;
}
*{box-sizing:border-box;margin:0;padding:0}
html,body{height:100%}
body{
background:var(--bg); color:var(--ink); overflow:hidden;
font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Inter,system-ui,sans-serif;
-webkit-font-smoothing:antialiased;
}
#stage{position:fixed;inset:0;display:flex;align-items:center;justify-content:center}
#deck{
width:1280px;height:720px;position:relative;transform-origin:center center;
background:var(--bg);
}
.slide{
position:absolute;inset:0;padding:46px 60px 62px;display:none;
flex-direction:column;
}
.slide.on{display:flex;animation:fade .28s ease}
@keyframes fade{from{opacity:0;transform:translateY(8px)}to{opacity:1;transform:none}}
.kicker{
font-size:13px;letter-spacing:.18em;text-transform:uppercase;
color:var(--accent);font-weight:700;margin-bottom:10px;
}
h1{font-size:52px;line-height:1.08;letter-spacing:-.02em;font-weight:800}
h2{font-size:34px;line-height:1.15;letter-spacing:-.015em;font-weight:750;margin-bottom:6px}
h3{font-size:17px;font-weight:700;margin-bottom:8px}
.sub{color:var(--dim);font-size:18px;margin-top:10px;line-height:1.45;max-width:1000px}
.body{flex:1;display:flex;gap:26px;margin-top:20px;min-height:0}
.col{flex:1;min-width:0;display:flex;flex-direction:column;gap:14px}
.card{
background:var(--panel);border:1px solid var(--line);border-radius:12px;
padding:16px 18px;
}
.card.good{border-color:#255c3b;background:#10241c}
.card.bad{border-color:#5c2828;background:#241214}
.card h3{display:flex;align-items:center;gap:8px}
.tag{font-size:11px;padding:2px 8px;border-radius:99px;font-weight:800;letter-spacing:.06em}
.tag.p{background:#153d2a;color:var(--good)}
.tag.c{background:#3d1717;color:var(--bad)}
ul{list-style:none;display:flex;flex-direction:column;gap:9px}
li{font-size:15.5px;line-height:1.42;color:#cdd7ee;padding-left:18px;position:relative}
li::before{content:"";position:absolute;left:0;top:8px;width:6px;height:6px;border-radius:2px;background:var(--accent2)}
.card.bad li::before{background:var(--bad)}
.card.good li::before{background:var(--good)}
li b{color:#fff}
.sm li{font-size:14px}
table{border-collapse:collapse;width:100%;font-size:14.5px}
th,td{padding:8px 10px;text-align:left;border-bottom:1px solid var(--line)}
th{color:var(--dim);font-size:12px;text-transform:uppercase;letter-spacing:.08em;font-weight:700}
td.num{text-align:right;font-family:var(--mono);font-variant-numeric:tabular-nums}
tr.hi td{background:#1c2547}
.ok{color:var(--good);font-weight:700}
.no{color:var(--bad);font-weight:700}
code,.mono{font-family:var(--mono);font-size:13px;color:var(--accent2)}
.note{
margin-top:auto;padding-top:12px;border-top:1px solid var(--line);
color:var(--dim);font-size:13.5px;line-height:1.4
}
.note b{color:var(--accent)}
/* tighter rhythm for field-heavy slides */
.dense .col{gap:10px}
.dense .card{padding:11px 15px}
.dense h3{font-size:15.5px;margin-bottom:5px}
.dense table td{padding:5px 10px}
.dense table th{padding:4px 10px}
.dense li{font-size:14.5px;line-height:1.38}
.dense .stat{padding:11px 14px}
.dense .stat .val{font-size:30px;margin-top:4px}
.big{font-size:46px;font-weight:800;font-family:var(--mono);letter-spacing:-.02em}
.big.g{color:var(--good)} .big.r{color:var(--bad)} .big.a{color:var(--accent)}
.statgrid{display:grid;grid-template-columns:repeat(4,1fr);gap:14px}
.stat{background:var(--panel);border:1px solid var(--line);border-radius:12px;padding:14px 16px}
.stat .lab{font-size:12px;color:var(--dim);text-transform:uppercase;letter-spacing:.08em;font-weight:700}
.stat .val{font-size:34px;font-weight:800;font-family:var(--mono);margin-top:6px}
.stat .cap{font-size:12.5px;color:var(--dim);margin-top:4px}
svg{display:block;max-width:100%}
.fig{display:flex;align-items:flex-start;justify-content:center;flex:1;min-height:0}
.fig img{width:100%;height:auto;max-height:100%;object-fit:contain;border-radius:10px;border:1px solid var(--line);background:#fff}
/* chrome */
#bar{position:fixed;left:0;bottom:0;height:3px;background:var(--accent);transition:width .25s;z-index:9}
#num{position:fixed;right:16px;bottom:12px;color:var(--dim);font-size:12px;font-family:var(--mono);z-index:9}
#hint{position:fixed;left:16px;bottom:12px;color:#4a5573;font-size:12px;z-index:9}
.title-slide{justify-content:center;align-items:flex-start}
.title-slide .meta{margin-top:34px;color:var(--dim);font-size:15px;line-height:1.7}
.rule{height:4px;width:80px;background:var(--accent);border-radius:2px;margin:22px 0}
.flow{display:flex;align-items:center;gap:10px;flex-wrap:wrap}
.node{background:var(--panel);border:1px solid var(--line);border-radius:10px;padding:10px 12px;font-size:13px}
.node .t{font-weight:700;color:#fff;font-size:13.5px}
.node .d{color:var(--dim);font-size:11.5px;margin-top:3px;font-family:var(--mono)}
.arrow{color:var(--accent);font-size:20px}
@media print{
body{overflow:visible;background:#fff}
#stage{position:static}
#deck{transform:none!important;width:1280px;height:720px}
.slide{display:flex!important;page-break-after:always;position:relative;height:720px}
#bar,#num,#hint{display:none}
}
</style>
</head>
<body>
<div id="stage"><div id="deck">
<!-- 1 ─────────────────────────────────────────── TITLE -->
<section class="slide title-slide">
<div class="kicker">Industrial Data Architecture · Proof of Concept</div>
<h1>Modeling the Factory<br/><span style="color:var(--accent)">Why Processes Trump Assets</span></h1>
<div class="rule"></div>
<div class="sub" style="font-size:21px;max-width:900px">
Connecting every sensor to a machine is the path of least resistance —
and it quietly closes the door on the most valuable questions.
I didn't argue this. I built a controlled experiment and measured it.
</div>
<div class="meta">
<b style="color:var(--ink)">process-modelling-poc</b> · synthetic brewery · fermentation + conditioning<br/>
Same algorithm · same hyperparameters · <span class="mono">SEED=42</span> — <b style="color:var(--accent)">only the data layer changes</b>
</div>
</section>
<!-- 2 ─────────────────────────────────────────── THE SCENE -->
<section class="slide">
<div class="kicker">The familiar scene</div>
<h2>A fully instrumented plant that cannot answer a simple question</h2>
<div class="body">
<div class="col" style="flex:1.05">
<div class="card" style="flex:1;display:flex;flex-direction:column;justify-content:center">
<div style="font-family:var(--mono);font-size:14px;line-height:2;color:var(--dim)">
<span style="color:var(--accent2)">FT-2043</span> flow 412.8 m³/h<br/>
<span style="color:var(--accent2)">TT-1180</span> temperature 18.4 °C<br/>
<span style="color:var(--accent2)">PT-0091</span> pressure 1.42 bar<br/>
<span style="color:var(--accent2)">VT-3310</span> vibration 2.1 mm/s<br/>
<span style="color:#4a5573">… 8,400 more tags</span>
</div>
<div style="margin-top:22px;padding:14px 16px;background:#241214;border:1px solid #5c2828;border-radius:10px">
<div style="font-size:16px;line-height:1.5">
<b>“That batch the customer complained about — which family of tanks
did it run in, on which recipe, on which temperature curve?”</b>
</div>
<div style="margin-top:10px;color:var(--bad);font-weight:700;font-size:15px">→ The system has no answer.</div>
</div>
</div>
</div>
<div class="col">
<div class="card">
<h3>The data is there. The meaning is not.</h3>
<ul>
<li>Temperature exists as <b>“tank 3's temperature”</b>, never as <b>“that Pilsner batch's fermentation curve.”</b></li>
<li>Every sensor faithfully watches its own machine.<br/><b>Nobody is watching the batch.</b></li>
</ul>
</div>
<div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The tension at the core</h3>
<div style="font-size:17px;line-height:1.5">
When we model the factory as a <b>list of machines</b>,
we never model <b>production itself</b> — the very thing we
actually care about.
</div>
</div>
<div class="card sm">
<ul>
<li>Failing to trace one batch is only the <b>symptom</b>.</li>
<li>The disease: being nailed to the bottom rung of the maturity ladder.</li>
</ul>
</div>
</div>
</div>
</section>
<!-- 3 ─────────────────────────────────────────── ASSET: WHAT -->
<section class="slide">
<div class="kicker">1 · Asset modelling</div>
<h2>The world as a tree of physical things <span style="color:var(--dim);font-size:20px;font-weight:500">ISA-95 physical hierarchy</span></h2>
<div class="body">
<div class="col" style="flex:1.15">
<svg viewBox="0 0 560 420" style="margin-top:4px">
<defs><style>
.bx{fill:#151d38;stroke:#26314f}
.lb{fill:#e8edf7;font:600 13px -apple-system,sans-serif}
.lb2{fill:#94a3c4;font:11px "SF Mono",monospace}
.ln{stroke:#26314f;stroke-width:1.5;fill:none}
.leaf{fill:#1b2547;stroke:#3a4a7a}
</style></defs>
<rect class="bx" x="10" y="8" width="150" height="34" rx="7"/>
<text class="lb" x="24" y="30">Site · Brewery</text>
<path class="ln" d="M30 42 V64 H30"/>
<rect class="bx" x="30" y="64" width="150" height="34" rx="7"/>
<text class="lb" x="44" y="86">Area · Cellar</text>
<path class="ln" d="M50 98 V120"/>
<rect class="bx" x="50" y="120" width="170" height="34" rx="7"/>
<text class="lb" x="64" y="142">ProcessCell · Line 2</text>
<path class="ln" d="M70 154 V176"/>
<rect class="bx" x="70" y="176" width="170" height="34" rx="7"/>
<text class="lb" x="84" y="198">Unit · Fermenter FV-3</text>
<path class="ln" d="M92 210 V236 H120"/>
<path class="ln" d="M92 236 V276 H120"/>
<path class="ln" d="M92 276 V316 H120"/>
<rect class="leaf" x="120" y="220" width="230" height="32" rx="7"/>
<text class="lb2" x="134" y="240">TT-1180 · temperature °C</text>
<rect class="leaf" x="120" y="260" width="230" height="32" rx="7"/>
<text class="lb2" x="134" y="280">ZT-1181 · valve position %</text>
<rect class="leaf" x="120" y="300" width="230" height="32" rx="7"/>
<text class="lb2" x="134" y="320">DT-1182 · specific gravity</text>
<text x="10" y="372" style="fill:#f5b544;font:700 13px -apple-system,sans-serif">The weight of the model sits in the leaves.</text>
<text x="10" y="394" style="fill:#94a3c4;font:13px -apple-system,sans-serif">Every datum is hung on a machine. There is no node called “Batch.”</text>
</svg>
</div>
<div class="col">
<div class="card">
<h3>What it answers well</h3>
<ul><li><b>“Which equipment is where, and how is it doing?”</b></li></ul>
</div>
<div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">Why is it so common?</h3>
<div style="font-size:15.5px;line-height:1.5;color:#cdd7ee">
Because <b>the sensors are already attached to the machines.</b>
PLC tags arrive machine-by-machine. This model isn't really <i>built</i> —
it's the shape the raw data already has on the wire.
</div>
<div style="margin-top:12px;font-size:15px;color:var(--accent)">
<b>It is the path of least resistance. Doing nothing, this is what you get.</b>
</div>
</div>
<div class="card sm">
<h3>What it structurally lacks</h3>
<ul>
<li>No <code>Batch</code> / <code>Recipe</code> / <code>ProcessSegment</code> entity</li>
<li>Vessel nodes carry identity & location only — <b>no volume, unit, thermal mass, cooling capacity</b></li>
</ul>
</div>
</div>
</div>
</section>
<!-- 4 ─────────────────────────────────────────── ASSET: PROS/CONS -->
<section class="slide">
<div class="kicker">1 · Asset modelling</div>
<h2>Pros and cons</h2>
<div class="sub">It is not <i>wrong</i>. It never gives wrong answers. The problem is more insidious — it is <b style="color:var(--accent)">insufficient</b>.</div>
<div class="body">
<div class="col">
<div class="card good" style="flex:1">
<h3><span class="tag p">PROS</span> Why teams pick it</h3>
<ul>
<li><b>Zero modelling cost.</b> Tags already arrive this shape; no semantic work needed.</li>
<li><b>Standard & familiar.</b> ISA-95 tree maps 1:1 to how maintenance and automation already think.</li>
<li><b>Excellent for machine health.</b> Vibration, runtime hours, predictive maintenance on a compressor — genuinely strong.</li>
<li><b>Unambiguous ownership.</b> Every tag has exactly one physical parent.</li>
<li><b>Immediate demo value.</b> Dashboards light up on day one.</li>
</ul>
</div>
</div>
<div class="col">
<div class="card bad" style="flex:1">
<h3><span class="tag c">CONS</span> The invisible ceiling</h3>
<ul>
<li><b>No run entity.</b> “Which machines did batch X pass through?” cannot be <i>derived</i> — the concept doesn't exist in the tree.</li>
<li><b>Schema follows topology.</b> Add a tank → the schema grows. A→B here: <b>5 vs 17 columns</b>, non-alignable.</li>
<li><b>Models are site-locked.</b> A model speaks the dialect of the plant it was born in.</li>
<li><b>Process blindness.</b> Is the liquid in the pump mashing or fermenting? No chemical context.</li>
<li><b>N×M definition explosion.</b> Every topology × equipment type = another copy of the same definition.</li>
<li><b>No provenance semantics.</b> Nothing says “this value can't be allocated per vessel.”</li>
</ul>
</div>
</div>
</div>
<div class="note">Honest framing: the asset model is not a mistake — it is a <b>default</b>. The real problem is never questioning the default.</div>
</section>
<!-- 5 ─────────────────────────────────────────── PROCESS: WHAT -->
<section class="slide">
<div class="kicker">2 · Process modelling</div>
<h2>The world as the work itself <span style="color:var(--dim);font-size:20px;font-weight:500">ISA-88 · the run is the unit</span></h2>
<div class="body" style="flex-direction:column">
<svg viewBox="0 0 1160 260">
<defs><style>
.ph{fill:#1b2547;stroke:#3a4a7a;rx:8}
.pt{fill:#e8edf7;font:600 13px -apple-system,sans-serif}
.pd{fill:#94a3c4;font:11px "SF Mono",monospace}
.eq{fill:#151d38;stroke:#26314f}
.ax{stroke:#f5b544;stroke-width:2.5;fill:none}
</style></defs>
<!-- run bar -->
<rect x="20" y="20" width="1120" height="46" rx="9" fill="#1e1809" stroke="#4a3a12"/>
<text x="38" y="41" style="fill:#f5b544;font:800 13px 'SF Mono',monospace">ProcessRun</text>
<text x="38" y="58" style="fill:#cdd7ee;font:12px 'SF Mono',monospace">run_id = Topology_B_Pilsner_HEALTHY_03 · product = Pilsner · one identity, start to finish</text>
<path class="ax" d="M20 78 H1140"/>
<!-- phases -->
<g>
<rect class="ph" x="20" y="94" width="250" height="50"/>
<text class="pt" x="36" y="116">Mashing</text><text class="pd" x="36" y="133">phase</text>
<rect class="ph" x="286" y="94" width="420" height="50" fill="#152e2b" stroke="#2c6a63"/>
<text class="pt" x="302" y="116">Fermentation</text><text class="pd" x="302" y="133">7 days · 1008 obs</text>
<rect class="ph" x="722" y="94" width="270" height="50" fill="#2b1f36" stroke="#5a3f70"/>
<text class="pt" x="738" y="116">Conditioning</text><text class="pd" x="738" y="133">different equipment type</text>
<rect class="ph" x="1008" y="94" width="132" height="50"/>
<text class="pt" x="1024" y="116">Packaging</text><text class="pd" x="1024" y="133">phase</text>
</g>
<!-- equipment actors -->
<text x="20" y="176" style="fill:#94a3c4;font:11px -apple-system,sans-serif;letter-spacing:.1em">EQUIPMENT ARE ONLY ACTORS ↓</text>
<g>
<rect class="eq" x="20" y="188" width="250" height="40" rx="7"/>
<text class="pd" x="36" y="213">Mash tun MT-1</text>
<rect class="eq" x="286" y="188" width="200" height="40" rx="7"/>
<text class="pd" x="302" y="213">FV-1 · FV-2</text>
<rect class="eq" x="500" y="188" width="206" height="40" rx="7"/>
<text class="pd" x="516" y="213">FV-3 · FV-4</text>
<rect class="eq" x="722" y="188" width="270" height="40" rx="7"/>
<text class="pd" x="738" y="213">Brite BT-1 · BT-2</text>
<rect class="eq" x="1008" y="188" width="132" height="40" rx="7"/>
<text class="pd" x="1024" y="213">Filler LN-2</text>
</g>
</svg>
<div class="body" style="margin-top:6px">
<div class="col"><div class="card">
<h3>The unit of analysis is a run, not a machine</h3>
<ul><li>“Batch X of Pilsner, from mashing to bottling.” The same run may pass through <b>different numbers and types of machines</b> at different plants — it is still the same run.</li></ul>
</div></div>
<div class="col"><div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The hospital analogy</h3>
<div style="font-size:15px;line-height:1.5;color:#cdd7ee">
Asset model = monitoring each <b>device</b> (is the MRI up?).
Process model = monitoring the <b>patient</b> (which path did they take, did they get better?).
<b style="color:var(--accent)">Nobody walks into a hospital asking how the MRI is doing.</b>
</div>
</div></div>
</div>
</div>
</section>
<!-- 6 ─────────────────────────────────────────── PROCESS: PROS/CONS -->
<section class="slide">
<div class="kicker">2 · Process modelling</div>
<h2>Pros and cons</h2>
<div class="sub">Stated honestly — including the costs, because an argument without limits is marketing.</div>
<div class="body">
<div class="col">
<div class="card good" style="flex:1">
<h3><span class="tag p">PROS</span> What it unlocks</h3>
<ul>
<li><b>Topology independence.</b> 1 tank or 4 tanks → one row. Count becomes an attribute, not a schema change.</li>
<li><b>Cross-plant comparability.</b> Fermentation is fermentation everywhere. Apples to apples.</li>
<li><b>Model portability.</b> Train once, run at the next site without a rewrite.</li>
<li><b>End-to-end traceability.</b> One key spans disjoint equipment types and stages.</li>
<li><b>Many-to-one lineage.</b> Several ferment runs → one blend lot; the asset tree cannot express this.</li>
<li><b>Provenance is first-class.</b> Coverage, scope and allocability travel inside the row.</li>
</ul>
</div>
</div>
<div class="col">
<div class="card bad" style="flex:1">
<h3><span class="tag c">CONS</span> What it costs</h3>
<ul>
<li><b>Requires domain expertise.</b> Someone must know what fermentation is and why diacetyl matters. Not just a software job.</li>
<li><b>Upfront modelling effort.</b> The asset model comes free; this one must be designed.</li>
<li><b>Curation dependency.</b> Volume, unit, thermal mass don't exist in the registry — they must be curated to compute intensive features.</li>
<li><b>Over-aggregation risk.</b> A volume-weighted mean can mask one hot vessel — that's why <code>spread</code> features are deliberately kept.</li>
<li><b>It does not learn physics.</b> An unseen topology still needs local calibration (<b>11.8% FPR on C</b>).</li>
<li><b>Weak payoff on a small static fleet.</b> No six topologies → no six adapters saved.</li>
</ul>
</div>
</div>
</div>
</section>
<!-- 7 ─────────────────────────────────────────── WHY: LADDER -->
<section class="slide">
<div class="kicker">3 · Why — the purpose</div>
<h2>The maturity ladder, and the ceiling the asset model imposes</h2>
<div class="body">
<div class="col" style="flex:1.25">
<svg viewBox="0 0 640 400">
<defs><style>
.rg{stroke:#26314f;fill:#151d38}
.rt{fill:#e8edf7;font:700 14px -apple-system,sans-serif}
.rs{fill:#94a3c4;font:12px -apple-system,sans-serif}
</style></defs>
<rect class="rg" x="10" y="330" width="380" height="52" rx="8"/>
<text class="rt" x="26" y="352">1 · What happened?</text>
<text class="rs" x="26" y="371">Descriptive — asset model is genuinely strong here</text>
<rect class="rg" x="10" y="268" width="430" height="52" rx="8"/>
<text class="rt" x="26" y="290">2 · Why did it happen?</text>
<text class="rs" x="26" y="309">Root cause — needs the end-to-end chain, stitched by hand today</text>
<rect class="rg" x="10" y="206" width="480" height="52" rx="8" fill="#241214" stroke="#5c2828"/>
<text class="rt" x="26" y="228">3 · What will happen?</text>
<text class="rs" x="26" y="247">Predictive — reaches “will this compressor fail?”, not product quality</text>
<!-- ceiling -->
<path d="M0 196 H640" stroke="#ff6b6b" stroke-width="2.5" stroke-dasharray="7 5"/>
<rect x="392" y="176" width="238" height="24" rx="6" fill="#3d1717"/>
<text x="404" y="193" style="fill:#ff6b6b;font:800 12px -apple-system,sans-serif">ASSET MODEL CEILING</text>
<rect class="rg" x="10" y="112" width="540" height="52" rx="8" fill="#10241c" stroke="#255c3b"/>
<text class="rt" x="26" y="134">4 · Benchmarking plants</text>
<text class="rs" x="26" y="153">Different brands, same job — comparable only at process level</text>
<rect class="rg" x="10" y="50" width="600" height="52" rx="8" fill="#10241c" stroke="#255c3b"/>
<text class="rt" x="26" y="72">5 · Learning at scale</text>
<text class="rs" x="26" y="91">Train once, deploy everywhere — impossible without a shared unit</text>
<text x="10" y="26" style="fill:#4ade80;font:800 13px -apple-system,sans-serif">↑ where the most valuable questions live</text>
</svg>
</div>
<div class="col">
<div class="card bad">
<h3>The mechanism of the ceiling</h3>
<ul>
<li>Rungs 4 and 5 both require a <b>shared unit of analysis</b>.</li>
<li>A machine-specific schema <b>imprisons the model in the site where it was born</b>.</li>
</ul>
</div>
<div class="card">
<h3>It doesn't give wrong answers…</h3>
<div style="font-size:15.5px;line-height:1.5;color:#cdd7ee">
…it simply <b>doesn't let you climb.</b> That is why this is an
architecture problem, not a modelling-accuracy problem.
</div>
</div>
</div>
</div>
</section>
<!-- 8 ─────────────────────────────────────────── WHY: THREE OUTCOMES -->
<section class="slide">
<div class="kicker">3 · Why — the purpose</div>
<h2>Three outcomes we are actually buying</h2>
<div class="body">
<div class="col">
<div class="card" style="flex:1">
<div style="font-size:26px;margin-bottom:8px">⏱️</div>
<h3 style="font-size:19px;color:var(--accent)">Shorter use-case development</h3>
<ul class="sm">
<li>Logic built once runs everywhere, because data is harmonized <b>at the source of meaning</b>.</li>
<li>Months-long site-specific projects → <b>weeks-long repeatable solutions</b>.</li>
<li>Measured proxy in this PoC: <b>6 adapters → 1 aggregator</b> for 3 topologies × 2 equipment types.</li>
</ul>
<div style="margin-top:auto;padding-top:12px;font-family:var(--mono);font-size:12px;color:var(--dim)">
ownership_cost.multi_type_reuse_ratio = <b style="color:var(--accent)">6.0</b>
</div>
</div>
</div>
<div class="col">
<div class="card" style="flex:1">
<div style="font-size:26px;margin-bottom:8px">🔁</div>
<h3 style="font-size:19px;color:var(--accent)">Model interoperability</h3>
<ul class="sm">
<li>An anomaly or quality model trained at one site <b>runs at the next without a rewrite</b>.</li>
<li>Requires the new site's normal behaviour to live in the <b>same analysis space</b>.</li>
<li>Measured: raw transfer A→B <b class="no">35.9%</b> vs. contract A+B→B <b class="ok">96.6%</b>.</li>
</ul>
<div style="margin-top:auto;padding-top:12px;font-family:var(--mono);font-size:12px;color:var(--dim)">
same OneClassSVM · same nu · same seed
</div>
</div>
</div>
<div class="col">
<div class="card" style="flex:1">
<div style="font-size:26px;margin-bottom:8px">📊</div>
<h3 style="font-size:19px;color:var(--accent)">Benchmarking lines & facilities</h3>
<ul class="sm">
<li>“Which plant makes the same product with less energy?”</li>
<li>“How efficiently does a German fermenter vs. a locally built reactor do the same job?”</li>
<li>Needs <b>intensive</b> features (per-HL), not extensive totals — a contract decision, not a query.</li>
</ul>
<div style="margin-top:auto;padding-top:12px;font-family:var(--mono);font-size:12px;color:var(--dim)">
cooling_intensity_c_per_min = total / Σ volume
</div>
</div>
</div>
</div>
<div class="note">All three require the <b>same prerequisite</b>: a stable analytical unit that survives a change of equipment. That is the whole thesis.</div>
</section>
<!-- 9 ─────────────────────────────────────────── HOW: THE SETUP -->
<section class="slide">
<div class="kicker">4 · How</div>
<h2>The experiment: one process, three physical realities</h2>
<div class="body" style="flex-direction:column">
<svg viewBox="0 0 1160 300">
<defs><style>
.tk{fill:#1b2547;stroke:#3a4a7a;stroke-width:1.5}
.tkl{fill:#cdd7ee;font:600 11px "SF Mono",monospace;text-anchor:middle}
.hd{fill:#f5b544;font:800 15px -apple-system,sans-serif}
.mt{fill:#94a3c4;font:12px -apple-system,sans-serif}
.pnl{fill:#151d38;stroke:#26314f;rx:12}
</style></defs>
<!-- A -->
<rect class="pnl" x="10" y="10" width="360" height="280" rx="12"/>
<text class="hd" x="30" y="40">Topology A</text>
<text class="mt" x="30" y="60">1 × 200 HL · analog valve · °C</text>
<rect class="tk" x="140" y="80" width="100" height="130" rx="14"/>
<text class="tkl" x="190" y="150">200 HL</text>
<text class="mt" x="30" y="240" style="fill:#4fd1c5;font-family:'SF Mono',monospace;font-size:12px">raw schema · 5 columns</text>
<text class="mt" x="30" y="262">temp_c · spread · gravity · valve_pct · cooling</text>
<!-- B -->
<rect class="pnl" x="390" y="10" width="380" height="280" rx="12"/>
<text class="hd" x="410" y="40">Topology B</text>
<text class="mt" x="410" y="60">4 × 100 HL parallel · solenoid · °F</text>
<g>
<rect class="tk" x="418" y="90" width="70" height="110" rx="12"/><text class="tkl" x="453" y="150">100</text>
<rect class="tk" x="500" y="90" width="70" height="110" rx="12"/><text class="tkl" x="535" y="150">100</text>
<rect class="tk" x="582" y="90" width="70" height="110" rx="12"/><text class="tkl" x="617" y="150">100</text>
<rect class="tk" x="664" y="90" width="70" height="110" rx="12"/><text class="tkl" x="699" y="150">100</text>
</g>
<text class="mt" x="410" y="240" style="fill:#ff6b6b;font-family:'SF Mono',monospace;font-size:12px">raw schema · 17 columns</text>
<text class="mt" x="410" y="262">vessel_1..4 × (temp_f, spread, gravity, solenoid) + cooling</text>
<!-- C -->
<rect class="pnl" x="790" y="10" width="360" height="280" rx="12"/>
<text class="hd" x="810" y="40">Topology C</text>
<text class="mt" x="810" y="60">2 × 150 HL + shared glycol · °C</text>
<rect class="tk" x="850" y="104" width="90" height="96" rx="12"/><text class="tkl" x="895" y="157">150</text>
<rect class="tk" x="960" y="104" width="90" height="96" rx="12"/><text class="tkl" x="1005" y="157">150</text>
<path d="M895 104 V90 H1005 V104" stroke="#f5b544" stroke-width="2" fill="none" stroke-dasharray="4 4"/>
<text x="950" y="84" style="fill:#f5b544;font:700 11px 'SF Mono',monospace;text-anchor:middle">shared utility — total only</text>
<text class="mt" x="810" y="240" style="fill:#4fd1c5;font-family:'SF Mono',monospace;font-size:12px">raw schema · 7 columns</text>
<text class="mt" x="810" y="262">allocation_to_equipment_available = <tspan style="fill:#ff6b6b;font-weight:700">False</tspan></text>
</svg>
<div class="body" style="margin-top:2px">
<div class="col"><div class="card sm">
<h3>Axis 1 — same equipment type, different layout</h3>
<ul><li>Chosen deliberately: this is the ground where the <b>opposing argument is strongest</b>. With one equipment type, “just aggregate on the asset side” genuinely sounds reasonable.</li></ul>
</div></div>
<div class="col"><div class="card sm">
<h3>Axis 2 — different equipment type, one run</h3>
<ul><li>The product matures in a <b>Conditioning (Brite) tank</b>: temperature + <b>pressure</b> + <b>diacetyl (VDK)</b>. Schema fully <b>disjoint</b> from the fermenter.</li></ul>
</div></div>
<div class="col"><div class="card sm" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The control that makes it evidence</h3>
<ul><li>Same <code>OneClassSVM</code>, <code>nu=0.01</code>, <code>gamma=scale</code>, <code>SEED=42</code>.<br/><b>The only variable is the data layer.</b></li></ul>
</div></div>
</div>
</div>
<div class="note">Each batch: <b>7 days at 10-minute intervals = 1008 observations</b> · 10 batches per (product × class) · healthy batches 0–6 train, 7–9 test · anomalous batches used for evaluation only.</div>
</section>
<!-- 10 ─────────────────────────────────────────── HOW: ARCHITECTURE -->
<section class="slide">
<div class="kicker">4 · How — architecture</div>
<h2>Four layers: from native tags to a canonical run row</h2>
<div class="body" style="flex-direction:column;gap:12px">
<svg viewBox="0 0 1160 372">
<defs>
<marker id="ar" markerWidth="9" markerHeight="9" refX="7" refY="4.5" orient="auto">
<path d="M0 0 L9 4.5 L0 9 z" fill="#f5b544"/>
</marker>
<style>
.lyr{fill:#151d38;stroke:#26314f;rx:10}
.lt{fill:#fff;font:800 13px -apple-system,sans-serif}
.ld{fill:#94a3c4;font:11.5px "SF Mono",monospace}
.lc{fill:#cdd7ee;font:12px -apple-system,sans-serif}
.fl{stroke:#f5b544;stroke-width:2;fill:none;marker-end:url(#ar)}
.n{fill:#f5b544;font:800 11px "SF Mono",monospace}
</style>
</defs>
<rect class="lyr" x="10" y="30" width="250" height="258" rx="10"/>
<text class="n" x="26" y="54">LAYER 0</text>
<text class="lt" x="26" y="76">Native per-vessel physics</text>
<text class="ld" x="26" y="100">topology_b.vessel_3.temp_f</text>
<text class="ld" x="26" y="120">topology_b.vessel_3.solenoid</text>
<text class="ld" x="26" y="140">topology_c.brite_1.vdk_ppb</text>
<text class="lc" x="26" y="170">Honest raw layer.</text>
<text class="lc" x="26" y="190">5 / 17 / 7 columns —</text>
<text class="lc" x="26" y="210">structurally unstackable.</text>
<path d="M26 232 H244" stroke="#26314f"/>
<text class="n" x="26" y="252" style="fill:#94a3c4">OWNED BY</text>
<text class="lc" x="26" y="272">the PLC / historian</text>
<path class="fl" d="M266 160 H296"/>
<rect class="lyr" x="302" y="30" width="270" height="258" rx="10" fill="#152e2b" stroke="#2c6a63"/>
<text class="n" x="318" y="54" style="fill:#4fd1c5">LAYER 1 · the interface</text>
<text class="lt" x="318" y="76">Role-tagged long form</text>
<text class="ld" x="318" y="100">equipment_role · active_volume_hl</text>
<text class="ld" x="318" y="120">temperature_value</text>
<text class="ld" x="318" y="140">temperature_unit</text>
<text class="lc" x="318" y="170">Every vessel of every plant</text>
<text class="lc" x="318" y="190">emits the SAME row shape.</text>
<text class="lc" x="318" y="210">Units & weights travel as data.</text>
<path d="M318 232 H556" stroke="#2c6a63"/>
<text class="n" x="318" y="252" style="fill:#4fd1c5">OWNED BY</text>
<text class="lc" x="318" y="272">the contract — this is the pivot</text>
<path class="fl" d="M578 160 H608"/>
<rect class="lyr" x="614" y="30" width="250" height="258" rx="10"/>
<text class="n" x="630" y="54">LAYER 2</text>
<text class="lt" x="630" y="76">Generic aggregator</text>
<text class="ld" x="630" y="100">1 · normalize (unit-driven)</text>
<text class="ld" x="630" y="120">2 · weight by volume</text>
<text class="ld" x="630" y="140">3 · group by run key</text>
<text class="ld" x="630" y="160">4 · divide → intensive</text>
<text class="lc" x="630" y="190">~60 lines, zero</text>
<text class="lc" x="630" y="210">site-specific branches.</text>
<path d="M630 232 H848" stroke="#26314f"/>
<text class="n" x="630" y="252" style="fill:#94a3c4">OWNED BY</text>
<text class="lc" x="630" y="272">one platform team, once</text>
<path class="fl" d="M870 160 H900"/>
<rect class="lyr" x="906" y="30" width="244" height="258" rx="10" fill="#1e1809" stroke="#4a3a12"/>
<text class="n" x="922" y="54">LAYER 3</text>
<text class="lt" x="922" y="76">ProcessRun contract</text>
<text class="ld" x="922" y="100">+ derived model features</text>
<text class="ld" x="922" y="120">temp_setpoint_error_c</text>
<text class="ld" x="922" y="140">gravity_attenuation</text>
<text class="lc" x="922" y="170">Product-relative, so</text>
<text class="lc" x="922" y="190">Pilsner 12 °C and Ale 20 °C</text>
<text class="lc" x="922" y="210">both sit near error ≈ 0.</text>
<path d="M922 232 H1134" stroke="#4a3a12"/>
<text class="n" x="922" y="252">OWNED BY</text>
<text class="lc" x="922" y="272">the data scientist</text>
<text x="10" y="322" style="fill:#94a3c4;font:12px 'SF Mono',monospace">simulator.py → topology.py::_build_source_frames → topology.py::aggregate_process_contract → analyzer.py::add_topology_features</text>
<text x="10" y="350" style="fill:#f5b544;font:700 13px -apple-system,sans-serif">Normalization happens ONCE, generically — because units and volumes are carried as data on a role-tagged row.</text>
</svg>
<div class="card" style="border-color:#4a3a12;background:#1e1809;padding:12px 18px">
<div style="font-size:15.5px;line-height:1.45">
<b style="color:var(--accent)">Layer 1 is the actual pivot.</b>
Not the maths in Layer 2 — anyone can average four numbers.
The move is that the value, <b>its unit</b> and <b>its weight</b> arrive together on a row whose shape does not depend on the plant.
</div>
</div>
</div>
</section>
<!-- 11 ─────────────────────────────────────────── HOW: GRAIN -->
<section class="slide">
<div class="kicker">4 · How — the grain</div>
<h2>Decision #1: what <i>is</i> one row?</h2>
<div class="body">
<div class="col" style="flex:1.1">
<svg viewBox="0 0 620 340">
<defs><style>
.r{fill:#241214;stroke:#5c2828;rx:8}
.g{fill:#10241c;stroke:#255c3b;rx:8}
.t{fill:#e8edf7;font:600 12.5px "SF Mono",monospace}
.h{fill:#94a3c4;font:700 11px -apple-system,sans-serif;letter-spacing:.1em}
</style></defs>
<text class="h" x="10" y="18">WITHOUT A GRAIN DECISION — TOPOLOGY B</text>
<rect class="r" x="10" y="30" width="600" height="30" rx="8"/>
<text class="t" x="26" y="50">FV-1 · 68.0 °F · 100 HL · solenoid=1</text>
<rect class="r" x="10" y="66" width="600" height="30" rx="8"/>
<text class="t" x="26" y="86">FV-2 · 67.8 °F · 100 HL · solenoid=0</text>
<rect class="r" x="10" y="102" width="600" height="30" rx="8"/>
<text class="t" x="26" y="122">FV-3 · 68.2 °F · 100 HL · solenoid=1</text>
<rect class="r" x="10" y="138" width="600" height="30" rx="8"/>
<text class="t" x="26" y="158">FV-4 · 67.9 °F · 100 HL · solenoid=0</text>
<text x="10" y="188" style="fill:#ff6b6b;font:700 13px -apple-system,sans-serif">4 rows → “which one is the batch?” → stacking problem</text>
<path d="M300 200 V228" stroke="#f5b544" stroke-width="2.5" marker-end="url(#ar)"/>
<text x="316" y="222" style="fill:#f5b544;font:700 12px 'SF Mono',monospace">group by run_id + timestamp</text>
<text class="h" x="10" y="256">WITH THE CONTRACT GRAIN</text>
<rect class="g" x="10" y="268" width="600" height="44" rx="8"/>
<text class="t" x="26" y="288">run_id=Topology_B_Pilsner_HEALTHY_03 · t=00:40</text>
<text class="t" x="26" y="306" style="fill:#4ade80">20.0 °C · 400 HL · active_equipment_count = 4</text>
<text x="10" y="332" style="fill:#4ade80;font:700 13px -apple-system,sans-serif">1 row — and Topology A's single 200 HL tank lands at the same coordinate.</text>
</svg>
</div>
<div class="col">
<div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The grain, stated</h3>
<div style="font-family:var(--mono);font-size:15px;line-height:1.6;color:#fff">
One analysis row =<br/><b style="color:var(--accent)">run_id × timestamp</b>
</div>
<div style="margin-top:10px;font-size:14.5px;color:#cdd7ee;line-height:1.45">
Note what is <b>absent</b> from the key: the equipment.
That absence is the entire design.
</div>
</div>
<div class="card sm">
<h3>What the grain buys you</h3>
<ul>
<li>Machine count no longer changes the schema — it becomes an <b>attribute</b> (<code>active_equipment_count</code>).</li>
<li>Different plants land in the <b>same analysis space</b>, so they can be trained together.</li>
<li><b>Product is an attribute too</b>, not the key — Pilsner and Ale share one model.</li>
</ul>
</div>
<div class="card bad sm">
<h3>“I can aggregate” ≠ a grain</h3>
<ul><li>Knowing how to compute the number does not tell you <b>where to put it</b>. Empirically: A-only canonical model still only reaches <b>74.0% / 35.9% FPR</b> on B.</li></ul>
</div>
</div>
</div>
</section>
<!-- 12 ─────────────────────────────────────────── HOW: FIELDS (FERMENTER) -->
<section class="slide dense">
<div class="kicker">4 · How — the contract</div>
<h2>The fields of one analysis row <span class="mono" style="font-size:15px;color:var(--dim)">data/topology_process_contract.csv · 21 columns</span></h2>
<div class="body" style="gap:18px">
<div class="col">
<div class="card" style="padding:12px 16px">
<h3><span style="color:var(--accent)">① IDENTITY</span> — what defines the row</h3>
<table>
<tr><td class="mono">timestamp</td><td style="color:var(--dim);font-size:13px">measurement instant</td></tr>
<tr><td class="mono">batch_id</td><td style="color:var(--dim);font-size:13px">run number within site</td></tr>
<tr class="hi"><td class="mono" style="color:var(--accent)">run_id</td><td style="font-size:13px"><b>the primary key</b></td></tr>
</table>
</div>
<div class="card" style="padding:12px 16px">
<h3><span style="color:var(--accent)">② CONTEXT</span> — attributes, never the key</h3>
<table>
<tr><td class="mono">site_id · topology</td><td style="color:var(--dim);font-size:13px">which plant / layout</td></tr>
<tr><td class="mono">product</td><td style="color:var(--dim);font-size:13px">Pilsner · Ale</td></tr>
<tr><td class="mono">split · anomaly_class</td><td style="color:var(--dim);font-size:13px">holdout · label</td></tr>
</table>
</div>
<div class="card" style="padding:12px 16px">
<h3><span style="color:var(--accent)">③ TIME</span> · <span style="color:var(--accent)">④ COMPOSITION</span></h3>
<table>
<tr><td class="mono">elapsed_step</td><td style="color:var(--dim);font-size:13px">0 – 1007</td></tr>
<tr><td class="mono">phase_progress_pct</td><td style="color:var(--dim);font-size:13px">topology-neutral progress</td></tr>
<tr><td class="mono">active_volume_hl</td><td style="color:var(--dim);font-size:13px">200 / 400 / 300</td></tr>
<tr><td class="mono">active_equipment_count</td><td style="color:var(--dim);font-size:13px">1 / 4 / 2</td></tr>
</table>
</div>
</div>
<div class="col">
<div class="card" style="padding:12px 16px;border-color:#2c6a63;background:#101f24">
<h3><span style="color:var(--accent2)">⑤ MEASUREMENT BODY</span> — all intensive</h3>
<table>
<tr><th>field</th><th>how</th><th>carries</th></tr>
<tr><td class="mono">process_temp_mean_c</td><td style="font-size:12.5px;color:var(--dim)">Σ(T·V)/ΣV</td><td style="font-size:12.5px">overheat, cold crash</td></tr>
<tr><td class="mono">process_temp_spread_c</td><td style="font-size:12.5px;color:var(--dim)">max−min <b>across</b></td><td style="font-size:12.5px">one hot vessel</td></tr>
<tr><td class="mono">intra_vessel_spread_c</td><td style="font-size:12.5px;color:var(--dim)">max <b>within</b></td><td style="font-size:12.5px">sensor drift</td></tr>
<tr><td class="mono">gravity_weighted_mean</td><td style="font-size:12.5px;color:var(--dim)">Σ(SG·V)/ΣV</td><td style="font-size:12.5px">stuck fermentation</td></tr>
<tr><td class="mono">gravity_spread</td><td style="font-size:12.5px;color:var(--dim)">max−min</td><td style="font-size:12.5px">vessel divergence</td></tr>
<tr><td class="mono">cooling_intensity_c_per_min</td><td style="font-size:12.5px;color:var(--dim)">total / ΣV</td><td style="font-size:12.5px"><b>scale-neutral</b></td></tr>
</table>
</div>
<div class="card" style="padding:12px 16px;border-color:#4a3a12;background:#1e1809">
<h3><span style="color:var(--accent)">⑥ PROVENANCE</span> — honesty inside the row</h3>
<table>
<tr><td class="mono">measurement_coverage_pct</td><td style="font-size:12.5px;color:var(--dim)">% of expected sensors reporting</td></tr>
<tr><td class="mono">utility_scope</td><td style="font-size:12.5px;color:var(--dim)">aggregated | <b style="color:var(--accent)">shared</b></td></tr>
<tr class="hi"><td class="mono">allocation_to_equipment_available</td><td style="font-size:12.5px"><b class="no">False</b> for Topology C</td></tr>
</table>
</div>
</div>
</div>
<div class="note">All of block ⑤ is <b>intensive</b> — divided by volume — which is exactly why a 1-tank plant and a 4-tank plant fall into the same numeric range.</div>
</section>
<!-- 13 ─────────────────────────────────────────── HOW: SECOND TYPE + DISJOINT -->
<section class="slide">
<div class="kicker">4 · How — a second equipment type</div>
<h2>Same skeleton, different body — and provably disjoint schemas</h2>
<div class="body">
<div class="col" style="flex:1.15">
<svg viewBox="0 0 600 350">
<defs><style>
.sec{fill:#151d38;stroke:#26314f;rx:8}
.st{fill:#cdd7ee;font:12px "SF Mono",monospace}
.sh{fill:#fff;font:800 12px -apple-system,sans-serif}
</style></defs>
<text x="10" y="16" style="fill:#94a3c4;font:700 11px -apple-system,sans-serif;letter-spacing:.1em">SHARED — IDENTICAL FOR BOTH ROLES</text>
<rect class="sec" x="10" y="26" width="580" height="74" rx="8" fill="#10241c" stroke="#255c3b"/>
<text class="st" x="26" y="48">① run_id · timestamp · batch_id</text>
<text class="st" x="26" y="68">② site_id · product · split · anomaly_class</text>
<text class="st" x="26" y="88">③ elapsed_step · phase_progress_pct ④ active_volume_hl · active_equipment_count</text>
<text x="26" y="118" style="fill:#4ade80;font:700 12px -apple-system,sans-serif">→ the same run_id binds both stages together</text>
<text x="10" y="152" style="fill:#94a3c4;font:700 11px -apple-system,sans-serif;letter-spacing:.1em">⑤ MEASUREMENT BODY — THE ONLY PART THAT CHANGES</text>
<rect class="sec" x="10" y="164" width="285" height="150" rx="8"/>
<text class="sh" x="26" y="186">Fermenter</text>
<text class="st" x="26" y="210">temperature_value</text>
<text class="st" x="26" y="230">temperature_intra_spread</text>
<text class="st" x="26" y="250" style="fill:#f5b544">gravity_value</text>
<text class="st" x="26" y="270" style="fill:#f5b544">actuator_value</text>
<rect class="sec" x="305" y="164" width="285" height="150" rx="8" fill="#2b1f36" stroke="#5a3f70"/>
<text class="sh" x="321" y="186">ConditioningTank</text>
<text class="st" x="321" y="210">cond_temp_value</text>
<text class="st" x="321" y="230">cond_temp_intra_spread</text>
<text class="st" x="321" y="250" style="fill:#ff6b6b">pressure_value</text>
<text class="st" x="321" y="270" style="fill:#ff6b6b">vdk_value (diacetyl)</text>
<text x="10" y="340" style="fill:#ff6b6b;font:800 13px 'SF Mono',monospace">raw_schemas_disjoint = True → zero shared measurement columns</text>
</svg>
</div>
<div class="col">
<div class="card sm">
<h3>Conditioning contract fields</h3>
<table>
<tr><td class="mono">cond_temp_mean_c</td><td style="font-size:12px;color:var(--dim)">°F→°C, volume-weighted</td></tr>
<tr><td class="mono">cond_temp_spread_c</td><td style="font-size:12px;color:var(--dim)">across vessels</td></tr>
<tr><td class="mono">intra_vessel_spread_c</td><td style="font-size:12px;color:var(--dim)">within vessel</td></tr>
<tr class="hi"><td class="mono">pressure_mean_bar</td><td style="font-size:12px"><b>psi→bar</b> · PRESSURE_LOSS</td></tr>
<tr><td class="mono">pressure_spread_bar</td><td style="font-size:12px;color:var(--dim)">across vessels</td></tr>
<tr class="hi"><td class="mono">diacetyl_ppb_mean</td><td style="font-size:12px"><b>DIACETYL_STALL</b></td></tr>
<tr><td class="mono">diacetyl_spread_ppb</td><td style="font-size:12px;color:var(--dim)">anti-masking in parallel</td></tr>
</table>
</div>
<div class="card bad">
<h3>Why this is a difference of <i>kind</i></h3>
<ul>
<li>Fermenter rows and conditioning rows <b>cannot be stacked</b> into one model — no common columns.</li>
<li>The only unit spanning the run end-to-end is <b>ProcessRun</b>.</li>
<li>Here the contract is not a preference. It is the <b>only possibility</b>.</li>
</ul>
</div>
</div>
</div>
</section>
<!-- 14 ─────────────────────────────────────────── HOW: DERIVED FEATURES -->
<section class="slide">
<div class="kicker">4 · How — model features</div>
<h2>The last step: making the row product-neutral</h2>
<div class="body">
<div class="col" style="flex:1.1">
<div class="card">
<h3>Derived in <code>analyzer.py::add_topology_features</code> — not stored in the contract</h3>
<table>
<tr><th>feature</th><th>formula</th></tr>
<tr><td class="mono">temp_setpoint_error_c</td><td class="mono" style="color:#cdd7ee">process_temp_mean_c − product_setpoint</td></tr>
<tr><td class="mono">gravity_attenuation</td><td class="mono" style="color:#cdd7ee">(OG − current) / (OG − target_FG)</td></tr>
<tr><td class="mono">cond_temp_setpoint_error_c</td><td class="mono" style="color:#cdd7ee">cond_temp_mean_c − setpoint</td></tr>
<tr><td class="mono">pressure_setpoint_error_bar</td><td class="mono" style="color:#cdd7ee">pressure_mean_bar − setpoint</td></tr>
</table>
</div>
<svg viewBox="0 0 600 150">
<defs><style>.ax2{stroke:#26314f;stroke-width:1.5}.lab3{fill:#94a3c4;font:11px "SF Mono",monospace}</style></defs>
<text x="0" y="14" style="fill:#94a3c4;font:700 11px -apple-system,sans-serif;letter-spacing:.1em">ABSOLUTE — TWO SEPARATE CLUSTERS</text>
<path class="ax2" d="M0 46 H600"/>
<circle cx="150" cy="46" r="9" fill="#4fd1c5"/><text class="lab3" x="126" y="70">Pilsner 12 °C</text>
<circle cx="430" cy="46" r="9" fill="#f5b544"/><text class="lab3" x="410" y="70">Ale 20 °C</text>
<text x="0" y="106" style="fill:#4ade80;font:700 11px -apple-system,sans-serif;letter-spacing:.1em">PRODUCT-RELATIVE — ONE CLUSTER</text>
<path class="ax2" d="M0 134 H600"/>
<circle cx="300" cy="134" r="9" fill="#4fd1c5"/>
<circle cx="308" cy="134" r="9" fill="#f5b544" opacity=".85"/>
<text class="lab3" x="322" y="138" style="fill:#4ade80">error ≈ 0 for both</text>
</svg>
</div>
<div class="col">
<div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The product policy</h3>
<div style="font-size:15px;line-height:1.5;color:#cdd7ee">
<code>product</code> stays <b>in the row</b> — but only to look up the setpoint.
It is <b>never the model key</b>.
</div>
<div style="margin-top:12px;font-family:var(--mono);font-size:12px;color:var(--dim)">
feature_contract.product_policy =<br/>“curated attribute, not the model key”
</div>
</div>
<div class="card good sm">
<h3>Measured consequence</h3>
<ul>
<li>Global model on Topology B: <b>Pilsner 96.9%</b> / <b>Ale 96.2%</b>.</li>
<li>One model, two products, three layouts, two equipment types.</li>
</ul>
</div>
<div class="card sm">
<h3>Full feature vector</h3>
<div class="mono" style="font-size:11.5px;line-height:1.7;color:var(--dim)">
temp_setpoint_error_c · process_temp_spread_c ·<br/>
intra_vessel_spread_c · gravity_attenuation ·<br/>
gravity_spread · gravity_drop_2h ·<br/>
cooling_intensity_c_per_min · phase_progress_pct · temp_rate
</div>
</div>
</div>
</div>
</section>
<!-- 15 ─────────────────────────────────────────── TRANSFORMATION ≠ CONTRACT -->
<section class="slide">
<div class="kicker">5 · Transformation ≠ Contract</div>
<h2>The most common objection — and why it dodges the question</h2>
<div class="card bad" style="margin-top:14px;padding:16px 20px">
<div style="font-size:19px;line-height:1.45;font-style:italic">
“In asset-centric modelling I can <b>also</b> do unit conversion, averaging, aggregation.
I'll convert Fahrenheit to Celsius, I'll sum the four tanks. So what do I need a process model for?”
</div>
</div>
<div class="body" style="margin-top:16px">
<div class="col">
<div class="card" style="flex:1">
<h3 style="font-size:20px;color:var(--dim)">TRANSFORMATION is a <span style="color:#fff">mechanism</span></h3>
<div style="font-size:16px;color:#cdd7ee;margin:8px 0 12px">“How do I compute this number?”</div>
<ul class="sm">
<li>°F → °C</li>
<li>volume-weighted average</li>
<li>solenoid 0/1 → cooling demand</li>
<li>multi-vessel roll-up</li>
</ul>
<div style="margin-top:14px;padding:12px;background:#10241c;border:1px solid #255c3b;border-radius:8px;font-size:14.5px;line-height:1.45">
<b class="ok">The objection is correct.</b> My PoC <b>proves this</b>, not the opposite —
and I concede it as a fair control on the next slide.
</div>
</div>
</div>
<div class="col">
<div class="card" style="flex:1;border-color:#4a3a12;background:#1e1809">
<h3 style="font-size:20px;color:var(--dim)">CONTRACT is a <span style="color:var(--accent)">decision</span></h3>
<div style="font-size:16px;color:#cdd7ee;margin:8px 0 12px">“What is my row? Who owns it? How trustworthy is it?”</div>
<ul class="sm">
<li><b>Grain</b> — what the analytical unit is</li>
<li><b>Single point of ownership</b> — where the definition lives</li>
<li><b>Provenance</b> — what the value is allowed to claim</li>
</ul>
<div style="margin-top:14px;padding:12px;background:#151d38;border:1px solid #26314f;border-radius:8px;font-size:14.5px;line-height:1.45">
A contract is not the <b>result</b> of a transformation —
it is the <b style="color:var(--accent)">decision that governs it</b>.
</div>
</div>
</div>
</div>
<div class="note">The objection reduces the whole debate to one sentence — <b>“I can compute this number”</b> — and never asks the three questions on the right.</div>
</section>
<!-- 16 ─────────────────────────────────────────── FAIR CONTROL + N×M -->
<section class="slide">
<div class="kicker">5 · Transformation ≠ Contract</div>
<h2>I conceded the transformation, then measured what's left</h2>
<div class="body">
<div class="col" style="flex:1.2">
<div class="card good" style="padding:14px 18px">
<h3>The fair control: I wrote the asset-side adapters myself</h3>
<table>
<tr><th>topology</th><th>custom adapter output vs. generic aggregator</th></tr>
<tr><td class="mono">Topology_A</td><td><b class="ok">numerically identical</b> <span class="mono" style="color:var(--dim)">np.allclose 1e-9</span></td></tr>
<tr><td class="mono">Topology_B</td><td><b class="ok">numerically identical</b> <span class="mono" style="color:var(--dim)">np.allclose 1e-9</span></td></tr>
<tr><td class="mono">Topology_C</td><td><b class="ok">numerically identical</b> <span class="mono" style="color:var(--dim)">np.allclose 1e-9</span></td></tr>
</table>
<div style="margin-top:10px;font-size:14.5px;color:#cdd7ee">
<b>3 / 3.</b> Transformation capability is <b>not</b> the point of distinction. Conceded, in code.
</div>
</div>
<svg viewBox="0 0 600 252">
<defs><style>
.ad{fill:#241214;stroke:#5c2828;rx:6}
.adt{fill:#ffb4b4;font:10.5px "SF Mono",monospace;text-anchor:middle}
.hh{fill:#94a3c4;font:700 11px -apple-system,sans-serif}
</style></defs>
<text class="hh" x="0" y="14" style="letter-spacing:.1em">ASSET SIDE — N × M EXPLOSION</text>
<text class="hh" x="0" y="42">Fermenter</text>
<rect class="ad" x="86" y="26" width="112" height="34"/><text class="adt" x="142" y="47">adapter A</text>
<rect class="ad" x="206" y="26" width="112" height="34"/><text class="adt" x="262" y="47">adapter B</text>
<rect class="ad" x="326" y="26" width="112" height="34"/><text class="adt" x="382" y="47">adapter C</text>
<text class="hh" x="0" y="86">Conditioning</text>
<rect class="ad" x="86" y="70" width="112" height="34"/><text class="adt" x="142" y="91">adapter A</text>
<rect class="ad" x="206" y="70" width="112" height="34"/><text class="adt" x="262" y="91">adapter B</text>
<rect class="ad" x="326" y="70" width="112" height="34"/><text class="adt" x="382" y="91">adapter C</text>
<text x="456" y="66" style="fill:#ff6b6b;font:800 26px 'SF Mono',monospace">6</text>
<text x="456" y="88" style="fill:#ff6b6b;font:11px -apple-system,sans-serif">definitions</text>
<text x="0" y="126" style="fill:#94a3c4;font:12px -apple-system,sans-serif">“process temperature” copied into every box. Add a plant → +2. Add a stage → +3.</text>
<text class="hh" x="0" y="158" style="letter-spacing:.1em">PROCESS SIDE</text>
<rect x="86" y="168" width="352" height="46" rx="6" fill="#10241c" stroke="#255c3b"/>
<text x="262" y="196" style="fill:#4ade80;font:700 12.5px 'SF Mono',monospace;text-anchor:middle">1 generic role-based aggregator</text>
<text x="456" y="200" style="fill:#4ade80;font:800 26px 'SF Mono',monospace">1</text>
<text x="0" y="242" style="fill:#f5b544;font:700 13px 'SF Mono',monospace">multi_type_reuse_ratio = 6.0 — measured, not claimed</text>
</svg>
</div>
<div class="col">
<div class="card">
<h3>Provenance: the part transformation never covers</h3>
<ul class="sm">
<li>Topology C's glycol loop is measured <b>in total only</b>.</li>
<li>The contract sets <code>allocation_to_equipment_available = False</code> and <b>refuses to invent</b> per-vessel granularity.</li>
<li>A spreadsheet transformation would just divide by two — and nobody would notice.</li>
</ul>
</div>
<div class="card" style="border-color:#4a3a12;background:#1e1809">
<h3 style="color:var(--accent)">The steel-man conclusion</h3>
<div style="font-size:15px;line-height:1.5;color:#cdd7ee">
Anyone doing <b>disciplined</b> asset-side harmonization — shared unit, invariant
semantics, role-based schema, single definition — has
<b style="color:var(--accent)">already built the process model under a different name.</b>
</div>
<div style="margin-top:10px;font-size:15px;color:#cdd7ee">
The real distinction isn't the tool. It's
<b>local/asset-specific meaning</b> vs. <b>shared/process-owned meaning</b>.
</div>
</div>