What AI Is Made Of

Run03 electrical material forensics

The frozen selected equipment counts are read from calculate_balance(...,snapshot='SNAP-RUN02') by build_handoff.py (repository-only reference). No count, topology, equipment selection, historical source or historical coefficient is changed. This workstream covers UPS, generators, busway, shelf whips and major UPS feeder conductors. Transformer/PDU work belongs to the separate transformer handoff.

Two validated ENGINEERED construction scenarios quantify the effect of unresolved busway conductor choice. Neither is an admitted plant coefficient. The source-specific point scenarios preserve the selected800A415V four-wire topology and3424×12m allowance. They do not select a vendor or manufacture a confidence interval for unknown cross-section.

Frozen busway construction sensitivityConductor geometry per poleConductor material massCatalog gross proxyUnresolved residual
Copper800A catalog constructionOne0.25×2.5inch bar;4 poles589.148t copper conductor material1027.200t438.052t
Aluminum800A catalog constructionOne0.25×3inch bar;4 poles214.718t aluminum conductor material739.584t524.866t

The same Schneider catalog supplies both actual800A sections, rather than scaling an unrelated ampere rating. Four poles means3 phases plus neutral; ground paths remain separate. The difference demonstrates potential hundreds-of-tonnes leverage from the genuinely unselected conductor specification. It does not demonstrate that these are universal minimum/maximum constructions or that all plant busway copper lies between0 and589t. Rail vendor, conductor section/grade, neutral sizing, joints, feeds, tap boxes and ground/enclosure details remain unresolved. The source gives nominal inches and rounded millimetres: calculations use exact inch conversion, never treat the rounded metric display as a measurement uncertainty endpoint.

The conductor equation is m_kg=rho_kg_m3*A_mm2*1e-6*L_m*n. Copper density8890kg/m³ is the CDA100%IACS reference; aluminum2700kg/m³ is supported independently by CDA and Hydro wrought bus-conductor data. Three copper bar cases from the CDA physical-property table and three6101 aluminum bar cases from Thyssenkrupp stock catalog validate the equation within0.257% and0.561%, respectively. The cause of the small aluminum overprediction is unresolved; density, catalog rounding and edge convention are possible contributors. This validates mass reconstruction when geometry is known, not target product identity or ampacity. Case-level observations, predictions, errors and1% acceptance criteria appear in proposed_forensics.json; transfer_admitted=false prevents promotion into the forensic base. Neither material's elemental purity is established, so the components retain copper_conductor_material and aluminum_conductor_material identities.

The approximate catalog assembly weights are proxy source-construction gross masses for arithmetic conservation checks only; they are not measured maxima or selected-plant ceilings. The residual includes enclosure, insulation, joints, supports and other unquantified mass. It is never assigned to steel or aluminum. Parent rail assembly and conductor composition are alternative views of one mass, not additive stocks. Root integration retains these scenarios separately from the primary and expanded forensic balances.

Exact selected UPS: identity improves; composition remains open

The current OEM Trinergy user/install manual, SL-80218/10068023P60CE-UL REVG06/26, is now captured and hashed. Its installation drawings explicitly identify copper core-junction, rectifier-input, bypass and output bars. It also establishes the transformer-free three-level IGBT topology. The diagrams and listed VTT terminal/internal-view drawing numbers provide precise future procurement targets, but public pages do not establish full bar thickness/length, sheet thickness or magnetics/component masses. Copper identity is stronger; copper kilograms remain unknown.

The selected UPS count856 and net4710kg source remain unchanged: 4031.760t exact selected non-battery assembly mass, with an unresolved residual of4710kg per UPS. Batteries are external separate cabinet stocks. Magnetics/ferrite, bars/cabling, heat sinks, enclosure/frame, PCBs, power electronics, capacitors, fans and polymers remain individually unknown. A gross-mass bound permits each material separately to lie between0 and4031.760t, without establishing a base or implying that those independent maxima can occur simultaneously.

The current400VCE PEP's denominator/packaging contradiction remains exactly the Run02 conflict. Neither its gross/net denominator alternatives nor percentages are used as UL480V low/high endpoints. Current manual Table8.2 lists an option envelope of up to14500lb (6577.089kg) for the2MWUL class and alternate widths. An option envelope can coexist with the4710kg selected brochure configuration; it cannot replace the selected mass or become an uncertainty interval without a component/configuration crosswalk.

The exact480V3phase2MW manual now supports a conditional physical feeder calculation. Table8.8 lists8×600kcmil on rectifier input and7×600kcmil on output and bypass. The calculation interprets those parallel quantities per phase in the three-phase schedule; this is a conventional engineering inference, not a verbatim manufacturer statement. A kcmil is1000 circular mils, giving600kcmil=304.024487mm². At8890kg/m³:

External copper feeder path per selected UPSParallel conductors across all3 phasesCopper kg per route metre
Rectifier input2464.866665
Output2156.758332
Bypass2156.758332

These are per-route-metre sensitivities, excluding grounds, insulation, terminations and supports. Actual routing is unknown; no common arbitrary10/20/50m length is selected. Shared/common input-bypass arrangements require an explicit circuit schedule before paths are added. The feeder mass is external to the UPS net assembly and must never be absorbed into, or added as internal decomposition of, the4710kg UPS. Table8.10 labeled aluminum repeats the copper gauges/counts while footnotes require recalculation for aluminum and temperatures. That table is retained as a source qualification concern, not accepted as an equal-ampacity Al transfer. Site temperature, breaker trip/continuous ratings and final cable schedules remain necessary for either conductor option.

Generators: matched parts and service mass; no false steel subtotal

The642 selected4MW480V60Hz diesel classes still have a 25000kg/unit power-family dry gross proxy, totaling16050t. The exact480V alternator and engine arrangement are unqualified. This proxy is not an unconditional selected gross upper bound and cannot turn into a steel or ferrous subtotal.

New OEM service-part captures confirm that crankshaft359-0941 fits C175-20 gensets/SR5 and that shortblock20R-6237 nests block, crankshaft, pistons and connecting rods. The OEM crankshaft capture supplies no mass/material attribute. A reseller1441.5kg weight lacks measurement provenance and net/shipping distinction; it remains a lead. The shortblock's “corrosion preventative fluid” metadata cannot identify the whole shortblock as fluid.

An accessible manufacturer-authored RENR7466 May2007 C175-16 WYB1-Up service manual, mirrored by PDFCoffee, supplies approximate68kg assembled-cylinder-head handling mass. Caterpillar's primary seismic engineering paper explicitly confirms shared C175-16/C175-20 head components. Twenty such heads therefore support a 1360kg/unit matched-family nested-head gross proxy (873.120t across642 units), kept source-only. The head assembly includes valves/injector and other parts; its entire mass is not cast iron. The service manual also lists748kg bare/1043kg counterweighted16-cylinder crank masses. Those do not transfer to a20-cylinder shaft; the longer crank geometry is unresolved.

The seismic paper's37t wet50Hz nuclear genset is a valuable engineering qualification case but cannot bound the selected60Hz480V dry system. Engine block/cast iron, forged crank steel, alternator conductor/electrical steel, radiator technology, skid/enclosure, operating fluids and fuel remain separate unknown subassemblies. Component gross mass without material identity and complete lineage does not produce elemental closure. The separately counted generator fuel-system stock and unknown fuel inventory remain outside the dry genset decomposition.

Whips and unbounded remaining exposure

Canonical whips are54784 assemblies with164352m aggregate3m allowance. Assembly-route metres are not summed conductor metres. Conductor count, grounding/neutral arrangement, metal, section, insulation and termination masses remain unknown. No section is inferred solely from60A. Additional major facility/IT feeders have no complete routed cable schedule. Therefore a finite whole-plant outer copper bound is still unsupported, even after bounding exact UPS assemblies. Unknown cable routes cannot be silently assigned zero or bounded by known rail/UPS gross mass.

UPS, generators and busway are high-leverage priorities for copper and ferrous material, with busway/feeder Cu-versus-Al sensitivity material enough to merit explicit procurement specification. Their remaining uncertainty can still dominate the known strong-evidence subtotal; none is declared immaterial or provisionally closed.

Run python research/run03/electrical/build_handoff.py to reproduce counts, source hash checks, dimensional conversions, all6 validation cases, scenario conservation/residuals and feeder sensitivities. sources.json/claims.json use additive registry proposals; proposed_forensics.json contains the2 engine-ready scenario rows. evidence.json also retains investigated/rejected cases and all unresolved subassemblies; search_log.json indexes13 targeted search captures and retrieval failures. Root owns canonical integration and the exact UPS gross-bound row to avoid duplicate overlap.

Source details

Archived Run 01-03 research. These statements retain their time-of-writing context; current material estimates use Run 04. Web prose uses current site terminology; downloadable records preserve original wording. Hashes identify the downloadable records.

Source: research/run03/electrical/HANDOFF.md
SHA-256: 275ca9c202f425ff93616e571b31221cc7a84c55c66ad65b4815530c68b312d4
Download research file

Optional: tell Sid you’re reading
The document is open; sharing your name/email is optional.