Operations 13
Assemblies
Layered build-ups, as distinct from single materials. The material ladder answers which board goes behind a door. An assembly answers what the whole stack is, in order, and which layer governs each property — and those two questions fail differently.
PROPOSED2 candidates5 blocking questions1 claim contradicted
WEAKEST-LAYER
the principle these two forcedA performance claim made for an assembly is only as good as the layer that governs that property — and the governing layer is not always the thickest, the most expensive, or the one the supplier is selling.
Worked example
A steel sandwich panel with a rock-wool core is non-combustible at the core. Cover it with bitumen felt shingle and the surface exposed to an external flame is bitumen. The core certificate is real and irrelevant to that exposure. The assembly needs its own answer, not the sum of its brochures.
Every performance claim on an assembly must name the layer it is claimed FOR, and be evidenced at that layer or for the assembly as tested. See RULE-MAT-CLAIM-001.
Reconciled against the canon
1 correctionBoth assemblies were drafted before six of the Books and the No-Hack Playbook were readable. Reading them changed the entries in three ways: two blocking questions now have an estate answer to cite, one recorded 'real advantage' turned out to be an existing ALUMAX product line rather than a new idea, and one figure written here was an invention that forked a rule the estate already owns.
The correction — a value this repo forked itself
The ultra-thin panel entry carried a substrate flatness tolerance expressed as a plus-or-minus figure over a two-metre straightedge.
Why wrong — The estate already owns wall flatness at Book 10 CARP-02, and the No-Hack Playbook applies it to exactly this overlay case. Writing a second number here is the precise defect this repo exists to prevent — and it was written by the repo itself, four hours after the register that forbids it.
Fixed — The figure is removed. The constraint now cites CARP-02 as the owner and records the product-tolerance relationship separately.
Precedent found — Book 14 SPC-SF-01 sets a tighter flatness tolerance than the general tile substrate rule and says so in the open, in the product book, with a note back to the owner. That is the correct shape for a thin-panel maker's tolerance: the general rule is cited, the tighter product figure lives in the product record, and the two cannot drift because only one is the owner.
Why this file exists
Two materials arrived on 2026-08-26 that the taxonomy could not hold: a roof build-up and an ultra-thin wall panel. Neither is a material in the sense materials.json means. Forcing them into the carcass ladder would have produced a wrong answer quietly.
The ladder answers which board goes behind this door. An assembly answers what the whole build-up is, in order, and which layer governs each property. Those are different questions and they fail differently.
The commercially dangerous difference: a single material carries its own properties, but an ASSEMBLY property is governed by its worst layer for that property. A non-combustible core under a bitumen shingle is not a non-combustible roof. Without a place to record layer order and the governing layer, that error is invisible on a spec sheet.
Not a catalogue
An entry here is a CANDIDATE under assessment, not an offer. Nothing in this file may be quoted, priced or shown to a customer until it has a rule at CANDIDATE or better and a supplier document behind every performance claim.
夹心彩钢板 + 保温棉 + 防水处理 + 油毛毡瓦
CANDIDATEInsulated sandwich steel panel + insulation wool + waterproofing + bitumen felt shingle
ASSY-ROOF-IMP-SHINGLE-001 · roof build-up
Displaces — Bare metal deck / zinc roofing — which in Sarawak is the incumbent, and whose two known complaints are rain noise and heat gain.
Layers, outer to inner
| # | Layer | Role | Governs |
|---|---|---|---|
| 1 | 油毛毡瓦 — bitumen felt shingle | weather surface, appearance | external fire exposure · appearance · rain impact noise (first stage) |
| 2 | 防水处理 — waterproofing treatment / underlay | secondary water barrier | water ingress |
| 3 | 保温棉 — insulation wool | thermal and acoustic absorption | thermal resistance · airborne and rain noise |
| 4 | 夹心彩钢板 — sandwich colour-coated steel panel | structure, span, inner face | structural span · reaction to fire OF THE CORE · inner appearance |
Claimed performance
隔音 — acoustic / rain noise reductionUNVERIFIED
The most likely of the three to be genuinely delivered. Mass, a fibrous absorbent layer and a discontinuous outer covering are the three things that actually reduce rain noise, and this stack has all three. It is also the claim that matches the real Sarawak complaint.
Needs — Rain-noise measurement to ISO 10140-5 (artificial rainfall), or airborne Rw to ISO 717-1. Failing a test, a Level A route exists: measure a completed roof against a bare-deck roof on the same street. That is cheap and it is ours to do.
Commercially — This is the claim to lead on once evidenced — it is the one the customer already feels.
隔热 — thermal insulationUNVERIFIED
Genuinely delivered in principle, but the AMOUNT is entirely a function of core type and thickness, neither of which is stated in the material as given. Insulated without a thickness and a lambda is not a specification.
Needs — Core lambda (W/mK), insulation thickness (mm), and an assembly U-value (W/m2K) to ISO 6946.
Commercially — Never quote a temperature drop. Quote the U-value or say nothing.
防燃 — fire resistant / flame retardantCONTRADICTED
This is the claim that must not be repeated as given, for two independent reasons. FIRST: 夹心彩钢板 does not name its core, and the core is the whole fire story. An EPS core is combustible, melts and drips; a PU/PIR core is better but organic; only a mineral or rock-wool core is genuinely non-combustible. The same words describe all three, and the cheapest is the most common. SECOND: even with a non-combustible core, the outermost layer here is bitumen felt — an organic, combustible surface. For external flame exposure the shingle governs, not the core. So the assembly as described cannot carry the claim even if the panel can.
Needs — Reaction-to-fire classification for the CORE (EN 13501-1 Euroclass, e.g. A1 / A2-s1,d0 / B / E), plus a surface-spread-of-flame or external-fire-exposure result for the FINISHED assembly. In Malaysia, BS 476 results are what local authorities and insurers commonly recognise.
Regulatory — Fire performance of a roof is a regulated matter under UBBL 1984, and the applicable requirement depends on building use, height and boundary distance. Confirm with the submitting architect or engineer — do not infer it from the product literature and do not let a salesperson infer it either.
Hard line
Until a core is named and a classification is on file, this claim is not to appear in any quotation, drawing note, brochure, WhatsApp message or verbal pitch. An unsubstantiated fire claim is the one commercial exposure in this file that does not stay commercial.
Open questions — 5 of 7 blocking
BLOCKING · What is the sandwich panel CORE — EPS, PU/PIR, glass wool or rock wool?
Governs the fire answer, the thermal answer and roughly half the cost. Nothing about this assembly can be settled without it.
BLOCKING · What are the bitumen shingles fixed INTO?
Asphalt shingle is a nailed product and normally needs a nailable deck such as plywood or OSB. Steel sandwich panel is not nailable. Either a deck layer is missing from the stack as described, or the shingles are bonded — and a bonded bitumen shingle in full Sarawak sun is a different, and worse, proposition.
BLOCKING · Is there a vapour control layer, and on which side?
Air-conditioned space below, hot humid air above, a steel skin between. 防水处理 stops rain from above and does nothing about vapour from below. The estate owns the principle — Book 11 WPC-VT-01 puts the vapour barrier on the WARM side behind a ventilated cavity, and Book 15 HV-RH-01 explains why cold surfaces are where it condenses. In an air-conditioned Sarawak building the warm side is OUTSIDE, which inverts the temperate-climate habit. Wet insulation loses its thermal value and corrodes the steel from inside, invisibly, for two to three years.
BLOCKING · What is the panel span and the supporting structure?
A re-roof usually inherits the existing purlin spacing. The allowable span of the panel must be checked against what is already there, not against the ideal in the datasheet.
How is the roof plane earthed and bonded?
A large continuous steel plane in a lightning-prone state. The estate already carries lightning surge as a Sarawak factor (Book 9 CB-01, Matrix #35 and #56) for electrical work; a steel roof raises the same question structurally.
What is the warranty, and does it survive a shingle repair?
A recovery question, and the answer decides what this roof costs over ten years rather than on the day.
BLOCKING · What is the steel skin's coating specification, and does it clear REF-M and REF-C?
REF-M's only steel row is permitted in the driest material zone and prohibited in the wetter two; REF-Z puts outdoor covered areas in the wettest. REF-C prohibits standard steel within the coastal band outright. The colour coating is what would earn the permission, and it is undocumented — which makes this a second unnamed specification sitting beside the unnamed core.
Failure modes
interstitial-condensationcorrosionthermal-movementflame-spreaddelaminationlightning-surge
Recovery — Layered assemblies recover badly. If a shingle fails, the repair reaches the waterproofing; if the waterproofing fails, the repair reaches the insulation. Establish before quoting whether any layer can be repaired without opening the ones below it, because that answer — not the install price — is what the assembly costs over ten years.
Proposed facet reading
- EXPOSURE_CONTEXT
- exterior
- VISUAL_IMPACT
- primary
- APPLICATION_EFFORT
- sequence-critical
- INSTALLER_REQUIREMENT
- specialist-required
- SOURCING
- unknown — establish before quoting
A facet reading here is the PROPOSED reading by this repo for an assembly it raised. Material facet assignments remain for ELS to make.
陶板 0.4 cm + 0.6 cm 超薄瓷砖墙
CANDIDATE4 mm terracotta panel and 6 mm ultra-thin porcelain tile wall
ASSY-WALL-THINPANEL-001 · wall cladding
As written this is TWO different systems, and treating them as one product is the first mistake available.
陶板 4 mm — terracotta panel — a ventilated rainscreen: dry-fixed to rails with a drained air cavity behind, mechanically held, exterior facade. Governed by the fixing system and the cavity, not the adhesive.
超薄瓷砖 6 mm — ultra-thin porcelain — adhesive-bonded to a prepared substrate, interior or exterior. Governed by the substrate, the adhesive and the coverage.
They share a thickness range and nothing else. Different substrate prep, different trade, different failure mode, different recovery, different price. Quote them as one line and the wrong one gets installed the cheap way.
The real advantage — It can go OVER existing tile.
This is the genuine commercial case and it is much stronger than the aesthetic one. No hacking, no debris, no screed rebuild, no skip, far less noise and dust, and a fraction of the programme. On an occupied home or a trading shoplot that is worth more than the finish is.
Caveat — Only where the existing tile is sound, well bonded and flat enough. A drummy tile under a bonded panel takes the new panel with it when it goes.
Evidence route — Percussion-test the existing surface and record it. This is a Level A route available to us on the very first job — programme days saved is a number we can measure and own.
Governing constraints
Substrate flatness
Large thin panels telegraph what is behind them and crack over a high spot. Wall flatness is owned by Book 10 CARP-02 and is NOT restated here. Where the panel maker states a tighter tolerance, that tighter figure is a product requirement recorded against the product — the same relationship Book 14 SPC-SF-01 has with the general tile substrate rule, declared openly rather than forked.
Sarawak — Local plaster is routinely outside the owned tolerance. The No-Hack Playbook is explicit that a wall beyond the limit is skim-coat corrected and never absorbed by oversized fillers. Price the correction as a visible line — it is frequently the larger half of the job, and a thin-panel quote with no substrate line has not been surveyed.
Adhesive and coverage
A flexible cementitious adhesive of the right class (EN 12004 / ISO 13007, typically C2TE with S1 or S2 deformability) and a solid bed. Dot-and-dab leaves voids, and a void under a 4 to 6 mm panel is where it cracks the first time something leans on it.
Confirm the required class against the instruction from the panel maker. Full coverage is the norm for large format and effectively mandatory externally.
Handling and cutting
Large thin panels are handled with a suction-cup frame and multiple people, and cut with the method specified by the panel maker. They break from bending during handling far more often than from anything that happens after installation.
Consequence — Breakage before it reaches the wall is a real cost line. Establish who carries it — supplier, installer or ALUMAX — before the first order, not after the first break.
Exterior fixing
Bonded-only fixing on an exterior wall must be treated with caution — sun, thermal cycling and driving rain work the bond, and a large panel letting go at height is a safety matter rather than a snag. The estate already rules on this for its own panel line: Book 11 WPC-FX-01 makes combined adhesive PLUS mechanical fixing mandatory above a stated panel height and weight, and Book 11 WPC-VT-01 requires a rear ventilated cavity behind panels.
Required — Establish the panel against WPC-FX-01's thresholds before choosing a fixing method. Where the height or the building warrants it, confirm with the engineer. Do not restate the thresholds — cite them.
Failure modes
delaminationimpact-chipthermal-movementsubstrate-telegraphingstaining
Recovery — Poor, and worse than it looks. A cracked large-format panel cannot be swapped the way a 300x600 tile can — reaching it usually means disturbing its neighbours. Add shade and batch variation: a panel replaced two years later may not match, and on a large-format surface a shade step is obvious in a way it never is on small tile. If the material is imported-far, treat an exact future match as unavailable rather than merely slow, and hold attic stock at the time of the original order.
Proposed facet reading
- EXPOSURE_CONTEXT
- interior or exterior — must be stated per job, and the answer changes the fixing method
- VISUAL_IMPACT
- primary
- APPLICATION_EFFORT
- sequence-critical
- INSTALLER_REQUIREMENT
- specialist-required
- SOURCING
- likely imported-regional or imported-far — confirm, because it sets the recovery answer
The PROPOSED reading by this repo only. ELS assigns.
Claim status vocabulary
what may be published, and what may not| Status | Means | Publishable |
|---|---|---|
| CERTIFIED | A test report or certificate exists, names a standard and a class, and is on file against this assembly or the governing layer. | yes |
| SUPPLIER-ASSERTED | The supplier says so in writing, but no test document has been produced. A brochure is marketing, not evidence. Level C at best. | no |
| UNVERIFIED | Claimed verbally or in passing, nothing on file. | no |
| CONTRADICTED | Something in the assembly defeats the claim as made. Record WHY. This is the most valuable state in the file — it is the error that would otherwise have been repeated. | no |
What these two exposed
7 gaps the taxonomy could not holdNo FIRE_PERFORMANCE facet anywhere in the estate.
material-facets.json has seven proposed facets and none of them ask how a material behaves in a fire. That was tolerable while the canon was kitchen carcasses. It is not tolerable now that envelope build-ups are arriving, and it is the facet with a regulator behind it.
Action — PROPOSED to ELS below.
No ASSEMBLY_ROLE facet — nothing records whether a material is a finish, a substrate, a core, a barrier or a fixing.
Without it there is no way to express the weakest-layer principle in data, and it stays prose that a spec sheet can ignore.
Action — PROPOSED to ELS below.
MATERIAL_RISK has no value for the two failure modes these introduce.
interstitial-condensation and substrate-telegraphing are both slow, invisible and expensive, which is exactly the profile of a failure worth naming. Neither is expressible today.
Action — PROPOSED as additions to the existing facet.
The FABRICATION_ROUTE gap already recorded on the MATERIALS axis now has a third value.
Neither of these is built-in, custom-made or modular. Both are SITE-ASSEMBLED — built up in layers in position, where the sequence is part of the specification and cannot be recovered later.
Action — Feeds the existing open gap rather than opening a new one.
The DELIVERY axis is still a whole-axis gap, and these two make it bite.
Large thin panels and roof panels are both damage-in-transit materials with awkward handling requirements. The axis that would carry that has nothing in it.
Action — Already recorded as the DELIVERY gap. These are the first two materials that make it concrete.
Roof rain noise has no acoustic owner
The estate's acoustic domain is split four ways and each part is owned — airborne (Book 11), impact through the floor (Book 14), partition build-up (Book 2 SP-PT-02), door performance (Book 6 DT-02). Rain on a roof plane is a fifth case with no book behind it, and it is the claim the roof assembly is most likely to deliver.
Action — Raise as a sub-domain to whichever book owns the roof plane.
No material in either assembly has a REF-M permission
The material permissions table is the estate's gate on what may be used in which zone, and it has been amended seven times as each book found a gap. Neither of these assemblies has a single row. Envelope materials as a class are absent from it.
Action — REF-M Amendment 8, proposed to ELS through the established pattern.
Proposed to ELS
proposed here · assigned there · never restatedFIRE_PERFORMANCE — How does it behave in a fire, and to which standard is that established?
The only facet in the estate with a regulator and an insurer behind it. It is also the one where a wrong answer is not recoverable by a credit note. untested is deliberately a value rather than a blank, so that an unknown reads as an unknown instead of as an omission.
non-combustible-certifiedlimited-combustibility-certifiedcombustibleuntested
Rule — A value other than untested requires a named standard and class on file. The facet records the CLASSIFICATION; the certificate itself stays with the supplier record.
ASSEMBLY_ROLE — What job does this layer do in a build-up?
Makes the weakest-layer principle expressible in data instead of prose. Once a layer carries a role, a surface can ask which layer governs a claimed property and refuse a claim whose governing layer is unevidenced.
finishsubstratecorebarrierfixingstandalone
standalone keeps every existing single material valid without a migration — the current ladder is all standalone.
Additions to MATERIAL_RISK
interstitial-condensationsubstrate-telegraphing
Both are slow, invisible and expensive, and neither is expressible with the current ten values. Condensation in particular fails late enough to land outside the warranty conversation, which is precisely why it needs to be visible at specification time.
Handoff
- Confirm or reject FIRE_PERFORMANCE and ASSEMBLY_ROLE as facets on /api/material-zones.
- Confirm the two MATERIAL_RISK additions.
- Decide whether assemblies belong in the material canon at all, or stay an ALUMAX-side construct that cites the canon for each layer. This file assumes the latter and defines nothing ELS owns.
- Neither assembly may be offered until its blocking open questions are answered. Four are blocking on the roof build-up, and the core question blocks the other three.