The claim under test
Can foaming void risk be predicted per cabinet, before the energy test, accurately enough to change the shot?
- Metric agreed in writing first
- Baseline window of 14 days minimum
- Partner owns the definition
Programme in progress · results publish as they are verified
The proof,publishedhonestly.
Duromex is pre-launch. Rather than invent customer logos and impressive percentages, this page publishes the design-partner programme in progress — the metric each partner chose, the method, and what has been measured so far.
How a Duromex case study is built
Every study follows the same method, because a result you cannot reproduce is marketing, not evidence.
1 · Agree the metric in writing
First-pass yield, functional-test failure rate, leak or warranty rate, scrap or downgrade rate. One metric, defined before anything is instrumented.
2 · Measure the baseline in shadow mode
The agent predicts while humans run the line. Nothing is written. The baseline and the accuracy gap are both published to the partner.
3 · Prove in advisory mode
The agent recommends; the engineer approves. The delta against the baseline is attributable and auditable.
4 · Publish with the partner
Numbers, method and caveats reviewed and approved by the partner before anything appears here.
Four studies under way
Segment, metric and current stage. Partner identities are withheld under NDA until each study publishes.
Refrigeration
Cabinet foam voids and energy-rating conformance
A refrigeration OEM instrumenting the foaming station to predict density distribution and void risk per cabinet, and to forecast thermal performance before the energy test.
LaundryWater leak and harness verification
A laundry OEM combining water-leak detection at fill, drain and seal stations with connector seating and polarity verification on the harness line.
CookingEnamel first-quality rate
A cooking OEM controlling enamel coverage, film build and cure while classifying chip, pinhole and orange-peel defects at takt.
Air conditioningBrazed-joint leak and charge accuracy
An AC manufacturer detecting brazed-joint and coil leaks before charge, and tracking refrigerant loss for F-gas evidence.
The metric lives somewhere in these seven stages
A case study is only credible if the metric is attributable. That is why instrumentation covers the whole sequence rather than the station where the failure was discovered.
Form & stampWeld & seamEnamel & paintFoam & insulateWire & assembleLeak & function testRate & pack
The metrics design partners choose
Each is something the plant already reports, so the result cannot be quietly redefined afterwards.
| Metric | Why it is chosen | Where the improvement comes from |
|---|---|---|
| First-pass yield | The number the plant director is judged on | Defect prevention upstream plus failure attribution |
| Functional-test failure rate | Directly observable at end of line | Prediction before test, plus automated triage |
| Leak / warranty rate | Highest cost per event | Micro-leak signatures below tester thresholds |
| Scrap and rework hours | Easy to cost accurately | Adaptive control instead of reactive rework |
| First-quality rate | Downgrades are pure margin loss | Enamel and surface control at takt |
| Energy-rating conformance | Regulatory and commercial exposure | Foam density and thermal-performance forecasting |
Inside one study, step by step
The instrumentation
Foaming machine telemetry, jig conditions, thermal imaging and downstream test results, all serialised to the cabinet.
- Shot weight, mix ratio, cure time
- Thermal and pressure signals fused
- Test outcome joined back to the unit
The shadow result
The agent predicts; humans run the line. We publish the accuracy gap against the tester, including where the model is wrong.
- Accuracy against the tester baseline
- False-positive and false-negative rates
- Failure modes documented, not hidden
The advisory result
Recommendations applied under human approval. The delta against baseline is the number that appears in the study.
- Attributable, auditable delta
- Engineer corrections logged as training signal
- Caveats published alongside the result
The caveats
Sample size, product mix, seasonality and anything else that would make a reader over-generalise the result.
- Single line, single product family
- Time-bounded observation window
- Partner-reviewed before publication
What partners say during the programme
“Our end-of-line functional testers told us what failed. They never told us which press stroke or foam shot caused it. That link is the whole product.”
“Foam voids are invisible until the energy rating comes back wrong. Predicting density distribution before the cabinet cures is the part I could not buy anywhere else.”
“If it writes to the press, it needs an audit trail a quality auditor accepts. Duromex started there instead of bolting it on.”
Rules we hold ourselves to
Written down so we can be held to them later.
0
results published without partner approval
100%
studies that publish their method
100%
studies that publish their caveats
14 d
minimum baseline window
Because the industry has been burned
Appliance plants have bought vision systems that detected everything and changed nothing, and analytics platforms that produced beautiful reports about losses that continued unabated.
The fastest way for Duromex to become another one of those is to publish a percentage we cannot defend. So we publish the method, the baseline and the caveats — and the number only when a partner signs off on it.
What you will never see here
- Customer logos we do not have permission to use
- Percentage uplifts with no baseline or window
- Results from a demo dataset presented as production
- A case study the partner has not read and approved
- Aggregate claims built from a single line
Segments represented in cohort one
Refrigeration OEM · EU
Laundry OEM · North America
Cooking OEM · Turkey
Dishwashing OEM · Italy
Air-conditioning OEM · Asia
Tier-1 cabinet supplier
Design-partner programme, 2026. Duromex is pre-launch — partner identities are withheld under NDA and shown by segment, not by logo.
How to read these
When will the first full case study publish?
When a design partner has completed advisory mode with a verified delta against a published baseline, and has approved the write-up. We will not publish earlier to hit a marketing date.
Can we see the raw data?
Not the partner's raw data — it is theirs and it is under NDA. The method, the metric definition, the baseline window and the caveats are always published.
Will you publish failures?
Yes. A pilot that misses its metric is a result, and hiding it would make every other number here worthless.
Be the study we publish first.
Design-partner terms are open for cohort one, with commercial terms that reflect the risk of working with a pre-launch company.
Duromex is pre-launch. Figures shown are design-partner targets and modelled economics, not audited results. Ask us for the methodology.