EV & battery cell manufacturing

Cell quality is discovered in final test,not predicted upstream.

Profile: Composite — 'Apex Cell'
Capacity: 12 GWh — ~480M cells / yr
Yield: 84% today — 92%+ target
Status: Target scenario — modelled

01Context

Apex Cell is a composite profile built from public yield benchmarks and battery-industry analyst data — representative of the mid-tier lithium-ion manufacturers we target. They built a $1.1B, 12 GWh plant eighteen months ago and lose money on every cell they make: 16 in every 100 are defective, against an industry average of 8. Each percentage point of yield is worth roughly $40M a year — the recoverable losses run to hundreds of millions annually on a $5.5B revenue line.

In lithium-ion cell manufacturing, quality is determined upstream — slurry viscosity and solids fraction in mixing, coat weight and oven zone temperatures at the coater, calendering pressure, dry-room dew point, electrolyte fill, and the CC-CV profiles of formation cycling.

The verdict, though, is delivered at end-of-line test: capacity, DC internal resistance, self-discharge during aging. By the time a cell fails there, every preceding step — materials, energy, machine time, days of aging — has already been paid for.

02The problem

The dominant losses are final-test failures whose root cause sits far upstream. A coating head drifts and one roll carries a coat-weight variance through to thousands of cells. The dry room takes a dew-point excursion overnight and a week later a cohort shows elevated self-discharge. A formation cycler's voltage profile deviates and DCIR shifts across a whole tray.

Operators can see the failure Pareto at end-of-line; what they cannot see is the causal chain — which roll, which oven zone, which shift, which cycler. Conventional dashboards add more trend views; they do not connect a final-test failure mode back to the upstream signature that predicted it. So scrap is discovered in bulk, after aging, at the most expensive possible point in the process.

And the scrap is only the visible half. Cells with marginal SEI formation — the protective layer built during the first charge — pass end-of-line testing, ship inside a customer's pack, and fail eighteen months later as warranty claims. The incumbent platforms this profile has already paid for produce dashboards and alarms; they do not reason about why coater two drifted, and they cannot fix the recipe.

We don't need a better factory. We need a brain that catches the defect at the moment it's made — not three days later, after we've spent $8 making a cell that's now worth $0.

03How Vertex-edge should help

01Connect to the coater, ovens, calender, dry-room sensors and formation cyclers over OPC UA; ingest coat-weight gauges, zone temperatures, dew point, calendering force, and the full current-voltage trace of every formation cycle.
02Add vision on the electrode web — pinholes, agglomerates, streaks — tied to roll and metre position, so a surface defect is traceable to the cells it ends up in.
03Train physics-informed models of the process itself: a coating-to-formation predictor that estimates final cell capacity within ±1.2% from upstream measurements; an SEI-formation model that catches the subtle defects end-of-line testing misses; a calender-roll mechanics model for cross-width porosity; and a drying-oven thermodynamics model that finds the optimal energy profile per zone.
04Stand up a twin of the cell build that flags a forecast failure before formation completes — "this tray's formation curves match the signature of last month's high-DCIR cohort; root indicator: coat-weight variance on roll 412" — with engineer-approved adjustments to oven zones and formation profiles inside set limits.

04What success should look like

Success would mean the bad roll stopped at the coater — not discovered as three thousand finished cells at final test.

Before formation completes

Prediction point

±1.2% from upstream data

Capacity prediction

Roll, zone, shift, cycler — named

Causal chain

SEI escapes caught before they ship

Latent defects

This is a problem brief, not a customer case study. It describes how Vertex-edge is designed to work on this problem — not a deployment that has already happened.

05Modelled economics

What this is worthto the factory.

Modelled, conservatively, for the Apex Cell profile above. Each capability targets a different bleeding point, and the recovery compounds:

Yield improvement (84% → 89%)

≈ $155M / yr

Formation cycle 12% faster

≈ $100M / yr

Unplanned downtime −40%

≈ $8.8M / yr

Warranty claims −50%

≈ $14M / yr

Drying & formation energy −25%

≈ $2.2M / yr

Conservative annual recovery

≈ $285M / yr

Composite target scenario built from public yield benchmarks and analyst data — not a completed customer deployment. At these volumes, the engagement pays for itself inside the first week of production.

06Other problems

Same system.Different floor.

Reference scenario

Automotive — body-in-white

One weld type, 12,000 recalled vehicles, €38M in warranty.

Discover

Problem brief

Surface finishing — ISO 9001

Defects are made now, found three days too late.

Discover

Problem brief

Precision manufacturing — CNC

Unplanned downtime is endured, not predicted.

Discover