Prove cell quality from coating through formation
Willowark builds measurement, data, and control systems for battery cell manufacturing and deployed energy storage. In the plant that means inline gauging at the coater, genealogy following material from electrode coil to finished pack, and a data path turning weeks of formation cycling into per-cell records. In the field it means edge telemetry from sites speaking Modbus, DNP3, and CAN.
The engineering follows the failure modes. Coat weight drift is a control problem, so measurement goes where feedback can still act and closes around the coater rather than reporting after calendering. Cell defects are a traceability problem, so identity carries through slitting, notching, and stacking, where it is easiest to lose. Thermal events are a sensing problem, so rate-of-rise, cell-to-cell divergence, and off-gas detection run at the edge.
First projects in cell plants tend to start at one measurement or one handoff: gauging at the coater, identity carried across slitting, or a formation data path that stops discarding channel data. Much of the sector is still commissioning lines, so we often work alongside equipment vendors whose machines expose data differently, and the hard constraint is not sensing but agreeing on what a cell identifier is before the line runs. For storage operators, the first project is usually a single site brought into one telemetry model, then repeated across the fleet.
Reviewed
Sound familiar?
If you've said any of these, we should talk.
“Coat weight drifts and we do not find out until the material is through calendering.”
Measurement belongs where the correction can still be made. We integrate inline thickness and basis-weight gauging at the coater, close the loop to the control system, and run SPC on the cross-web profile rather than one lane. Scrap found by the roll becomes an adjustment inside the run.
“Formation runs for days across hundreds of channels and we are guessing which cells fail grading.”
Cycler data is the richest signal in the plant and the least used. We build the ingestion path at native resolution, extract capacity, coulombic efficiency, and resistance features per cell, and train classification against your own grading outcomes.
“A pack failed in the field and we cannot trace it back to the electrode coil.”
Identity has to survive the transitions where it is easiest to drop: slitting one roll into many, notching, stacking, and pack build. We carry coil, roll, and lot identity forward through each step to cell, module, and pack.
“Every storage site reports something different and we have three vendor portals open.”
We build edge gateways speaking Modbus TCP, DNP3, SunSpec, and CAN at the site, normalize into one telemetry model, buffer through outages, and publish to a single fleet view. Alarms get consistent severity and routing regardless of whose inverter raised them.
“Every machine on the line came from a different vendor and each one has its own idea of what a batch is.”
That is normal on a new line, and it is a data model problem rather than a vendor problem. We define one identity scheme — coil, roll, lot, cell — and map each machine's native records onto it at the gateway, so the genealogy stays consistent even when the equipment never agrees.
How this industry actually runs
The operation as we usually find it.
Cell manufacturing is a continuous front end bolted to a discrete back end. Slurry is mixed, slot-die coated onto foil, dried, calendered to density, slit, and notched — a web process where coat weight, thickness, and edge quality determine everything downstream. Then it turns discrete: stacking or winding, tab welding, enclosure, electrolyte fill in a dry room held below one percent relative humidity, and sealing. Formation follows, cycling every cell across hundreds or thousands of channels for days and producing per-channel time series most plants archive and rarely mine. Aging, end-of-line test, and grading come next, then module and pack assembly with laser-welded busbars and a BMS. Certification pressure runs through all of it, with UL 1973 and IEC 62619 for cells and systems and UL 9540A data behind installation approvals. Deployed storage inherits the same concerns differently: fleets of sites with inverters, PCS, and BMS from different vendors, an EMS dispatching against market signals, and warranties written against degradation curves someone has to measure.
Machine signals to the people who decide
Components:
- PLCs & sensors (counts, states, current)
- Legacy machine (dry contact / clamp)
- Edge gateway (normalize, buffer)
- Production dashboard (downtime, OEE)
- Alerts & reports (who acts, when)
Connections:
- PLCs & sensors to Edge gateway (EtherNet/IP, Modbus)
- Legacy machine to Edge gateway
- Edge gateway to Production dashboard (MQTT)
- Edge gateway to Alerts & reports
What we build
Starting projects that fit Battery & Energy Storage.
- Inline coating thickness and basis-weight measurement with closed-loop control and profile SPC
- Cell genealogy from electrode coil and slit roll through cell, module, and pack
- Formation and cycler pipelines with per-cell feature extraction and early-scrap classification
- Dry room and clean utility monitoring: dew point, differential pressure, temperature, alarms
- Vision inspection for electrode edges, tab alignment, and laser weld seams
- Edge telemetry for storage sites across Modbus, DNP3, SunSpec, and CAN
- Fleet degradation and warranty analytics built from cycle and dispatch history
- End-of-line test and grading integration writing capacity, resistance, and grade into the cell record
Capabilities we bring
Working in Battery & Energy Storage?
Tell us the line.
What runs by hand, what is not connected, what you are trying to build. An engineer replies within one business day with whether and how we would approach it.
Common questions
What Battery & Energy Storage teams ask first.
Can you handle the data volume that formation and cycling produce?
Yes, and the design decision is where to keep resolution. Full-rate channel data is worth retaining during formation and for a defined window after; beyond that, extracted features and summary curves carry most of the analytical value at a fraction of the storage cost.
Do you work on cell manufacturing or on deployed storage sites?
Both, and they share more than they appear to. A plant and a storage fleet are both problems of getting reliable measurements out of heterogeneous equipment, keeping identity attached to the data, and making alarms mean something. The protocols differ; the engineering does not.
How do you approach thermal runaway monitoring?
As a layered sensing problem that supports, never replaces, certified protection systems. That means temperature rate-of-rise instead of absolute thresholds, cell-to-cell voltage and temperature divergence, off-gas sensing where the enclosure allows, and local logic that acts without a network round trip.
Can you integrate with the BMS, PCS, and EMS vendors we already have on our sites?
Usually, yes. Most of these systems expose Modbus TCP, DNP3, SunSpec, CAN, or a vendor API, and the work is mapping each register map or message set into a common model rather than replacing anything. The harder part is typically documentation quality and firmware differences between sites of the same vendor, so we build per-device profiles and validate them on site before trusting the data fleet-wide.
We are still building our first line. Is it too early to bring you in?
Usually it is the best time, because the cheap decisions are still open. Identity schemes, what each machine must expose, where gauging sits relative to control, and how formation data is written are much easier to specify in equipment purchase orders than to retrofit after acceptance. We typically work alongside your process and equipment engineers on those specifications, then build the data and control layers as the line is commissioned.
Strategy. Software. Systems.
Engineering for Battery & Energy Storage.
Describe the problem in your own words. An engineer reads it — not a sales script — and tells you plainly what it would take.