Perception layer
Multi-camera vision, inline X-ray/CT fusion, coat-weight and thickness gauges, weld monitors, formation curves and MES tags fused at line speed.
Perception, control, simulation and governance — built for the tolerances, chemistries and qualification bar of battery cell manufacturing, and deployed on the factory edge.
Deployed on electrode, assembly and formation lines
Multi-camera vision, inline X-ray/CT fusion, coat-weight and thickness gauges, weld monitors, formation curves and MES tags fused at line speed.
Fine-tuned defect, contamination, weld and formation models plus a battery-process reasoning agent, served on-site via Triton and NIM.
Bounded setpoint write-back into coating, calendering, welding, fill, formation and robot controllers, with fail-safe stop behaviour.
An Omniverse-based cell-line twin that validates every envelope and predicts as-built yield, quality and takt before the run.
Inspection-to-decision has to happen in milliseconds on a moving web. Anodela puts inference where the electrode is, and uses the cloud only for training and fleet governance.
A hardened runtime on Jetson and IGX clusters running DeepStream, TensorRT and Holoscan pipelines for 8–24 cameras per line plus X-ray/CT fusion, targeting 50–250 ms inspection-to-decision latency. [ASPIRATIONAL targets.]
High-volume perception steps run on fine-tuned open models to control cost; premium reasoning steps route to frontier models. The router is abstracted so the best and cheapest model serves each step.
Chemistry and recipe IP is the most sensitive asset in a cell plant. Per-tenant data and model scoping is enforced end to end, with an on-prem option for producers who cannot let process data leave the site.
Vision, inline X-ray/CT, coat-weight gauges, weld monitors, formation curves and MES tags fuse into one live model of the electrode, the jelly roll and the cell.
The orchestrator plans the coat, calender, stack/wind, weld, fill, seal and formation moves inside your qualified process window and the twin's validated envelope.
Bounded setpoints are written back to coating, calendering, welding, fill, formation and robot controllers — or proposed for engineer approval in assist mode.
Every action is scored against downstream yield, defect, contamination and grading outcomes, logged immutably, and fed back into training.
Continuous coat weight, thickness and porosity estimation across the full web, with closed-loop correction of pump rate, die gap, web tension and drying zones inside the qualified window.
Metal-particle, burr and debris signatures fused from vision and CT.
Nugget depth, spatter, misalignment, seal integrity and fill-void detection.
Per-cell protocols, channel scheduling and grading accuracy.
Line balancing, WIP routing, rework and quarantine sequencing recomputed at minute granularity when a bottleneck or quality risk changes.
Every perception event, proposal and executed action is available over the API and streamed to your historian or data platform.
from anodela import Line
line = Line("plant-2/electrode-3", region="eu")
# Stream fused perception events at line speed
for event in line.stream(kinds=["coat_weight", "metal_particle", "weld"]):
if event.risk > 0.8:
line.quarantine(event.cell_ids, reason=event.signature)
line.propose(setpoint="die_gap", delta=event.suggested_delta,
envelope="qualified", approver="process_engineer")
“Formation was 24% of our cell cost. Anodela re-sequenced channels and trimmed the protocol per cell, and we took 17% of the formation hours out without touching a single capacity spec.”
“Metal-particle escapes were our nightmare. The contamination model flags a signature our inspection vendor simply does not see, and it quarantines the affected cells automatically.”
“We commissioned line four with the twin first. It predicted the calendering window that would hold porosity, and the real line matched it within tolerance on week two.”
Pre-built OT and API connectors read process, vision, X-ray/CT, formation and MES data — and write bounded setpoints back. Nothing rips and replaces; Anodela runs alongside the equipment you have qualified.
Cell manufacturing is safety-critical and IP-dense. Anodela is designed so quality, IT/OT and EHS approvers can say yes.
SOC 2 Type II program with encryption in transit and at rest, SSO/SAML, SCIM and RBAC down to the line.
Chemistry, coat-weight and formation recipes stay in your tenant. Optional fully on-prem deployment with no egress.
Every perception, proposal, approval and executed setpoint change is immutably logged for IATF 16949 and IEC 62660 evidence.
EU, US, Korea, Japan and China-domestic regions, or air-gapped factory edge for sovereign programs.
Standardize coating, weld and formation autonomy across every plant with central model governance and per-site envelopes.
Managed Jetson/IGX clusters with OTA updates, versioned rollbacks and 99.9% availability targets.
Named process-AI engineer, qualification support, PPAP-aligned documentation and 24/7 escalation.
Multi-year agreements with pricing tied to first-pass yield, scrap and formation cost.
On your factory floor. The control path never depends on an internet connection; cloud is used for training, fleet governance and reporting, and can be removed entirely in on-prem mode.
Every autonomy envelope is defined against your qualified limits, validated in the cell twin, and approved by your quality team before it is armed. Actions outside the envelope are proposals, never executions.
Yes — most sites land on a single wedge workflow, often coating-defect and coat-weight sensing or metal-contamination inspection, then expand to control once accuracy and ROI are proven.
We will scope the wedge workflow, the connectors and the measurable success metric in one call.