Perception layer
EL, PL, and RGB vision fused per cell and string into a live defect and geometry model.
Platform overview
The factory brain for solar module lines
Panevex is not another inspection camera or a dashboard. It is a closed-loop autonomy layer that senses every cell, decides in milliseconds, and controls the machine — station by station, module by module.
7coordinated agents
<10 msedge decision latency
1shared module twin
The stack
Perception, decision, and control — as one system
Decision layer
Physics-aware models score crack-risk, cure state, and power impact and choose set-points.
Control layer
Closed-loop actuation on stringers, laminators, and handling — corrected in real time.
The data flywheel is the moat
Each module produces perception data that trains better models, which set better policy, which produces better modules. Ribbix, Elumnex, Lamira, Binova, and Twinvex all read and write to the same module twin — so improvement compounds across the whole line.
- Shared twin, not siloed point tools
- Every run improves the next
- Policy that transfers across lines and fabs
The agents
Each agent owns a station
String & interconnectRibbix
Sound, low-loss bonds without cracks or cold joints.
Explore EL & defect sensingElumnexSee inside every cell — predict microcracks before they cost you watts.
Explore Lamination & cureLamiraCured right the first time — no delamination, bubbles, or cell shift.
Explore Binning & matchingBinovaCell binning and power-class matching that recovers every watt.
Explore Module digital twinTwinvexHit target power-class and yield before the physical run.
Explore
Capabilities
What the platform does, end to end
Cell-level EL/PL
Microcrack, broken-finger, and solder segmentation on every cell.
Cure control
Encapsulant cure-state prediction and closed-loop lamination.
Bond control
Real-time thermal, pressure, and alignment control on the stringer.
Power-class matching
Mismatch-minimizing binning on true defect state.
Module twin
Simulate stringing, cure, and IV before the physical run.
Yield analytics
Root-cause traces from field power back to the exact station.
| Layer | Runs on | Latency | Owner agent |
|---|---|---|---|
| Stringer control | Jetson Orin / Thor edge | < 10 ms | Ribbix |
| EL / vision perception | Jetson Orin + DGX/HGX | in-line | Elumnex |
| Lamination loop | Jetson Orin + OVX | closed-loop | Lamira |
| Binning & matching | DGX / HGX + edge GPU | per lot | Binova |
| Module digital twin | OVX / RTX + DGX/HGX | pre-run | Twinvex |
Integration
Fits the line you already run
01
Sense
Add Panevex cameras and edge units at each station — no rip-and-replace of your machines.
02
Shadow
Run in observe-only mode to build trust and baseline your current escape and mismatch rates.
03
Close the loop
Enable closed-loop control station by station, with human-set guardrails on every actuator.
04
Scale
Roll proven policy to every line and fab from one control plane.
5
Line stations covered
99.4%
Defect segmentation recall
<10 ms
Edge control latency
24/7
Autonomous operation
“Panevex behaves like a process engineer that never sleeps and never misses a cell. It's the first system that closed the loop instead of just charting the problem.”
DMDirector of Manufacturing TechnologyGlobal module manufacturer
Map your line to the agent stack
Bring your station layout. We'll show you exactly where perception and closed-loop control plug in — and what it's worth.