CamTwin builds digital twins of cathode active material production lines that execute the plant's original robot and PLC programs. On top of that we add a monitoring and failure-prediction layer trained on real alarm, trend and maintenance data.
Cathode material for EV batteries goes through mixing, kilns, saggar handling robots, conveyors and fillers. These are long, tightly coupled lines controlled by thousands of PLC rungs and robot jobs.
Changing a PLC rung or robot job means testing on a live line. Vendor simulators replay hand-made motions, not the program that actually runs.
When a robot freezes holding a saggar, finding the cause depends on whoever is on shift. The knowledge lives in people's heads, not in the system.
Alarm logs, heater and motor trends, and maintenance records already exist. They are rarely connected, and failures are found only after they happen.
Each layer is working software today, built as a reference installation for one robot cell and its conveyor line.
Our Unity twins move material the way the real process does. Equipment is driven by control logic, and powder follows mass balance and physics instead of keyframes.
We inject faults into the twin, such as a stuck cylinder, a broken sensor cable, a noisy distance sensor or a paper sheet of the wrong thickness, and record what the operator would actually see.
A facility monitoring system that brings alarms, process trends and maintenance history into one place, then predicts equipment failure in the next N hours. Operators confirm or reject each prediction, and the model learns from their answers.
Interpreters for robot and PLC languages, a 3D simulator, an ML pipeline and a full-stack monitoring app: CamTwin was built by its founder working with Claude. Next, Claude becomes part of the product itself.
Claude Code wrote and refactored most of the codebase: the robot job interpreter, the MELSEC-style PLC scanner, the fault-injection harness and the monitoring app (about 37k lines in the monitoring system alone).
Claude cross-checks PLC comments and robot jobs to work out what each signal means. That is how the robot ↔ PLC handshake was reconstructed and verified.
A technician describes the symptom in plain Korean or English. Claude reads the live twin state and the Fault Lab knowledge base, then answers: "Check the anti-drop cylinder's down-position sensor first, 55% likely."
Ask "why did the robot stop waiting at this step?" and get the answer traced through the actual PLC logic and robot job, not a generic manual.
Claude turns a day of alarms, trends and predictions into a short shift-handover summary with prioritized actions.
We are turning the reference builds into a demo we can sell. Then we build customized twins for each plant, starting with the equipment every cathode line has.
Line Twin, Fault Lab and Line Monitor packaged as one demo on reference line data.
Meet cathode and battery-material plants; scope twins customized to their lines.
First paid twin built on a customer's own programs and data.
Maintenance copilot and program explainer shipped to customers in Korean and English.
Sunghun Heo · Founder
CamTwin's founder works on a cathode active material production line in Korea, alongside the kilns, robots and conveyors this software models. Every feature comes from a real stop, a real alarm or a real troubleshooting session on the floor.
This domain knowledge is our moat. We know which signals matter, what the field data actually looks like, and what a maintenance technician needs at 3 a.m.
We're looking for cathode and battery-material plants to partner on pilots, and for people who want to build industrial AI with us.