Industrial Black Box
A flight recorder for industrial operations: enough continuously captured state to reconstruct exactly what happened before a failure.
- Program
- Concept
- Started
- Aug 2026
- Field test
- Not yet
- Commercial
- No
01 Problem
The evidence is gone before the question is asked.
When production equipment fails, the record of what led up to the failure usually does not exist. The machine stops, the loss begins, and the investigation starts with almost nothing.
PLC logs are shallow. Maintenance history lives in another system. The operator context that would explain the sequence is verbal, so it is never written down.
02 Why existing approaches fail
Every existing record stops short.
Each of the tools that should answer the question captures a fragment of it.
- Historians
Process historians capture process values, but not the context around them. The trend survives while the reason for it does not.
- Downtime codes
Downtime codes compress an eleven-day causal chain into a single entry that reads 'machine fault'.
- Reconstruction from memory
Once the failure has happened, the evidence needed for maintenance, insurance and disputes is reconstructed from what people remember.
03 System
Capture everything, seal what matters.
The recorder captures continuously into a ring buffer and keeps almost nothing. A failure trigger decides which window of state is sealed and preserved for reconstruction.
- 01Sensors + PLC events + operator notes + schedules
- 02Continuous capture ring buffer
- 03Failure trigger
- 04Sealed incident record
- 05Reconstruction timeline
04 What exists today
Design work, not deployments.
What exists today is system design and capture-model work. Component selection for the edge recorder is in progress.
Nothing is deployed. There is no hardware in the field and no recorded incident to show. The status table below is the honest state of the program.
05 Evidence
What we can show.
No published evidence yet. This section fills as experiments produce results worth showing.
06 Status
The program, stated plainly.
This table is the actual state of the program. It changes when the state changes, not before.
- Status
- Concept
- Started
- Aug 2026
- Field test
- Not yet
- Commercial
- No
- Open components
- Planned
07 What comes next
A bench prototype, then an induced failure.
The next step is a bench prototype recording a small machine cell. After that comes a reconstruction exercise against a deliberately induced failure: break a known thing on purpose, then test whether the record alone can explain it.
08 Work with us
Have failures nobody can reconstruct?
Send us the machine architecture and the data you currently record. We will tell you what a recorder would have to capture to answer your last unexplained stop.