Interactive demo
A simple machine-health story.
One fictional factory. One cooling-water pump that matters to the line. Choose a scenario, then walk through what the maintenance team would see and decide.
-
Step 1
The machine is running normally
For seven days the pump is observed. Its usual vibration pattern, including the quieter night-time state, becomes the learned normal.
-
Step 2
Something changes
-
Step 3
An amber notice appears
-
Step 4
The maintenance team investigates
-
Step 5
A decision is recorded
The example report shows what changed, what the team checked and what action was chosen. The evidence and the decision stay together.
-
Machine Health Report
Cooling-water pump CWP-2, fictional plant, Singapore
Example output. Simulated values.
Normal versus unusual behaviour
Maintenance review
Why the third scenario matters
False alarms are the biggest credibility risk in this field.
A system that flags every product change as a fault gets switched off within a month. The concept treats the operating state as part of normal, asks people to label what they find, and learns from the labels. The demo shows the honest limit: a baseline method knows that something changed, not why.
Normal operation
The system can establish a reference pattern. A quiet result is still a result.
Unusual vibration
An earlier warning can prompt an investigation while a planned repair is still possible.
Operating conditions change
A warning does not necessarily mean a machine fault. Review by people is part of the design.