The alarm flood
In large engineered systems, thousands of linked sensors go haywire at once when something fails. One fault triggers a chain reaction, and hundreds of alarms go off in a control room.
Operators are overwhelmed and can miss the one warning that matters. Alarm floods have contributed to real industrial accidents. Afterwards, engineers spend days or weeks untangling what started it.
One burning pan
Picture fire, smoke, and car alarms all blaring when one pan is burning. Every one of those alarms points back to the pan.
A causal digital twin gives the control room the same view. It says: ignore these 200 alarms, they are consequences; this one thing caused everything.
How the graph separates cause from consequence
The twin knows the cause-and-effect links between components, so it traces backwards automatically, from the alarms to where they started. Alarms downstream of the root are marked as consequences.
This is root cause analysis: find the original reason something went wrong, not the symptoms. A car overheats leads to coolant is low leads to there is a leak leads to a hose cracked. The cracked hose is the root cause.
Order in time helps
A cause always comes before its effect, so time-series sensor data makes causal analysis stronger. It reveals recurring cause-and-effect patterns and the time lags between actions and outcomes, and it helps detect hidden (unmeasured) factors that influence the system.
Then look forward
The forward version of root cause analysis is fault propagation forecasting: once a fault starts, predict which subsystems it will hit next and how fast. In the power grid, one tripped line overloads the next. Operators get time to act before the fault spreads, instead of reacting after.
See the pattern
The demo on the home page shows the pattern on a diesel engine on a test stand: twelve alarms, one coolant leak. It is an illustrative example with synthetic data.