Rhizant
Book a call

Thousands of alarms. One cause. Find it before it spreads.

Causal Intelligence for Complex Engineering Systems

Rhizant builds a causal digital twin of your engineered system from the sensor data and engineering knowledge you already have. It traces faults to their root, predicts what a change will do, and tells you how sure it is.

Book a call See capabilities

Illustrative demo

Root cause in one view.

A diesel engine on a test stand. An alarm flood begins: twelve alarms in two and a half minutes. The causal twin traces them back to one cause.

Diesel engine on a test stand Illustrative example with synthetic data

What the control room sees

12 alarms
  1. 0:00 Coolant level low root: coolant leak
  2. 0:20 Coolant pressure low consequence
  3. 0:45 Coolant temperature high consequence
  4. 1:05 Cylinder head temperature high consequence
  5. 1:10 Oil temperature high consequence
  6. 1:20 Charge air temperature high consequence
  7. 1:30 Exhaust temperature high consequence
  8. 1:40 Oil pressure low consequence
  9. 1:50 NOx high consequence
  10. 2:05 Fuel consumption high consequence
  11. 2:15 Engine derate active consequence
  12. 2:30 Power output low consequence

What the twin sees

Tracing causes1 root cause, 11 consequences

Coolant level alarm root cause Coolant pressure alarm consequence Coolant temp alarm consequence Cyl. head temp alarm consequence Oil temp alarm consequence Charge air temp alarm consequence Exhaust temp alarm consequence Oil pressure alarm consequence NOx alarm consequence Fuel use alarm consequence Engine derate alarm consequence Power output alarm consequence Root cause: coolant leak Confidence: high What-if: leak repaired Consequences clear Coolant level alarm Root cause: coolant leak Confidence: high What-if: leak repaired Consequences clear Coolant pressure alarm consequence Coolant temp alarm consequence Cyl. head temp alarm consequence Oil temp alarm consequence Charge air temp alarm consequence Exhaust temp alarm consequence Oil pressure alarm consequence NOx alarm consequence Fuel use alarm consequence Engine derate alarm consequence Power output alarm consequence

With the leak repaired, the twin predicts that the eleven consequence alarms clear.

Ask the twin

Engineer

Why did exhaust temperature spike?

Rhizant

Exhaust temperature rose because coolant temperature rose, because coolant level dropped. The most likely root cause is a coolant leak.

  1. Coolant leak
  2. Coolant level low
  3. Coolant temp high
  4. Exhaust temp high

Eleven of the twelve alarms are consequences of it.

Recommended action: inspect the coolant circuit before changing anything else.

Suggested follow-up Is the exhaust temperature sensor itself trustworthy?

How it works

Sensor data and manuals in. A twin you can ask why out.

01

Connect

Stream sensor and operating-condition data from historians, SCADA, telematics, or an existing digital twin. Add manuals, schematics, and maintenance logs. Nothing has to be centralized first.

How it's built
02

Model

Rhizant drafts a causal graph of your system from its documents, confirms and corrects it against your time-series data, and builds in the physics you already know. Every link carries a confidence score and its evidence.

How it's built
03

Act

Ask why a fault happened, what will happen if you change a setting, and which action to take. Review it yourself, or let the twin intervene in closed loop.

Capabilities
WhyWhy did this fault happen?
What ifWhat will happen if we change this setting?
What to doWhich action should we take?
Human in the loopClosed loop

Ladder of causation

Three levels of questions

Correlation has no direction; causation does. Standard digital twins and normal machine learning stop at the first level. A causal digital twin also answers the two above it.

3

Counterfactual

“What if I had done something differently?”

Why did this specific failure happen?

Causal twin
2

Intervention

“What if I do this?”

Change an operating setting, perform a maintenance action.

Causal twin
1

Association

“What if I see this sensor reading?”

Standard digital twins and normal machine learning stop here.

Standard ML and digital twins
Where the causal twin worksInterventions and counterfactuals
Where standard ML and digital twins stopPatterns in what is observed

Industries

Where this applies

For engineers and operations teams responsible for complex, sensor-instrumented systems, starting with the people who untangle faults and alarms every day: reliability engineers, controls engineers, plant operators, maintenance managers, and field-quality and warranty engineers.

01

Engines and powertrainsDiesel engines, heavy-duty equipment, OEM fleets
The pain
Field failures across thousands of units. Emissions and fuel targets. Warranty costs.
What the twin does
Root cause analysis, emissions and fuel-efficiency interventions, warranty and field-failure analysis, remaining useful life.
Ask it
“Which design or usage factors are causing these field failures?”
02

Power generation and gridPower plants and grids
The pain
Alarm floods in the control room. Cascading failures. Safety margins.
What the twin does
Alarm flood suppression, fault propagation forecasting, operational safety.
Ask it
“A line tripped. What fails next, and how soon?”
03

Process and semiconductor manufacturingSemiconductor and chemical manufacturing
The pain
Faulty product, and nobody knows which of hundreds of settings is to blame. Sensors that cry wolf.
What the twin does
Process optimization, sensor fault vs machine fault, sensor placement.
Ask it
“Yield dropped. Which setting is actually to blame?”
04

Wind, rotating equipment, and fleets across sitesWind farms, other rotating equipment, vehicle fleets
The pain
Models that break when moved to a new site or season. Maintenance that predicts but does not explain.
What the twin does
Robustness across sites, prescriptive maintenance, learning from past decisions.
Ask it
“This turbine will fail. Why, and what exactly do we fix?”

Trust

Built for the questions engineers will ask.

Every answer carries a confidence estimate.

You see how sure the model is before you act on it.

Every causal link is inspectable.

The graph is not a black box. Each edge shows its evidence, and your engineers can correct it.

Works with or without a digital twin.

Start from the sensor data you have. If you already run a digital twin, the causal twin sits on top of it.

The science

Built on causal science

Rhizant is built on structural causal models: a graph of which variables drive which, plus the equations behind each link.

That is what lets it answer the questions correlation-based AI cannot: why this happened, what happens if you intervene, and what would have happened if you had acted differently. The methods come from decades of causal-inference research. Rhizant's work is making them automatic, scalable, and trustworthy enough to run on live engineering systems.

How it's built
ABCDEFGHIJ

See why it failed. Decide what to change. Before you touch the machine.

A 30-minute conversation with the founder about your system, your data, and whether a causal twin fits.

Book a call Contact

Looking for design partners

  • A painful problem

    A fault, alarm, or reliability problem your team needs explained.

  • The sensor data to go with it

    Existing sensor data and documents. No new experiments needed to start.

  • Book a call

    [PLACEHOLDER: scheduling link]

  • Email

    [PLACEHOLDER: contact email]

Book a call