Start from what we know about the asset, not from a blank page.
A Blueprint describes how a failure mode is observed on a type of asset: required signals, parameters, the expected degradation pattern (rising trend, residual drift, tracking error, asymmetry, decay rate, flat signal), linked failure modes and suitable analysis engines. It is cloned across many units and adapted to each plant. Today: hundreds of Blueprints for wind, solar and hydro, defined on the fault-tree base and on a taxonomy of 192 component types in the ISO 14224 style.
Example. The “generator bearings” Blueprint of a wind turbine already knows which temperatures are needed, against which load and ambient conditions, and what the expected degradation looks like. Applying it to fifty turbines takes tag mapping, not fifty projects.
Pumps, engines, compressors and other industrial machines: the same analysis engines and the Blueprint wizard build their Blueprint on your data, as a project.
Tell an emergency from a manageable risk.
Every degradation signal is linked to its failure mode and to the asset's fault tree (271 failure nodes in the catalogue). Risk matrices weigh safety, environment and continuity of service; minimal cut sets show which basic events lead to system failure.
Example (illustrative). On a hydro unit, the temperature of a guide bearing rises out of its band. Sibilla shows which failure mode it relates to, where it sits in the tree up to the unit trip, and its weight on safety, environment and continuity.
From algorithms to a traffic light any maintainer understands.
Key Diagnostic Indicators (KDIs) turn model results into a green, amber or red state for each model, with interactive history, predicted versus actual values, confidence bands, a similarity score to the expected degradation pattern and a log of every run.
Example (illustrative, simulated data). A turbine gearbox turns amber. Open the history: the oil temperature has been outside its predicted band for ten days, at the same wind and power.
The right model for each signal. And data you can trust.
A catalogue of 14 analysis engines, from statistical models to deep learning, validated before field use: in the field we only run what has passed validation (today regression, baseline departure and peer comparison between twin units). Every model reports confidence bands and error metrics. Before the models, the Data Quality dashboard checks every measurement point: missing, frozen, duplicated or out-of-range data.
Example. A “frozen” temperature sensor on an inverter has returned the same value for three days. Sibilla flags it as a data problem, not a failure, and avoids a false alarm.
Every alarm leads to an informed decision.
The predictive maintenance loop: acquisition → data quality → models → KDI → failure mode and risk → the expert's decision. The worst state among an asset's models feeds the Asset Health dashboard and the event calendar.
Example. A bearing KDI turns red; the asset page shows the linked failure mode, the out-of-band signals and the risk. The maintenance lead decides the inspection.
- Available
KDI, fault trees, Asset Health
- Project-based
Inspections suggested from the fault tree; opening notifications in your CMMS/ERP (SAP PM and others)
AI explains, the expert decides.
Sibilla exposes its functions through a REST API and an MCP server, so an AI assistant can read results already computed and describe them in plain words (why the KDI is amber, which signals weighed most, what to check). Detection stays deterministic and verifiable; the AI explains, it does not decide.
Example (illustrative). “Why has the generator bearing been amber since Tuesday?” → an answer naming the signals that contributed most, with links to the charts.
Available
