Observe
Continuously evaluate telemetry, derived features, physics residuals and operating context.
NeIO Predictive IntelligenceNeIO PIE · Predictive engine
NeIO PIE · Predictive Intelligence Engine · by Codvo
Know what's degrading before it's an incident.
Physics-first predictive intelligence for data centres and critical infrastructure. NeIO PIE combines equipment physics, synthetic fault trajectories and live telemetry to find developing faults, explain the evidence and help your teams act earlier.
Site A · Chiller plant · CH-02Observing
Fault probability4%
Anomaly score0.08
Est. RUL—
Stage 2 · developingProbable: condenser fouling
Contributing evidence
Condenser approach Δ
kW / ton residual
CW ΔT trend
RE
Work order proposedCondenser tube inspection · awaiting approval
REVIEWIllustrative values● Advisory · no plant command sent
Reads what you already run
BMSSCADAOPC UA / DAHistoriansEPMS and DCIMTelemetry hubsEquipment interfaces
NeIO
PIE
PIE
Hands back what your teams use
Ranked asset riskProbable diagnosisFault stage and RULContributing evidenceWork-order hand-offs
Your systems show what's happening. Not what's developing.
Data centres already produce plenty of telemetry. Dashboards, threshold alarms and history-hungry models still leave operators reacting after degradation is visible.
01
Thresholds arrive late
Rules fire when one signal crosses a limit. Slow, interacting degradation stays hidden while every tag still looks acceptable.
02
Failure history is scarce
New assets have no years of labelled site failures to train a conventional data-only model on.
03
Diagnosis stays manual
Operators still cross-reference trends, alarms, maintenance history and engineering judgement before they can act.
04
The missing layer
Not more monitoring. Early, equipment-aware intelligence that tells normal variation from a developing fault.
See it in action
Same asset, same data. Weeks of extra warning.
Scrub the timeline. Threshold alarms wait for a limit; NeIO PIE flags the drift, names the probable fault and proposes the next step.
Day −30
Status
NORMALAll tags within limits
PIE compares live readings with the physics-expected band. No significant residual yet.
Illustrative interface · assets, values and warning horizons are for demonstration. Real lead time depends on signals, data quality, operating conditions and fault type.
The NeIO PIE Workbench
One view from fleet to feature.
Rank risk across sites, move from an alert to its probable diagnosis, then inspect the signals and features behind it.
Fault probabilityLikelihood of a recognised fault signature.
Anomaly scoreEvidence of deviation from expected behaviour.
Probable diagnosisFault type, stage and operating context.
Contributing evidenceFeature importance and deviation from baseline.
Platform views shown for illustration. Dataset breadth and available models depend on the configured release and customer scope.
Cold-start intelligence
Protect a new asset before it learns to fail.
PIE starts with the physics that governs the equipment. Synthetic fault trajectories give it a foundation; live telemetry calibrates it to your site.
01PhysicsFirst-principles and physics-informed models
Operating equations, engineering constraints and causal relationships define what healthy behaviour looks like for each equipment family.
Equipment-awareExplainable by design
02SynthesisSynthetic fault trajectories
Known degradation patterns are generated across operating conditions and failure stages, so the models see faults your site has not had yet.
Faults across stagesMany operating points
03ModelsPretrained asset-family models
A fault classifier and an anomaly detector per family, ready without waiting for a local failure corpus.
Known faults classifiedNovel anomalies flagged
04CalibrationSite calibration
Live telemetry, operator feedback and governed model refinement tune the baselines to your equipment and operating envelope.
Operator feedback loopGoverned updates
Coverage from day one. Physics and synthetic fault data mean the asset is protected while site history is still being built.
Conceptual illustration, not measured data.
Data-centre coverage
Coverage across the facility, not one machine.
An extensible starting library for 12+ equipment families, configured to the signals and environment you have.
Equipment family
Chiller
Typical signals
Approach tempsKW/tonRefrigerant pressures
Example fault signatures
Condenser foulingLow chargeCompressor wear
Signals and signatures are configured to the telemetry available at your site.
01Live telemetryStreams at configurable cadence, e.g. 5-second*
02Derived featuresTrends, residuals, statistics
03Model evidenceClassifier plus anomaly detector
04Operator outputRisk, diagnosis, RUL, action
*Ingestion cadence depends on source-system access, telemetry availability and implementation design.
Governed agentic layer
Observe, diagnose, explain. With human approval.
Model output becomes an operating workflow, while control stays with your facilities, reliability and maintenance teams.
Diagnose
Correlate fault probability and novel anomalies into the most probable developing condition.
Explain
Show the contributing signals, feature deviations, stage, confidence and recommended next step.
Approval gate
Orchestrate
Route approved intelligence to maintenance, operator response and work-order workflows.
Human-governed by design. Approval gates, traceability and role-based policies are configured to your environment.
Who it's for
Same engine, different reasons to say yes.
Facilities, reliability, OT security and finance each see a different return from earlier, explainable warning.
Facilities and operations heads
Running live halls, chiller plants and electrical rooms
Top concerns
Unplanned downtimeSLA exposureNight-shift coverage
In a pilotRanked risk for your priority assets on one screen.
Reliability and maintenance engineers
Owning PM plans, root cause and spares
Top concerns
Time to isolate faultsFalse alarmsNew-asset blind spots
In a pilotDiagnosis, stage and evidence for every flag.
CIO and OT security
Approving anything that touches operational data
Top concerns
Data sovereigntyNo write pathUpdate governance
In a pilotOn-prem or air-gapped, inside your boundary.
Finance and risk leaders
Weighing uptime risk against spend
Top concerns
Avoided incidentsMaintenance costMeasurable pilot
In a pilotValue exposure estimated for each detected condition.
Sovereignty and integration
Runs inside your boundary. Extends what you run.
Deploy in customer-controlled infrastructure and add intelligence to systems your teams already trust, with no rip and replace.
Where people work
NeIO PIE WorkbenchExisting HMI and dashboardsCMMS / work ordersMCP-enabled connectors
Added by Codvo
Canonical tag mappingPhysics-aware asset schemaPIE models and inferenceApproval and audit
Your estate, unchanged
BMSSCADAOPCHistoriansTelemetry hubsEquipment interfaces
Air-gappedFully disconnected
Sensor data and inference stay inside the facility.
On-premisesInside your OT/IT boundary
Designed for OT security review and operational continuity.
ContainerisedIncluding Red Hat OpenShift
Containerised deployment on the platform you already run.
GovernanceCustomer-controlled
You own model, access and update governance.
Privacy & security. Runs on NeIO inside your boundary, under NeIO Govern controls. Advisory by design: no command is sent to your plant.
Read moreA controlled path to proof
A six-week pilot, measured on your operations.
Pick a focused site and priority equipment families. We validate telemetry, set baselines and measure whether predictions are early, explainable and useful.
Accelerated pilotIndicative plan
0 weeks to a decision
- ScopeSite, asset families and success criteria agreed
- ConnectTelemetry validated, tags mapped to the asset schema
- BaselineOperating baselines set, models calibrated
- RunLive predictions reviewed with your team
- DecideResults against criteria, scale plan
Scope, criteria and warning horizons are agreed per asset. Results depend on signal availability, data quality, operating conditions and fault type.
- Warning lead timeHow early actionable degradation is surfaced.
- Prediction usefulnessPrecision, relevance and operator acceptance.
- Asset coveragePriority equipment and the usable signal set confirmed.
- Investigation effortTime saved in fault isolation and triage.
- Workflow readinessAcknowledgement, escalation and approved hand-offs tested.
- Value exposureOperational risk addressed by detected conditions.
FAQ
Questions we hear most.
Does NeIO PIE replace our BMS, SCADA, historian or CMMS?
No. It is an intelligence layer that reads from approved data sources and hands results to your existing workflows. Your systems remain the systems of record and control.
Can it work on assets with no failure history?
Yes. Physics-informed models and synthetic fault trajectories provide the starting point, and live telemetry calibrates them to your site over time.
What data does it need?
Telemetry from the selected equipment families through BMS, SCADA, OPC, historians or telemetry hubs. During the pilot we confirm which signals are available and usable.
Can it run air-gapped?
Yes. Deployment can be on-premises or fully air-gapped, including containerised patterns such as Red Hat OpenShift, so sensor data and inference stay in your facility.
How do we know why something was flagged?
Every prediction shows fault probability, anomaly score, probable diagnosis and stage, plus the contributing features and their deviation from baseline.
Does it act on the plant?
No. Outputs are advisory. Hand-offs to maintenance or work-order systems go through approval gates you configure.
See what's developing on your own assets.
Bring one site and your priority equipment families. We'll scope a six-week pilot around them.