AMI Operational Correlation Engine
Eight classes of source system, one ordered timeline per device — and cause derived from sequence rather than assumed from symptom.
A dozen dialects, none of them comparable
A smart metering estate produces events in a dozen incompatible dialects. A meter reports a power-fail with its own timestamp and clock drift. A head-end reports a communication failure minutes later, after its retry policy is exhausted. A network operator reports a cell sector degradation on a different clock entirely. MDMS reports an estimated read the following day. None of them are wrong, and none of them are comparable.
The correlation engine is the layer that makes them comparable. It ingests events from every source, normalises identity, time and severity, orders them into a single sequence per device and per group of devices, and then derives cause from that sequence — which event preceded which, across which boundary, affecting how many devices in the same place.
What the engine correlates
Operational evidence is only conclusive when it spans every boundary a fault can hide behind.
Six stages from raw event to retained evidence
Each stage is deterministic and auditable, so a conclusion drawn today can be re-examined and defended months later.
The same device is known by different identifiers in each system: meter serial, device ID, SIM ICCID, MPAN/MPRN, asset number. The engine resolves them to one canonical device so evidence from eight systems lands on one object.
Timestamps are reconciled across source clocks, time zones, drift and reporting delay, separating when something happened from when a system noticed. Ordering survives late-arriving data.
Vendor event codes and severities are mapped onto common operational meaning, so a comparable fault from two manufacturers is genuinely comparable.
Devices are grouped by concentrator, cell sector, feeder, site and tenant, turning individual symptoms into population-level signals.
Ordered, grouped, normalised events are chained into a cause hypothesis with the supporting evidence attached — not a score without a reason.
The chain is retained as an auditable record, so the conclusion can be reviewed months later by an operator, an auditor or a supplier.
What correlation changes in the operations console
The engine exists to change what happens next, not to produce a richer report.
Remote-first resolution
Because cause is established before dispatch, the remote diagnostic and remote action path is exhausted first and the truck roll becomes the residual case, not the default.
Correlation needs consistent meaning
Ordering events is only half the problem. The semantic information model is what makes a register or an event code from one vendor genuinely comparable to another's.
See your estate in smoc.cloud.
A demo walks through your utility types, your vendors and your compliance obligations — not a generic slide deck.