Symbolic Engineering

Home / Product / Sensor intelligence

Module

Sensor Intelligence

Contour ingests instrumentation continuously and resolves every tag to the equipment it measures. That resolution is the join a plant has never had, and it is what makes the time series usable by the asset tree and the reliability models at all. This module is less a report than the ground the other two stand on.

A tag is not an asset

Instrumentation is addressed by tag, but failures happen to equipment. The tags were named by the control engineers at commissioning; the functional locations were named by the maintenance department, separately and usually some years later. No table mapping one vocabulary onto the other has ever existed, which is why the historian and the maintenance system describe the same machine in two languages and no analysis can span them.

Contour builds that mapping from the sources that already encode part of it: the naming convention itself, the asset framework in the historian where one has been maintained, the equipment references carried on work orders, and the loop and interlock structure in the control system configuration. The result is a proposed mapping with a confidence against each row, presented for your engineers to correct rather than asserted as fact. Tags that cannot be resolved are listed as unresolved and excluded from anything downstream, because a tag attached to the wrong asset is worse than a tag attached to nothing.

What the resolved history is for

Once readings are attached to equipment, a decade of instrumentation can be read against a decade of failures, and that is the condition everything else depends on. Resolved readings feed the reliability models, moving remaining-useful-life estimates from history alone to history conditioned on the equipment's current observed state. They feed the asset tree the same way, so a component that is degrading can be evaluated for what it threatens downstream rather than only for its own condition. Condition data is worth most where it changes a decision already being made.

One event instead of a dozen alarms

The same resolution repairs alarm handling as a side effect. Thresholds are configured per tag, in isolation, so one physical event trips a dozen related tags and produces a dozen alarms describing a single problem. Operators learn to discount them, which is a rational response to a system that cries wolf and a dangerous one when the wolf arrives. Correlated deviations across tags belonging to the same subsystem collapse into one event, described in terms of the asset rather than the instrument.

TI-4021bearing temp highPI-4030oil pressure lowVI-4101vibration highFI-4044oil flow lowOne eventlube oil starvation, P-101Named cause,ranked by consequence
Figure 1Four alarms, one failure. Raised per tag and in isolation they are four tag numbers and no diagnosis, which is what teaches operators to discount them. Collapsed onto the equipment they measure, they are a named cause that can be ranked against everything else waiting.

Drift is the failure nobody sees

A dead sensor is obvious. A drifting one is not: it keeps producing plausible numbers that are quietly wrong, and every model, report and decision downstream inherits the error without any signal that something has gone bad. Contour monitors instruments against their own history and against physically related instruments, so divergence surfaces as a data quality event before it propagates into conclusions.

Common questions

What is sensor drift?

The gradual divergence of an instrument's readings from the true value it measures, dangerous because the output continues to look valid.

Does this replace our historian or SCADA?

No. Contour reads from the systems you already run and adds structure on top of them.

What sampling rates are supported?

From sub-second process instrumentation to daily manual readings. Rates are reconciled against the asset model rather than requiring a common frequency.

Related: Model context layer · Failure propagation · Reliability engineering · Use cases