How Do You Standardise Engineering Units in Plant Data?
To standardise engineering units in plant data, give every measurement a defined unit, a clear meaning and an explicit rule for converting its source value. Separate raw signal scaling from later unit conversion, and assign an owner to each step. Verify the instrument, controller, historian and dashboard against the same agreed examples before accepting the data.
What should you define for each measurement?
Start with a measurement register. Treat it as the agreement that every system handling the value must follow.
Do not rely on a tag name alone. Require enough information to explain what the number measures, where it came from and how it should be interpreted.
| Field | What to record |
|---|---|
| Measurement identity | A stable identifier and the equipment or process location |
| Quantity | Pressure, temperature, flow, level or another defined quantity |
| Measurement basis | Any required reference, such as gauge or absolute pressure |
| Source representation | Raw input, signal value or an already scaled engineering value |
| Source unit and range | The unit and configured endpoints at the source |
| Stored unit | The agreed engineering unit used for downstream calculations |
| Scaling and conversion rules | Equations, input units, output units and applicable conditions |
| Validity | Quality state, timestamp and agreed handling of missing data |
| Rule version | Which configuration produced the stored value |
For flow measurements, require the register to distinguish mass flow from volume flow. Where a reference condition matters, record it explicitly. Leave no reader to infer these details from a shortened label.
How should raw signal scaling work?
Define scaling as the step that turns the incoming signal representation into an engineering value.
Before writing a formula, inspect what the receiving system actually gets. Ask whether the source sends a raw count, an electrical signal value or an engineering value. Record the answer and verify it against the configuration.
For a linear mapping, specify both endpoints and the intended behaviour outside them. Use a different, documented rule where the instrument requires one.
Illustrative worked example
Assume an input delivers a signal value from 4 to 20 mA, and the agreed instrument range is 0 to 1,000 kPa. For this example, define:
Pressure in kPa = (signal in mA − 4) × 1,000 ÷ 16
The expected checks are:
| Illustrative input | Expected scaled value |
|---|---|
| 4 mA | 0 kPa |
| 12 mA | 500 kPa |
| 20 mA | 1,000 kPa |
These are test values for the stated mapping, not a default range for other instruments.
Also specify what happens below and above the configured signal range. Decide whether the system retains the calculated value with a quality flag, rejects it or follows a documented instrument fault rule. Do not silently replace an invalid reading with zero.
Where should unit conversion happen?
Choose one accountable conversion point for each data path.
A practical arrangement is to scale the source into an agreed engineering unit, store that value with its unit metadata, and let displays convert it only when required. If your existing system uses another arrangement, document it with the same precision.
For every conversion, write down:
- The input quantity, unit and measurement basis.
- The output quantity, unit and measurement basis.
- The equation and any required parameters.
- The component responsible for applying it.
- The behaviour when metadata or parameters are missing.
Keep scaling and unit conversion as separate named steps, even when the same component performs both.
Continuing the illustrative example, suppose the agreed display rule is bar = kPa ÷ 100. The stored value of 500 kPa should then appear as 5 bar. Check that the dashboard receives 500 kPa before applying the rule; do not assume the incoming value still needs conversion.
Do not approve a conversion based only on compatible-looking labels. For a proposed mass-to-volume calculation, require the density input, its unit and its applicable conditions to be specified. Treat that as a defined calculation with additional inputs.
What should the historian retain?
Require each stored engineering value to remain interpretable after configuration changes.
Keep the timestamp, quality state, measurement identity and applicable unit definition associated with the reading. Where metadata is stored separately, require an effective period or version that identifies which definition applied.
Retain the source value where it is available and useful for investigation. Give source and converted values distinct identities so nobody mistakes one for the other.
Set an explicit policy for range or unit changes. Either preserve historical values under their original definitions or perform a controlled conversion with a traceable record. Do not change a unit label and assume the earlier records now share the new meaning.
Include downloaded data in this policy. Require exports to identify the unit in a column heading or a separate field, with the measurement basis available where needed.
Which checks catch mismatches across systems?
Use expected values agreed before commissioning. Test the whole path, then inspect each boundary to locate any disagreement.
| Check | Evidence to collect |
|---|---|
| Source configuration | Instrument range, unit and measurement basis |
| Scaling | Results at the lower endpoint, midpoint and upper endpoint |
| Interface metadata | Value, unit, quality and timestamp at each handover |
| Historian storage | Stored value and the definition that applies to it |
| Dashboard display | Displayed value, unit label, axis and selected conversion |
| Failure behaviour | Results for missing units, bad quality and out-of-range input |
| Historical behaviour | Readings from before and after a controlled configuration change |
For the illustrative midpoint, the acceptance record should show 12 mA at the input, 500 kPa after scaling, 500 kPa in storage and 5 bar on the selected display.
Test a mismatch deliberately. Send a value with an unexpected unit and confirm that the receiving system follows the agreed rejection or exception behaviour. A check should be capable of rejecting incorrect input.
Where you compare two instruments, first agree whether they measure the same quantity under comparable conditions. Define the expected tolerance with the responsible engineer.
What should dashboards and calculations display?
Put the unit beside the value and on chart axes. Make the measurement basis visible where it affects interpretation.
Require calculations to accept defined units, rather than unlabelled numbers. Before combining readings from several sources, check that their quantity, basis and time meaning match the calculation specification.
Keep full working precision through the agreed calculation path and define rounding at the output. Test the rounded display separately from the stored value.
Apply the same review to report totals and production indicators. Our guide to what an OEE number can hide provides related questions for reviewing a headline measure.
What common mistakes should you prevent?
Changing the label without changing the value. Review the displayed number and its conversion rule together.
Applying the same conversion twice. Assign ownership and record the value at every handover.
Using a familiar range without checking the instrument. Confirm the installed configuration before approving scaling.
Treating missing metadata as permission to guess. Route the reading to an exception process with a named owner.
Using one formula for unlike measurements. Require the quantity, basis and conditions to match the rule.
Testing only the live operating point. Include endpoints, invalid inputs and configuration changes in acceptance checks.
For control, alarm or protective functions, agree the scope with the responsible engineers. Any safety-related design should be reviewed with a qualified assessor.
How do you introduce this into an existing plant?
Start with one data path and give it clear acceptance criteria. Inventory its measurements, confirm the source configurations and agree the stored units before changing mappings.
Then verify the path through storage, dashboards and exports. Review alarms and dependent calculations before switching the approved configuration into use.
Record conversion ownership and acceptance examples in the project specification. The guide to what a functional design specification should contain can help structure that agreement.
Complete the handover with the measurement register, verified examples and a process for approving later instrument or range changes.
Related reading: measuring energy use by production line and how to detect sensor drift in plant data.
Frequently asked questions
Must every dashboard use the same unit?
Choose a common stored unit for shared calculations. Allow another display unit where users need it, provided the conversion and label are explicit and tested.
Is a unit in the tag name enough?
Require structured unit metadata as well. Use the name as a helpful label, with the register defining the quantity, basis and conversion rule.
Should raw readings be deleted after conversion?
Set a retention policy based on investigation and operational needs. Where raw readings are retained, keep them clearly separate from engineering values.
What should happen when a unit is unknown?
Flag the reading as unsuitable for the affected calculation or display. Resolve its meaning at the source before adding an approved mapping. Unit checks are one step in cleaning plant data before building dashboards.