The refinery asset is running, yet calculated effectiveness remains low or changes sharply between reporting periods. The number that matters is useful, specification-compliant output during scheduled production time—not motor run status, nameplate capacity, or total calendar output. OEE combines three losses: time unavailable, output rate below the defined reference rate, and production that fails the quality criterion.
Wrong fixes and misleading shortcuts
This is heat, flow, pressure, mechanical load, and time—not a single PLC run bit. Several common shortcuts conceal the physical loss instead of measuring it.
| Attempted fix | Why it fails | Better basis |
|---|---|---|
| Use equipment running hours as OEE | Running hours measure state or utilization. They omit reduced throughput and off-spec output. | Calculate availability, performance, and quality separately. |
| Divide actual output by nameplate capacity | Nameplate capacity may not represent the approved rate for the current feed, product, temperature, pressure, or equipment configuration. | Define a documented reference rate for the operating mode. |
| Count every calendar hour as planned production | This mixes commercial scheduling, maintenance policy, and equipment loss into one denominator. | Freeze a planned-production-time rule before comparing assets. |
| Give every asset its own production total | A pump, exchanger, heater, compressor, and turbine may support one process stream. Crediting the same output to every machine creates duplicate production claims. | Assign process output at the constraint or custody boundary; use asset-level service units elsewhere. |
| Average component OEE values | An arithmetic average ignores series dependence and the process bottleneck. | Calculate process OEE from process-level time, rate, and quality records. |
OEE terms for a continuous refinery process
Use a consistent calculation boundary and reporting period:
OEE = Availability × Performance × Quality
Availability = Run Time / Planned Production Time
Run Time = Planned Production Time − Counted Downtime
Performance = Actual Throughput / Reference Throughput During Run Time
Quality = Accepted Output / Total Output
Keep all ratios dimensionless and express throughput on the same mass, volume, energy, or service basis. If volume changes materially with temperature or pressure, use the plant-approved compensated quantity or a mass basis. Cap neither unexplained over-performance nor bad data silently. A performance result above unity usually points to a conservative reference rate, mismatched units, compensation differences, inventory movement, or timestamp misalignment.
OEE and utilization answer different questions. OEE evaluates effectiveness during planned production; utilization compares operating time or output with a broader calendar opportunity. Report both when management needs commercial and equipment views, but never substitute one for the other.
Asset boundaries and physical service units
A process-level OEE normally belongs at a production boundary where planned time, throughput, and accepted product can be reconciled. Asset-level OEE works only when the asset has a measurable service output and a defensible reference rate.
| Asset | Useful service quantity | Rate-loss indicators | Where to read the basis |
|---|---|---|---|
| Pump | Qualified liquid flow during required service | Flow, differential pressure, speed, recirculation, valve position, motor load | Approved operating envelope, flow measurement, historian, and maintenance record |
| Heat exchanger | Process duty or qualified stream throughput | Inlet and outlet temperatures, flow, pressure drop, bypass position | Heat balance, instrument records, and approved performance basis |
| Crude oil heater | Qualified heated-stream throughput or useful heat duty | Coil inlet and outlet conditions, firing demand, fuel use, draft, process constraint | Process historian, heat balance, and operating envelope |
| Compressor | Qualified gas flow delivered at required suction and discharge conditions | Corrected flow if approved, pressure ratio, recycle position, speed, driver load | Performance map, approved calculation method, and historian |
| Turbine | Useful shaft power or driven-equipment service | Load, speed, inlet and exhaust conditions, control-valve position | Driver and driven-equipment records, performance basis, and historian |
An exchanger has no independent production count unless its duty is the declared service output. A turbine driving a compressor should not receive a second copy of the compressor's product volume. Measure the turbine by delivered driver service, or treat it as a loss contributor inside the compressor train boundary.
Calculation procedure
- Declare the boundary. State whether the record covers one asset, a train, or the complete process. Identify the inlet, outlet, and quality measurement points.
- Define planned production time. Document which scheduled outages, turnaround periods, feed shortages, and downstream constraints are excluded or retained. Apply the same rule to every comparison period.
- Build the state model. Convert timestamps into planned, running, counted downtime, and excluded time. Use process evidence such as valid flow and operating conditions with the run command; a run bit alone can count idle circulation as production.
- Select the reference rate. Use the approved achievable rate for the defined feed, product, and configuration. Record its units, operating-mode key, effective date, and approval source.
- Calculate availability. Subtract only downtime falling inside planned production time. Split events at reporting boundaries so minutes are neither lost nor duplicated.
- Calculate performance. Integrate actual qualified throughput over run time and compare it with the reference throughput for those same intervals. For changing operating modes, calculate expected output interval by interval before summing.
- Calculate quality. Count accepted output at the chosen boundary. Treat off-spec production, downgraded product, reprocessing, and inventory transfer according to one written disposition rule.
- Multiply the factors. Preserve the unrounded component values for calculation and round only displayed results.
Practical symbolic example
For a compressor train, let P be planned production time, D counted downtime, M measured qualified gas delivered during run time, R the approved reference flow rate, AQ accepted delivered quantity, and TQ total delivered quantity.
Run Time = P − D
Availability = (P − D) / P
Performance = M / [R × (P − D)]
Quality = AQ / TQ
OEE = [(P − D) / P] × [M / (R × (P − D))] × [AQ / TQ]
The same structure can serve a pump, heater, exchanger, or turbine after replacing M and R with the asset's qualified service quantity and reference service rate. A heat exchanger may require calculated duty from measured flow and inlet/outlet state rather than a raw production counter. Use the site's approved property calculation when density, heat capacity, or gas condition correction affects the result.
Verification and recurring data traps
Reconcile each factor before publishing the product. Planned time must equal run time plus counted downtime. Integrated flow should agree with independent production or inventory accounting within the site's accepted reconciliation tolerance. Accepted plus rejected or reprocessed output must reproduce total output under the declared quality rule.
Trend the three factors beside raw quantities. A sudden availability change with unchanged event history points to state classification or timestamp trouble. A performance jump without a physical change points to the reference rate, units, compensation, meter scaling, or duplicated integration. A quality change with stable laboratory results points to delayed sample timestamps or incorrect product attribution.
Check daylight-saving transitions, historian gaps, bad-quality tags, counter rollover, maintenance overrides, zero-flow circulation, parallel-train allocation, and events crossing shift or month boundaries. Keep the raw timestamps, quantities, state classifications, reference-rate version, and quality disposition so another engineer can reproduce the result.
Frequently asked questions
What happens if an asset runs with zero production flow?
The interval remains run time only if the documented state model classifies it that way, but it contributes no actual throughput and lowers performance. If zero-flow operation represents a failure to produce, classify the event consistently as downtime instead of switching rules between periods.
What happens if measured throughput exceeds the reference rate?
Investigate the reference-rate version, unit conversion, flow compensation, timestamp alignment, and meter scaling. Retain the raw result until the cause is identified; automatic capping hides a configuration or measurement problem.
What happens if off-spec product is reprocessed later?
Apply the written quality-disposition rule at the original process boundary. Avoid crediting the same material once as reprocessed input and again as new accepted output.
What happens if the pump, heater, and compressor share one process rate?
Calculate process OEE once at the common production boundary. Use asset availability, service-rate loss, condition indicators, and downtime contribution to identify which component constrained the process.
What happens if historian data cannot be reconciled?
Stop publishing the OEE value when missing intervals, conflicting clocks, invalid meter data, or an undocumented reference rate can change the engineering decision. Preserve the raw records and calculation version, then escalate through the equipment or control-system manufacturer's official support channel when diagnostics indicate a product-specific measurement, firmware, or calculation fault.