- October 20, 2026
- Posted by: editor
- Category: RECD & Emission Control Devices
Last Monday, the differential pressure across your Retrofit Emission Control Device read 21 mbar. On Thursday it read 48. The following week it read 31, then 64, then 21 again. Is the filter loading up, clearing itself, or behaving erratically?
You can’t tell from these numbers alone, because each was taken at a different generator load. Differential pressure across an RECD depends heavily on how much exhaust is flowing through it, and exhaust flow depends on load. Comparing raw readings from different loads is like comparing a person’s heart rate while sitting and while running, and concluding they have a heart problem.
This guide explains why pressure readings change with load, and gives three practical methods for comparing them fairly. The main one is a load-based baseline and ΔP index that turns scattered readings into a clear trend.
Why Pressure Changes With Load
Exhaust flow drives pressure drop
Pressure drop across a filter depends on how much gas is pushed through it. As load rises:
- The engine burns more fuel and, especially in turbocharged engines, moves more air
- Exhaust temperature rises, so the gas expands and its volume flow increases further
- Higher volume flow through the same filter means higher pressure drop
Pressure drop isn’t simply proportional to flow
For a wall-flow filter, part of the pressure drop comes from gas passing slowly through the porous walls and soot cake, which rises roughly in proportion to flow. Another part comes from gas entering, leaving and travelling along the channels, which rises faster, roughly with the square of flow. Total ΔP therefore rises more than proportionally with flow.
The practical consequence is that you can’t simply divide ΔP by load to normalise it. A filter in the same condition can show a ΔP at full load several times the value at light load.
Soot and ash add to it
On top of the flow effect, ΔP rises as soot and ash accumulate. That’s the signal you want to track. See can RECD efficiency change between new and aged conditions. The challenge is separating the filter condition signal from the load effect.
Engine back-pressure behaves the same way
Engine exhaust back-pressure includes the RECD’s ΔP plus losses in the rest of the exhaust system (pipework, bends, silencer and stack). It also rises strongly with load. See how exhaust back-pressure affects a DG set.
Method 1: Standard Check Condition
The simplest approach is to compare readings only when taken under the same conditions:
- The same load, for example 50% of rated kW (±5%)
- Fully warm, for example at least 45 minutes into a run
- No regeneration active
Readings taken at that condition can be compared directly over weeks and months.
Strengths: simple; no calculation needed.
Weaknesses: you only get a usable reading when the DG happens to run at that load for long enough. At sites where the DG mostly runs at light load during short outages, the standard condition may rarely occur. A load bank during monthly test runs can create it deliberately.
This method suits sites with manual readings and regular test runs. See why RECD test results may differ from day-to-day site readings.
Method 2: Load-Based Baseline Curve and ΔP Index
This method uses every valid reading, at any load, by comparing it with what a clean, healthy device would show at that same load.
Step 1: Build the baseline curve
Soon after commissioning, or after a cleaning, collect ΔP readings across the range of loads the set normally sees. Use only readings that meet these conditions:
- DG fully warm
- Load steady for at least a few minutes
- No regeneration active or recently completed
- Sensor checked and in calibration. See installed RECD sensor calibration: what should facility teams document
Group them by load band and take the typical value for each band. A simple baseline table might look like this (illustrative figures for a 250 kVA set):
| Load (% of rated kW) | Baseline ΔP (mbar) |
|---|---|
| 25% | 18 |
| 50% | 32 |
| 75% | 49 |
| 100% | 70 |
Acceptance-test readings at known loads are a good starting point. See how to set up a witnessed RECD performance demonstration.
Step 2: Find the baseline at any load
For a reading taken between two baseline points, estimate the baseline by linear interpolation:
Baseline at load L = Baseline at lower point + (L − lower load) ÷ (upper load − lower load) × (Baseline at upper point − Baseline at lower point)
For example, at 60% load:
Baseline = 32 + (60 − 50) ÷ (75 − 50) × (49 − 32) = 32 + 0.4 × 17 = 38.8 mbar
Linear interpolation between points 25% apart is accurate enough for trend tracking. More baseline points improve it.
Step 3: Calculate the ΔP index
ΔP index = Measured ΔP ÷ Baseline ΔP at the same load
- An index of 1.00 means the filter behaves exactly as it did when the baseline was set
- An index of 1.20 means ΔP is 20% higher than baseline at that load, typically from soot or ash build-up
- An index below baseline, especially a sudden drop, may indicate a crack, seal failure or bypass, or a sensor fault. See how to separate a sensor fault from an RECD performance problem
Worked example: six weeks of readings
The figures are illustrative. Weekly readings were taken at whatever load the DG was running:
| Week | Load | Raw ΔP (mbar) | Baseline at that load | ΔP index |
|---|---|---|---|---|
| 1 | 30% | 21.2 | 20.8 | 1.02 |
| 2 | 70% | 47.9 | 45.6 | 1.05 |
| 3 | 45% | 31.5 | 29.2 | 1.08 |
| 4 | 85% | 64.3 | 57.4 | 1.12 |
| 5 | 25% | 20.7 | 18.0 | 1.15 |
| 6 | 60% | 46.2 | 38.8 | 1.19 |
The raw readings jump around between 21 and 64 mbar, with no visible pattern. The ΔP index rises steadily from 1.02 to 1.19, a clear sign of progressive soot loading, perhaps because light-load running isn’t allowing enough passive regeneration. See choosing an RECD for a generator running at low load.
Without normalisation, nobody would have spotted the trend until a high ΔP alarm triggered. With it, the maintenance team had several weeks of warning and could plan a response calmly.
Step 4: Set index-based thresholds
Once you work with an index, you can set alarms and maintenance triggers in index terms, which apply at any load. For example (illustrative; agree actual values with your supplier):
| ΔP index | Interpretation | Action |
|---|---|---|
| 0.85 – 1.25 | Normal range | Monitor |
| 1.25 – 1.50 | Elevated soot or ash | Plan regeneration or cleaning; check duty cycle |
| > 1.50 | High loading | Investigate promptly; follow alarm matrix |
| < 0.85 (sudden) | Possible crack, seal failure or sensor fault | Inspect device and sensor |
Some RECD controllers calculate a load-compensated ΔP or index automatically. Where yours doesn’t, the BMS or a spreadsheet can. See integrating RECD monitoring with a building management system.
Step 5: Update the baseline after cleaning or major changes
After ash cleaning, substrate replacement or significant engine or exhaust changes, the “clean” behaviour of the system may change. Rebuild the baseline curve, keep the old one for reference, and mark the change in the event log. See what should an RECD event log record and when should an RECD be retested after changes to the DG set.
Using the index to plan cleaning
Because the index trend is smooth, it can be used to forecast when a maintenance threshold will be reached. In the worked example, the index rose from 1.02 to 1.19 over five weeks, about 0.034 per week. If the planning threshold is 1.50:
Weeks to threshold ≈ (1.50 − 1.19) ÷ 0.034 ≈ 9 weeks
That gives time to schedule a regeneration support run, such as a higher-load test with a load bank, or an ash cleaning, without waiting for an alarm. Recalculate the forecast as new readings arrive, because the rate can change with duty cycle and season.
Building a baseline when the DG mostly runs at one load
Some sets run at almost the same load every time, for example a site with a fixed process load. Then you may only collect baseline points over a narrow range. That’s fine for tracking, because nearly all readings will be at that load too. Use Method 1 (a standard condition) or a two-point baseline around the usual load. Run a load bank test occasionally to extend the baseline if you need to understand behaviour at higher loads.
Combining ΔP and Back-Pressure Indices for Diagnosis
Calculating an index for both RECD ΔP and engine back-pressure, each against its own load-based baseline, helps locate where a restriction is:
| RECD ΔP index | Back-pressure index | Likely meaning |
|---|---|---|
| Rising | Rising similarly | Restriction is in the RECD (soot or ash build-up) |
| Steady | Rising | Restriction is elsewhere in the exhaust, such as a silencer, stack, damaged pipe or rain cap |
| Falling suddenly | Falling | Possible crack, seal failure or bypass in the RECD |
| Falling suddenly | Steady | Possible ΔP sensor or impulse line problem |
| Rising | Steady | Possible ΔP sensor drift or impulse line issue. Verify before acting |
This comparison is one of the quickest ways to tell device problems from sensor problems and from problems elsewhere in the exhaust. See how to separate a sensor fault from an RECD performance problem.
Method 3: Flow-Based Normalisation (Advanced)
If exhaust flow can be estimated, for example from engine data, intake air flow or fuel flow together with exhaust temperature, ΔP can be normalised against flow rather than load. A simple model fitted to clean-filter data might take the form:
Expected ΔP = a × Q + b × Q²
where Q is the volume flow of exhaust at the filter, and a and b are constants found from baseline data. The measured ΔP is then compared with the model’s prediction.
This approach handles changes in exhaust temperature and ambient conditions better than a load-only baseline. It needs reliable flow estimates and some data analysis skill, so it’s usually done by RECD suppliers or specialist monitoring platforms. For most facility teams, Method 2 gives most of the benefit with much less effort. For background on exhaust flow data, see using OEM exhaust flow data when specifying an RECD.
Comparing Back-Pressure With the Engine Limit
Engine makers specify a maximum exhaust back-pressure, usually at rated output. If your set rarely runs at full load, you may never see back-pressure close to that limit, even if the RECD is quite heavily loaded.
Be careful about extrapolating a light-load reading to full load. Because pressure rises more than proportionally with flow, and exhaust flow doesn’t rise in a simple fixed ratio with load, simple scaling can give misleading answers. Better options:
- Test at high load periodically with a load bank, and measure back-pressure directly
- Track a back-pressure index against a load-based baseline, in the same way as ΔP. A rising index warns that full-load back-pressure will also be higher
- Ask your supplier for the expected back-pressure at rated load for a given ΔP index
A set that suddenly has to carry full load during a long grid outage, with an RECD that has loaded up during months of light running, is exactly the situation where back-pressure problems appear.
Things That Distort Load Comparisons
Even with an index, check for these before drawing conclusions:
| Factor | Effect | How to handle |
|---|---|---|
| Regeneration active or just finished | ΔP lower than normal for that load | Exclude these readings |
| Cold engine or device | Different gas temperature and flow | Use only fully warm readings. See how to account for warm-up time when testing an RECD |
| Rapidly changing load | Pressure lags load changes | Use readings after load has been steady for several minutes |
| Very light load | ΔP small; sensor error becomes a large share | Treat light-load indices with caution; prefer readings above ~25% load |
| Hot vs cool ambient | Slight changes in gas temperature and flow | Usually minor; note season if comparing across the year |
| Sensor drift or impulse line problem | False trend | Check sensor calibration and impulse lines |
| Load measured in kVA instead of kW | Wrong load used for baseline lookup | Use kW. See why generator load must be matched in RECD comparison tests |
| Clock offset between DG and RECD data | Readings paired with wrong load | Align timestamps. See how to align RECD sensor timestamps with generator runtime logs |
Applying the Method With Manual Readings
You don’t need a BMS to use a ΔP index. A paper log or simple spreadsheet works:
- Keep the baseline table on the log sheet or in the spreadsheet
- At each reading, record date, time, kW load, minutes since start, ΔP, back-pressure and regeneration status
- Look up or interpolate the baseline for that load
- Calculate the index
- Plot the index over time
A spreadsheet can do steps 3 to 5 automatically once the baseline table is entered.
A spreadsheet layout that works
A simple layout for a ΔP index spreadsheet:
| Column | Content |
|---|---|
| A | Date and time (ISO format, IST) |
| B | Load (% of rated kW) |
| C | Minutes since DG start |
| D | Regeneration status (Y/N) |
| E | Measured ΔP (mbar) |
| F | Baseline ΔP at that load (looked up and interpolated from the baseline table) |
| G | ΔP index (E ÷ F) |
| H | Valid? (Y if C ≥ 45 and D = N, otherwise N) |
| I | Notes (cleaning, sensor check, engine work) |
Plot column G against column A, using only rows where column H is “Y”. Add a horizontal line at each threshold, and mark cleaning or regeneration support runs as events. The same layout works for back-pressure, with its own baseline table.
Showing the Index on a Dashboard
On a dashboard, the ΔP index works best as:
- A single number with colour bands on the operating view, such as “ΔP index 1.12 (normal)”
- A long-term trend on the maintenance view, with cleaning events marked
- A monthly value on the management view
See what should a useful RECD operating dashboard show.
Related Reading
- Installed RECD sensor calibration: what should facility teams document?
- Pressure sensors vs pressure switches in RECD monitoring
- RECD temperature sensor placement: what should be measured?
- How to separate a sensor fault from an RECD performance problem
- DG load logs to assess RECD suitability
Frequently Asked Questions
Can I just divide ΔP by load to compare readings?
No. ΔP rises more than proportionally with exhaust flow, and exhaust flow doesn’t rise in a simple fixed ratio with load. Dividing by load over-corrects at high load and under-corrects at light load. A baseline curve built from your own clean-filter data is far more reliable.
How many baseline points do I need?
Three or four load points covering the normal operating range, such as 25%, 50%, 75% and 100%, are enough for most sites. More points improve accuracy, especially if the set operates over a wide load range.
What if my DG never runs above 40% load?
Build the baseline over the range you actually use, and track the index within it. To understand full-load back-pressure, test periodically at higher load with a load bank. Don’t extrapolate light-load readings to full load.
Does the ΔP index replace alarm limits?
It complements them. Absolute limits for back-pressure, based on engine protection, should always remain. The index gives earlier, load-independent warning of filter loading and helps plan maintenance.
My RECD controller already shows ΔP. Why calculate an index too?
The controller’s raw ΔP is correct, but it can’t be compared across loads. The index can. If your controller already offers load-compensated ΔP, that serves the same purpose, so check with your supplier how it’s calculated before duplicating it in a spreadsheet.
Does ambient temperature affect the comparison?
Slightly. Hotter intake air and exhaust change gas density and volume flow, so ΔP at the same load can differ a little between seasons. For most sites, the effect is small compared with soot and ash loading. If you see a seasonal pattern in the index, compare like with like, for example this summer against last summer.
Can I compare ΔP indices between two different DG sets?
Only roughly. Each set has its own engine, exhaust layout and RECD size, so each needs its own baseline. Indices are comparable in the sense that 1.30 means “30% above its own clean baseline” for both, which is useful for prioritising maintenance across a fleet.
When should I rebuild the baseline?
After ash cleaning, substrate replacement, significant exhaust modifications or engine changes that affect exhaust flow, and whenever readings suggest the old baseline no longer represents a clean device.
Turn Scattered Readings Into a Clear Trend
Raw pressure readings at different loads hide the very trend you’re trying to see. A simple baseline curve and ΔP index make filter loading visible weeks earlier. Aceget supplies CPCB type-approved Retrofit Emission Control Devices for DG sets from 25 kVA to 2500 kVA, and we can help you build baseline curves and index thresholds for your installation. See our certifications, or contact our team.
[…] load should be in the same range, allowing for filter loading since the baseline. Our guide on comparing RECD pressure readings at different generator loads explains why load matching […]