HVAC
Plate heat exchanger fouling diagnosis: a field guide
Separate deposits from bypass and changing load with four temperatures, two-side pressure trends, clean-model normalization and measurable cleaning acceptance.
Direct answer
Diagnose plate heat exchanger fouling by recording all four connection temperatures, measuring flow and pressure drop on each side, and comparing thermal and hydraulic performance with a clean baseline at the same operating conditions. Xylem's Plateflow manual, revision B of September 2018, identifies reduced temperature performance together with increased pressure drop as evidence of fouling or clogging. Measure at exchanger connections and test bypass flow before ordering cleaning.
On this page
How do you diagnose plate heat exchanger fouling?What must you measure before changing the plant?Does the four-temperature heat balance close?What does approach temperature actually tell you?How do you normalize approach to a clean baseline?What do pressure drops on both sides reveal?Is it scale, debris, low flow, air or incorrect assembly?How do you prove bypass instead of fouling?Should you backflush, use CIP or open the plate pack?What counts as successful cleaning?How often should you trend and clean the exchanger?What should the diagnosis and cleaning scope cost?Worked example: normalized performance and cleaning acceptanceOpen the related tool for the calculation, checklist, or packet.
How do you diagnose plate heat exchanger fouling?
Diagnose fouling from four connection temperatures, both branch flows and both pressure drops compared with clean performance at the same conditions. Xylem's Plateflow manual, revision B of September 2018, identifies declining temperature performance with rising pressure drop as evidence of fouling or clogging. Test bypass before ordering cleaning: mixed header temperatures can make a clean exchanger appear fouled.[2]
Separate three boundaries: exchanger core, external piping with strainers and valves, and the bypass mixing junction. Pump head includes other losses; it is not exchanger pressure drop. Retrieve the chiller plant commissioning record before treating today's readings as deterioration.
Alfa Laval's undated troubleshooting guide distinguishes gradual deterioration, abrupt changes and cyclic problems: investigate deposits and chemistry for gradual decline, debris or operating changes for sudden decline, and controls or seasonal conditions for repeating decline.[1]
This method covers single-phase liquid-to-liquid exchangers. Refrigerant evaporators and condensers need saturation and enthalpy procedures. Equipment instructions govern physical work; the selection predicts performance; the project sets acceptance.
What must you measure before changing the plant?
Measure the exchanger branches before changing valves, pump speeds or treatment. Label hot and cold by inlet temperature: in cooling service, the load-return loop is the thermodynamic hot stream. Photograph the measurement boundaries.
Place insulated temperature sensors between each exchanger connection and any mixing junction. Danfoss's operating guide AQ13708645598701-000503, printed 2020.02, locates the outlet sensor close to the exchanger; Alfa Laval's Performance Audit uses four temperature gauges and flow logging.[5][8] Independently verify exchanger branch flow; deriving flow from a clean core pressure curve assumes away the restriction being investigated.
| Item | Required location/value | Clean example input |
|---|---|---|
| Hot inlet temperature | Before hot connection, after upstream mixing | 30°C |
| Hot outlet temperature | At exchanger outlet before bypass junction | 24°C |
| Cold inlet temperature | At exchanger inlet | 20°C |
| Cold outlet temperature | At exchanger outlet before downstream mixing | 26°C |
| Hot and cold mass flow | Actual exchanger branch on each side | 10 kg/s each |
| Fluid/property basis | Product, concentration, mean temperature | Water; cp 4.18 kJ/(kg·K) |
| Pressure difference | Inlet/outlet taps for each side | Hot 40 kPa; cold 30 kPa |
| External strainer/valve loss | Separate taps outside core boundary | Record independently, excluded from 40/30 kPa |
| Controls and configuration | Mode, pump speed, valve travel, plate/pass identity | One-pass counterflow; bypass shut |
| Instrument and timestamp record | Model, serial, calibration, sample interval | 1-minute logging; 15-minute stable window |
The clean values are calculation inputs. Preserve the plant's specified minimum flows, freeze protection, relief paths and redundancy throughout testing.
How stable and accurate must the readings be?
Anvilfield recommends synchronized one-minute samples over a 15-minute window, no operating-mode or valve change, and each exchanger flow within ±5% of its window mean. Temperature drift must be smaller than the effect being diagnosed. These are field recommendations, not laboratory rating requirements.
Specify uncertainty at the installed sensor and meter, including mounting, insulation, pipe conduction, fluid properties and meter installation. A bath comparison reveals mismatch, not installed accuracy. Repeated samples reduce random scatter; they do not remove calibration or mounting bias.
NIST Technical Note 1297, 1994 edition, section 5 combines standard uncertainties with covariance; section 6 expresses expanded uncertainty using a stated coverage factor. Section 4 explains converting calibration intervals and tolerance bounds to standard uncertainties using their coverage factors or assumed distributions.[14][11][12]
Anvilfield estimates that independent standard uncertainties of 0.1°C per temperature channel and 2% per flow reading give 0.14°C uncertainty in a temperature difference and 4.06% duty uncertainty at a 4°C stream change, with fluid properties fixed. At a 1°C approach, the temperature uncertainty alone is 14.14% of the signal. Improve installation or test at a permitted condition with a larger signal.
Does the four-temperature heat balance close?
Calculate duty on both sides before interpreting thermal deterioration. Multiply each stream's mass flow by its specific heat and its inlet-to-outlet temperature change, using positive changes for heat released and absorbed. This stream delta-T differs from the terminal approach between two streams.
For hot inlet/outlet Thi/Tho and cold inlet/outlet Tci/Tco: Qh = mh × cph × (Thi − Tho); Qc = mc × cpc × (Tco − Tci). Imbalance (%) = 100 × |Qh − Qc| / [(Qh + Qc)/2]. Mass flow is kg/s, specific heat is kJ/(kg·K), and duty is kW; multiply volumetric flow by the actual fluid density first.
ASHRAE's 2024 HVAC Systems and Equipment Handbook, chapter 47, provides the stream heat-balance framework; use fluid-specific properties or enthalpy differences when constant specific heat is inadequate.[6] The water calculations here use 4.18 kJ/(kg·K).
Anvilfield recommends investigating duty imbalance above 10% before interpreting UA, with a tighter screen where calibrated measurement supports it. At equal 4°C stream changes, the stated independent uncertainties give about 5.74% standard uncertainty in the signed duty difference relative to mean duty. The 10% screen is therefore an investigation trigger, not a 95% confidence test.
A failed balance directs attention to wrong sensor boundaries, flow-meter setup, fluid properties, thermal storage or unaccounted heat loss. Use the mean duty only after balance validation and when both sides have comparable uncertainty, as in the worked case. With unequal measurement quality, select the better-supported side or document an uncertainty-weighted reconciliation. A passing balance establishes consistency; it does not exclude bypass.
What does approach temperature actually tell you?
Report both terminal temperature differences and name the one serving the plant setpoint. For counterflow cooling duty, hot outlet minus cold inlet is the usual leaving-load approach; the opposite gap is hot inlet minus cold outlet. Unequal heat-capacity rates can make the opposite terminal smaller.
A cold outlet warmer than the hot outlet is possible in counterflow: those outlets sit at opposite ends. That temperature crossover differs from a nonpositive terminal gap, which requires checking the readings and arrangement before calculating a logarithm.
The logarithmic mean temperature difference, LMTD, represents the effective thermal driving difference along the exchanger. Divide duty by that driving difference to calculate apparent conductance, UA (overall heat-transfer coefficient times area), for the single-pass counterflow arrangement used here.
Let a = Thi − Tco and b = Tho − Tci. LMTD = (a − b)/ln(a/b); when a = b, use their common value. Apparent UA = Q/LMTD, in kW/K. Both terminal gaps must be positive for this calculation.
ASHRAE 2024, chapter 47, distinguishes arrangement effects in thermal analysis.[6] For a multipass unit, use its actual correction or manufacturer model rather than applying the single-pass equation unchanged. Verify the pass drawing before trusting a calculated UA.
How do you normalize approach to a clean baseline?
Normalize approach by predicting the clean outlet temperatures at today's inlet conditions, actual flows and fluid properties, then subtracting the predicted approach from the measured connection approach. This approach residual, in K or °C of temperature difference, is zero at the clean prediction and positive for warmer leaving-load delivery.
First seek a matched clean run with the same fluids, flows, inlet temperatures and valve path. Otherwise rerate the exact clean selection at current conditions. Alfa Laval's undated Performance Audit describes unit-specific thermal software and benchmarks; Veolia's Water Handbook, chapter 23, supports controlled-flow heat-transfer comparisons.[8][9]
Divide measured conductance by its current clean prediction for the thermal ratio; calculate duty recovery separately.
Thermal conductance ratio R_UA = UAmeasured / UAclean,predicted, with both values at current conditions. Duty recovery is separately Qmeasured / Qclean,predicted. UA removes the immediate driving-temperature difference but still changes with flow, viscosity, conductivity and distribution.
Anvilfield estimates that the dirty case retains 83.33% of clean duty but 66.67% of clean conductance. Keep these ratios distinct; neither is simply exchanger efficiency.
| Quantity | Clean core | Dirty core | After cleaning | Bypass header case |
|---|---|---|---|---|
| Hot inlet, °C | 30 | 30 | 30 | 30 |
| Hot outlet used, °C | 24 | 25 | 24.1 | 25.2 mixed; core remains 24 |
| Cold inlet, °C | 20 | 20 | 20 | 20 |
| Cold outlet, °C | 26 | 25 | 25.9 | 26 |
| Hot flow used, kg/s | 10 | 10 | 10 | 12.5 total; 10 core plus 2.5 bypass |
| Cold flow, kg/s | 10 | 10 | 10 | 10 |
| cp both streams, kJ/(kg·K) | 4.18 | 4.18 | 4.18 | 4.18 |
| Hot duty, kW | 250.8 | 209 | 246.62 | 250.8 |
| Cold duty, kW | 250.8 | 209 | 246.62 | 250.8 |
| Duty imbalance, % | 0 | 0 | 0 | 0 |
| Hot-outlet/cold-inlet approach, °C | 4 | 5 | 4.1 | 5.2 header; 4 at core |
| Opposite terminal gap, °C | 4 | 5 | 4.1 | 4 |
| LMTD calculated from listed temperatures, K | 4 | 5 | 4.1 | 4.57 |
| Apparent UA, kW/K | 62.7 | 41.8 | 60.15 | 54.83 from header; 62.7 at core |
| Apparent UA/clean core UA, % | 100 | 66.67 | 95.93 | 87.45 from header; 100 at core |
| Action | Retain baseline | Isolate cause and clean if deposits confirmed | Point target met; guarded thermal decision HOLD, see worked example | Correct bypass/boundary before diagnosing core |
Method: Anvilfield estimates, November 5, 2026; listed inputs, constant water properties, single-pass counterflow and mean duty. The bypass column uses header temperatures, not core boundaries.
What if the load changes or there is no clean baseline?
Predict clean duty from the smaller stream heat-capacity rate, inlet-temperature difference and exchanger effectiveness. Obtain clean UA for the current flow and properties from the exact selection or a validated clean curve. A fixed design UA does not itself correct for changed flow.
Ch = mh × cph; Cc = mc × cpc; Cmin = min(Ch, Cc); Cmax is the larger value; Cr = Cmin/Cmax; NTU = UAclean/Cmin. For counterflow, ε = [1 − exp(−NTU × (1 − Cr))]/[1 − Cr × exp(−NTU × (1 − Cr))]; at Cr = 1, ε = NTU/(1 + NTU). Qclean = ε × Cmin × (Thi − Tci). Subtract Qclean/Ch from Thi and add Qclean/Cc to Tci for the predicted outlets.
Lower duty alone is not fouling: the next table changes entering temperature and flow together, using a separately supplied clean UA for those conditions.
| Input or result | Original clean point | Current clean prediction | Current connection readings |
|---|---|---|---|
| Hot/cold inlet temperatures, °C | 30 / 20 | 28 / 20 | 28 / 20 |
| Hot/cold mass flow, kg/s | 10 / 10 | 8 / 10 | 8 / 10 |
| Clean UA supplied for these conditions, kW/K | 62.7 | 56 | 56 reference |
| Hot outlet, °C | 24 | 22.68 | 23.00 |
| Cold outlet, °C | 26 | 24.26 | 24.00 |
| Duty from either side, kW | 250.8 | 178.02 | 167.2 |
| Hot-outlet/cold-inlet approach, °C | 4 | 2.68 | 3.00 |
| Approach residual against current clean prediction, K | 0 | 0 | +0.32 |
| Apparent UA, kW/K | 62.7 | 56 | 48.10 |
| R_UA at current conditions, % | 100 | 100 | 85.89 |
| Point-estimate disposition | Baseline | Lower duty expected | Investigate the repeatable normalized deficit |
Anvilfield estimates in the changed-condition table use 56 kW/K as the current clean-model output. The assumed connection approach improves against the old baseline even though conductance falls to 85.89% of its current clean prediction. Calculate before rounding outlets.
Without commissioning data, retrieve the serial-number selection, plate/pass arrangement and clean rerating. Verify glycol concentration and treatment before comparing properties. Retain the selection's clean versus fouled rating distinction; a design fouling allowance is not a measured deposit. If neither a clean model nor a verified clean run is available, document the condition and investigate competing faults; a percentage of clean performance cannot yet be assigned. Establish a post-clean baseline only after acceptance.
What do pressure drops on both sides reveal?
Measure pressure drop separately across each stream's exchanger connections, excluding external strainers and control valves. Correct for elevation, impulse-line heads and differing tap-section velocities where applicable. Zero differential instruments and purge trapped air from liquid sensing lines; subtracting two large-range gauges can overwhelm a small differential with uncertainty.
Compare each measured loss with its own clean prediction at actual flow, concentration and temperature. Xylem Plateflow 104-70 Rev B requires actual-flow comparison, while AHRI's liquid-to-liquid certification program treats heat rate and the two pressure drops as separate ratings.[2][7] Balancing valve readings describe a different component boundary.
Divide measured core pressure drop by the clean prediction at that flow and fluid condition to obtain the hydraulic ratio.
Hydraulic ratio = measured core pressure drop / predicted clean core pressure drop at the same flow and properties. Over a narrow validated range, pressure drop scales approximately with flow squared. Anvilfield's screening approximation is ΔPcurrent ≈ ΔPreference × (flowcurrent/flowreference)², with unchanged properties and roughly constant resistance coefficient.
| State | Hot/cold flow, kg/s | Hot expected/measured, kPa | Hot ratio | Cold expected/measured, kPa | Cold ratio | Interpretation |
|---|---|---|---|---|---|---|
| Clean baseline | 10 / 10 | 40 / 40 | 1.00 | 30 / 30 | 1.00 | Retain clean curve point |
| Dirty at full flow | 10 / 10 | 40 / 60 | 1.50 | 30 / 31 | 1.03 | Hot-side restriction indicated; exclude port/strainer issue |
| Dirty at reduced flow | 8 / 8 | 25.6 / 32 | 1.25 | 19.2 / 19.2 | 1.00 | Raw hot drop below 40 kPa still exceeds current-flow prediction |
| Post-clean full flow | 10 / 10 | 40 / 41 | 1.03 | 30 / 30.5 | 1.02 | Point ratio inside band; verify uncertainty |
Reduced-flow predictions use the approximate square law; the dirty reduced-flow reading is a separate test input. OEM software or measured clean curves supersede this approximation when viscosity, Reynolds number or geometry changes the relationship.
Anvilfield recommends fault isolation when repeatable valid readings show R_UA below 0.90 or either hydraulic ratio above 1.20. These investigation bands are Anvilfield recommendations, not AHRI limits. A 32 kPa reading can indicate restriction even when the old clean reading was 40 kPa: the clean expectation at that reduced flow is only 25.6 kPa.
Is it scale, debris, low flow, air or incorrect assembly?
Use timing, normalized thermal loss and separate hydraulic behavior to select the next test. Veolia's Water Handbook, chapter 23, treats deposit resistance separately from fluid-film resistance.[9] A poorly conducting film can therefore reduce heat transfer before it appreciably narrows the channels. Normal pressure drop alone does not establish cleanliness.
| Fault | Thermal signature | Hot/cold hydraulic signature | Timing | Discriminating test/action |
|---|---|---|---|---|
| Surface scale/biofilm | R_UA falls; required approach widens | Affected side can rise; thin deposits need not create a large rise | Usually gradual | Validate properties/bypass, sample deposits, inspect or approved cleaning and matched retest; [1]/[9] |
| Debris/port obstruction | Duty can remain adequate initially | Affected-side ratio rises | Often abrupt | Check port/filter and external-strainer taps; backflush only if allowed; [1]/[2]/[4] |
| External strainer restriction | Duty falls because branch flow falls | Strainer differential rises; raw core drop can fall | Abrupt or progressive | Measure separate strainer and core losses, restore branch flow; [2]/[4] |
| Insufficient pump head or throttled valve | Duty fails against design but can match actual-flow clean prediction | Branch flow and raw core drop low | Follows controls or equipment change | Check valve travel, pump state and actual flow; [2]/[5] |
| Air binding | Unsteady duty and temperatures | Flow/differential fluctuations or reduced wet channels | Startup, refill or air ingress | Vent using unit instructions, then rerun stable window; [5] |
| External bypass/three-way leakage | Core model can pass while header supply fails | Core-flow drop can be low if flow diverts | Follows valve mode or leak path | Measure branch flows and three junction temperatures; mixing calculation, bypass test |
| Wrong glycol/property basis | Apparent duty/UA departs from former baseline | Pressure drop changes with viscosity and density | After fill/top-up or temperature shift | Verify concentration and rerate exact fluid; [6]/[7] |
| Wrong plate order/pass connection | Persistent duty loss after service | One or both losses depart from selection | Immediately after reassembly | Verify plate hanging list, drawing and connections; [1]/[2] |
| Plate leak/cross-contamination | Duty behavior alone inconclusive | Differential alone inconclusive | Corrosion, fatigue or damage | Fluid analysis and model-specific isolated cross-leak test; [1]/[2] |
| Sensor/boundary error | Heat balance fails, or false UA trend | Hydraulic evidence can remain normal | After instrument work or under low signal | Channel comparison, meter setup and junction-location audit; measurement checks; [8] |
The matrix combines manufacturer troubleshooting with the measurement checks above; confirm the cause with the indicated test. Start with the changed variable: refill, reassembly or a dirty-water event. Thermal loss alone does not locate the fouled side; use separate hydraulic readings, water history and deposit findings.
Separate a persistently changed configuration from progressive deterioration. Added plates, changed passes or a new glycol product require rerating. Fewer parallel channels can increase pressure loss at fixed total flow. If both hydraulic ratios rise together, verify shared flow and property inputs before concluding both sides fouled.
Use cooling tower water treatment records to investigate open-loop deposit drivers, liquid cooling loop chemistry records for coolant changes, and air and pump cavitation diagnosis for unstable supply. Stabilize operation before interpreting the trend.
Treat suspected cross-contamination separately. Fluid chemistry or conductivity changes justify investigation; an equipment-specific isolated cross-leak test establishes leakage. Alfa Laval and Xylem address plate damage and mixing paths.[1][2] Use the specified equipment test procedure and limits rather than borrowing a piping hydrotest multiplier.
How do you prove bypass instead of fouling?
Prove bypass by measuring core branches before mixing and checking temperatures and flows through the junction. Trace three-way valves, parallel exchangers, warm-return connections and the minimum-flow bypass and decoupler. Compare commanded position with actual stem travel and shutoff behavior.
For the same fluid at approximately equal properties, mixed temperature is the mass-weighted average. Subtract core outlet temperature from mixed temperature and divide by bypass-to-core temperature separation to estimate the bypass share of total flow.
Tmix = (mcore × Tcore + mbypass × Tbypass)/(mcore + mbypass). Bypass fraction = (Tmix − Tcore)/(Tbypass − Tcore). Use an enthalpy balance where properties differ materially; nearly equal core and bypass temperatures make this temperature-based fraction poorly resolved.
Anvilfield estimates that mixing 10 kg/s at a clean core outlet of 24°C with 2.5 kg/s bypass at 30°C gives 25.2°C at the header, with bypass equal to 20% of total header flow. The header approach is 5.2°C against a true core approach of 4°C.
The heat balance still closes: 12.5 kg/s losing 4.8°C releases the same 250.8 kW gained by the cold stream. Header temperatures nevertheless produce apparent UA of 54.83 kW/K, only 87.45% of the clean core value. A balanced pair of duties cannot validate the core measurement boundary.
What is the controlled bypass test?
Use the plant's approved operating sequence with sufficient cooling capacity available. Change the suspect path gradually while watching core flow, junction temperatures, supply setpoint and branch differential together. Watch the mixed header move toward core outlet temperature; physically verify valve travel and shutoff.
Keep required minimum flow, relief paths and standby capacity available. Leave a bypass required for minimum flow in service; isolate a suspect path only when the approved sequence maintains flow and pressure protection. Danfoss's 2020.02 guide calls for slow valve changes.[5]
Should you backflush, use CIP or open the plate pack?
Choose cleaning from the deposit, construction and installed accessories. Approved backflush addresses loose debris; compatible cleaning in place, or CIP, addresses soluble deposits; opening a gasketed pack permits inspection and manual cleaning. Xylem's brazed IOM 03241 Rev B, October 2024, suggests replacement when commercial cleaning cannot recover the unit and recommends against chemical cleaning of its refrigerant circuit.[3]
Identify plate, braze and gasket materials before choosing concentration, temperature or contact time. Record the deposit finding and approved chemical instructions in the service scope. Apply each instruction to its named assembly.
Before opening either circuit, isolate and secure all energy sources under the site's energy-control procedure, cool the unit, drain both sides and verify pressure relief. For covered general-industry servicing, OSHA 29 CFR 1910.147(d) requires energy isolation, lockout/tagout, control of stored energy and verification before work; continue verification where energy can reaccumulate.[13] Xylem's Plateflow opening procedure requires stopping pumps, closing both circuits and draining both sides.[2]
| Manufacturer/document | Equipment | Verified instruction/value | Conflict or limit | Use in scope |
|---|---|---|---|---|
| Xylem Plateflow 104-70 Rev B, September 2018, p. 25[2] | Gasketed plate-and-frame | Backflush at 1.5 times process flow | Do not transfer to a unit/accessory prohibiting reverse flow | Size compatible backflush arrangement within design and differential-pressure limits |
| Alfa Laval 200001927-2-EN-GB, p. 16, undated issue[4] | GPHE port filter, port sizes 200 mm or larger | No reverse flow with port filter installed | Direct conflict with applying a general backflush instruction | Use assembly-approved filter service method |
| Xylem Plateflow 104-70 Rev B, September 2018, p. 25[2] | Same gasketed equipment | Initial rinse 100 to 110°F, CIP approximately 180°F for 3 to 6 hours | Temperature, agent and seals must suit actual unit | Separate circulation time from total service outage |
| Xylem Plateflow 104-70 Rev B, September 2018, p. 23[2] | Stainless thermal plates | No hydrochloric acid; cleaning-solution preparation water no more than 300 ppm chloride | Not an operating coolant specification | State water supply and chemical compatibility in scope |
| Xylem 03241 Rev B, October 2024, p. 5[3] | Brazed plate | Suggested 5% phosphoric or oxalic acid for stubborn deposits, followed by fresh-water rinse | Materials/cleaner instructions govern; refrigerant circuit chemical cleaning not recommended | Identify braze and deposit before selecting chemistry |
| Danfoss AQ13708645598701-000503, printed 2020.02, pp. 5 to 6[5] | Brazed plate | Depressurize and cool below 40°C before maintenance; compatible backwash/CIP and fresh-water flushing | No fixed universal chemical concentration or circulation duration | Use installed equipment and selected cleaning product instructions |
| Danfoss AQ13708645598701-000503, printed 2020.02, p. 6[5] | Hard-water service | Scaling risk warning above 60°C | Not a zero-scaling threshold below 60°C | Link treatment and operating-temperature history to deposits |
Which instruction controls the service?
An installed filter's reverse-flow prohibition controls even when the bare exchanger permits backflushing. Check chemical compatibility across the entire cleaning circuit, including pumps, hoses and seals. Danfoss's 2020.02 guide requires gloves and eye protection when using cleaning agents.[5]
Xylem's brazed manual gives a suggested acid percentage without defining a mass or volume basis in that procedure. Obtain the selected cleaner's dilution instructions rather than inventing a mixing ratio. Select the rinse endpoint from the cleaner's instructions, then verify restored coolant concentration and treatment before restart.
APV's current service resource describes equipment-specific cleaning, reconditioning and regasketing.[10] After opening, follow the unit's plate hanging list and specified pack dimension.[2] Damaged plates require repair disposition before reassembly.
What counts as successful cleaning?
Successful cleaning restores verified thermal and hydraulic performance, plant output and equipment integrity. Agree the criteria and measurement decision rule before shutdown, using consistent boundaries, fluid properties and an actual-condition clean model before and after service.
Anvilfield proposes these worked-case contract targets, separate from manufacturer warranty criteria and AHRI rating tolerances. The owner or designated commissioning authority approves output requirements and the decision rule; the tester documents uncertainty; the contractor records service; the operator authorizes return to the plant lineup.
| Item | Proposed pass criterion | Worked result | Disposition/source |
|---|---|---|---|
| Configuration | Correct plate/pass identity, routing and documented valve lineup | One-pass counterflow, core boundaries, bypass shut | Required setup, [1]/[2] |
| Stable measurement window | 15 minutes, flow within ±5% of mean, no mode/valve change | Specified test condition | Anvilfield field recommendation |
| Heat balance | Imbalance ≤10% with calibrated uncertainty reviewed | 0% | Pass; Anvilfield screening target |
| Thermal restoration | Measured UA/clean UA at actual conditions ≥0.95, with agreed decision rule | 95.93%; expanded uncertainty 6.31 percentage points at k = 2; lower bound 89.62% | Point target met; HOLD under proposed guarded rule, worked example |
| Hot hydraulic restoration | Actual drop/clean actual-condition drop 0.90 to 1.10 | 1.03 | Point target met; verify hydraulic uncertainty against band |
| Cold hydraulic restoration | Same 0.90 to 1.10 band | 1.02 | Point target met; verify hydraulic uncertainty against band |
| Plant output | Meets stated plant leaving-temperature and duty requirement | Scope requires ≤24.2°C hot outlet and ≥245 kW at case inputs; achieved 24.1°C and 246.62 kW | Point targets met; apply their uncertainty budgets |
| Integrity | No external leakage or detected cross-leak under applicable documented test | Record actual result; calculation cannot establish integrity | Verify separately to equipment/project procedure; [1]/[2]/[5] |
| Rinse/fluid restoration | Rinse endpoint to approved chemical instructions; restored operating coolant to equipment specification | Record rinse results, concentration and treatment sample | Verify separately; [3]/[5] |
| Record complete | Raw trends, calibration, model inputs, chemical log, valve lineup and acceptance signoff retained | Attach test package | Anvilfield acceptance record; [8] benchmark practice |
Under Anvilfield's proposed guarded rule, accept a minimum when the result minus expanded uncertainty meets it, a maximum when the result plus expanded uncertainty meets it, and a band when the full interval lies inside it. An overlapping interval requires a hold for improved measurement or corrective work and retesting. It is not proof that cleaning failed.
Investigate unexpectedly low normalized pressure loss for measurement errors, wrong properties or altered routing. Retain unrounded values for decisions.
NIST TN 1297 supplies uncertainty terminology and propagation, not exchanger acceptance bands.[11][12] Include clean-model or reference-run uncertainty and shared calibration effects in actual ratio uncertainty.
Field diagnosis and acceptance checklist
- Record nameplate, serial, plate/pass drawing, materials and clean selection; resolve configuration errors.
- Sketch core, strainer, valve and bypass boundaries; correct sensors contaminated by mixing.
- Compare temperature channels; verify flow-meter configuration and fluid concentration; correct bias.
- Record four temperatures, both flows and separate core/strainer differentials during one stable window.
- Calculate both duties and imbalance; resolve a failed balance before interpreting UA.
- Rerate clean performance at current conditions; retain expected/measured approach, UA and both pressure drops.
- Check deterioration timing, actual valve travel, pump state, air and assembly against the fault matrix.
- Quantify bypass flow and mixing; preserve minimum flow and cooling capacity during valve tests.
- Identify deposits and equipment/accessory restrictions before selecting backflush, CIP or manual cleaning.
- Compare before/after performance with agreed targets; apply the uncertainty decision rule.
- Verify integrity and rinse endpoints; restore treatment, venting, strainers and operating valve lineup.
- Retain raw trends, calibrations, photos, cleaning log and signoff; reset the baseline after acceptance.
How often should you trend and clean the exchanger?
Review trends weekly during initial operation and monthly once stable under Anvilfield's proposed schedule; clean when verified condition or equipment instructions call for it. Xylem's brazed manual links cleaning frequency to fouling conditions, and Alfa Laval's port-filter instruction links filter service to debris loading.[3][4]
Use continuous BMS sampling where available; for manual-only plants, collect stable-window snapshots at the review interval. Recheck the baseline after cleaning or fluid, plate or control changes.
Before seasonal changeover, verify the clean curve and bypass operation against the waterside economizer operating sequence; repeat the two-side check after dirty tower-water events. For a CDU, retain coolant composition and branch boundaries beside the CDU commissioning baseline.
Set shutdown timing from deterioration rate, equipment instructions and capacity consequence. If the trend will cross an agreed investigation band before the next shutdown, bring inspection forward.
What should the diagnosis and cleaning scope cost?
Anvilfield estimates $1,180 for a two-person diagnostic visit, $2,910 for accessible CIP and $5,320 for an accessible gasketed-pack opening and cleaning under the November 2026 assumptions below. Price labor-hours separately from elapsed outage and chemical circulation time.
| Scope | Crew/elapsed assumption | Labor | Other included inputs | Anvilfield estimate |
|---|---|---|---|---|
| Diagnostic visit | 2 technicians × 4 hours each; 4 elapsed crew hours | 8 labor-hours × $110 = $880 | Instrument/mobilization allowance $300 | $1,180 |
| Accessible CIP | 2 technicians × 8 hours each; 8 elapsed crew hours | 16 labor-hours × $110 = $1,760 | Chemicals $550; rig/mobilization $350; disposal $250 | $2,910 |
| Accessible gasketed-pack opening/cleaning | 2 technicians × 16 hours each; 16 elapsed crew hours | 32 labor-hours × $110 = $3,520 | Consumables/gasket allowance $1,200; rig/mobilization $350; disposal $250 | $5,320 |
Method: Anvilfield estimates dated November 5, 2026, calculated by multiplying stated labor-hours by the assumed $110 hourly rate and adding the listed allowances. The CIP crew window includes four assumed circulation hours within Plateflow's specified range.
Alfa Laval's Performance Audit usually collects data for one hour per unit.[8] That collection period is only part of a diagnostic visit; isolation, cooling, draining, rig setup, rinsing, refill, venting and acceptance testing also determine the cleaning outage.
Price deposit testing, exchanger and CIP circuit volume, chemical disposal, gasket and plate condition, access and required repeat tests explicitly. The gasket allowance is a consumable budget rather than a complete replacement pack; damaged plates, temporary cooling, tax and major rigging require separate quantities.
Estimators find public HVAC solicitations and past award amounts for exchanger service scopes through PorchRocket Bid Control.
Use plant sequencing and measured power to establish energy consequences. Measure plant electrical demand separately from lost exchanger thermal duty when estimating operating savings.
Worked example: normalized performance and cleaning acceptance
Use the water-to-water inputs in the thermal table: one-pass counterflow, unchanged routing and area, constant specific heat and negligible external heat exchange. All case inputs are Anvilfield calculation assumptions dated November 5, 2026.
First validate the balance: dirty outlets at 25°C give equal 209 kW duties. Both terminal gaps are 5°C, producing 41.8 kW/K conductance.
Next localize restriction. The hot-side loss rises from its 40 kPa clean expectation to 60 kPa, while the cold-side loss is 31 kPa against 30 kPa expected. Exclude external strainer loss and bypass mixing, then confirm the deposit and clean the indicated side by its equipment instructions.
Repeat the matched test after cleaning. Outlets of 24.1°C and 25.9°C yield 246.62 kW duty, 4.1 K LMTD and 60.15 kW/K conductance. Point estimates meet the acceptance table's targets.
Why does the post-clean result remain on hold?
Propagate all four temperatures and both flows jointly through mean duty and LMTD: the same temperature readings occur in both, so treating calculated Q and LMTD as independent loses their covariance. Using the earlier independent standard uncertainties and a fixed clean reference, Anvilfield estimates post-clean UA standard uncertainty of 1.98 kW/K and expanded ratio uncertainty of 6.31 percentage points at coverage factor k = 2.
The 95.93% conductance estimate therefore has a lower bound of 89.62%, below the proposed guarded 95% minimum. Hold thermal acceptance, improve the measurement budget or correct remaining performance loss, then retest. Overlap alone does not justify another chemical clean. The hot-outlet target also remains on hold: adding 0.2°C expanded uncertainty at k = 2 gives 24.3°C, above the 24.2°C limit. Verify hydraulic and duty uncertainty, integrity and rinse results before signoff.
First-order propagation uses standard uncertainties of 0.1°C per temperature and 0.2 kg/s per flow. Temperature sensitivities for hot inlet, hot outlet, cold inlet and cold outlet are −2.24, −12.43, 2.24 and 12.43 kW/K²; each flow sensitivity is 3.01 kW/K per kg/s. Combine their uncertainty contributions by root-sum-of-squares for these independent inputs. Use full precision for decisions; NIST TN 1297 sections 5 and 6 [11] and [12] support the propagation and coverage terminology.
At the unchanged 95.93% result, expanded ratio uncertainty must be no greater than its 0.93 percentage-point margin above the target to pass the guarded minimum. Specify instruments and installation around that margin. A scope using point-estimate acceptance must name that different rule and report uncertainty with the result.
Sources checked and reviewed November 5, 2026. Update log: November 5, 2026, reviewed manufacturer limits and calculations; clarified deposit diagnosis, service isolation and guarded acceptance; added source links and removed repeated explanations.
How to cite this page: Anvilfield, Plate heat exchanger fouling diagnosis: a field guide, 2026, https://anvilfield.com/field-guides/hvac/plate-heat-exchanger-fouling-diagnosis/.
FAQ
How do temperature and pressure drop diagnose plate heat exchanger fouling?
Falling thermal performance together with increased pressure drop at comparable flow supports a fouling or obstruction diagnosis, after bypass, external restriction and instrument errors are checked.
How do you normalize exchanger approach?
Predict the clean outlet temperatures at the current inlet conditions, actual flows and fluid properties, then compare the measured connection approach with that prediction.
How can you tell fouling from bypass?
Measure exchanger-branch temperatures and flow before the mixing junction, then use the junction heat balance and actual valve travel to identify bypass.
What are heat exchanger cleaning acceptance criteria?
Require agreed actual-condition UA and pressure-drop targets, the plant's duty and temperature targets, an uncertainty decision rule, integrity and rinse verification, and a complete before/after record.
Can a fouled plate heat exchanger have normal pressure drop?
Yes; a thermally important deposit need not create a large hydraulic restriction, so normal pressure drop alone does not establish cleanliness.
Does a larger approach always mean fouling?
No; compare the same thermal boundary and correct for changing inlet temperatures, flows, properties and bypass mixing.
Should both sides have the same pressure drop?
No; each side has its own flow, fluid properties and channel arrangement, and must be compared with its own clean prediction.
Can you backflush every plate heat exchanger?
No; Alfa Laval's Port filter instruction 200001927-2-EN-GB prohibits reverse flow while that filter is installed.
How often should a plate heat exchanger be chemically cleaned?
Clean when verified deterioration or the installed equipment's instructions call for it. Anvilfield recommends weekly trend review at startup and monthly review once stable to establish that service interval.
How long does plate heat exchanger cleaning take?
Xylem's Plateflow 104-70 Rev B lists 3 to 6 hours of CIP circulation, with isolation, cooling, setup, rinsing, refill and testing added to the outage.
People also ask
Codes cited in this guide
This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.
Sources checked
- [1] Alfa Laval, Troubleshooting for plate heat exchangers Undated resource; checked November 5, 2026.
- [2] Xylem Standard Xchange, Plateflow gasketed plate-and-frame IOM, 104-70 Rev B, September 2018 Checked November 5, 2026.
- [3] Xylem Standard Xchange, Brazed Plate Heat Exchangers IOM, 03241 Rev B, October 2024 Checked November 5, 2026.
- [4] Alfa Laval, Port filter instruction 200001927-2-EN-GB Undated resource; checked November 5, 2026.
- [5] Danfoss, Brazed Plate Heat Exchanger Operating Guide AQ13708645598701-000503, printed 2020.02 Checked November 5, 2026.
- [6] ASHRAE, 2024 Handbook, HVAC Systems and Equipment, SI, chapter 47, Heat Exchangers Checked November 5, 2026.
- [7] AHRI, Liquid to Liquid Heat Exchangers certification program Undated resource; checked November 5, 2026.
- [8] Alfa Laval, Plate heat exchanger Performance Audit Undated resource; checked November 5, 2026.
- [9] Veolia, Water Handbook, chapter 23, Cooling Water Systems: Heat Transfer Undated resource; checked November 5, 2026.
- [10] SPX FLOW/ITT Flow Technologies, APV plate heat exchanger service entry Undated resource; checked November 5, 2026.
- [11] NIST Technical Note 1297, 1994 edition, section 5, Combined Standard Uncertainty Checked November 5, 2026.
- [12] NIST Technical Note 1297, 1994 edition, section 6, Expanded Uncertainty Checked November 5, 2026.
- [13] OSHA, 29 CFR 1910.147, Control of hazardous energy, paragraphs (a) and (d) Current regulatory text checked November 5, 2026; covered general-industry servicing.
- [14] NIST Technical Note 1297, 1994 edition, section 4, Type B Evaluation of Standard Uncertainty Checked November 5, 2026.