Failure Modes We’ve Engineered Out
The failures that quietly take gas analyzers out of compliance in the field — and the design moves we put in front of them, traced from R&D bench to your serial number.
Most analyzer failures are old failures. We treat them that way.
The failure modes that take gas analyzers out of compliance — wavelength drift, sample-line condensation, NOx converter ageing, optical fouling, cross-interference, Ex-seal ingress, vibration drift — are not exotic. They are well-documented in the field, year after year, on stacks across the industry.
What separates one analyzer build from another is whether those patterns were treated as known engineering problems and pre-empted at design time, or quietly inherited from the data sheet. This page is our register: the failure modes our R&D and field-validation teams have studied, reproduced, and engineered against — with the design move dated, version-locked, and traceable to the unit you receive.
Seven failure modes, seven design moves
Each entry: the industry pattern as it shows up in the field, what our R&D or pilot-deployment work found when we reproduced it, and the specific engineering response that ships in every analyzer today.
Case 01
Wavelength drift past the compliance window
Industry Pattern
TDLAS analyzers in continuous stack service often drift 2–5% over the first 12 months. The numbers stay plausible, so the drift is invisible until the next CEMS audit — by which point the plant is already out of band.
What R&D / Field Testing Found
2,000-hour accelerated-ageing runs on 14 production-grade DFB lasers showed wavelength shift accelerating past spec under thermal cycling above 60 °C — a regime every in-situ stack head sees.
The Design Move
Closed-loop wavelength locking against an internal reference cell, plus laser-diode binning at incoming QC. Drift is corrected per measurement cycle, not per service call.
Evidence with the unitDrift trace per serial number is logged to FAT report.
Case 02
Sample-line condensation eating soluble gases
Industry Pattern
Heated sample lines lose temperature uniformity at the cold-end joint. Water condenses, soluble species (SO₂, NH₃, HCl) dissolve, and the analyzer reads low without alarm.
What R&D / Field Testing Found
Wet-stack scrubber-outlet trials with five commercial heated lines: three failed thermal uniformity at the elbow, with cold spots 12–18 °C below set-point.
The Design Move
Two-zone heat tracing with redundant RTDs, alarm at ≥5 °C deviation, and a documented elbow-routing rule shipped with every sampling kit.
Evidence with the unitHeated-line uniformity certificate ships with the unit.
Case 03
NOx converter ageing past the warranty edge
Industry Pattern
Catalytic NO₂ → NO converters silently lose efficiency below 95% at 16–24 months in service. CEMS readings under-report NO₂; nobody notices until the regulator does.
What R&D / Field Testing Found
Field-validation campaign across 14 sites identified an efficiency cliff between months 16 and 20, well before most published service intervals.
The Design Move
Quarterly conversion-efficiency check protocol, replacement cadence keyed to converter lot and analyzer serial, and a span-gas reference calibrated to ISO 17025 traceability.
Evidence with the unitConverter lot number traced to your serial number on the audit pack.
Case 04
Optical fouling in dust-loaded stacks
Industry Pattern
Cement, lime, and steel stacks coat optical windows with sub-millimetre particulate film. The transmitted intensity drops, the inversion algorithm reads it as a concentration shift, and operators chase a fault that is not in the gas.
What R&D / Field Testing Found
A cement-plant FAT showed 70 mbar instrument-air purge insufficient during stack negative-pressure transients — windows fouled within 90 days.
The Design Move
Pressure-compensated purge gas with closed-loop flow control, plus a dirty-window self-diagnostic that fires before signal loss reaches measurement-relevant levels.
Evidence with the unitPurge-flow setpoints and self-diagnostic thresholds keyed to your stack profile.
Case 05
Cross-interference in single-line NH₃ measurement
Industry Pattern
Single-line NH₃ TDLAS systems show false positives when flue moisture climbs and CO₂ overlaps the chosen line. De-NOx slip monitoring fails exactly when the plant most needs it.
What R&D / Field Testing Found
In-house cross-talk mapping against HITRAN line databases revealed unmodelled interference at the conventional NH₃ pick-line under high-H₂O / high-CO₂ matrices typical of post-SCR sampling points.
The Design Move
Multi-line fitting algorithm with a CO₂ / H₂O matrix correction, plus a pre-deployment matrix-screening step that confirms the chosen line set against the customer’s actual flue chemistry.
Evidence with the unitMatrix-correction coefficients dated and version-locked to the firmware build.
Case 06
Ex-proof seal failure at the cable gland
Industry Pattern
Hazardous-area enclosures are rated to Ex db IIC and IP67 in the lab — but cable glands torqued by hand at field commissioning routinely fail ingress in 6–12 months. Water reaches the electronics; the analyzer dies before the warranty ticks.
What R&D / Field Testing Found
Returns analysis on hazardous-area units: 85% of seal-related failures traced to over- or under-torque at the cable gland during site commissioning.
The Design Move
Torque-spec stickers on every gland, calibrated tools shipped with hazardous-area units, and a commissioning sign-off step that requires recorded torque values.
Evidence with the unitTorque-checked commissioning record returns to the audit trail.
Case 07
Mirror-alignment drift under stack vibration
Industry Pattern
Stack-mounted in-situ heads accumulate small alignment shifts under continuous vibration. Signal-to-noise quietly degrades, and the unit limps into its next maintenance cycle delivering noisier numbers than the spec sheet promised.
What R&D / Field Testing Found
ANSYS modal analysis plus 500-hour shaker-table testing identified a resonance band at 80–120 Hz — exactly where induced-draft fans on coal-fired and waste-to-energy stacks live.
The Design Move
Vibration-damped mounting frame, dual-mirror retroreflector geometry that tolerates micro-misalignment, and an on-board alignment-stability self-test that flags drift before noise reaches the measurement budget.
Evidence with the unitModal-test report and self-test log attached to the FAT package.
Four steps from field pattern to engineered counter-move
The register above is the output. The discipline that produces it runs every quarter, against new returns data and the previous quarter’s audit findings.
Stage 01
Pattern intake
Failure modes are gathered from returns analysis, field-service tickets, third-party audit findings, and published industry incident reports. Each pattern is logged with frequency, severity, and the cost a customer would carry.
Stage 02
Reproduce in the lab
The pattern is reproduced under accelerated conditions — thermal cycling, vibration, dust loading, matrix mixing, accelerated ageing — until the failure mechanism is isolated and quantified, not just observed.
Stage 03
Engineer the counter-move
A specific design change is committed: hardware revision, firmware path, sourcing rule, or commissioning protocol. The change is dated, version-locked, and tied to the failure mode it pre-empts.
Stage 04
Verify in the field
The counter-move ships in the next production batch and is monitored across pilot deployments before becoming default. Effectiveness is reported back into the failure-mode register on a quarterly cadence.
A failure register is only useful if it is honest.
We hold the register to four rules — the same ones procurement and compliance teams should expect of any analyzer maker that claims to have engineered the field out of its product.
- A failure mode is documented, not denied. The register is open to qualified buyers under NDA.
- A counter-move is dated and version-locked. We can show you when it entered the build.
- Field evidence beats lab evidence. Pilot deployments validate before the rule becomes default.
- The same failure register feeds R&D, Quality Inspection, and the Approved Vendor List.