What Happens When a Fluid Bed Dryer Probe Drifts Out of Tolerance?

The inlet temperature probe on a fluid bed dryer is one of the most consequential instruments in solid dose manufacturing — and one of the most likely to drift quietly out of tolerance between calibrations. By the time anyone notices, the question isn't whether drift happened. It's how many batches it affected.

This is one of the most common findings we see on Glatt, ACG, Vector, and Aeromatic fluid beds that have been running production for several years without a full calibration review. Here's how it happens, how we catch it, and what an OOT investigation actually looks like once you have one on your hands.

How fluid bed probes actually drift

Drift isn't usually a single event. It's a slow accumulation of small effects that, taken together, push a probe past its tolerance band.

• Thermal cycling. A fluid bed cycles from ambient to 80°C or higher dozens of times per batch. Over years, repeated expansion and contraction loosens the wire-to-element connection inside an RTD, slowly raising its resistance baseline.
• Process contamination. Even with a sheath, fines and binder solution find their way into the well over time. A layer of residue insulates the probe and slows its response — usually showing up first as lag, then as a steady-state offset.
• Mounting and immersion changes. A probe that gets pulled for cleaning and reinstalled to a slightly different depth will read differently. We've seen 0.8°C shifts from a probe that someone rotated an eighth-turn during reassembly.
• Transmitter age. Don't blame the probe. Half the drift we trace ends up at the transmitter, especially older 4–20 mA units that have been in service through multiple validation cycles.

How drift gets caught

Most drift is found in one of two places: at the scheduled calibration, or as an unexplained process anomaly during a batch.

The first path is the better one. A proper as-found measurement against a NIST-traceable reference, taken before any adjustment, captures exactly where the instrument was at the moment the technician walked up to it. If that as-found reading is out of tolerance, you have a documented OOT event the moment it shows up.

The second path is harder. The operator reports the inlet won't hold setpoint, or the product temperature looks low even though the batch record says otherwise, or two batches in a row require unusual airflow adjustments to dry out. By the time you're investigating from a process anomaly, you've already produced material on a suspect instrument — which means an OOT investigation that has to look backward.

What an OOT investigation actually costs

Once you have a documented out-of-tolerance event on a process-critical instrument, the work doesn't stop at "adjust and recalibrate." Depending on your QMS, the chain of impact assessment typically includes:

1. Quarantining any product manufactured since the previous in-tolerance calibration
2. Reviewing batch records for that span against the magnitude and direction of the drift
3. Determining whether the drift could plausibly have affected critical quality attributes
4. Documenting the rationale for release or rejection of affected batches
5. CAPA on the root cause, with verification of effectiveness

None of that work is optional. Your QA team will spend days on it whether the impact assessment ends in release or rejection. Catching drift earlier — through tighter calibration intervals on critical probes, or through periodic verification between full calibrations — is almost always cheaper than the investigation it prevents.

Adjust, replace, or escalate?

When we find a probe out of tolerance on a fluid bed, the decision tree usually goes like this:

• Within 2× tolerance, no contamination evident, transmitter clean → adjust at the transmitter, recalibrate, restore to service. Document the as-found and as-left.
• More than 2× tolerance, or a non-linear deviation across the range → the probe itself is suspect. Replace with a calibrated spare. Send the original out for bench characterization if the failure mode matters.
• Drift accompanied by lag or sluggish response → pull the probe and inspect the well. Clean the well, replace the probe, and consider whether the SOP for thermowell cleaning needs to be revised.

In every case, the as-found data is the most important record you generate. It's what makes the OOT investigation defensible to an FDA inspector. Skip it — or adjust the instrument before taking it — and the calibration certificate becomes useless for retrospective impact assessment.