Why Thermowells Fail: Wake Frequency, Immersion and the Trade-off Nobody Explains

9 min read · updated 2026-09-12

A thermowell is the most mechanically stressed component most instrument engineers ever specify, and the two things it needs — deep immersion for accuracy, short length for survival — pull in exactly opposite directions.

The failure mechanism

A thermowell sticks into a flowing stream as a cantilever beam. Fluid passing around it does not flow smoothly past — it sheds vortices alternately from one side then the other, and each shed vortex gives the well a sideways kick. Those kicks arrive at a regular frequency set by velocity and tip diameter:

Meanwhile the well has its own natural frequency, the rate at which it would ring if you struck it. When the shedding frequency approaches that natural frequency, each kick arrives in step with the previous oscillation and amplitude builds — resonance. The stress concentrates at the root, the metal fatigues, and eventually the tip snaps off and travels downstream with the sensor inside it, leaving an open hole into the process.

This is why ASME PTC 19.3 TW exists as a standard in its own right rather than a clause in a general code. Thermowell failures have caused serious incidents, including releases of hot and hazardous fluid.

Length is the variable that catches people

Natural frequency depends on stiffness and mass, and for a cantilever it scales with 1 / L². That inverse square is the whole problem. Take one well in one service and change nothing but immersion:

(316 stainless, 22 mm root, 6.6 mm bore. The acceptance criterion is a frequency ratio below 0.8 — run your own numbers through the wake frequency calculator.)

Somebody with good intentions saying "insert it further into the flow for a better reading" can convert a safe design into a broken one without touching anything else. It is the most common route to a thermowell failure.

The trade-off with accuracy

Here is the uncomfortable part. A short thermowell is mechanically safe and thermally compromised. The well conducts heat along its stem toward the pipe wall and flange, which sit at a different temperature from the process — so a sensor sitting close to the wall reads somewhere between process and ambient. This is stem conduction error, and it can easily be several degrees.

The conventional guidance is at least ten tip diameters of immersion, with the sensor tip reaching past the boundary layer into the bulk flow. On a 22 mm tip that means 220 mm — which, as the table above shows, is already into failing territory at high velocity.

So the two requirements genuinely conflict, and the resolution is not to maximise either. It is to run the wake frequency check first, establish the maximum safe length, then check whether that length gives acceptable immersion. If it does not, something in the design has to change.

What to change when it fails

1. Shorten it. Because of the inverse square relationship, this is by far the most effective lever — and free.
2. Increase the root diameter or use a tapered or stepped shank. A tapered well is stiffer at the root where bending stress lives, and thinner at the tip where shedding is set, which improves both sides of the ratio at once.
3. Add a support collar to shorten the unsupported length without reducing immersion. Often the cheapest fix on an existing installation.
4. Move to a lower velocity location — a larger line, or a position away from a reducer where velocity peaks.
5. Use a velocity collar or reduce the bore where the design allows.

Field notes

• Check at maximum velocity, including upsets. Start-up, blowdown and relief conditions produce velocities the normal operating case never sees, and the well only has to resonate once for long enough.
• Steam and gas are the dangerous services. High velocity with low density is exactly where shedding frequencies climb. Water lines rarely fail this check, which is why people who work mostly on liquids are sometimes unaware the check exists.
• Get velocity from the true bore — the pipe schedule lookup and velocity calculator. Nominal diameter understates velocity and flatters the result.
• A well that has been in service for years is not proven safe. Fatigue is cumulative, and a process change that raised throughput can start a clock that was not running before.
• The sensor inside matters too. A loosely fitted sensor in an oversized bore responds slowly and reads low; see RTD vs thermocouple for the measurement side.

Calculators used in this guide: thermowell wake frequency · pipe velocity · pipe schedule · PT100 conversion