Type K Thermocouple Failure


Thread Starter


Hello everybody,

Looking for any information or ideas to the possible failures. I will give as much info as possible. The Thermocouples are type K. I believe they 316 stainless steel sheath, 1/4" OD, mineral insulated, Hot junction is bead welded, 48" long w/ 15ft armored extension.

Now the install, thermocouples are in furnace, used as skin temp on 288 individual tubes. (score furnace, You can google it)

There is a protective tube that is completely closed off that allows thermocouple to contact tube. (a thermowell of sorts). temp of operation is 1600f to 1800f continuous. out of all 288 elements 24 are dual element with each going to 2 different transmitter's. The rest are taken to the DCS via thermocouple extension cable. (failure is element related, keep reading).

The problem, The thermocouples have only been is service about a year. Operations had been complaining about variations in readings. sometimes 5degf to as high as 100f. we tried venting the couples for a while, due to the contractor had sealed tubing farrels on some of the couples with some water in the extension. this had been going on for some time so we changed out some of the couples. all couples changed came up 30 to 100deg f. we tested some of the couples pulled on a Joffra dry block calibrator. set at 800degf the 'bad' thermocouples read 575f while any other new one read 800f. proving the calibrator and making a statement that thermocouples are failing.

but why?'Green Rot' is a theory I've bounced around. Don't know how to prove it if it is. I mean these are newer thermocouples. any ideas on thermocouple element failure or any other ideas? I can provide further info if needed.

Sorry guys for the long spill. just have to do that do people don't give you stupid answers like "its induced voltage". Thanks for the input.
Type K's do experience higher failure rates at those temperatures.

You can use larger wire guage, it will buy you a bit of service life, but does not stop the decalibration. The purity of the oxide insulation can also play a role.

We've used Type N at 2000F+ with out difficulty.
The thermocouples are drifting.

Drift is caused by chemical change to the thermocouple junction. This is referred to in the literature as "inhomogeneity", meaning the junction is no longer an combination of its two 'pure' alloys, Chromel {90% nickel and 10% chromium} and Alumel {95% nickel, 2% manganese, 2% aluminium and 1% silicon}, for type K. The junction is polluted with additional elements or compounds so it is no longer a type K thermocouple.

The theory, as I understand it, is that ions migrate from the sheath to the junction through the mineral insulation (MI). The higher the exposure temperature, the greater the chemical activity, the greater the chemical changes, the faster the pollution occurs with resulting drift.

The commonly used insulation material, mineral insulation (MI) MgO, will absorb and hold water (hygroscopic). The presence of water increases the electrical conductivity through the insulation material, or inversely, reduces the resistivity of the insulation material. Overall lower insulation resistance aids the ability of ion traverse through the MI and the subsequent pollution of the junction.

Green Rot is only one form of chemical pollution, frequently seen on type K elements without mineral insulation, on the heavy gauge elements that are inserted into 1/2" diameter protection tubes. If you carefully grind the end of the sheath away, you might see green discoloration of the junction tip, or you might not, depending on what chemical change your T/C experiences.

If you put your mis-reading thermocouple on a bench and let both ends equilibrate to room temperature so it's isothermal it will not generate an EMF since both ends are at the same temperature. Measure its resistance with an ohm meter and I'll bet it's well above 100 ohms. Under the same conditions, measure the resistance of a 'new' thermocouple. Its resistance will be well under an ohm. The increase in resistance is an indication of a polluted junction. Honeywell UDC and HC-900 temperature controllers measure T/C loop resistance and will alarm on high resistance as an indication that the thermocouple is drifting.

There is no method of correcting for drift because the chemical composition of the polluted junction is an unknown. If the chemical composition is unknown, there is no reference table to refer to in order convert EMF/mV to temperature. The type K table does not apply, as you've seen, because your bastard thermocouple produces an EMF/mV output equivalent to 575 Deg F when it is at 800 Deg F. You need a different reference table, which does not exist, because its chemical composition is an unknown.

The presence of water is a serious problem. Water exacerbates drift.

There is a vendor, Accutru, who uses "MI dry", a proprietary mineral insulation that is hydrophobic (little or no affinity to absorb water) and has a much higher resistance than MgO. I didn't believe the resistance claim, until I cut open an MGO thermocouple open and tested the MgO resistance (100sK ohms), compared to 'dry MI' (Megohms). Accutru's thermocouples are used in process heaters, furnaces and boilers that have long run times between shutdowns, where T/C stability is critical because of inability to change out a T/C until a scheduled outage.


You might consider inconel as a sheath material at those temperatures. If you solve your drift problem, you might find SS sheath burning out quicker than inconel, and the cost difference is negligible.

Curt Wuollet

At the risk of sounding stupid :^) A year at 1600-1800F is a long hard life for all but the least reactive metals. All kinds of things could be causing this.

Impurities, slag or oxides in the weld, gas infusion. At the foundry, K thermocouples used at temperature were considered consumables and the cost of doing business. Those were bare assemblies though (because they were expendable). I would expect the TC manufacturer (or competitor) might be the ones to consult on what reasonable life is and what might be done to improve it. Omega has always been most helpful if you frame the questions properly.

Thanks guys.

I have read about type N couples but have never had the pleasure of seeing one in action. They sound awesome, But all the thermocouples extension wire is type J and it is a nice jog to marshaling room plus I would need new LLMUX cards for DCS. (Honeywell TDC 3000 running Experion). My thoughts were to go to Inconel sheath, From my research 316 SS max temp is like 1650. I do believe that the couples are being poisoned, I will check resistance of the couples, and see if both or either lead is magnetic. I have read that in certain contamination that both end up magnetic in some sort of metallurgy phenomenon.

Thanks again for the good input.
Misc comments

Type K sensors with Type J extension wires, you do need to be cautious about that.

The magnet test you describe is not reliable, temperature/millivolt tests are the only accepted test.

The resistance test is only for continuity of your t/c junction/wiring; you can only test the insulation from t/c to sheath with ungrounded sensors.

Try to use the largest gauge t/c wire you can in the sensor at you temperatures.