Strange results when putting thermistor in boiling water.

I am extremely new to Arduino, bought my first board today. My first project is to measure temperature using a 100k thermistor. It works completely fine under normal conditions, submerged in a cup of water. I used epoxy to seal the connections so the thermistor does not short when submerged.

The problem is when I boil the water. The reading increases as expected but then gradually decreases over time. When I start the readings from the beginning I get a lower boiling point temp reading.

Must I use another sensor? This one is rated for +41C to +120C (According to the distributor)

Are you submerging the wire leads in the water?

Yes

Perhaps water has migrated into connections.
The water is probably messing with your resistance measurement.
Only submerge the bulb part.
.

Thank you. I will be sure to try that next.

Another interesting (and probably obvious) observation is that do not use an induction stove to boil the water. It interferes with the thermistor.

Good observation.
Try a .1uf ceramic capacitor across the thermistor leads to see if the induction effect disappears.
.

FrancoisStevens:
Thank you. I will be sure to try that next.

Another interesting (and probably obvious) observation is that do not use an induction stove to boil the water. It interferes with the thermistor.

The pan will block the induction!

Epoxy (like most plastics) is permeable to moisture, and steam is particularly aggressive,
epoxy the sensor inside a metal tube or similar for a proper hermetic seal.

MarkT:
The pan will block the induction!

Epoxy (like most plastics) is permeable to moisture, and steam is particularly aggressive,
epoxy the sensor inside a metal tube or similar for a proper hermetic seal.

The induction definitely had an effect on the sensor. The readings were extremely erratic. I did not log the data from that. I moved to a plain electric stove for the rest of my testing.

Epoxying the sensor in a metal tube does sound like a good idea though. I will have more of a delay, but I may get more accurate data. Thanks.