How can I estimate real data from sensor data

I have a thermistor sensor and need to measure temperature data. When I try to measure temperature, I saw thermistor data is not realistic at high temperatures. I want to use Matlab to find coefficients or function to estimate real value. To do that, I have real temperature log data and thermistor log data. Which toolbox or method do I need to use?

[DGM]

Without knowing exactly what sensor you have, how you have it configured, and how you're reading it, I can only guess. Are all readings off expectations, or are only high temperature readings off? What temperature is "high temperature"? Is the rest of the circuit exposed to the same temperature as the thermistor itself?

[Dogukan ARAT]

I am reading the temperature -50C deg to +170C deg. My measurements are diverging after 50C deg. Actually, I just applied curve fitting sum of sine method before this process, my lower temperature measurement was also bad. Right now, I am reading low values but I cannot read high values correctly. I think I need to change my technique that I apply after analog reading.

Answer 1

[DGM]

I mentioned the circuit specifics, as a lot of thermistors aren't intended for temperatures that high. Even if the thermistor itself is, any sort of linearizing circuit may not be linear outside its intended region of operation. If the circuit itself is exposed to the same temperature, the thermal coefficients of the biasing resistors may become significant contributors to error if no attempt has been made for compensation.

If you can't change the circuit or how it's being read, then model the sensor as it's read. Expose it to known temperatures, collect the corresponding values, and make a LUT translate the readings.