I think the key lies in the curve fitting process. After sensor aging, the input-output curve will change, and some input-output relationships will become unstable. For example, when the input temperature is 300 degrees Celsius, the output voltage changes greatly and cannot be stabilized in a certain range, so these temperatures can no longer be measured. Secondly, due to the aging of the sensor, the curve will change accordingly, sometimes it is obviously 200 degrees Celsius. The sensor shows 250 degrees Celsius, so the curve needs to be re-fitted. For example, the old sensor only needs to multiply the output voltage value by 15.36 when displaying the temperature value, while the new fitting curve needs to be multiplied by 14.5, and the result needs to be added with 1.04 to get close to the real value.