Abstract: The unbalanced DC bridge method is an important foundation for non-electrical quantity measurement technology and is widely used in university physics experimental teaching and engineering measurement practice． Addressing the issue that the traditional experimental method using a linear approximation formula for NTC thermistors( such as the MF51type) introduces significant principle errors overa wide temperature range( especially at the high-temperature end) ，this paper proposes an improved method based on an accurate model and carries out a standardized uncertainty evaluation．Based on the JJF 1059．1—2012Evaluation and Expression of Uncertainty in Measurement specification，an accurate mathematical
inversion model for the voltage output method based on Thevenin' s theorem was established，effectively
eliminating the method errors introduced by linear approximation．The combined standard uncertainty arising
from Type A and Type B evaluations was determined using the GUM method．Furthermore，by comparing 1．5 V
low-voltage supply with 5． 0 V conventional supply conditions，the impact of the self-heating effect was
quantitatively analyzed，verifying the necessity of low-voltage supply in suppressing systematic errors． The
experimental results show that in the temperature range of 25．0 ℃ to 65．0 ℃，after adopting the nonlinear
inversion model，the relative expanded uncertainty of the measurement system is stable within the range of
0．12% to 0．13%( Note: this evaluation excludes the uncertainty of temperature measurement) ; in contrast，the
systematic deviation produced by the traditional linear approximation method exceeds 40% at 65．0 ℃． The
material constant of the thermistor obtained based on the improved model is 3 946 K，with a relative deviation
from the nominal value of only －0．1%．This study provides an operable teaching demonstration for nonlinear
model correction and standardized uncertainty processing in unbalanced bridge experiments．
