In order to overcome the shortcoming of artificial neural networks and other conventional methods

a novel method to non-linearity compensation for thermocouple sensor using least squares support vector machine（LS-SVM） is put forward.In this method

the non-linear compensator

with power series expansion-based model

is established based on the principle of positive feedback.Meanwhile

the coefficients of power series of the compensator are identified by LS-SVM linear regression algorithm.Through this compensator

the desired linear characteristics of thermocouple can be obtained.Finally

platinum-rhodium 30-platinum-rhodium 6 thermocouple（B） sensor is taken as an example to non-linearity compensation

and experimental results show that the linearity of compensated sys

tem is less than 0.035 3 in the temperature range from 0

1

820 ℃.So the presented method is effective and can be also used in other similar systems..