In order to improve the problem that the change of a single physical quantity has too much impact on the final result due to the linear relationship between physical quantities in the process of indirect torque measurement, a nonlinear multi model soft sensing method based on weighted K-means clustering and LSSVM is proposed in this paper. Firstly, multiple easily measured variables are selected as auxiliary parameters, and the data are preprocessed by using subjective and objective comprehensive weighting theory. Secondly, K-means clustering algorithm is used to form clusters of data with similar physical characteristics. Finally, multi model of data cluster is established and measured based on least squares support vector machine algorithm. The results show that under the same experimental conditions, the root mean square error of the proposed model is reduced by 0.484 and 0.263 respectively, and the average absolute percentage error is reduced by 1.003 and 0.292 respectively, which effectively improves the accuracy and stability of the measurement.