In order to improve the kinematics calibration accuracy of robot arm,the effects of different D-H kinematics error models on identification accuracy of kinematic errors were studied for a theoretical basis of accurate modeling.The performance and linearization error of variants of SD-H and MD-H kinematic error models were compared.And their comprehensive effect on the kinematic calibration accuracy was analyzed.The results show that the x-z-y Euler transform breaks the minimality of the MD-H model and is not recommended;the performances of other full-rank MD-H model is superior to SD-H model,and the linearization error of MD-H modeling is the same with that of SD-H;with the effect of model performance and linearization error,the calibration accuracy (from 0.1% to 39.3%) using SD-H model under strong noise is better than that using MD-H model (from 0.1% to 71.2%);a significant difference between the identification accuracy of angular and positional errors is also observed.The results provide a theoretical basis for the calibration modeling methods and selection of identification methods of different types of kinematic errors.