The dynamic errors analysis and compensation of articulated arm measuring machine has become the focus of domestic and foreign scholars. In this paper, by analyzing the thermal deformation error, the measurement force error and the angle coding error, the optimal temperature value of the articulated arm measuring machine is obtained, and the three error factors representing its dynamic error, namely maximum positioning error (MPE), residual positioning error (RPE) and joint angle value (Ja), are obtained. In view of the above error factors, this paper proposes a new compensation method which combines the advantages of T-S fuzzy neural network (T-S FNN) in self-learning ability and large-scale operation, and the ability of simulated annealing (SA) to global optimization, and establishes the model. The orthogonal experiment shows that the error of dynamic process is reduced by 88.8%, 80.2% and 71.3% respectively by the proposed compensation method, which proves that the model can effectively improve the dynamic measurement accuracy of the measuring machine.