Based on the time domain method (TDM) theory, a truncated generalized singular value decomposition(TGSVD) method for moving force identification has been developed by introducing a regularization matrix. The influence of TGSVD identification can be identified by comparing it with different regularization matrixes. Then, when comparing the TGSVD method with the TDM method, the numerical simulation of two-axle moving force loads shows that regularization matrixes have great influence on such properties as identification accuracy, robustness, etc. With appropriate regularization matrixes, the proposed TGSVD method has many good properties compared with TDM, such as better identification accuracy and robust noise with bending moment responses or acceleration responses. TGSVD has prominent advantages with single responses type or less responses number, which is beneficial for the application of TGSVD in the field identification of dynamic axle loads on bridges.