It should be noted that the dynamic substructure method is mainly convenient to establish the coupling dynamic model of the whole gear system, and the property and accuracy of the coupling model still depend on the property and accuracy of each substructure model. For example, transmission system modeling can adopt transfer matrix method or lumped parameter method, box modeling can adopt plate element or solid element, and gear meshing excitation modeling can adopt time-varying meshing stiffness or slicing method. In addition, the dynamic substructure method can be subdivided into generalized finite element method and modal synthesis method. If the finite element model of the box is not condensed, the generalized coordinates are physical coordinates, and the process of establishing the coupling model can be classified as generalized finite element method; if the finite element model of the box is condensed (this is most cases), the generalized coordinates are modal coordinates, and the process of establishing the coupling model can be classified as modal synthesis method.

To sum up, there are some problems in the dynamic modeling of transmission system box coupling

① When the full finite element model is used to solve the dynamic response of the gearbox, the dynamic meshing force applied on the model is calculated by the lumped parameter method in advance, but the dynamic meshing force does not consider the influence of the dynamic deformation of the gearbox on the meshing state of the gear pair, and does not consider the influence of the coupling gearbox on the natural vibration characteristics of the transmission system, so the transmission system and the gearbox are not truly coupled.

② When the dynamic substructure method is used to establish the dynamic model of transmission system box coupling, the substructure coupling is mainly carried out by manpower, that is, the substructure models are manually integrated into the system model according to the connection relationship of components. This is not difficult for the simple system in the example, but for the complex multi-stage gear system, the modeling is not only time-consuming and laborious, but also easy to make mistakes. Therefore, in the study of complex objects, many scholars have to rely on commercial multi-body dynamics software, such as Adams, simpac, romax, dads, RecurDyn and so on. Although the operation of commercial software is simple, it is lack of pertinence, and the effect often can not meet the requirements of specific tasks; and the program code is closed, so it is difficult to analyze the reasons and improve the algorithm when the application effect is poor.