An efficient treatment of parameter identification in the context of multibody system dynamics using the adjoint method

In multibody system dynamics, a wide range of parameters can occur where some of them may not be known a priori. Therefore, this work presents an efficient adjoint method for parameter identification that can be utilized in multibody simulation software. Compared to standard system sensitivity based approaches the adjoint method has the major advantage of being independent on the number of parameters to identify. Especially when dealing with large and probably flexible multibody systems this characteristic is crucial. Formulating parameter identification as an automatable procedure, of course, leads to a complicated structure of the involved matrices and equations. However, during a forward simulation of the system many of the matrices needed for solving the so called “adjoint system equations” are already evaluated. Adopting the functionality of the forward solver for the adjoint system solver therefore results in little additional effort. In order to illustrate the performance of the adjoint method two examples are presented. A planar example shows the possibility of identifying non-linear control parameters and a three-dimensional example is presented for identifying time-invariant inertia parameters.