Improved Simulation of Models with Complex User-Subroutines in MSC.ADAMS 2006-01-0158
The vast majority of the analysis types available in MSC. ADAMS, (dynamics, statics, kinematics, etc.) require computation of partial derivatives of all quantities that define the model, including forces, variables, differential equations, etc., specified in external user-subroutines. Computing accurate partial derivatives (sometimes referred to as Jacobian information, or sensitivities) is an essential requirement for fast and accurate simulation. The paper first presents the context in which the Jacobian information is used and the default mechanism through which MSC.ADAMS obtains this information. A new functional interface available in the new-generation MSC.ADAMS/Solver(C++) is discussed, along with the approach for using this communication channel to provide Jacobian information back to the solver. Also presented are three approaches to generate required Jacobian information given an arbitrary user-subroutine. The paper reports CPU timing results and accuracy comparisons for simulations run with and without the proposed methodology for Jacobian computation.