Digital Twin of a Tribology Test Bench: The Adjoint Gradient Computation for Parameter Identification

This work discusses the application of an optimization algorithm
to determine the parameters for the system dynamics of a tribometer.
To determine the parameters, an optimization algorithm is used that
compares a simulation signal with a measurement signal from the real
test bench. The optimization problem is solved using an indirect method
in which the gradient of a cost function is computed with respect to
a parameter set. To reduce the computational effort for the gradient
determination, the adjoint gradient method is introduced. The method
presented in this work avoids the numerical gradient calculation and can
be used for the iterative determination of optimal parameters. These parameters form the basis for a digital twin of such a tribometer, e.g. to
investigate the coupling between the friction in the model contact and
the vibrations caused by the experimental setup.