The adjoint method is a very powerful tool for parameter identification or optimal control in time domain. In most cases the results lead to some kind of best-fit solution, which means that high frequency components with low amplitudes are not considered. However, in this contribution we present an approach for the identification of parameters which influences the system at special frequencies or frequency ranges. The basic idea is to compute the Fourier coefficients for the relevant oscillations. Then, the cost function consists of the difference of the amplitudes from the simulation and a measured value e.g., from a test bench. The derivation and computation of the gradient of the cost function with the adjoint method is presented and applied to a complex four-cylinder engine model in order to identify the parameters of a torsional vibration damper (TVD).