No label defined (Q4247435)

The author discusses the well-established methods of ``equivalent linearization'' (approximation of a nonlinear random dynamical system by a linear one which is nearest in a mean square sense) and ``stochastic averaging'' (Stratonovich-Krylov-Bogolyubov-Fatou) in mechanical engineering. In particular, these methods are useful to study characteristics of second order nonlinear stochastic differential equations \(\ddot{x}+f(x,\dot{x}) = \sigma \dot{W}\) with additive noise. It is interesting to note that the linearization methods fail to fit the power spectral density of stationary response of the dynamical system, whereas stochastic averaging can be successful for a Duffing-type oscillator and stochastic oscillators with impacts. However, it is still an open problem how to extend and implement practically those methods for multidimensional systems (i.e. mechanical systems with more than one degree of freedom). The paper is descriptive and easily readable without mathematical proofs, but with hints where the reader can find some rigorous mathematical justifications as with Freidlin-Wentzell theory for random perturbations. This work can be recommended to those who want to meet interesting crossrelations between mathematical theory and applications in engineering. For the entire collection see [Zbl 0914.00073].