A probabilistic linearization method for non-linear systems subjected to additive and multiplicative excitations

A general method to obtain approximate solutions for the random response of non-linear systems subjected to both additive and multiplicative Gaussian white noises is presented. Starting from the concept of linearization, the proposed method of "Probabilistic Linearization" (PL) is based on the replacement of the Fokker-Planck equation of the original non-linear system with an equivalent one relative to a linear system subjected to additive excitation only. By means of the general scheme of the weighted residuals, the unknown coefficients of the equivalent system are determined. Assuming a Gaussian probability density function of the response process and by choosing the weighting functions in a suitable way, the equivalence of the proposed method, called "Gaussian Probabilistic Linearization" (GPL), with the "Gaussian Stochastic Linearization" (GSL) applied to the coefficients of the Ito differential rule is evidenced. In addition, the generalization of the proposed method, called "Generalized Gaussian Probabilistic Linearization" (GGPL), is presented. Numerical applications show as, varying the choice of the weighting functions, it is possible to obtain different linearizations, with a variable degree of accuracy. For the two examples considered, different suitable combinations of the weighting functions lead to different equivalent linear systems, all characterized by the exact solution in terms of variance.