Harmonic Chaos
Abstract We introduce Harmonic Chaos, a semi-analytical method for computing stationary statistics of stochastic differential equations driven by Gaussian white noise. The method is based on the combined application of the Karhunen-Lo\'eve expansion of the Wiener process, harmonic balance, and polynomial chaos, and permits to transform stochastic differential equations into systems of algebraic equations, with deterministic coefficients as unknowns. The procedure can be summarized as follows: on a finite time interval, the stochastic differential equation is transformed into a family of periodically forced ordinary differential equations parametrized by independent Gaussian variables by means of a truncated Karhunen-Lo\'eve expansion applied to the Wiener process feeding the stochastic system. This enables use of harmonic balance for their solution. The amplitudes of the harmonics depend on the random variables, in turn expanded through Hermite polynomial chaos. Applying a coupled temporal-stochastic Galerkin projection, we obtain a system of algebraic equations for the coefficients of the polynomial chaos representation. Once the algebraic system has been solved, stationary statistical moments can be calculated directly, while probability density functions can be obtained by sampling the resulting representation without repeatedly solving the dynamical equations. The method is first validated applying it to the Ornstein-Uhlenbeck process, for which the Harmonic Chaos equations can be solved analytically. The application of the method to nonlinear problems is demonstrated by considering a stochastic Duffing oscillator. We show that, by exploiting the symmetries of the system, a reduced spectral representation can be found that allows for a computationally tractable numerical implementation. The resulting stationary marginal distributions of displacement and velocity show good agreement with the corresponding analytical distributions.
Authors
- Fabrizio Bonani (ORCID: https://orcid.org/0000-0002-4964-0100)
- Fabio Traversa
- Kailing Song (ORCID: https://orcid.org/0000-0003-0564-3905)
- Michele Bonnin (ORCID: https://orcid.org/0000-0002-9106-563X)
Institutions
- Politecnico di Torino (IT)
- Memcomputing (United States) (US)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-12
- DOI
- https://doi.org/10.21203/rs.3.rs-11409714/v1
- Primary Topic
- Probabilistic and Robust Engineering Design
- Type
- preprint