Basin stability for updating system uncertainties.

Abstract

In this paper, we propose an application of the basin stability tool which allows us to update the information on the system properties under parameter uncertainties. The concept is presented using a classical mechanical setup of coupled pendula, exchanging the energy via the supporting structure. Depending on the support parameters, the model can exhibit different types of coexisting synchronous patterns as well as remaining desynchronized. We calculate basin stability maps of particular behaviors and combine them with prior parameter distributions using Bayesian inference. The obtained posterior distributions, based on the attractor occurrence, update our knowledge on the system properties in the terms of probabilities. We also underline the problem of evaluating basin stability close to the existence borders, comparing the classical approach of fixed parameters with the one involving variations. The differences between the estimation methods can have a crucial meaning for the discussed application and should be considered carefully. The results presented in this paper uncover ways of applying the basin stability concept, which can be used to study the properties of complex dynamical systems from a probability perspective.