Estimation of the \(f(\alpha)\) spectrum from simulated and measured time series (Q1196036)

The chaotic dynamics of dissipative nonlinear systems is characterized by the investigation of their strange attractors as sets formed by the sampled orbit in phase space. Multifractal measures can also be characterized by the ``generalized dimensions'' \(D(q)\) or by the ``singular spectrum'' \(f(\alpha)\), the two being related via Legendre transformation. The paper compares several global methods for estimating the \(f(\alpha)\) spectrum from chaotic time series including box-covering and correlation- integral algorithms. Data series obtained from two-dimensional dissipative Baker's transformation were used. Calculations confirm the observation that the needed number of points depends sensitively on the embedding dimension. The authors report on estimations of the \(f(\alpha)\) spectrum for the chaotic motion of a periodically disturbed discharge plasma. The conclusion are that generalized correlation-integral methods are well suited for the estimation of the generalized dimensions \(D(q)\) for not too large \(| q|\).