Narrow Results

Glass networks have been proposed as a model framework for gene regulation, chemical kinetics and neural networks. Their main distinguishing feature is that although the network variables evolve continuously in time, interactions between them depend discontinuously on their sign (i.e. above or below a threshold). While this is a simplification, it has tremendous analytic advantages if the approximation is reasonable in an application. This study explores and classifies bifurcations in Glass networks, and relates them to bifurcations of smooth systems. These bifurcations can often not be studied with traditional bifurcation theory, as the vector fields are discontinuous. However, the theory that has been developed for periodic orbits of Glass networks allows a natural classification for bifurcations of periodic orbits. Some of these are shown to correspond to smooth-system bifurcations, others are shown to fit into the framework of "C-bifurcations" or "border-collision bifurcations" and others are shown to allow truly ambiguous behavior, for which Filippov's theory for discontinuous vector fields is an appropriate tool. Routes to chaos are also explored, and it is demonstrated that period-doubling cascades do not occur. However, sudden transitions to chaos, which are common in Glass networks, can result in a limiting sense from compression to a point of a period-doubling cascade in corresponding networks with sigmoidal interactions as the sigmoid's gain is increased. Other phenomena such as intermittency and multistability are also discussed.

In this paper, we propose a qualitative and quantitative dynamical study about the evolution of nonsmooth torus in a Digital Delayed Pulse-Width Modulator (PWM) switched buck converter. We explain the birth and destruction of the torus by successive discontinuity induced bifurcations (DIBs). The Digital-PWM control is based on Zero Average Dynamics (ZAD) strategy and a one-period delay is included in the control law. The control parameter (k_s) of the ZAD strategy can be varied in a large range, ideally (-∞, ∞). On these borders, the dynamical behavior is the same and thus an annulus-like parameter space is considered. Under variation of k_s, the system gets closer to a codimension-two bifurcation point, where two simultaneous border-collision bifurcations (BC) meet. The system leaves the high-order periodic behavior and a high-order band torus appears. Wrinkles and nonsmoothness due to more and more successive border-collisions bifurcations cause the torus destruction in high-order band chaos with tent-map-like structures and Mandelbrot-like sets. Finally, the chaotic bands merge in one-band chaos. The switched converter is modeled as a piecewise linear system where an analytical expression of the Poincaré map is available. Characteristics as duality, symmetry and recurrent behavior are determined using additional numerical methods for the Poincaré map decomposition in periodic sequences.

In this paper, we report Mandelbrot-like bifurcation structures in a one-dimensional parameter space of real numbers corresponding to a dc-dc power converter modeled as a piecewise-smooth system with three zones. These fractal patterns have been studied in two-dimensional parameter space for smooth systems, but for nonsmooth systems has not been reported yet. The Mandelbrot-like sets we found are created in transition from the torus band to chaos band scenarios exhibited by a dc-dc buck power converter controlled by Delayed Pulse-Width Modulator (PWM) based on Zero Average Dynamics (or ZAD strategy), which corresponds to a piecewise-smooth system (PWS). The real parameter is provided by the PWM control strategy, namely ZAD strategy, and it can be varied in a large range, ideally (-∞, +∞). At -∞ and +∞ the dynamical behavior is the same, and thus we will describe the synamics in an ring-like parameter space. Mandelbrot-like borders are built by four chaotic bands, therefore these structures can be thought as instability islands where the state variables cannot be located. Using the Poincaré map approach we characterize the bifurcation structures and we describe recurrent patterns in different scales.

The paper describes some aspects of sudden transformations of closed invariant curves in a 2D piecewise smooth map. In particular, using detailed numerically calculated phase portraits, we discuss transitions from smooth to piecewise smooth closed invariant curves. We show that such transitions may occur not only when a closed invariant curve collides with a border but also via a homoclinic bifurcation. Furthermore, we describe an unusual transformation from a closed invariant curve to a large amplitude chaotic attractor and demonstrate that this transition occurs in two steps, involving a small amplitude closed-invariant-curve-like chaotic attractor.