Barbara Cannas

Hyperchaotic behaviour of two bi-directionally coupled Chua's circuits

In this paper, a non-linear bi-directional coupling of two Chuas circuits is presented. The coupling is obtained by using polynomial functions that are symmetric with respect to the state variables of the two Chuas circuits. Both a transverse and a tangent system are studied to ensure a global validity of the results in the state space. First, it is shown that the transverse system is an autonomous Chuas circuit, which directly allows the evaluation of the conditions on its chaotic behaviour, i.e. the absence of synchronization between the coupled circuits. Moreover, it is demonstrated that the tangent system is also a Chuas circuit, forced by the transverse system; therefore, its dynamics is ruled by a time-dependent equation. Thus, the calculus of conditional Lyapunov exponents is necessary in order to exclude antisynchronization along the tangent manifold. The properties of the transverse and tangent systems simplify the study of the coupled Chuas circuits and the determination of the conditions on their hyperchaotic behaviour. In particular, it is shown that hyperchaotic behaviour occurs for proper values of the coupling strength between the two Chuas circuits. Finally, numerical examples are given and discussed. Copyright

River flow forecast for reservoir management through neural networks

Abstract River flow forecasts are required to provide basic information for reservoir management in a multipurpose water system optimisation framework. An accurate prediction of flow rates in tributary streams is crucial to optimise the management of water resources considering extended time horizons. Moreover, runoff prediction is crucial in protection from water shortage and possible flood damages. In this paper, a neural approach is used to model the rainfall-runoff process when different time step durations have to be considered in reservoir management. Numerical comparisons with observed data are provided for runoff prediction in the Tirso basin at the S.Chiara section in Sardinia (Italy).

Learning of Chua's circuit attractors by locally recurrent neural networks

Abstract Many practical applications of neural networks require the identification of strongly non-linear (e.g., chaotic) systems. In this paper, locally recurrent neural networks (LRNNs) are used to learn the attractors of Chuas circuit, a paradigm for studying chaos. LRNNs are characterized by a feed-forward structure whose synapses between adjacent layers have taps and feedback connections. In general, the learning procedures of LRNNs are computationally simpler than those of globally recurrent networks. Results show that LRNNs can be trained to identify the underlying link among Chuas circuit state variables, and exhibit chaotic attractors under autonomous working conditions.

Disruption forecasting at JET using neural networks

Neural networks are trained to evaluate the risk of plasma disruptions in a tokamak experiment using several diagnostic signals as inputs. A saliency analysis confirms the goodness of the chosen inputs, all of which contribute to the network performance. Tests that were carried out refer to data collected from succesfully terminated and disruption terminated pulses performed during two years of JET tokamak experiments. Results show the possibility of developing a neural network predictor that intervenes well in advance in order to avoid plasma disruption or mitigate its effects.

An adaptive real-time disruption predictor for ASDEX Upgrade

In this paper, a neural predictor has been built using plasma discharges selected from two years of ASDEX Upgrade experiments, from July 2002 to July 2004. In order to test the real-time prediction capability of the system, its performance has been evaluated using discharges coming from different experimental campaigns, from June 2005 to July 2007. All disruptions that occurred in the chosen experimental campaigns were included with the exception of those occurring in the ramp-up phase, in the ramp-down phase (if the disruption does not happen in the first 100 ms), those caused by massive gas injection and disruptions following vertical displacement events. The large majority of selected disruptions are of the cooling edge type and typically preceded by the growth of tearing modes, degradation of the thermal confinement and enhanced plasma radiation. A very small percentage of them happen at large beta after a short precursor phase. For each discharge, seven plasma diagnostic signals have been selected from numerous signals available in real-time. During the training procedure, a self-organizing map has been used to reduce the database size in order to improve the training of the neural network. Moreover, an optimization procedure has been performed to discriminate between safe and pre-disruptive phases. The prediction success rate has been further improved, performing an adaptive training of the network whenever a missed alarm is triggered by the predictor.

Neural network-based analog fault diagnosis using testability analysis

A fault diagnosis procedure for analog linear circuits is presented. It uses an off-line trained neural network as a classifier. The innovative aspect of the proposed approach is the way the information provided by testability and ambiguity group determination is exploited when choosing the neural network architecture. The effectiveness of the proposed approach is shown by comparing with similar work that has already appeared in the literature.

Neural networks for power system condition monitoring and protection

Abstract To maintain a good voltage quality level to customers in power distribution system, it is essential to minimise transients, line voltage dips and spikes due to a variety of causes, including fault occurrence, power interruption and large load changes. In case of private generating systems it is essential that ultra-rapid switching devices be used which cut off the customer plant from the utility system so quickly that the presence of voltage dips is not perceived by the industrial plants sensitive loads. These devices require very fast acquisition and control systems which permit to diagnose and, possibly, predict abnormal events. In this paper, a control methodology based on a locally recurrent-globally feed-forward neural network and on a neural classifier is proposed. It will be shown that it is possible to predict with good accuracy the value of the control variables based on previously acquired samples and use these values to recognise the kind of abnormal event that is about to occur on the network.

Performance analysis of locally recurrent neural networks

Many practical applications of neural networks require the identification of non‐linear deterministic systems or chaotic systems. In these cases the use of a network architecture known as locally recurrent neural network (LRNN) is often preferable in place of standard feedforward multi‐layer perceptron (MLP) networks, or of globally recurrent neural network. In this paper locally recurrent networks are used to simulate the behaviour of the Chua’s circuit that can be considered a paradigm for studying chaos. It is shown that such networks are able to identify the underlying link among the state variables of the Chua’s circuit. Moreover, they are able to behave like an autonomous Chua’s double scroll, showing a chaotic behaviour of the state variables obtainable through a suitable circuit elements choice.

Algorithms for the Automatic Identification of MARFEs and UFOs in JET Database of Visible Camera Videos

MARFE instabilities and UFOs leave clear signatures in JET fast visible camera videos. Given the potential harmful consequences of these events, particularly as triggers of disruptions, it would be important to have the means of detecting them automatically. In this paper, the results of various algorithms to identify automatically the MARFEs and UFOs in JET visible videos are reported. The objective is to retrieve the videos, which have captured these events, exploring the whole JET database of images, as a preliminary step to the development of real-time identifiers in the future. For the detection of MARFEs, a complete identifier has been finalized, using morphological operators and Hu moments. The final algorithm manages to identify the videos with MARFEs with a success rate exceeding 80%. Due to the lack of a complete statistics of examples, the UFO identifier is less developed, but a preliminary code can detect UFOs quite reliably.

Neural reconstruction of Lorenz attractors by an observable

Abstract In this paper, Locally Recurrent Neural Networks (LRNNs) are used to learn the chaotic trajectory of the Lorenz system starting from measurements of an observable. An analysis of the Lorenz system based on algebraic observability ensures, in principle, the feasibility of the reconstruction. Thus, algebraic observability of a non-linear system determines the sufficient conditions for the existence of a solution for a state reconstruction problem. Moreover, this procedure can be directly implemented by LRNNs. Results show that LRNNs can be trained to identify and to predict the link among the Lorenz state variables starting from a proper observable, i.e., the LRNNs reconstruct the Lorenz attractor. Furthermore, the LRNNs exhibit generalization properties, i.e., they are characterized by a chaotic Lorenz-like strange attractor under autonomous working conditions.