C. Kambhampati

A Generic Strategy for Fault-Tolerance in Control Systems Distributed Over a Network

This paper provides a tutorial overview of a number of aspects and approaches to Control over the Network for Network Control Systems (NCS) that are likely to lead to good fault-tolerant control properties, subject to network faults. In order to analyze and derive the best strategies for fault tolerant NCS, it is initially assumed that the network communication bandwidth is infinite. This gives a simpler way to map the NCS structure in terms of computing nodes and control subsystems/components. Two Fault-tolerant NCS architectures have been described, analyzed and compared with view to demonstrating that the classical concepts of Fault-tolerant Control (FTC), namely of active and passive FTC can be related to equivalent (although more complex) concepts in NCS. The study confirms that the de-centralized approach to fault-tolerant control of NCS suffers from a difficult challenge as to how to compensate for fault effects occurring throughout the NCS. On the other hand, the distributed hierarchical structure, requires a coordination function which is able to manage (a) the local control task, (b) the compensation of faults, and (c) the network reconfiguration, if required, subject to significant network subsystem faults.

A stable one-step-ahead predictive control of non-linear systems

In this paper stability of one-step ahead predictive controllers based on non-linear models is established. It is shown that, under conditions which can be fulfilled by most industrial plants, the closed-loop system is robustly stable in the presence of plant uncertainties and input-output constraints. There is no requirement that the plant should be open-loop stable and the analysis is valid for general forms of non-linear system representation including the case out when the problem is constraint-free. The effectiveness of controllers designed according to the algorithm analyzed in this paper is demonstrated on a recognized benchmark problem and on a simulation of a continuous-stirred tank reactor (CSTR). In both examples a radial basis function neural network is employed as the non-linear system model.

Dynamic Neural Networks

From the quest for artificial intelligence by emulating the brain and behaviour of living organisms, there have been several encouraging outcomes. The study of the mechanics and structure of the brain has led to the development of new computational models, inspired by the connectionism of the nervous system, for solving complex numerical problems such as pattern recognition, system modelling and information processing. These mathematical models are known as artificial neural networks (ANNs) or simply neural networks (NNs) and they consist of a set of interconnected processing units or artificial neurons [76, 14]. Depending on the internal structure of the neuron and the interconnections between them, the neural network can be static or dynamic.

Reconfiguration in Networked Control Systems: Fault Tolerant Control and Plug-and-Play

: The last decade has seen a rapid increase and convergence in the use of (a) digital network technology, (b) embedded systems and (c) fault analysis and on-line diagnosis. This has resulted in an increase in complexity of the resultant system which must not only ensure performance but also monitor and compensate for faults and hence avoid total failure. The advances made have also facilitated a plug-and-play (PnP) characteristic of networked embedded systems. This paper provides (i) a suitable architecture for the fault tolerant operation and (ii) enables the above PnP feature.

Fault-tolerance as a key requirement for the control of modern systems

Abstract Networks of Embedded Systems are becoming ubiquitous today. The performance of these networks is measured in terms of the Quality of Service (QoS) delivered. This has been taken on board by the Computer Scientists, who have developed concepts like “Ubiquitous” and “Pervasive” Computing. In the world of Control, there has always been an “implicit” QoS, in that the quality or level of performance has been measured using a cost function, often the error between the reference signals and the system outputs. However, such “point-to-point” notions of QoS are fast becoming redundant in the networked, information-rich world. This paper outlines a new way of formulating the Control problem which is suitable for the networked world, enabling Fault Tolerance to become a natural consequence to ensure that the system performance is maintained under all eventualities. Thus Control has to become more ubiquitous, pervasive, and intelligent. To facilitate this outcome, this paper proposes a new research direction which could be termed “Embedded Cognitive Control”, bringing together the various fields of Cognitive Science, Embedded Systems, and Control.

Dynamic Systems in Neural Networks

Many schemes for the employment of neural networks in control systems have been proposed [9] and some practical applications have also been made [2]. It is possible to apply a neural network to just about every conceivable control problem, however in many cases, although of interest, the network might not be the best or even a good solution, due to its relatively complex nonlinear operation. A neural network is in essence a nonlinear mapping device and in this respect, at the present time, most of the reported work describing the use of neural networks in a control environment is concerned solely with the problem of process modelling or system identification.

Robust FDI for FTC Coordination in a Distributed Network System

Abstract This paper focuses on the development of a suitable Fault Detection and Isolation (FDI) strategy for application to a system of inter-connected and distributed systems, as a basis for a fault-tolerant Network Control System (NCS) problem. The work follows a recent study showing that a hierarchical decentralized control system architecture may be suitable for fault-tolerant control (FTC) of a network of distributed and interacting subsystems. The main idea is to use robust FDI methods to facilitate the discrimination between faults acting within one subsystem and faults acting in other areas of the network, so that a powerful form of active FTC of the NCS can be implemented, through an autonomous network coordinator. By using a robust form of the Unknown Input Observer (UIO), fault effects in each subsystem are de-coupled from the other subsystems, thus facilitating a powerful way to achieve local FDI in each subsystem under autonomous system coordination. Whilst the autonomous distributed control system provides active FTC under learning control, the FDI-based Reconfiguration Task enhances the network fault-tolerance, so that more significant subsystem faults can be accommodated in order to achieve a suitable standard of Quality of Performance (QoP) of the NCS.

A comparative study of missing value imputation with multiclass classification for clinical heart failure data

Clinical data often contains missing values. Imputation is one of the best known schemes to overcome the drawbacks associated with missing values in data mining tasks. In this work, we compared several imputation methods and analyzed their performance when applied to different classification algorithms. A clinical heart failure data set was used in these experiments. The results showed that there is no universal imputation method that performs best for all classifiers. Some imputation-classification combinations are recommended for the processing of clinical heart failure data.

Handling missing values in data mining - A case study of heart failure dataset

In this paper, we investigate the characteristics of a clinical dataset using feature selection and classification techniques to deal with missing values and develop a method to quantify numerous complexities. The research aims to find features that have high effect on mortality time frame, and to design methodologies which will cope with the following challenges: missing values, high dimensionality, and the prediction problem. The experimental results will be extended to develop prediction model for HF This paper also provides a comprehensive evaluation of a set of diverse machine learning schemes for clinical datasets.

An interaction predictive approach to fault-tolerant control in network control systems

Abstract This paper illustrates some of the capabilities of previously proposed network control system (NCS) architectures to carry on functioning in the event of faults, without recourse to system reconfiguration. The principle of interaction prediction is used to set up a coordination strategy that encapsulates an ability to withstand or tolerate certain faults, thereby allowing the system to continue functioning. It is also shown that the coordination strategy can be made more effective if a learning agent is allowed to learn the coordination functions. This facilitates the use of different types of agent at the local level, together with recurrent networks and genetic algorithms (GAs) at the coordination level. The experimental test-bed system is a benchmark three-tank system that has some of the main features of an industrial process control system.