Don Price

On Thermodynamic Interpretation of Transfer Entropy

We propose a thermodynamic interpretation of transfer entropy near equilibrium, using a specialised Boltzmann’s principle. The approach relates conditional probabilities to the probabilities of the corresponding state transitions. This in turn characterises transfer entropy as a difference of two entropy rates: the rate for a resultant transition and another rate for a possibly irreversible transition within the system affected by an additional source. We then show that this difference, the local transfer entropy, is proportional to the external entropy production, possibly due to irreversibility. Near equilibrium, transfer entropy is also interpreted as the difference in equilibrium stabilities with respect to two scenarios: a default case and the case with an additional source. Finally, we demonstrated that such a thermodynamic treatment is not applicable to information flow, a measure of causal effect.

Self-Organizing Hierarchies in Sensor and Communication Networks

We consider a hierarchical multicellular sensing and communication network, embedded in an ageless aerospace vehicle that is expected to detect and react to multiple impacts and damage over a wide range of impact energies. In particular, we investigate self-organization of impact boundaries enclosing critically damaged areas, and impact networks connecting remote cells that have detected noncritical impacts. Each level of the hierarchy is shown to have distinct higher-order emergent properties, desirable in self-monitoring and self-repairing vehicles. In addition, cells and communication messages are shown to need memory (hysteresis) in order to retain desirable emergent behavior within and between various hierarchical levels. Spatiotemporal robustness of self-organizing hierarchies is quantitatively measured with graph-theoretic and information-theoretic techniques, such as the Shannon entropy. This allows us to clearly identify phase transitions separating chaotic dynamics from ordered and robust patterns.

On connectivity of reconfigurable impact networks in ageless aerospace vehicles

The research results presented in this paper were obtained as part of the joint CSIRO-NASA Ageless Aerospace Vehicle (AAV) project. We describe the underlying principles, methodology, and preliminary results of modelling and simulating a multi-cellular sensor and communication network in a dynamic decentralised setting, motivated by a self-monitoring, self-repairing AAV. Such networks are expected to detect and react to multiple impacts and damage over a wide range of impact energies. In particular, we address the problem of forming a reconfigurable network (a minimum spanning tree) connecting cells that detected non-critical impacts, in presence of connectivity disruptions caused by critical impacts. The presented algorithm is based on the ant colony metaphor and may be complemented by gradient-based techniques. In addition, we measure the robustness of impact networks and present quantitative metrics that clearly identify phase transitions in network connectivity, separating chaotic dynamics from ordered and robust patterns.

Sensor network for structural health monitoring

Structural health monitoring (SHM) uses an array of sensors to continuously monitor a structure to provide an early indication of problems such as damage to the structure from fatigue, corrosion or impact. The use of such a system enables maintenance costs to be reduced, and new structures can be designed to be lighter and more efficient. CSIRO has developed an SHM system for detecting high-velocity impacts in the skin of a structure, such as may occur to space vehicles. The system is a large sensor network containing about two-hundred nodes, each of which contains multiple sensors. The system has been built as a flexible testbed for undertaking research in the use of sensor networks in a wide range of SHM applications. This paper outlines the testbed that has been developed and the research that has been conducted using this testbed.

Self-organising impact boundaries in ageless aerospace vehicles

Self-monitoring, self-repairing aerospace vehicles require modular, flexible and adaptive sensing and communication networks. In general, a modular (multi-cellular) sensing and communication network is expected to detect and react to impact location, energy and damage over a wide range of impacts. It is critical that global response emerges as a result of interactions involving transfer of information embedded locally, avoiding single points-of-failure. This work presents mechanisms ensuring self-organisation of autonomous cells into robust and continuously connected impact boundaries. The spatiotemporal stability is demonstrated for a variety of cell shapes in a dynamic environment with varying energy dissipation and damage probability models.

Development of a sensor-based learning approach to prognostics in intelligent vehicle health monitoring

This paper reports the development and testing of an intelligent sensing network for monitoring corrosion of aircraft structures. The agent-based system uses sensed microclimate and corrosion data with an innovative analysis approach to diagnose corrosion and to infer the presence of corrosion in locations such as crevices and fasteners where it cannot be sensed directly. An important feature of the system is its ability to provide prognostic information to enable timely corrective maintenance scheduling.

Towards adaptive clustering in self-monitoring multi-agent networks

A Decentralised Adaptive Clustering (DAC) algorithm for self-monitoring impact sensing networks is presented within the context of CSIRO-NASA Ageless Aero-space Vehicle project. DAC algorithm is contrasted with a Fixed-order Centralised Adaptive Clustering (FCAC) algorithm, developed to evaluate the comparative performance. A number of simulation experiments is described, with a focus on the scalability and convergence rate of the clustering algorithm. Results show that DAC algorithm scales well with increasing network and data sizes and is robust to dynamics of the sensor-data flux.

A self-organising sensing system for structural health management

This paper describes a new approach to structural health monitoring and management (SHM) that aims to diagnose and respond to damage using the self-organization of a complex system of distributed sensors and processing cells. To develop and evaluate the approach, an experimental SHM test-bed system has been developed, with the aim of detecting and characterising the damage from high-velocity impacts such as those due to micrometeoroids on a space vehicle. An important new feature of the system is an ability to support mobile (robotic) agents that can roam the exterior surface of the test-bed, obtaining additional damage information and providing a crude repair capability. The focus of this paper is the development of a self-organised approach to the operation of such a robotic agent, for which it obtains local information by direct communication with the fixed agents embedded in the underlying structure.

An Intelligent Sensor System for Detection and Evaluation of Particle Impact Damage

An integrated vehicle health monitoring system is being developed, with its initial application being to detect and locate impacts of fast particles, and characterise and respond to the damage such collisions may cause. Large numbers of sensors are used, with the continuous monitoring and processing of signals being performed by autonomous local agents that communicate only essential information with their nearest neighbours. This multi‐agent system is completely scalable and generates emergent and intelligent responses, making possible the detection and diagnosis of damage even in the presence of existing damage in the system or in the structure it is monitoring.

Optimising Sensor Layouts for Direct Measurement of Discrete Variables

An optimal sensor layout is attained when a limited number of sensors are placed in an area such that the cost of the placement is minimized while the value of the obtained information is maximized. In this paper, we discuss the optimal sensor layout design problem from first principles, show how an existing optimization criterion (maximum entropy of the measured variables) can be derived, and compare the performance of this criterion with three others that have been reported in the literature for a specific situation for which we have detailed experimental data available. This is achieved by firstly learning a spatial model of the environment using a Bayesian Network, then predicting the expected sensor data in the rest of the space, and finally verifying the predicted results with the experimental measurements. The development of rigorous techniques for optimizing sensor layouts is argued to be an essential requirement for reconfigurable and self-adaptive networks.