Neural network-based wind vector retrieval from satellite scatterometer data

Obtaining wind vectors over the ocean is important for weather forecasting and ocean modelling. Several satellite systems used operationally by meteorological agencies utilise scatterometers to infer wind vectors over the oceans. In this paper we present the results of using novel neural network based techniques to estimate wind vectors from such data. The problem is partitioned into estimating wind speed and wind direction. Wind speed is modelled using a multi-layer perceptron (MLP) and a sum of squares error function. Wind direction is a periodic variable and a multi-valued function for a given set of inputs; a conventional MLP fails at this task, and so we model the full periodic probability density of direction conditioned on the satellite derived inputs using a Mixture Density Network (MDN) with periodic kernel functions. A committee of the resulting MDNs is shown to improve the results.

Outlier detection in scatterometer data:Neural network approaches

Satellite-borne scatterometers are used to measure backscattered micro-wave radiation from the ocean surface. This data may be used to infer surface wind vectors where no direct measurements exist. Inherent in this data are outliers owing to aberrations on the water surface and measurement errors within the equipment. We present two techniques for identifying outliers using neural networks; the outliers may then be removed to improve models derived from the data. Firstly the generative topographic mapping (GTM) is used to create a probability density model; data with low probability under the model may be classed as outliers. In the second part of the paper, a sensor model with input-dependent noise is used and outliers are identified based on their probability under this model. GTM was successfully modified to incorporate prior knowledge of the shape of the observation manifold; however, GTM could not learn the double skinned nature of the observation manifold. To learn this double skinned manifold necessitated the use of a sensor model which imposes strong constraints on the mapping. The results using GTM with a fixed noise level suggested the noise level may vary as a function of wind speed. This was confirmed by experiments using a sensor model with input-dependent noise, where the variation in noise is most sensitive to the wind speed input. Both models successfully identified gross outliers with the largest differences between models occurring at low wind speeds. © 2003 Elsevier Science Ltd. All rights reserved.

The n-tuple network: coping with sub-optimality

The subject of this thesis is the n-tuple net.work (RAMnet). The major advantage of RAMnets is their speed and the simplicity with which they can be implemented in parallel hardware. On the other hand, this method is not a universal approximator and the training procedure does not involve the minimisation of a cost function. Hence RAMnets are potentially sub-optimal. It is important to understand the source of this sub-optimality and to develop the analytical tools that allow us to quantify the generalisation cost of using this model for any given data. We view RAMnets as classifiers and function approximators and try to determine how critical their lack of' universality and optimality is. In order to understand better the inherent. restrictions of the model, we review RAMnets showing their relationship to a number of well established general models such as: Associative Memories, Kamerva's Sparse Distributed Memory, Radial Basis Functions, General Regression Networks and Bayesian Classifiers. We then benchmark binary RAMnet. model against 23 other algorithms using real-world data from the StatLog Project. This large scale experimental study indicates that RAMnets are often capable of delivering results which are competitive with those obtained by more sophisticated, computationally expensive rnodels. The Frequency Weighted version is also benchmarked and shown to perform worse than the binary RAMnet for large values of the tuple size n. We demonstrate that the main issues in the Frequency Weighted RAMnets is adequate probability estimation and propose Good-Turing estimates in place of the more commonly used :Maximum Likelihood estimates. Having established the viability of the method numerically, we focus on providillg an analytical framework that allows us to quantify the generalisation cost of RAMnets for a given datasetL. For the classification network we provide a semi-quantitative argument which is based on the notion of Tuple distance. It gives a good indication of whether the network will fail for the given data. A rigorous Bayesian framework with Gaussian process prior assumptions is given for the regression n-tuple net. We show how to calculate the generalisation cost of this net and verify the results numerically for one dimensional noisy interpolation problems. We conclude that the n-tuple method of classification based on memorisation of random features can be a powerful alternative to slower cost driven models. The speed of the method is at the expense of its optimality. RAMnets will fail for certain datasets but the cases when they do so are relatively easy to determine with the analytical tools we provide.

The fatigue ageing of vulcanised rubbers

The cause of the respective rough and smooth fatigue failure surfaces of Neoprene GS : Neoprene W and Neoprene GS : natural rubber vulcanisates is investigated. The contrasting morphology of the vulcanisates is found to be the major factor determining the fatigue behaviour of the blends. Neoprene GS and Neoprene W appear to form homogeneous blends which exhibit physical properties and fatigue failure surfaces intermediate between those of the two horropolymers. Neoprene GS and natural rubber exhibit heterogeneity when blended together. The morphology of these blends is found to influence both the fatigue resistance and failure surface of the vulcanisates. Exceptional uncut and cut initiated fatigue lives are observed for blends having an interconnecting network morphology. The network structure and cross-link density of the elastomers in the blends and the addition of carbon black and antioxidant are all found to influence the fatigue resistance but not the failure mechanism of the vulcanisate.

Fatigue in rubber vulcanisates

A study has been made of the effect of single extensions and continuous fatigue on the structures of various natural rubber networks. The change in network structure of a conventional vulcanisate on a single extension manifests itself as permanent set. The change in network structure has been assessed by the use of the chemical probes, propan-2-thiol/piperidine, hexane-thiol/piperidine and triphenyl phosphine, which determine the polysulphide and disulphide crosslink densities and main chain modification respectively. The permanent set induced on a single extension of a conventional sulphur vulcanisate has been shown to result from the destruction and reformation of polysulphide crosslinks. The magnitude of the effect was dependent upon the degree of extension and showed a maximum at extensions corresponding to the onset of stress-induced crystallisation. The incorporation of a reinforcing filler, HAF-carbon black, magnified the effect. Vulcanisates that possessed only mono and disulphide crosslinks did not show any significant permanent set. The continuous changes in network structure during fatigue have also been determined, and the effects of carbon black and antioxidants on these changes and the fatigue life have been assessed. During fatigue the overall crosslink density increased slightly, which resulted from the destruction of polysulphide crosslinks. and their replacement by principally disulphide crosslinks. Antioxidants reduced the rate of destruction of polysulphide crosslinks and increased the fatigue life of the rubber network. The fatigue life of the network also depended upon the concentration of free chain ends. These chain ends were incorporated into the network by masticating rubber under nitrogen in the presence of bis (diisopropyl)thiophosphoryl disulphide, which improved the fatigue resistance by up to 9%.

A network coding based interference cancelation scheme for wireless ad hoc networks

The performance of wireless networks is limited by multiple access interference (MAI) in the traditional communication approach where the interfered signals of the concurrent transmissions are treated as noise. In this paper, we treat the interfered signals from a new perspective on the basis of additive electromagnetic (EM) waves and propose a network coding based interference cancelation (NCIC) scheme. In the proposed scheme, adjacent nodes can transmit simultaneously with careful scheduling; therefore, network performance will not be limited by the MAI. Additionally we design a space segmentation method for general wireless ad hoc networks, which organizes network into clusters with regular shapes (e.g., square and hexagon) to reduce the number of relay nodes. The segmentation methodworks with the scheduling scheme and can help achieve better scalability and reduced complexity. We derive accurate analytic models for the probability of connectivity between two adjacent cluster heads which is important for successful information relay. We proved that with the proposed NCIC scheme, the transmission efficiency can be improved by at least 50% for general wireless networks as compared to the traditional interference avoidance schemes. Numeric results also show the space segmentation is feasible and effective. Finally we propose and discuss a method to implement the NCIC scheme in a practical orthogonal frequency division multiplexing (OFDM) communications networks. Copyright © 2009 John Wiley & Sons, Ltd.

Dynamic causal modeling of load-dependent modulation of effective connectivity within the verbal working memory network

Neuroimaging studies have consistently shown that working memory (WM) tasks engage a distributed neural network that primarily includes the dorsolateral prefrontal cortex, the parietal cortex, and the anterior cingulate cortex. The current challenge is to provide a mechanistic account of the changes observed in regional activity. To achieve this, we characterized neuroplastic responses in effective connectivity between these regions at increasing WM loads using dynamic causal modeling of functional magnetic resonance imaging data obtained from healthy individuals during a verbal n-back task. Our data demonstrate that increasing memory load was associated with (a) right-hemisphere dominance, (b) increasing forward (i.e., posterior to anterior) effective connectivity within the WM network, and (c) reduction in individual variability in WM network architecture resulting in the right-hemisphere forward model reaching an exceedance probability of 99% in the most demanding condition. Our results provide direct empirical support that task difficulty, in our case WM load, is a significant moderator of short-term plasticity, complementing existing theories of task-related reduction in variability in neural networks. Hum Brain Mapp, 2013. © 2013 Wiley Periodicals, Inc.

A simulation study of service differentiation for the bandwidth request scheme in fixed IEEE 802.16 network

Various flexible mechanisms related to quality of service (QoS) provisioning have been specified for uplink traffic at the medium access control (MAC) layer in the IEEE 802.16 standards. Among the mechanisms, contention based bandwidth request scheme can be used to indicate bandwidth demands to the base station for the non-real-time polling and best-effort services. These two services are used for most applications with unknown traffic characteristics. Due to the diverse QoS requirements of those applications, service differentiation (SD) is anticipated over the contention based bandwidth request scheme. In this paper we investigate the SD with the bandwidth request scheme by means of assigning different channel access parameters and bandwidth allocation priorities at different packets arrival probability. The effectiveness of the differentiation schemes is evaluated by simulations. It is observed that the initial backoff window can be efficient in SD, and if combined with the bandwidth allocation priority, the SD performances will be better.

Role of reinforcement/matrix interfacial strength in fatigue crack propagation in particulate SiC reinforced aluminum alloy 8090

A study has been made of the influence of the reinforcement/matrix interfacial strength on fatigue crack propagation in a powder metallurgy aluminum alloy 8090-SiC particulate composite. The interfacial region has been altered by two separate routes, the first involving aging of the 8090 matrix, with the subsequent formation of precipitate free zones at the boundaries, and the second consisting of oxidizing the surface of the SiC particles before their incorporation into the composite. In the naturally aged condition, oxidation of the SiC leads to a reduction in fatigue crack growth resistance at higher values of stress intensity range ΔK. This is due to a proportion of the crack growth occurring through voids formed in association with many of the weak SiC interfaces which have retained a layer of thick surface oxide after processing. On overaging no difference in crack growth rate is discernible between the oxidized and unoxidized SiC composites. It is proposed that this is due to similar levels of interfacial weakening having occurred in both composites, indicating that this is an important factor in the reduction of the high ΔK crack growth resistance of the unoxidized SiC composite on aging.

Fatigue crack propagation in nickel-base superalloys:effects of microstructure, load ratio, and temperature

The current state of knowledge and understanding of the long fatigue crack propagation behavior of nickel-base superalloys are reviewed, with particular emphasis on turbine disk materials. The data are presented in the form of crack growth rate versus stress intensity factor range curves, and the effects of such variables as microstructure, load ratio, and temperature in the near-threshold and Paris regimes of the curves, are discussed.

Fatigue crack growth behaviour in molten-metal processed SiC particle-reinforced aluminium alloys

Fatigue crack initiation and subsequent short crack growth behaviour of 2014-5wt%SiC aluminium alloy composites has been examined in 4-point bend loading using smooth bar specimens. The growth rates of long fatigue cracks have also been measured at different stress ratios using pre-cracked specimens. The distributions of SiC particles and of coarse constituent particles in the matrix (which arise as a result of the molten-metal processing and relatively slow cooling rate) have been investigated. Preferential crack initiation sites were found to be SiC-matrix interfaces, SiC particles associated with constituent particles and the coarse constituent particles themselves. For microstructurally short cracks the dispersed SiC particles also act as temporary crack arresters. In the long crack growth tests, higher fatigue crack growth rates were obtained than for monolithic alloys. This effect is attributed to the contribution of void formation, due to the decohesion of SiC particles, to the fatigue crack growth process in the composite. Above crack depths of about 200 μm 'short' crack growth rates were in good agreement with the long crack data, showing a Pris exponent, m = 4 in both cases. For the long crack and short crack growth tests little effect of specimen orientation and grain size was observed on fatigue crack growth rates, but, specimen orientation affected the toughness. No effect of stress ratio in the range R = 0.2-0.5 was seen for long crack data in the Paris region.

Effects of dynamic strain ageing on fatigue crack propagation and threshold behaviour in a Ni-base superalloy

In the temperature range 200-400 degree C the Ni-base superalloy, N901, develops marked dynamic strain ageing effects in its tensile behavior. These include inverse strain rate sensitivity, especially in UTS values, strongly serrated stress-strain curves and a heavily sheared failure mode at the higher test-temperatures. As for steels these effects seem to be due to interactions between the dislocations and the interstitial carbon atoms present. The results of tensile and fatigue threshold tests carried out between 20 degree C and 420 degree C are reported and the fatigue behavior is discussed in terms of the effects of surface roughness induced closure, temperature and strain aging interactions.

Study on fatigue and energy-dissipation properties of nanolayered Cu/Nb thin films

Energy dissipation and fatigue properties of nano-layered thin films are less well studied than bulk properties. Existing experimental methods for studying energy dissipation properties, typically using magnetic interaction as a driving force at different frequencies and a laser-based deformation measurement system, are difficult to apply to two-dimensional materials. We propose a novel experimental method to perform dynamic testing on thin-film materials by driving a cantilever specimen at its fixed end with a bimorph piezoelectric actuator and monitoring the displacements of the specimen and the actuator with a fibre-optic system. Upon vibration, the specimen is greatly affected by its inertia, and behaves as a cantilever beam under base excitation in translation. At resonance, this method resembles the vibrating reed method conventionally used in the viscoelasticity community. The loss tangent is obtained from both the width of a resonance peak and a free-decay process. As for fatigue measurement, we implement a control algorithm into LabView to maintain maximum displacement of the specimen during the course of the experiment. The fatigue S-N curves are obtained.

Modular ESS with second life batteries operating in grid independent mode

This work is part of a bigger project which aims to research the potential development of commercial opportunities for the re-use of batteries after their use in low carbon vehicles on an electricity grid or microgrid system. There are three main revenue streams (peak load lopping on the distribution Network to allow for network re-enforcement deferral, National Grid primary/ secondary/ high frequency response, customer energy management optimization). These incomes streams are dependent on the grid system being present. However, there is additional opportunity to be gained from also using these batteries to provide UPS backup when the grid is no longer present. Most UPS or ESS on the market use new batteries in conjunction with a two level converter interface. This produces a reliable backup solution in the case of loss of mains power, but may be expensive to implement. This paper introduces a modular multilevel cascade converter (MMCC) based ESS using second-life batteries for use on a grid independent industrial plant without any additional onsite generator as a potentially cheaper alternative. The number of modules has been designed for a given reliability target and these modules could be used to minimize/eliminate the output filter. An appropriate strategy to provide voltage and frequency control in a grid independent system is described and simulated under different disturbance conditions such as load switching, fault conditions or a large motor starting. A comparison of the results from the modular topology against a traditional two level converter is provided to prove similar performance criteria. The proposed ESS and control strategy is an acceptable way of providing backup power in the event of loss of grid. Additional financial benefit to the customer may be obtained by using a second life battery in this way.

Bayesian neural network approach for determining the risk of re-intervention after endovascular aortic aneurysm repair

This article proposes a Bayesian neural network approach to determine the risk of re-intervention after endovascular aortic aneurysm repair surgery. The target of proposed technique is to determine which patients have high chance to re-intervention (high-risk patients) and which are not (low-risk patients) after 5 years of the surgery. Two censored datasets relating to the clinical conditions of aortic aneurysms have been collected from two different vascular centers in the United Kingdom. A Bayesian network was first employed to solve the censoring issue in the datasets. Then, a back propagation neural network model was built using the uncensored data of the first center to predict re-intervention on the second center and classify the patients into high-risk and low-risk groups. Kaplan-Meier curves were plotted for each group of patients separately to show whether there is a significant difference between the two risk groups. Finally, the logrank test was applied to determine whether the neural network model was capable of predicting and distinguishing between the two risk groups. The results show that the Bayesian network used for uncensoring the data has improved the performance of the neural networks that were built for the two centers separately. More importantly, the neural network that was trained with uncensored data of the first center was able to predict and discriminate between groups of low risk and high risk of re-intervention after 5 years of endovascular aortic aneurysm surgery at center 2 (p = 0.0037 in the logrank test).