Project neptune: Improved operation of water distribution networks

Water service providers (WSPs) in the UK have statutory obligations to supply drinking water to all customers that complies with increasingly stringent water quality regulations and minimum flow and pressure criteria. At the same time, the industry is required by regulators and investors to demonstrate increasing operational efficiency and to meet a wide range of performance criteria that are expected to improve year-on-year. Most WSPs have an ideal for improving the operation of their water supply systems based on increased knowledge and understanding of their assets and a shift to proactive management followed by steadily increasing degrees of system monitoring, automation and optimisation. The fundamental mission is, however, to ensure security of supply, with no interruptions and water quality of the highest standard at the tap. Unfortunately, advanced technologies required to fully understand, manage and automate water supply system operation either do not yet exist, are only partially evolved, or have not yet been reliably proven for live water distribution systems. It is this deficiency that the project NEPTUNE seeks to address by carrying out research into 3 main areas; these are: data and knowledge management; pressure management (including energy management); and the associated complex decision support systems on which to base interventions. The 3-year project started in April of 2007 and has already resulted in a number of research findings under the three main research priority areas (RPA). The paper summarises in greater detail the overall project objectives, the RPA activities and the areas of research innovation that are being undertaken in this major, UK collaborative study. Copyright 2009 ASCE.

Free-space optoelectronic switching cores with MPLS for SANs over WDM ring networks

With increasing demands on storage devices in the modern communication environment, the storage area network (SAN) has evolved to provide a direct connection allowing these storage devices to be accessed efficiently. To optimize the performance of a SAN, a three-stage hybrid electronic/optical switching node architecture based on the concept of a MPLS label switching mechanism, aimed at serving as a multi-protocol label switching (MPLS) ingress label edge router (LER) for a SAN-enabled application, has been designed. New shutter-based free-space multi-channel optical switching cores are employed as the core switch fabric to solve the packet contention and switching path conflict problems. The system-level node architecture design constraints are evaluated through self-similar traffic sourced from real gigabit Ethernet network traces and storage systems. The extension performance of a SAN over a proposed WDM ring network, aimed at serving as an MPLS-enabled transport network, is also presented and demonstrated. © 2012 OSA.

Short-term in vitro responses of human peripheral blood monocytes to ferritic stainless steel fiber networks.

Beneficial effects on bone-implant bonding may accrue from ferromagnetic fiber networks on implants which can deform in vivo inducing controlled levels of mechanical strain directly in growing bone. This approach requires ferromagnetic fibers that can be implanted in vivo without stimulating undue inflammatory cell responses or cytotoxicity. This study examines the short-term in vitro responses, including attachment, viability, and inflammatory stimulation, of human peripheral blood monocytes to 444 ferritic stainless steel fiber networks. Two types of 444 networks, differing in fiber cross section and thus surface area, were considered alongside austenitic stainless steel fiber networks, made of 316L, a widely established implant material. Similar high percent seeding efficiencies were measured by CyQuant® on all fiber networks after 48 h of cell culture. Extensive cell attachment was confirmed by fluorescence and scanning electron microscopy, which showed round monocytes attached at various depths into the fiber networks. Medium concentrations of lactate dehydrogenase (LDH) and tumor necrosis factor alpha (TNF-α) were determined as indicators of viability and inflammatory responses, respectively. Percent LDH concentrations were similar for both 444 fiber networks at all time points, whereas significantly lower than those of 316L control networks at 24 h. All networks elicited low-level secretions of TNF-α, which were significantly lower than that of the positive control wells containing zymosan. Collectively, the results indicate that 444 networks produce comparable responses to medical implant grade 316L networks and are able to support human peripheral blood monocytes in short-term in vitro cultures without inducing significant inflammatory or cytotoxic effects.

Investigation of multilingual deep neural networks for spoken term detection

The development of high-performance speech processing systems for low-resource languages is a challenging area. One approach to address the lack of resources is to make use of data from multiple languages. A popular direction in recent years is to use bottleneck features, or hybrid systems, trained on multilingual data for speech-to-text (STT) systems. This paper presents an investigation into the application of these multilingual approaches to spoken term detection. Experiments were run using the IARPA Babel limited language pack corpora (∼10 hours/language) with 4 languages for initial multilingual system development and an additional held-out target language. STT gains achieved through using multilingual bottleneck features in a Tandem configuration are shown to also apply to keyword search (KWS). Further improvements in both STT and KWS were observed by incorporating language questions into the Tandem GMM-HMM decision trees for the training set languages. Adapted hybrid systems performed slightly worse on average than the adapted Tandem systems. A language independent acoustic model test on the target language showed that retraining or adapting of the acoustic models to the target language is currently minimally needed to achieve reasonable performance. © 2013 IEEE.

Avoiding pathologies in very deep networks

Choosing appropriate architectures and regularization strategies of deep networks is crucial to good predictive performance. To shed light on this problem, we analyze the analogous problem of constructing useful priors on compositions of functions. Specifically, we study the deep Gaussian process, a type of infinitely-wide, deep neural network. We show that in standard architectures, the representational capacity of the network tends to capture fewer degrees of freedom as the number of layers increases, retaining only a single degree of freedom in the limit. We propose an alternate network architecture which does not suffer from this pathology. We also examine deep covariance functions, obtained by composing infinitely many feature transforms. Lastly, we characterize the class of models obtained by performing dropout on Gaussian processes.