No description available.

No description available.

New event detection, also known as first story detection (FSD), has become very popular in recent years. The task consists of finding previously unseen events from a stream of documents. Despite the apparent simplicity, FSD is very challenging and has applications anywhere where timely access to fresh information is crucial: from journalism to stock market trading, homeland security, or emergency response. With the rise of user generated content and citizen journalism we have entered an era of big and noisy data, yet traditional approaches for solving FSD are not designed to deal with this new type of data. The amount of information that is being generated today exceeds by many orders of magnitude previously available datasets, making traditional approaches obsolete for modern event detection. In this thesis, we propose a modern approach to event detection that scales to unbounded streams of text, without sacrificing accuracy. This is a crucial property that enables us to detect events from large streams like Twitter, which none of the previous approaches were able to do. One of the major problems in detecting new events is vocabulary mismatch, also known as lexical variation. This problem is characterized by different authors using different words to describe the same event, and it is inherent to human language. We show how to mitigate this problem in FSD by using paraphrases. Our approach that uses paraphrases achieves state-of-the-art results on the FSD task, while still maintaining efficiency and being able to process unbounded streams. Another important property of user generated content is the high level of noise, and Twitter is no exception. This is another problem that traditional approaches were not designed to deal with, and here we investigate different methods of reducing the amount of noise. We show that by using information from Wikipedia, it is possible to significantly reduce the amount of spurious events detected in Twitter, while maintaining a very small latency in detection. A question is often raised as to whether Twitter is at all useful, especially if one has access to a high-quality stream such as the newswire, or if it should be considered as sort of a poor man’s newswire. In our comparison of these two streams we find that Twitter contains events not present in the newswire, and that it also breaks some events sooner, showing that it is useful for event detection, even in the presence of newswire.

No description available.

The search for clean and efficient combustors is motivated by the increasingly stringent emissions regulations. New gas turbine engines are designed to operate under lean conditions with inhomogeneous reactants to ensure cleanliness and stability of the combustion. This ushers in a new mode of combustion, called the inhomogeneous reactants premixed combustion. The present study investigates the effects of inhomogeneous reactants on premixed combustion, specifically on the interactions of an initially planar flame with field of inhomogeneous reactants. Unsteady and unstrained laminar methane-air flames are studied in one- and two-dimensional simulations to investigate the effects of normally and tangentially (to the flame surface) stratified reactants. A three-dimensional DNS of turbulent inhomogeneous reactants premixed combustion is performed to extend the investigation into turbulent flames. The methaneair combustion is represented by a complex chemical reaction mechanism with 18 species and 68 steps. The flame surface density (FSD) and displacement speed S_d have been used as the framework to analyse the inhomogeneous reactants premixed flame. The flames are characterised by an isosurface of reaction progress variable. The unsteady flames are compared to the steady laminar unstrained reference case. An equivalence ratio dip is observed in all simulations and it can serve as a marker for the premixed flame. The dip is attributed to the preferential diffusion of carbon- and hydrogen- containing species. Hysteresis of S_d is observed in the unsteady and unstrained laminar flames that propagate into normally stratified reactants. Stoichiometric flames propagating into lean mixture have a larger S_d than lean flames propagating into stoichiometric mixtures. The cross-dissipation term contribution to S_d is small (~~10%) but its contribution to the hysteresis of S_d is not (~~50%). Differential propagation of the flame surface is observed in the laminar flame that propagates into tangentially stratified reactants. Stretch on the flame surface is induced by the differential propagation, which in turn increases the flame surface area.

Fossil fuel reserves are projected to be decreasing, and emission regulations are becoming more stringent due to increasing atmospheric pollution. Alternative fuels for power generation in industrial gas turbines are thus required able to meet the above demands. Examples of such fuels are synthetic gas, blast furnace gas and coke oven gas. A common characteristic of these fuels is that they are multi-component fuels, whose composition varies greatly depending on their production process. This implies that their combustion characteristics will also vary significantly. Thus, accurate and yet flexible enough combustion sub-models are required for such fuels, which are used during the design stage, to ensure optimum performance during practical operating conditions. Most combustion sub-model development and validation is based on Direct Numerical Simulation (DNS) studies. DNS however is computationally expensive. This, has so far limited DNS to single-component fuels such as methane and hydrogen. Furthermore, the majority of DNS conducted to date used one-step chemistry in 3D, and skeletal chemistry in 2D only. The need for 3D DNS using skeletal chemistry is thus apparent. In this study, an accurate reduced chemical mechanism suitable for multi-component fuel-air combustion is developed from a skeletal mechanism. Three-dimensional DNS of a freely propagating turbulent premixed flame is then conducted using both mechanisms to shed some light into the flame structure and turbulence-scalar interaction of such multi-component fuel flames. It is found that for the multi-component fuel flame heat is released over a wider temperature range contrary to a methane flame. This, results from the presence of individual species reactions zones which do not all overlap. The performance of the reduced mechanism is also validated using the DNS data. Results suggest it to be a good substitute of the skeletal mechanism, resulting in significant time and memory savings. The flame markers commonly used to visualize heat release rate in laser diagnostics are found to be inadequate for the multi-component fuel flame, and alternative markers are proposed. Finally, some popular mean reaction rate closures are tested for the multi-component fuel flame. Significant differences are observed between the models’ performance at the highest turbulence level considered in this study. These arise from the chemical complexity of the fuel, and further parametric studies using skeletal chemistry DNS would be useful for the refinement of the models.

The influence of stratification on the structure of turbulent methane/air combustion is investigated using experimental data from laboratory scale burners: a weakly turbulent slot burner, and a higher turbulence co-annular swirl burner. The degree of stratification can be controlled independently of the overall fuel/air flow rate. The resulting measurements of scalar and velocity fields provide detailed test cases for existing and emerging turbulent flame models, covering a range of u'/sL from 1 to 10, turbulence intensities from 5% to 60%, and stratification ratios from 1 to 3. Simultaneous Rayleigh/Raman/CO-LIF measurements of temperature and major species concentrations - CH4, CO2, CO, H2, H2O and O2 - along a line are used to investigate the structure of a series of flames in both the slot and swirl burners. Concurrent cross-planar OH-PLIF allows thermal gradients to be angle corrected to their three-dimensional values. Finally, non-reacting and reacting velocity fields complete the flame database. The behavior of major species concentrations in the slot and swirl burner with respect to temperature is found to agree well on the mean with unstrained premixed laminar flame calculations. Scalar means conditioned on stoichiometry also show good agreement, aside from hydrogen which is enhanced under stratified conditions. Surface density function and scalar dissipation are lower than calculated values in all cases, suggesting that turbulence-induced thickening dominates the effect of increased strain. Metrics commonly used to derive flame surface density (FSD) were investigated. FSD may be determined using a statistical method based on measurements of temperature and its gradient, or a geometric method based on 2D temperature or LIF imaging. A third metric, an extension of the geometric method, is proposed. Good agreement is observed between the three metrics. The current database provides the first detailed high resolution scalar measurements for premixed and stratified flames. The data analysis provides insight into the physics of stratification: for the flames considered, the effects of stratification appear to be surprisingly small compared to those of turbulence, even at significant stratification ratios. The datasets provide a means of validating current and future computational turbulent combustion models.