Transient ventilation dynamics induced by heat sources of unequal strength

We examine theoretically the transient displacement flow and density stratification that develops within a ventilated box after two localized floor-level heat sources of unequal strengths are activated. The heat input is represented by two non-interacting turbulent axisymmetric plumes of constant buoyancy fluxes B1 and B2 > B1. The box connects to an unbounded quiescent external environment of uniform density via openings at the top and base. A theoretical model is developed to predict the time evolution of the dimensionless depths λj and mean buoyancies δj of the 'intermediate' (j = 1) and 'top' (j = 2) layers leading to steady state. The flow behaviour is classified in terms of a stratification parameter S, a dimensionless measure of the relative forcing strengths of the two buoyant layers that drive the flow. We find that dδ1/dτ α 1/λ1 and dδ2/dτ α 1/λ2, where τ is a dimensionless time. When S 1, the intermediate layer is shallow (small λ1), whereas the top layer is relatively deep (large λ2) and, in this limit, δ1 and δ2 evolve on two characteristically different time scales. This produces a time lag and gives rise to a 'thermal overshoot', during which δ1 exceeds its steady value and attains a maximum during the transients; a flow feature we refer to, in the context of a ventilated room, as 'localized overheating'. For a given source strength ratio ψ = B1/B2, we show that thermal overshoots are realized for dimensionless opening areas A < Aoh and are strongly dependent on the time history of the flow. We establish the region of {A, ψ} space where rapid development of δ1 results in δ1 > δ2, giving rise to a bulk overturning of the buoyant layers. Finally, some implications of these results, specifically to the ventilation of a room, are discussed. © Cambridge University Press 2013.

Language model combination and adaptation using weighted finite state transducers

In speech recognition systems language model (LMs) are often constructed by training and combining multiple n-gram models. They can be either used to represent different genres or tasks found in diverse text sources, or capture stochastic properties of different linguistic symbol sequences, for example, syllables and words. Unsupervised LM adaptation may also be used to further improve robustness to varying styles or tasks. When using these techniques, extensive software changes are often required. In this paper an alternative and more general approach based on weighted finite state transducers (WFSTs) is investigated for LM combination and adaptation. As it is entirely based on well-defined WFST operations, minimum change to decoding tools is needed. A wide range of LM combination configurations can be flexibly supported. An efficient on-the-fly WFST decoding algorithm is also proposed. Significant error rate gains of 7.3% relative were obtained on a state-of-the-art broadcast audio recognition task using a history dependently adapted multi-level LM modelling both syllable and word sequences. ©2010 IEEE.

Bayesian separation and recovery of convolutively mixed autoregressive sources

In this paper we address the problem of the separation and recovery of convolutively mixed autoregressive processes in a Bayesian framework. Solving this problem requires the ability to solve integration and/or optimization problems of complicated posterior distributions. We thus propose efficient stochastic algorithms based on Markov chain Monte Carlo (MCMC) methods. We present three algorithms. The first one is a classical Gibbs sampler that generates samples from the posterior distribution. The two other algorithms are stochastic optimization algorithms that allow to optimize either the marginal distribution of the sources, or the marginal distribution of the parameters of the sources and mixing filters, conditional upon the observation. Simulations are presented.