Simulating dynamic covariance structures for testing the adaptive behavior of variable selection algorithms (Invited Paper)

Variable selection for regression is a classical statistical problem, motivated by concerns that too large a number of covariates may bring about overfitting and unnecessarily high measurement costs. Novel difficulties arise in streaming contexts, where the correlation structure of the process may be drifting, in which case it must be constantly tracked so that selections may be revised accordingly. A particularly interesting phenomenon is that non-selected covariates become missing variables, inducing bias on subsequent decisions. This raises an intricate exploration-exploitation tradeoff, whose dependence on the covariance tracking algorithm and the choice of variable selection scheme is too complex to be dealt with analytically. We hence capitalise on the strength of simulations to explore this problem, taking the opportunity to tackle the difficult task of simulating dynamic correlation structures. © 2008 IEEE.

The effect of hole size upon the strength of metallic and polymeric foams

Tensile and compressive tests have been performed on centre-hole panels, made from three types of metallic foams and two polymeric foams. In compression, the foams fail in a ductile, notch-insensitive manner, in support of a "net section strength" criterion. In tension, a ductile-brittle transition is observed for some of the foams at sufficiently large specimen sizes: for a small hole diameter the net section strength criterion is obeyed, whereas for a large hole a local stress criterion applies and the net section strength is reduced. For a number of the foams, the panel size was not sufficiently large to observe this ductile-brittle switch in behaviour. The predictions of a cohesive zone model are compared with the measured strengths and are found to be in good agreement. © 2001 Elsevier Science Ltd. All rights reserved.

Effect of inclusions and holes on the stiffness and strength of honeycombs

A finite element study has been performed on the effects of holes and rigid inclusions on the elastic modulus and yield strength of regular honeycombs under biaxial loading. The focus is on honeycombs that have already been weakened by a small degree of geometrical imperfection, such as a random distribution of fractured cell walls, as these imperfect honeycombs resemble commercially available metallic foams. Hashin-Shtrikman lower and upper bounds and self-consistent estimates of elastic moduli are derived to provide reference solutions to the finite element calculations. It is found that the strength of an imperfect honeycomb is relatively insensitive to the presence of holes and inclusions, consistent with recent experimental observations on commercial aluminum alloy foams.

Transient Ventilation Dynamics Following a Change in Strength of a Point Source of Heat

We investigate the transient ventilation flow within a confined ventilated space, with high- and low-level openings, when the strength of a low-level point source of heat is changed instantaneously. The steady-flow regime in the space involves a turbulent buoyant plume, which rises from the point source to a well-mixed warm upper layer. The steady-state height of the interface between this layer and the lower layer of exterior fluid is independent of the heat flux, but the upper layer becomes progressively warmer with heat flux. New analogue laboratory experiments of the transient adjustment between steady states identify that if the heat flux is increased, the continuing plume propagates to the top of the room forming a new, warmer layer. This layer gradually deepens, and as the turbulent plume entrains fluid from the original warm layer, the original layer is gradually depleted and disappears, and a new steady state is established. In contrast, if the source buoyancy flux is decreased, the continuing plume is cooler than the original plume, so that on reaching the interface it is of intermediate density between the original warm layer and the external fluid. The plume supplies a new intermediate layer, which gradually deepens with the continuing flow. In turn, the original upper layer becomes depleted, both as a result of being vented through the upper opening of the space, but also due to some penetrative entrainment of this layer by the plume, as the plume overshoots the interface before falling back to supply the new intermediate layer. We develop quantitative models which are in good accord with our experimental data, by combining classical plume theory with models of the penetrative entrainment for the case of a decrease in heating. Typically, we find that the effect of penetrative entrainment on the density of the intruding layer is relatively weak, provided the change in source strength is sufficiently large. However, penetrative entrainment measurably increases the rate at which the depth of the draining layer decreases. We conclude with a discussion of the importance of these results for the control of naturally ventilated spaces.

Strength of nanotubes, filaments, and nanowires from sonication-onduced scission

A model to describe the cavitation-induced breakage of nanofilaments during their sonication in solution is proposed. The model predicts a limiting length below which scission no longer occurs, and accurately describes experimental results for materials ranging from carbon nanotubes to protein fibrils. Sonication-induced breakage can now be used as a probe for the strength of nanostructures. © 2009 WILEY-VCH Verlag GmbH & Co. KGaA.

Robust approach to dynamic statistical process control

Statistical Process Control (SPC) technique are well established across a wide range of industries. In particular, the plotting of key steady state variables with their statistical limit against time (Shewart charting) is a common approach for monitoring the normality of production. This paper aims with extending Shewart charting techniques to the quality monitoring of variables driven by uncertain dynamic processes, which has particular application in the process industries where it is desirable to monitor process variables on-line as well as final product. The robust approach to dynamic SPC is based on previous work on guaranteed cost filtering for linear systems and is intended to provide a basis for both a wide application of SPC monitoring and also motivate unstructured fault detection.