Modelling soot formation in a DISI engine

In this work, the formation of soot in a Direct Injection Spark Ignition (DISI) engine is simulated using the Stochastic Reactor Model (SRM) engine code. Volume change, convective heat transfer, turbulent mixing, direct injection and flame propagation are accounted for. In order to simulate flame propagation, the cylinder is divided into an unburned, entrained and burned zone, with the rate of entrainment being governed by empirical equations but combustion modelled with chemical kinetics. The model contains a detailed chemical mechanism as well as a highly detailed soot formation model, however computation times are relatively short. The soot model provides information on the morphology and chemical composition of soot aggregates along with bulk quantities, including soot mass, number density, volume fraction and surface area. The model is first calibrated by simulating experimental data from a Gasoline Direct Injection (GDI) Spark Ignition (SI) engine. The model is then used to simulate experimental data from the literature, where the numbers, sizes and derived mass particulate emissions from a 1.83 L, 4-cylinder, 4 valve production DISI engine were examined. Experimental results from different injection and spark timings are compared with the model and the qualitative trends in aggregate size distribution and emissions match the exhaust gas measurements well. © 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Structure learning in a sensorimotor association task.

Learning is often understood as an organism's gradual acquisition of the association between a given sensory stimulus and the correct motor response. Mathematically, this corresponds to regressing a mapping between the set of observations and the set of actions. Recently, however, it has been shown both in cognitive and motor neuroscience that humans are not only able to learn particular stimulus-response mappings, but are also able to extract abstract structural invariants that facilitate generalization to novel tasks. Here we show how such structure learning can enhance facilitation in a sensorimotor association task performed by human subjects. Using regression and reinforcement learning models we show that the observed facilitation cannot be explained by these basic models of learning stimulus-response associations. We show, however, that the observed data can be explained by a hierarchical Bayesian model that performs structure learning. In line with previous results from cognitive tasks, this suggests that hierarchical Bayesian inference might provide a common framework to explain both the learning of specific stimulus-response associations and the learning of abstract structures that are shared by different task environments.