Search

The sense of smell is a powerful biological tool although it is the least understood. Attempts to mimic this feature over the last two decades have resulted in the creation of the electronic nose. In comparison to the biological system, its ability to distinguish complex odours is poor. This has mainly been attributed to the lack of sensors and their diversity compared to the human in the order of 105 and 102 respectively. In our efforts to improve the performance of the electronic nose, here we have used a different approach using a unique feature of the biological olfactory system. This technique is analogous to a multi-dimensional gas chromatography (MD-GC) technique that is capable in generating spatial and temporal signals to aid odour discrimination. As the physical realisation requires expensive and time consuming micro- nano fabrication processes, finite element method simulations have been used to validate the proposed design and aid optimisation. This paper describes the finite element modelling process and compares these simulation results to that of the well-established analytical model. Preliminary results of the optimised system are also presented; these results are in good agreement to the simulated outputs. [ABSTRACT FROM AUTHOR]

In this chapter we deal with the use of Support Vector Machines in gas sensing. After a brief introduction to the inner workings of multisensor systems, the potential benefits of SVMs in this type of instruments are discussed. Examples on how SVMs are being evaluated in the gas sensor community are described in detail, including studies in their generalisation ability, their role as a valid variable selection technique and their regression performance. These studies have been carried out measuring different blends of coffee, different types of vapours (CO, O2, acetone, hexanal, etc.) and even discriminating between different types of nerve agents. [ABSTRACT FROM AUTHOR]