Detection algorithms for the Nano nose

The Nano nose is an instrument with an array of nano sized optical sensors that produces digital patterns when exposed to radiation passing through a gaseous mixture. The digital patterns correspond to the amount of photocurrent registered on each of the sensors. The problem is to find the gas constituents in the gaseous mixture and estimate their concentrations. This thesis outlines an algorithm using a combination of a mixed gas detector and a gas concentration predictor. The mixed gas detector is an array of neural networks corresponding to the number of gases. There are two techniques outlined for the implementation of the gas concentration predictor which are the partial least squares regression (PLS) and the Kalman filter. The output of the developed algorithm would not only show the detection of the individual constituents in the gaseous mixture but also provide the prediction of their concentrations. The algorithm designed is entirely re-configurable providing greater amount of flexibility and has detected the constituents along with the prediction of their concentrations of a mixture of three gases.