Optimization of temperature programmed sensing for gas identification using micro-hotplate sensors

Micro-hotplate chemical gas sensors, such as those being developed at the National Institute of Standards and Technology (NIST) by micromachining Si, can be operated in a temperature-pulsed mode, due to their small size and mass. In the temperature-pulsed mode of operation, different gases give different dynamic responses (i.e. signatures) depending on the temperature program used. In this paper a new methodology is presented to optimize the operation of micro-hotplate gas sensors for discriminating volatile organic compounds at a fixed concentration, while minimizing the detection time. The extension of the methodology to cases where concentrations vary is currently under investigation. The Wavelet Network method is applied to accurately predict the sensor's response for a given temperature profile. Once a dynamic model is obtained, it is used for off-line optimization of the temperature profile, i.e. the maximization of the difference between two gas signatures. The difference between two response curves was initially measured by a metric based on the Euclidean distance. This metric was then modified using the Haar wavelet transformation. The methodology was implemented in a case study in which either methanol or ethanol had to be detected in air, but the methodology is generic, and it can be applied to any two gases. (C) 1998 Elsevier Science S.A. All rights reserved.