2004 Volume 54 Issue 1 Pages 1-8
This paper proves that the hierarchy neural network application is effective in simplifying the distinction process and accommodating dispersion of the gas sensitivity by each sensor and the deterioration of gas sensitivity by the aged degradation. I first transform the gas sensor's heater voltage to sawtooth waveforms and replace the sensor's output response for ethanol and methane mixed gases and ethanol and propane mixed gases with a quadratic experimental equation using regression analysis. I show the gas kinds distinction results obtained from three types of input elements—the regression coefficients, difference voltage, and output voltage. I then show how to obtain a reliable distinction result by using jointly distinction results obtained from the three different input elements.