Fuzzy理論を用いたガス判別

抄録

Numerous detecting systems equipped with gas sensors of sintered powders are commercially available, and they are considerably used in usual domestic affairs. But it is difficult to give them good selectivity of gas species. In order to improve the selectivity, recent studies seem to be directed towards intelligent sensors. For example many researchers used simultaneously several sensors of having different characteristics for identification of the gas species. Their method is to discriminate the gas species from response pattern of the many gas sensors of different characteristics by using pattern recognition.
Considering the recent direction towards the intelligent sensors, we have introduced the discrimination method of gases by using fuzzy theory.
Noisy data obtained from the sensors are corresponded to a membership function and thus the Fuzzy discrimination is achieved. Following three performance indices were used, namely (1) gas concentration (2) output ratio of sensors and (3) distance from a characteristic curve. The gas discrimination is done by selecting a gas species which totally most satisfy these performance indices.
In order to test the usefulness of our present method, we have discriminated these gases by both computer simulation and experimental results for four gas species (methane, propane, city gas, hydrogen). The obtained results are as follows.
(1) Methane and propane among four gas species can be discriminated from the other gases.
(2) City gas, whose main component is H₂, can not be discriminated from hydrogen.
The gas, which has distinctly different characteristics, can be discriminated from other gases in the present study. However, the gases having similar characteristics can not.