Multisensor system for toxic gases detection generated on indoor environments

2016 2nd International Conference on Mechanical Engineering and Automation Science

机械制造;无线电电子学

Durán, C.M.^1 ; Monsalve, P.A.G.^1 ; Mosquera, C.J.^2

Multisensor System and Pattern Recognition Research Group, Electronic Engineering Program, Engineering and Architecture Faculty, Universidad de Pamplona, Km 1 Via Bucaramanga, Pamplona-N.S, Colombia^1

Grilex Research Croup, Faculty of Education, Universidad de Pamplona, Km 1 Via Bucaramanga, Pamplona-N.S, Colombia^2

关键词: Cross-validation methods;    Discriminant function analysis;    Multisensory systems;    Pattern recognition method;    Probabilistic neural networks;    Statistical parameters;    Underground coal mine;    Wireless communications;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/157/1/012029/pdf
DOI  :  10.1088/1757-899X/157/1/012029

来源: IOP

【 摘 要 】

This work describes a wireless multisensory system for different toxic gases detection generated on indoor environments (i.e., Underground coal mines, etc.). The artificial multisensory system proposed in this study was developed through a set of six chemical gas sensors (MQ) of low cost with overlapping sensitivities to detect hazardous gases in the air. A statistical parameter was implemented to the data set and two pattern recognition methods such as Principal Component Analysis (PCA) and Discriminant Function Analysis (DFA) were used for feature selection. The toxic gases categories were classified with a Probabilistic Neural Network (PNN) in order to validate the results previously obtained. The tests were carried out to verify feasibility of the application through a wireless communication model which allowed to monitor and store the information of the sensor signals for the appropriate analysis. The success rate in the measures discrimination was 100%, using an artificial neural network where leave-one-out was used as cross validation method.