Metal Oxide Co Sensors Intelligent Study

Traditional metal oxide CO sensors are lack of selectivity for theircross-sensitivity to reductive gases, especially hydrogen gas. Besides, theirperformances are prone to be affected by ambient humidity and temperature.Increasing the intelligence level of gas sensor has been the mainstream in gas sensingtechnology. This paper aims to improve the H₂ resistibility of traditional CO sensor bymeans of dynamic working temperature control and self-learning algorithm, andcompensation for the effect of ambient humidity and temperature is also considered.After a large number of experiments, effective methods of temperature modulation,self-learning and temperature compensation are exploited, and a smart CO sensormodule is designed based on a high-performance rnicrocontroller.A testing apparatus for gas sensor is developed and temperature modulationexperiments are conducted for the SnO₂-based CO sensor MQ307A and TGS2442 inturn. Various heating voltages for MQ307A are studied, and sinusoidal voltage isfound to be preferable. MQ307A's responses to CO, H₂ and CO/H₂ mixtures areacquired and Discret Wavelet Transform (DWT) is found to be better than FastFourier Transform (FFT) in extracting important features from the sensor's responsesignals. Support Vector Machine (SVM) and BP Artificial Neural Network (BP ANN)are then applied respectively to classify these three gases and furthermore, estimatethe concentrations of CO in CO/H₂ binary mixtures. The results demonstrate thatSVM gives higher success rate of classification and prediction accuracy than BP ANNdoes. TGS2442's responses to CO, H₂ and CO/H₂ mixtures are acquired and featuresextracted either from the original responses or wavelet coefficients are then processedby SVMs. Following this procedure, CO, H₂ and CO/H₂ mixtures are classifiedqualitatively, and the concentrations of CO in CO/H₂ mixtures are predicted.Effects of ambient humidity and temperature on the performance of CO sensorare investigated. While MQ307A's output response is affected severely by relativehumidity, TGS2442 shows low humidity dependency due to its special encapsulation, however, the baseline of TGS2442 rises with the ambient temperature rising.Knowledge-based temperature compensation method has been adopted for TGS2442,and this method is validated in predicting the concentrations of CO at 20℃, 40℃and60℃respectively.A microcontroller-based smart CO sensor module is designed using TGS2442 asthe sensing element, and the feature extraction and SVM algorithms are transportedinto the microcontroller. The results of validation experiments show that the smart COmodule is able to classify CO, H₂ and CO/H₂ mixtures qualitatively, and moreover,predict the concentration of CO in the CO/H₂ binary mixtures. Additionally, thismodule is characteristic of self-diagnose and ambient temperature compensation.