In order to accurately and quickly detect the concentration of each kind of composition in automobile exhaust

an automobile exhaust detection system is designed based on sensor array and cornbined with back propagation（ BP） neural network model. According to five kinds of exhaust gas needed to be detected regulated by GB18352. 5—2013 specified

corresponding sensors and temperature and humidity sensors are selected to form sensor array. To overcome the shortcomings of low prediction precision

slow convergence rate and easy to fall into local extremes

an improved network model is proposed by using additional momentum method and adaptive learning rate method. The improved neural network model is used for regression analysis of acquired sensor signal.The results show that the predicted average relative error of the improved BP neural network model is less than4 %

which can efficiently process the exhaust gas detection data.