| Along with the swift and violent development of automobile industry, traffic accidents caused by driving after drunk are steadily increasing, so it is an extremely urgent and important research project to monitor whether the driver has drunk and make measure. As a innovative technology in 90's, Electronic nose (E-nose) syncretizes many technology about sensor, electronics, signal processing and computer et al, which can overcome faults in traditional single sensor and recognize mixed gas qualitatively and quantitatively. E-nose has been widely applied in food industry, medicine, quality control (QA) and deleterious gas detecting. Therefore, in this paper, an E-nose system for detecting alcohol odor qualitatively and quantitatively is developed.In order to distinguish alcohol odor from petrol and smog in driver's cab, an array of chemical sensors has been formed with five gas sensors based on metal-oxid semiconductor, one smog sensor, and one integrated module for temperature and humidity. Hardware of E-nose system etc consists of preprocessing circuit, ADC, amplifier, AT89C52, DAC and LCD etc. Software compiled with C51 in single chip can supply online and offline work, and software compiled with Viual C++ 6.0 in PC has the capability of experiment control, real-time display, data processing and parameters of neural network etc.BP algorithm is optimized by genetic algorithm, which greatly avoided local minimum of energy function by BP algorithm. BP NN (Backpropagation neural network) and RBF NN (Radial Basis Function neural network) are compared for recognition in quality. BP and multi-BP neural network are used to recognize alcohol odor in quantity, which results proved that mBP neural network is capable of better recognition and smaller error. Besides the optimal array dimension is fixed on by the genetic algorithm. On-line compensation about temperature and humidity as two main factors can resolve the drift of sensor to a great extent.
|
PDF Full Text Request