【作者】 周洪林；

【导师】 王兢；

【作者基本信息】 大连理工大学 ， 电路与系统， 2007， 硕士

【摘要】 近几年来，我国传感技术正在蓬勃发展，应用领域已渗入到国民经济的各个部门以及人们的日常生活之中，因此对传感器新理论的探讨、新技术的应用、新材料和新工艺的研究已成为传感器总的发展方向。声表面波（Surface Acoustic Waves，SAW）气体传感器具有精度高、分辨率高、体积小，易于集成化等特点，成为国际上器件研究的新热点之一。本论文的主要内容是关于声表面波甲醛气体传感器的设计及其气湿敏特性和选择性的研究，最后采用BP神经网络（Artificial Neural Networks）识别混合气体。首先基于声表面波谐振器（SAWR），采用射频放大器和必要的LC移相网络，设计出输出频率稳定的振荡器。将未涂敷敏感膜的声表面波振荡器的输出作为混频器的本机振荡频率，和涂敷敏感膜的声表面波振荡器的输出进行混频，把频率降到可以使用单片机测量的范围。通过整形电路，使混频后的正弦信号变为方波信号。设计了基于单片机的数据采集系统对传感器输出频率进行采集和显示，并送入PC机中进行存储和作图等操作。研究敏感膜的涂敷方法，最后选用旋涂法对敏感膜进行涂敷。研究敏感材料薄膜的厚度对灵敏度及响应时间的影响，综合考虑后选择180nm厚的敏感膜。选用乙基纤维素（ethyl cellulose）、聚异丁烯（Polyisobutene）和聚环氧氯丙烷（Polyepichlorohydrin）三种敏感材料，涂敷于声表面波传感器上对甲醛及干扰气体（乙醇、丙酮、甲苯）进行测量。使用涂敷NaCl和BaTiO₃敏感膜及未涂敷敏感膜的声表面波传感器对湿度进行测量。针对传感器对各种气体的交叉敏感特性，采用BP神经网络的方法对混合气体的成分进行分析。网络为3层网络，输入层8个节点，隐形层10个节点，输出层4个节点，输入层数据来自8个声表面波气体传感器组成的传感器阵列。BP神经网络用传感器阵列输出的数据进行学习，直至网络的样本输出误差小于期望值。使用训练过的BP网络对一些保留样本进行预测。结果表明，使用这种方法可以成功地用现有选择性较差的气敏传感器元件，分析出甲醛、乙醇、丙酮、甲苯混合气体中各气体的成分和浓度。更多还原

【Abstract】 In recent years, sensing technology is advancing rapidly. Sensor applications have been extended to people’s daily life and various departments. Therefore, research in new theories, new materials and new technology of sensors has received more and more attention from scientists. SAW （Surface Acoustic Waves） gas sensor has many excellent properties such as high precision, high resolution, small size and easy integration. It has become one of the new focus of sensor technology. The main contents of this thesis are the design of SAW formaldehyde gas sensor, the property of gas and humidity responses, the selective property and the recognition of the gas mixture using a array of 8 sensors and a BP neural network.In this work, a stable SAW oscillator based on SAWR （surface acoustic wave resonator） was designed. It contains an RF amplifier and a necessary LC phase shifter network. Then it was coated with sensitive film. A mixer was used to low down the frequency, so that the MCU （Micro Control Unit） can measure it. The SAW oscillator with sensitive film was the input of the mixer, and another SAW oscillator without sensitive film was taken as the reference input. The sine signal from the mixer was turned to the square signal by a reshape circuit. A data acquisition system based on MCU was designed to measure and display the frequency. It sends the data to a PC to draw the response curves and store the data.Spin-coating method was used in coating films. The sensor achieved its optimal performance when the thickness of film was about 180nm. Three sensitive materials （ethyl cellulose, polyisobutylene, polyepichlorohydrin） were chosen to coating on the surface acoustic wave sensors, respectively. The concentration of formaldehyde and the interference gases （ethanol, acetone and toluene） were measured by the SAW gas sensors. Sensors composed of NaCl and BaTiO₃ were used to measure the humidity, respectively.The cross-sensitivity characteristic of the sensor is a big trouble for formaldehyde gas detection. A three-level BP neural network was use to identify the gases and measure the concentrations. There were 8 nods in input layer, 10 nods in hide layer and 4 nods in output layer. A sensor array with 8 sensors was used to collect the data which were the input of the network. Then the network learned by itself based on the data circularly until the error of output was below that you set before. Then the network was ready to distinguish quantificational formaldehyde, ethanol, acetone, and toluene from mixed air.更多还原