【作者】 潘胤飞；

【导师】 赵杰文；

【作者基本信息】 江苏大学 ， 农产品加工及贮藏工程， 2003， 硕士

【摘要】 本文研究了气体传感器阵列原理和构成，重点研究了遗传算法和神经网络等非线性模式识别技术在电子鼻中的应用。本课题在大量的实验基础上，建立了一个较完善的电子鼻试验装置，可以依据气味对苹果进行分类和分级。主要研究内容及方法如下： 1．给出了电子鼻试验装置的气体传感器阵列组成的基本形式，改进了电子鼻试验装置，可以人为控制反应室的温度、湿度和载气的气流大小，同时设计了新的试验方案； 2．利用MATLAB软件编程，对所采集的花牛苹果的实验数据进行预处理后，从每条数据曲线中提取了最大值、时间、相对值和积分值4个特征值，对每个样本(有4条曲线)我们得到了16个有效的定量化参数； 3．建立了基于二叉树编码的遗传算法的模式识别模型，提出了一种叫做稳态轮盘赌选择法的选择方法，并在VC++平台上实现，利用这种算法对三种鸢尾花数据进行了模式分类，还对花牛苹果的特征值数据进行了识别，都取得了很理想的效果； 4．建立了遗传算法优化BP神经网络和遗传算法优化RBF神经网络的模式识别模型，在VC++平台上实现，用这两种算法对三种鸢尾花数据进行了模拟判别，同时对花牛苹果的特征值数据进行了识别，结果令人满意；在此基础上还对两种算法的性能进行了比较。 根据上述研究得出：改进后的电子鼻试验装置，能够控制载气的气流大小，能人为的设置反应室的温度，保证环境恒湿；试验证明该传感器阵列能对苹果的气味进行识别；基于二叉树编码的遗传算法模式识别模型、遗传算法优化BP神经网络模式识别模型和遗传算法优化RBF神经网络模式识别模型都能对花牛苹果进行理想的分类和分级。更多还原

【Abstract】 In this thesis, our attention was paid to the principle and construction of gas sensor arrays, and the new methods of pattern recognition (PR) achieved by genetic algorithms (GA) and neural networks (NN) in the processing of electronic nose. On the study of the extensive reference in the field of electronic nose, we had developed an electronic nose equipment. Such equipment could classify the apples by odor. The main contents and methods of this thesis are:1. The basic form of gas sensor array in electronic nose was given. The electronic nose equipment was developed and the new test processing was given too. The temperature, humidity and velocity of flow of gas carrier was controlled in this new equipment.2. Using MATLAB, four characters (max, time, max-gradient and integral) from each curve of gas sensors had been picked-up. And we got 16 characters data from each sample.3. The PR model of GA based on Binary-Tree Coding was established, and was programmed with VC++. A new method named steady routlette wheel selection was created. Good result was obtained.4. The model of BP optimized by GA and the model of RBF optimized by GA wase established, and were programmed with VC++. Good result was obtained.The following conclusions were obtained according to the research described above: the PR model of GA based on Binary- Tree Coding, the model of BP optimized by GA and the model of RBF optimized by GA had obtained good result in classifying the apples.更多还原