范文编号:HG182 范文字数:20263,页数:37 摘要:现代近红外光谱(NIR)分析技术在很多行业得到广泛应用。而在近红外光谱分析使用的化学计量学方法中,波长选取方法是实际应用的基础。本文综述了波长选择方法及其在近红外光谱数据中的应用的发展,介绍了相关系数法、方差分析法、逐步回归法、无信息变量的消除法(UVE)、间隔偏最小二乘法(interval PLS, iPLS)、遗传算法(genetic algorithms,GA)等主要的波长选择方法及相应的原理,并进行了一些必要的数学描述。在以上提到的多种选择方法基础上,本范文详细介绍了遗传算法,进行了GA在近红外光谱的应用实现,并与多元线性回归法的应用实现的结果进行比较分析。遗传算法是采用二进制编码,通过对选择、交叉和变异对波长进行选择并选出最优子集。由150个训练数据生成的参数,去预测25个测试样本,并由此计算它们的平均相对偏差。本文通过平均相对偏差作为评价模型预测效果的标准。结果表明,采用遗传算法进行波长及预处理选择是可行的,平均相对偏差较小,结果较为理想,但仍需进一步改进。 Abstract: The technology of Near-Infrared spectroscopy(NIR)analysis has great application in many industry. Wavelength selection method is the base of application that is used in chemical metrology method of Near-Infrared spectroscopy analysis. This article summarizes wavelength selection methods and the development of the application on Near-Infrared Spectroscopy, and introduce wavelength selection and principle of correlation connection, which includes square difference analytical method, multivariate linear regression, uninformative variable elimination, multiple linear regression,principal component regression,interval partial least square,genetic algorithm and so on.. Meanwhile, the article introduces some essential mathematics description. Based on the selection methods above, this article particularly introduces genetic algorithms, and makes GA apply and realize in NIR. Then, put the results that use GA and MLR to compare and analyze. GA selects the best class by selection, crossover, mutation with binary strings, then calculate average comparatively warp. It is shown that this method is feasible, but it also needs to ameliorate more.
|
上一篇:长效防雾滴农膜的研制 | 下一篇:吡硫醇锌检测方法的建立 |
点击查看关于 波长 选择 方法 及其 红外 光谱 数据 应用 的相关范文题目 | 【返回顶部】 |