【作者】 陈国顺；

【导师】 李井会； 于洪梅；

【作者基本信息】 辽宁科技大学 ， 应用化学， 2006， 硕士

【摘要】 本论文将小波分析、傅立叶变换、二阶样条小波卷积、神经网络、蚁群算法、模糊聚类、遗传算法等数学方法用于化学信号的处理，完成论文5篇，其中一篇已发表。径向基函数网络(RBF)是一种性能优良的前传型网络。本文利用径向基网络解析铋和锆的重叠光谱，建立了一种新的化学计量学方法。蚁群算法(ACA)是一种新的仿生算法。本文将蚁群算法和模糊聚类相结合确定径向基网络的隐含层节点数，建立一种新的网络(ACA-FCM-RBF)模型，并用于铋和锆的重叠光谱的解析，建立了一种新的化学计量学方法。利用遗传算法的全局优化能力，将遗传算法和BP网络结合建立遗传神经网络模型，解决了ANN-BP的过拟合、样本容量大以及各组分测定结果误差相差较大等问题。将遗传神经网络用于铋和锆的重叠光谱的解析，建立了一种新的化学计量学方法。依据傅立叶变换有关的帕塞瓦尔定理和小波变换的能量守恒以及能量压缩的大拇指规则，在对红外光谱数据压缩去噪和重建时，提出确定阈值的新方法，该方法准确、简便、快速。并将该方法用于L-色氨酸的红外光谱数据处理，获得满意结果。本文建立了两种集去噪、扣背景和解析重叠光谱为一体的ICP-AES光谱分析新方法。两种方法首先采用小波变换滤除噪声，然后通过样条小波变换扣除背景，最后分别利用傅立叶自去卷积和二阶样条小波卷积分离重叠的ICP-AES谱线。将这两种方法分别用于模拟ICP-AES和钢中As的ICP-AES数据的处理，并进行了比较，获得了满意的结果。论文所有程序均采用C／C++语言和MATLAB语言编写。引用160余篇文献，对近5年来小波分析、傅立叶变换、神经网络、蚁群算法、遗传算法在化学中的新进展作了评述。更多还原

【Abstract】 In this dissertation, Wavelet Analysis(WA), Fouier Transform(FT), Spline Wavelet Convolution, Artifical Neural Networks(ANN), Ant Colony Algorithms(ACA), Fuzzy Clustering(FC), Genetic Algorithms(GA) were applied to chemical signal processing and five papers were developed.Radial basis function artificial neural network is a fore-propagation model with excellence capability. It was applied to resolve overlapping spectra of Bismuth and Zirconium, and a new chemometric method was developed.Ant Colony Algorithms is a new bionic algorithm. Ant Colony Algorithms and Fuzzy Clustering were applied to train the RBF model, and a new network model was developed. The new chemometrics method was applied to resolve overlapping spectra of Bismuth and Zirconium.Genetic Algorithms(GAs) was applied to optimize the structure and parameters of ANN-BP with its capability of global optimization. GA-ANN-BP was developed to solve overfitting, overtraining of ANN-BP. The new method was applied to resolve overlapping spectra of Bismuth and Zirconium.A new threshold method was developed in compressing, denoising and reconstruction of infrared spectrum. The threshold value was determined by energy sequence according to Parseval’s theorem for Fourier transformation and conservation of energy for Wavelet transform and rule of thumb for energy compaction. The new method was used to deal with the L- tryptophan infrared spectra, the results were satisfactory.Two new chemometrics method that have abblity of denoising, background correction and resolution of overlapping spectra in ICP-AES were developed. In two methods, wavelet transform was applied to denoise and spline wavelet transform was applied to correct background, Fourier Deconvolution and second-order Spline Wavelet Convolution were applied to resolve overlapping spectra respectively. The two methods were used to deal with simulated ICP-AES and ICP-AES spectra of As in steel, and the results were compared.The programs of these methods were written with C/C++ and MATLAB language.By citing 160 literatures, the author reviewed the applications of WA、FT、ANN、GA、ACA in chemistry.更多还原