论文部分内容阅读
针对UV-Vis分光光度多组分同时测定数据处理中经常遇到的变量多、样本少问题,提出了用小波包变换-线性插值- RBF网络解析多组分分光光度同时测定数据的方法。该方法采用小波包变换对测定数据滤噪,用线性插值处理使训练集样本对待辩识空问形成较好的覆盖,从而使RBF网络能够提取到更多的特征信息。改善网络性能、提高预测准确性。对系列实验数据的解析应用证明,该方法获得的测定结果优于常用的解析变量多、样本少数据的其它方法。由于该方法预测误差小,易在Matlab上实现编程,因此具有广泛的推广应用前景。本文以解析与炼油厂催化裂化装置生产相关的Fe、Ni、V的同时测定数据为例,对该方法的应用做了详细介绍,并与PLS法、RBF网络、小波压缩数据-RBF网络、小波压缩数据-PLS法、线性插值-RBF网络、小波变换-线性插值-RBF网络、小波包变换-线性插值-PLS法等方法傲了应用比较。
Aiming at the problem of many variables and few samples frequently encountered in data processing of UV-Vis spectrophotometric multi-component simultaneous determination, a method of simultaneously analyzing data by multi-component spectrophotometry using wavelet packet transform-linear interpolation-RBF network is proposed. In this method, wavelet packet transform is used to filter the measured data. Linear interpolation is used to make the sample of training set to be better covered by the recognition space, so that the RBF network can extract more characteristic information. Improve network performance and improve forecast accuracy. The analytical application of the series of experimental data proves that the method obtained better than the commonly used analytical variables, sample less data and other methods. Because of the small prediction error of this method, it is easy to realize the programming in Matlab, so it has a wide range of popularization and application prospects. In this paper, the analytical data of Fe, Ni and V related to the production of FCCU in refinery are taken as an example to describe the application of this method in detail. And with the PLS method, RBF network, wavelet compression data - RBF network, wavelet Compressed data-PLS method, linear interpolation-RBF network, wavelet transform-linear interpolation-RBF network, wavelet packet transform-linear interpolation-PLS method proud of application comparison.