Multi-models Combined Water Quality Analyzing Based On Multi-Spectra

Compared with conventional chemical methods, detection of organic pollutants in water based on spectroscopy is a kind of green testing technology with higher analysis speed, no chemical reagent pollution and easy operation. The existing spectrum based methods generally use single spectrum for analysis and the analysis accuracy is relatively low at the present stage. This thesis deals with the issues on the optical analysis method of the comprehensive organic contaminant indexes of water. In order to improve the accuracy and robustness of the analysis based on spectroscopy, due to the complementary advantages of multi-spectrums, a model combination method based on the ultra-violet absorption spectrum and the fluorescence emission spectrum with variable weighted coefficient was adopted for rapid measure of six water quality parameters such as TOC, COD, etc.. The main work of this thesis is organized as follows:1. To evaluate the performance of the method based on Spectroscopy, a set of TOC,COD—ultraviolet (UV) spectral data, fluorescence spectral data and spectral fusion data are measured, and two common approaches in spectral analysis, such as Partial Least Squares and Least Squares Support Vector Machine, were used in modeling and prediction experiments respectively. And through the root mean square error of prediction and correlation coefficient to assess the models' forecasting performance.2. As the presenctly fusion-modelling methods are unsuitable for our research object. In order to improve the accuracy and robustness of the analysis based on multi-spectra, we put forward a model combination method based on multi-spectra with variable weighted coefficient. At first builds sub-models for multi-spectra and water quality index. then,using the spectral angle to calculate the matching degree between the forecasting sample's multi-spectra and the training samples' multi-spectra. And finally, Calculated the combined weights based on the matching degree,which realizing the variable weighted combination forecasting model with better analysis accuracy and robustness.3. A multi-spectral variable weight combination forecasting software was designed for portable realization to perform the field mobile tasks and rapid scene measurements, and application Testing was maked.