COD Detecting Technique Research Based On UV Analysis

The UV-Vis spectroscopy detection method of chemical oxygen demand (COD) has already become research hotspot in the field of chemical oxygen demand detecting for unique advantages, such as, no secondary polluting, fast test rapidity of detecting process, real time and easily actualize measuring on-line. Ultraviolet spectroscopic data processing and analysis is an important content on the research of detection system of COD detecting by UV. How to establish the mathematic model between ultraviolet absorbency data and organic pollutant and enhance extrapolating ability of the model is the key question. The research is significant for the in-depth research and realization of the method detecting COD by UV on-line measuring.On the base of summing-up the methods of COD existent by at present, in allusion to the limitation of that the general UV method is sensitive to turbidity and water quality variety, combined with artificial neural network's excellent ability identifying non-linear models, a rapidly estimating method based on artificial neural network detecting COD is put forward. The feasibility and validity of the method has been validated through testing system establishing and interrelated software design. Using multi-wavelength technique obtaining spectral data of wastewater by scanning on-line, through feature extract technique simplifying sample data, an algorithm model forecasting COD based on general regression neural network technology is proposed. Synchronously, adopting different models founded in different algorithms to water samples' spectral data with different COD values, the algorithm model forecasting COD based on general regression neural network can give nicer integrative performance in network training rate and detecting precision and correlative aspects. The testing model has been proved to be having good and generalization performance with experiment results.The theoretical analysis and our experimental results show clearly that, the COD detection method based on GRNN is provided with those advantages good predicting performance and generalization performance and non-linear disposing ability and fast convergence velocity. Through founding predicting model for COD values swatch data, could be gained corresponding COD values quickly according to UV absorbency data of water swatch, providing an effective method for sewage COD values detecting on-line.