Spectrum COD Measurement Technology And Research Based On Water Sample Identification

COD (Chemical Oxygen Demand, COD) is an important indicator in water quality monitoring.charactering the amount of reducing substance, which is need to be oxidized.We usually use chemical and physical methods to detect COD.Chemical methods has the advantages of high detection accuracy, but the detecting process requires reagent, witch is easy to cause secondary pollution.The detecting process may coast a long time.The physical method,witch is an optical method, has advantages of quick-measurement, simple operation, free-agent, but the physical method is an indirect measurement of chemical indicators and its applicability and accuracy is lower.In this paper, a new method based on dynamic identification of water sample type is proposed. Firstly, the algorithm identifies the type of a water sample.Secondly, the algorithm chose the matched calculation module to get a COD result with high accuracy. The main results of the work and achievements include:1.Analysis of machine learning methods commonly used in the spectrum identification applications, including support vector machines (SVM), artificial neural networks (ANN). We finally chose BP neural network algorithm considering of the instrument’s hardware resources and other options. According the ultroviolet-visable (UV-Vis) spectrum characteristics in COD measurement,5 morphology parameters are chosen as input parameters of water samples to identify the model type.2.Studied on water sample’s correlation in the time domain.Introduce the concept of historical data queue and historical recognition factor based on this feature.We also add a second level artificial neural networks in,witch forming a cascade of neural network pattern recognition structure.The method has good robustness and high accuracy.3. Program for the COD measurement monitor based on the water sample type identification methods.Software includes graphical user interface (GUI) module, data storage module, artificial neural networks (ANN), module,COD calculation script module. etc.Through the actual test, the expected function of the spectral method COD measuring instrument is realized.4.Using a variety of real water samples to verify the accuracy of the identification of water samples. The experiments show that the recognition accuracy of the water sample recognition algorithm is good. The structure of the cascade neural network can be optimized based on the single level neural network to improve the recognition accuracy. The recognition accuracy of the water sample recognition algorithm based on cascade neural network is more than 98%.5.Did the test of measurement accuracy of COD monitor with Potassium biphthalate solution and actual water sample.The result shows that Experimental results show that compared with the conventional spectral COD measurement method can obtain good results, which the samples are first identified and then call the corresponding "absorbance (Auv) -COD" algorithm model.This method makes the applicability and accuracy of the spectral COD measurement method greatly improved...