Study On Data Processing Technology About Near Infrared Spectroscopy Of Organic Dirt In Industrial Water Treatment System

In industrial water treatment system, the biofilm which produced by someorganic dirt will increase the energy consumption and maintenance cleaning costs ofequipments, resulting in the waste of energy and raw materials. At present ourresearch in the dirt monitoring filed, not only the innovative design of the monitoringprinciple, but also the development and application of monitoring devices, be laggingfar behind of the other industrial countries in the world.This article describes the common fouling monitoring methods and devices athome and abroad, and analyses the principle, advantages and disadvantages of eachmonitoring device and the scope of application.This article analyzes the current situation of the near-infrared spectroscopy, andapplied the near-infrared spectroscopy techniques to organic fouling monitoring inindustrial water treatment system greatly enhanced the efficiency of the organic dirtmonitoring.This paper studies the formation of organic fouling in industrial water treatmentsystem and its optical characteristics, built the sample data and spectral data used tothe near infrared spectroscopy. Studied the partial least squares regression ofChemometrics, established a quantitative calibration model which can describe of therelationship between the ingredients in organic dirt with its Near InfraredSpectroscopy. Studied the radial basis function network compensation mechanism ofartificial neural network, solved the nonlinear problem of partial least squaresmethod in the measurement parameters and output to expand the measurement range,and improve the measurement accuracy.This article using Visual C++software dialog-based architecture completedthe near-infrared spectral data processing software system. This software systemachieved display the spectral data in table, easy to search for the spectral data of eachsample at each wavelength point, also can draw the function figure of spectral data,visually reflect each sample absorbance changes in the spectral region, timely detectthe abnormal data in the spectra. This system implements the pretreatment of thesample spectral data and the nature of data to achieve the best optimization of the modeling data. This system completed the partial least squares modeling, theartificial neural network modeling and the hybrid modeling of artificial neuralnetwork to compensate for the partial least squares, through the correlationcoefficient R of the model, the standard error of prediction SEP reliable evaluate themodel reliability. After validation, the correlation coefficient between the modelpredicted values and the true value R can be up to0.9, the model standard error ofprediction SEP can reach0.6.