Rapid Detection Of COD And DIN In Water By Near-infrared Spectroscopy

As the development of industry, water pollution becomes social problems. Real-timeanalysis is the future direction in water quality testing to grasp the water dynamic changes.Chemical oxygen demand (COD) and dissolved inorganic nitrogen (DIN) were importantparts of the detection of water indicators. In this paper, a real-time detection method ofCOD and DIN was built by near-infrared spectroscopy combined with multivariatecalibration analysis methods.The main studies’ contents as following:1. The feasibility of COD detection in urban lake was verified by near infraredspectroscopy. Then the model was built by principal component regression (PCR) andpartial least squares (PLS). The result showed correlation coefficient of calibration andprediction set of PCR model were respectively0.8124and0.8591, while correlationcoefficient of calibration and prediction set of PLS model were respectively0.9381and0.9290. The result showed that PLS model had a better effect than PCR model.2. The feasibility of simultaneous detections of nitrate nitrogen, nitrite nitrogen andammonia nitrogen was discussed by near infrared spectroscopy. Simulated water sampleswere prepared containing potassium nitrate, ammonium nitrate and sodium nitrite. Thenpredicted residual sum of squares method and the number of system independent variablesmethod were used to determine the number of system principal components in thesimulated water, and the result of predicted residual sum of squares method was two,while the result of the number of system independent variables method was four. Thenumber of system independent variables method judged accurately from the aspect ofchemical composition which the system mainly contained. When the number of systemprincipal components was four, the model was built to simultaneously detect potassiumnitrate, ammonium nitrate and sodium nitrite. Correlation coefficient of potassium nitrate,ammonium nitrate and sodium nitrite calibration sets were0.9798,0.9928and0.9788,while correlation coefficient of potassium nitrate, ammonium nitrate and sodium nitriteprediction sets were0.9808,0.9940and0.9899. The result was satisfied.3. The Effect of modeling samples selection on the detection of DIN in water wasdiscussed. Modeling samples were selected by concentration gradient method andprincipal component analysis combined with Euclidean distance method. Representativecalibration samples were obtained, and PLS model was built to detect DIN in simulatedwater. Correlation coefficient of calibration and prediction set of concentration gradientmethod were respectively0.9839and0.9798, and correlation coefficient of calibration andprediction set of principal component analysis combined with Euclidean distance modelwere respectively0.9839and0.9849. The result showed both of the two selecting sample methods were good.