A rapid determination method for chemical oxygen demand in aquaculture wastewater using the ultraviolet absorbance spectrum and chemometrics
Abstract
The ultraviolet (UV) absorbance spectrum was studied for the fast determination of chemical oxygen demand (COD) in aquaculture wastewater. Four calibration methods were investigated and compared in building the model, including multiple linear regression, partial least squares, least squares support vector machine and back-propagation artificial neural network (BP-ANN). In order to eliminate the superfluous information within the UV spectrum, a variable selection method of successive projections algorithm (SPA) was taken into consideration to select efficient wavelengths (EWs), which was applied for the first time in the UV spectroscopic technique for the COD determination in aquaculture wastewater. Six different spectral pretreatment methods were compared. The best prediction result was obtained by the BP-ANN model based on the 11 EWs with the optimum spectral pretreatment method of standard normal variate, and the coefficient of determination of prediction was 0.90, the root mean squared error of prediction was 10.96 mg Lā1 and the residual predictive deviation was 5.06. The overall results obtained revealed that SPA is a favourable method to select UV spectral EWs, the usage of EWs in model calibration could obtain better prediction results in comparision with the full range UV spectrum, and the combination of UV absorbance spectral EWs with chemometric analysis could be applied for the rapid determination of COD in aquaculture wastewater successfully.