Application of neural networks in online oil content monitors

Four major types of oils were examined to obtain both fluorescence and light scattering spectra as a function of oil concentrations. The large variations in fluorescence and scattering intensity with oil types and sub-types make it difficult to calibrate the analytical instrument using traditional methods. We implemented a multivariate, nonlinear calibration of instrumental response through an artificial neural network. We demonstrated that the combined use of fluorescence and scattering data significantly improves the quantitative prediction accuracy. The trained backpropagation neural network was used successfully to predict the concentrations of single oils and their mixtures, and appears well suited for the calibration of an online oil content monitor. The newly developed technique permits the online monitoring of oil concentrations in wastewater discharged from ships and oil refinery industry