| Chemical Oxygen Demand(COD)is one of the indicators for organic pollution in water bodies.The rapid and accurate detection of COD is of great significance for environmental protection,pollution source supervision,and water treatment process monitoring.Compared to chemical detection methods,spectroscopic methods for water quality COD detection have the advantages of fast detection speed,simple operation,and no pollution,making it suitable for on-site rapid water quality testing.In this paper,we focus on fluorescent organic substances in surface water and aim to achieve rapid detection of absorbance-fluorescence spectra COD.We addressed the problems of slow spectral detection response under fast flowing water sample scenarios,the large impact of water turbidity on the absorbance spectra COD inversion method,and the large COD inversion error of absorbance spectra for fluorescent organic substances.We studied the scintillation xenon lamp pulse absorption full-spectrum fast low-noise acquisition technology,developed a water absorbance full-spectrum COD inversion algorithm that is resistant to turbidity interference,and studied the absorbance-fluorescence spectra synchronous acquisition technology.We also developed a characteristic fusion COD inversion algorithm for absorbance-fluorescence spectra,improving the speed and accuracy of spectroscopic COD detection.The main research contributions of this paper are as follows:① We studied the high-speed acquisition technology of single pulse light absorption full-spectrum based on the CMOS sensor,and developed a UV-Vis absorption full-spectrum acquisition system suitable for rapid flow water sample COD detection,realizing the rapid and accurate measurement of water COD absorbance fullspectrum.We designed a miniaturized dual-path structure,developed a spectral acquisition circuit based on CMOS image sensor and its driver,and achieved singlepulse UV-Vis absorption full-spectrum acquisition.We used FPGA parallel computing technology to realize real-time digital filtering of spectral data,improving the processing speed and measurement accuracy of spectral data.The system test results showed that the high-frequency noise component of the signal was suppressed between-80～-140dB,and the spectral signal amplitude at around 50kHz was almost not attenuated.The acquisition and processing time of a single pulse light absorption spectrum data was 516 microseconds,the repeatability of the spectral measurement result was 1.74%,the signal-to-noise ratio of the absorbance light measurement result of the 10mg/L COD standard solution was 426.97,and the R-Square value between the COD value and its absorbance spectral characteristic peak value was above 0.97.The detection range of COD reached 0～150mg/L,and the detection limit reached 0.1mg/L.②We studied an absorbance spectrum COD inversion method that is resistant to turbidity interference and designed a one-dimensional dual-channel convolutional neural network(1D-CNN)full-spectrum COD resolution algorithm based on fractal structure,achieving accurate inversion of water COD under random turbidity interference.We extracted the absorbance full-spectrum data features through the dilation convolution method,reduced the training parameters of the convolutional neural network by separating the convolution kernel and Adam optimization function,optimized the training speed of the convolutional neural network,improved the robustness of the model under turbidity interference through data augmentation,and quantified the solution COD value by solving the absorbance spectrum feature data through a fully connected neural network.The research results showed that the 1DCNN method could successfully extract the absorbance spectrum feature information and quantify the solution COD value from the absorbance spectrum under random turbidity interference,and the COD prediction accuracy was significantly improved compared with the traditional method.③A combined absorption-fluorescence spectrum acquisition system was constructed,and the absorption-fluorescence fusion spectral detection technology for water COD was studied in this paper.The system achieved high-speed acquisition of absorption and fluorescence spectra through a combined absorption-fluorescence spectral acquisition path.A FPGA-based high-speed scanning acquisition module with a multi-channel photomultiplier tube array was designed,and a UV-visible absorption spectrum-multi-wavelength fluorescence spectrum high-speed acquisition technology was developed based on CMOS sensors and photomultiplier tube arrays.By combining the absorption-fluorescence integrated acquisition path,the UV-visible absorption spectra and three-dimensional fluorescence spectra of organic solutes in water were successfully obtained.The test results showed that the system can complete the fluorescence emission spectrum acquisition in the 200-700nm wavelength range under a single excitation wavelength within 13 microseconds.The relative deviation between the system’s fluorescence spectrum acquisition Stokes displacement calculation results and the F-7000 measurement results was 0.56%.The obtained fluorescence and absorption spectra were consistent with the results from the F-7000 and UV2550 tests,respectively.④A spectral feature fusion method was studied in this paper,and a fusion neural network algorithm for absorption-fluorescence spectral feature extraction and COD inversion was developed based on a one-dimensional dual-channel convolutional neural network and 2D Gabor transform,which improved the accuracy of water COD inversion.The one-dimensional convolutional neural network algorithm was used to extract the UV-visible absorption spectrum features of water COD,and the 2D Gabor transform algorithm was used to extract the three-dimensional fluorescence spectrum features of water COD.A fusion neural network algorithm for absorption and fluorescence spectral feature data was developed to achieve rapid and accurate COD inversion.The research results showed that the RRMSEP of amino acid water solution COD inversion was 0.32%,which reduced by 84%compared to the single absorption spectrum method.The accuracy of COD inversion was 98%,which was improved by 15.3%compared to the single absorption spectrum method. |
PDF Full Text Request