Monitoring of Membrane Integrity Based on Electrical Measurement and Deep Learning

Membrane module integrity monitoring is essential in the water treatment process. Problems such as high cost and low sensitivity limit the development of existing detection methods. An intelligent detection method for membrane integrity based on array impedance measurement is proposed in this paper. The boundary voltage data are collected in real time through the designed electrical sensor array. A deep learning algorithm is used to analyze the degree of damage of the membrane based on the collected voltage data. Membrane integrity testing experiments are conducted for different water qualities (lake water and domestic sewage) under different aeration intensities and membrane fluxes. Detection models based on a convolutional neural network (CNN) and deep neural network (DNN) are built, and the identification results are compared with those of the average voltage method. The experiments show the following: 1) For the tests based on random samples of membranes under detection, the overall sensitivities of the CNN in the lake water experiment and domestic sewage experiment reach 97.3% and 98.5%, respectively, which are significantly higher than those of the DNN method (94.1% and 93.6%) and the average voltage method (87.4% and 77.4%). 2) When the membrane process is affected by the variation in the aeration intensity and membrane flux, the CNN still has the best robustness. Hence, the new method could stably and accurately reflect the level of membrane breakage, even mild damage to the membrane, under flow disturbance.