Texture analysis of UTDR images for enhancement of monitoring and diagnosis of membrane filtration

•Collecting fouling status easily in different conditions via non-invasive UTDR•WT and GLCM can represent the observed features of UTDR sample well.•Texture analysis is good for early detecting anomaly of membrane fouling.•The proposed method is effectively applied to a real membrane filtration system.

Membrane fouling is unavoidable in a membrane filtration process; fouling caused by recalcitrant gel and cake layers on the membrane always results in a decrease of permeate flux. In this research, the non-destructive ultrasonic time-domain reflectometry (UTDR) signals from the block-scanning mode are used to investigate membrane fouling in various conditions. Through Hilbert transform on UTDR signals, direct visualization of the membrane fouling distribution is developed. Although ultrasonic data acquirements are automatically achieved, the analysis of ultrasonic images is achieved manually by an operator. To build up the automatic fouling monitoring system, in this research, the texture analysis and the wavelet transform applied to UTDR images are used to extract the spatial statistic features of the fouling distribution. Then the principal component analysis of multivariate statistics is applied to statistic features. The research results show that it is more convenient to detect faults using the proposed method than the commonly used visualized approaches.