Application of wavelet analysis and decision tree in UTDR data for diagnosis of membrane filtration

Fouling is readily acknowledged to be one of the most critical problems with respect to wider applications of membranes in liquid separation. It is a significant hindrance to successful membrane operation. It may also result in extensive time and maintenance costs. The detection of the defects is crucial to preventing the filtration system from malfunction that could cause damage or entire system halt. This study presents the technical detection results using the ultrasonic frequency domain reflectometry (UTDR), which is capable of diagnosing membrane fouling in various operation conditions. The detection technique is the combination of the wavelet transforms (WT) and the decision tree (DT). It is proposed to exploit the virtues from UTDR signals. WT is used to represent all the possible types of transients in generated vibration signals for feature extraction. Its relative effectiveness in feature extraction is compared. DT is used for feature selection as well as classification. This work is the first-ever attempt to develop UTDR for direct and unambiguous diagnosis of membrane fouling in the practical operating condition. Experimental results show that the algorithms are indeed efficient and effective.

► Non-invasive UTDR for the fouling status in different filtration conditions. ► Successful application of wavelet transform to extracting the features of UTDR. ► A decision tree that represents different operational modes for UTDR signals