ANOVA based pipeline scale formation detection using vibration estimates and minimum number of feedback sensors. (July 2019)

• Record Type: Journal Article
• Authors: El-Sinawi, Ameen; Al Ghailani, Latifa; Wang, Yongxiang
• Is Part Of: Measurement. Volume 141(2019)
• Page Start: 302
• Page End: 312
• Publication Date: 2019-07
• Subjects: Pipe scaling -- Nondestructive testing -- System identification -- Failure detection -- ANOVA -- LQG -- Kalman filter; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.04.047
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: The method presented was able to predict scale formation with over 97% accuracy. Scale formation is detected from acceleration estimates. Thus, the need for sensor placement is minimized. Method requires minimum accelerometer measurements. Therefore it is a nondestructive and practical approach. Proposed method was automated using Matlab and dSPACE, thus technique is accurate, with low numerical and economical cost. Abstract: Scale formation is large deposits of minerals inside pipelines that obstructs fluid flow and causes serious operational and structural damage. Hidden inside the pipe, detection of its location presents a challenge and can be costly, tedious, and time consuming. In this work a new vibration-based method for scale formation (i.e. damage) detection in pipes is presented. The approach uses acceleration estimates, rather than actual measurements, to detect damage. Estimates are generated by a dynamic model constructed using two methods, namely (a) finite element, and (b) system identification. Acceleration estimates generated by the dynamic model are improved using Linear Quadratic Gaussian (LQG) servo-controller that requires minimal acceleration feedback from few accessible points on the pipeline. Power spectral densities of acceleration estimates (PSD's) of both, healthy and damaged pipe are compared to quantify shifts in resonant frequencies. Such shifts are treated as the primary indication of damage. Analysis of Variance (ANOVA) of shifts was implemented to construct a statistical damage detection model that was able to identify the location of the scale formed within the pipeline. Experimental results have shown that the presented method is able to detect the scale location with good accuracy.