Evaluation of Piezodiagnostics Approach for Leaks Detection in a Pipe Loop

Pipe leaks detection has a great economic, environmental and safety impact. Although several methods have been developed to solve the leak detection problem, some drawbacks such as continuous monitoring and robustness should be addressed yet. Thus, this paper presents the main results of using a leaks detection and classification methodology, which takes advantage of piezodiagnostics principle. It consists of: i) transmitting/sensing guided waves along the pipe surface by means of piezoelectric device ii) representing statistically the cross-correlated piezoelectric measurements by using Principal Component Analysis iii) identifying leaks by using error indexes computed from a statistical baseline model and iv) verifying the performance of the methodology by using a Self Organizing Map as visualization tool and considering different leak scenario. In this sense, the methodology was experimentally evaluated in a carbon-steel pipe loop under different leaks scenarios, with several sizes and locations. In addition, the sensitivity of the methodology to temperature, humidity and pressure variations was experimentally validated. Therefore, the effectiveness of the methodology to detect and classify pipe leaks, under varying environmental and operational conditions, was demonstrated. As a result, the combination of piezodiagnostics approach, cross-correlation analysis, principal component analysis, and Self Organizing Maps, become as promising solution in the field of structural health monitoring and specifically to achieve robust solution for pipe leak detection. Introduction Leaks are the main source of incidents and damage types reported by fluid transportation companies, which cause abnormal operation and possible adverse consequences. For instance, the US Department of Transportation Pipeline and Hazardous Materials Safety Administration (PHSA) has reported around 11.192 incidents in the last 20 years period, which has generated costs by approximately MMUS$ 6.800, corresponding to hazardous liquid incidents [1]. Bashtransgaz Ltd has documented another accidents, where people were affected due to ignition of natural gas transported by Urengoy-Petrovsk GP. Since these accidents affect the integrity of surrounding structures (i.e. buildings, farms, etc), human safety and have great economical impacts, it is necessary to obtain a leak detection system with high reliability, in order to reduce maintenance costs and minimize accidents occurrence. In this sense, structural health monitoring (SHM) methodologies can be implemented in order to achieve robustness, precision and sensitivity characteristics specified on international standards as API 1155. Key Engineering Materials Online: 2016-09-30 ISSN: 1662-9795, Vol. 713, pp 107-110 doi:10.4028/www.scientific.net/KEM.713.107 © 2016 Trans Tech Publications, Switzerland All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans Tech Publications, www.ttp.net. (#74700351, Iowa State University, Ames, USA-21/02/17,20:27:32) According to above requirements, in this paper a methodology for SHM based on piezodiagnostics and principal component analysis was used to detect and classify pipe leaks by means of self-organizing maps. Experimental data from a carbon-steel pipe loop was used to validate the effectiveness of the methodology, and its robustness considering environmental and operational variations. Pipe leak damage assessment methodology Pipe leaks detection and classification are achieved by implementing three stages (See Fig 1.): i.) Data acquisition, ii.) Statistical feature extraction and iii.) Clustering. Figure 1. Components of the pipe leak detection and classification methdoology In the data acquisition stage is used the piezo-diagnostics principle, by bonding piezoelectric devices to the surface structure, where one of them operates as actuating element and the other ones as sensors. Several works have demonstrated the high sensitivity of guided waves generated by PZT devices to scattering, reflection, and mode conversion due to structural discontinuities, which makes possible to identify structural damages [2, 3]. After piezoelectric records are collected, crosscorrelation is computed between actuating and sensing signals with the objective of reducing common noise and minimizing data trends effects. The statistical feature extraction stage comprises two procedures: data reduction through principal component analysis (PCA) and statistical indexes computation. By means of PCA a minimal representation of cross-correlated signals is obtained to represent the data variance in a reduced space with minimal redundancy. If records from undamaged state of the structure are unfolded in the 2D matrix X, then a baseline statistical model corresponding to pristine measurements can be expressed through PCA by means of Eq. 1 [4, 5]. X = TP + E = model + noise (1) Where, P is a linear transformation matrix that relates the data matrix X in the new coordinates and it is known as the principal components. T is the projected matrix to the reduced space and the noise E-matrix describe the residual variance neglected by the statistical model. Once the PCA model is built, and statistical indexes (defined by Eq. (2)) are used to determine the existence or absence of pipe leaks [6], where differences between statistical indexes regarding to baseline and current state are attributed to leak presence. The Q-statistic is defined as a lack of fit measure between the analyzed experiment and the baseline records. Likewise, the Hotelling T statistic indicates how far each trial is from the center (T = 0) of the reduced space of coordinates. = ∑ and = (2) N o d e 1 N o d e 2 N o d e 3 COM PRESO R Correlated signal Graphical Interpretation Index and Scores Actuator s ignal