Pipeline leakage detection and isolation: An integrated approach of statistical and wavelet feature extraction with multi-layer perceptron neural network (MLPNN) Morteza Zadkarami, Mehdi Shahbazian * , Karim Salahshoor Department of Instrumentation and Automation, Petroleum University of Technology, Ahwaz, Iran article info Article history: Received 18 January 2016 Received in revised form 12 May 2016 Accepted 27 June 2016 Available online 11 July 2016 Keywords: Pipeline leakage Fault detection and isolation (FDI) system Multi-layer perceptron neural network (MLPNN) classifier Wavelet transform Statistical features abstract Leakage diagnosis of hydrocarbon pipelines can prevent environmental and financial losses. This work proposes a novel method that not only detects the occurrence of a leakage fault, but also suggests its location and severity. The OLGA software is employed to provide the pipeline inlet pressure and outlet flow rates as the training data for the Fault Detection and Isolation (FDI) system. The FDI system is comprised of a Multi-Layer Perceptron Neural Network (MLPNN) classifier with various feature extrac- tion methods including the statistical techniques, wavelet transform, and a fusion of both methods. Once different leakage scenarios are considered and the preprocessing methods are done, the proposed FDI system is applied to a 20-km pipeline in southern Iran (Goldkari-Binak pipeline) and a promising severity and location detectability (a correct classification rate of 92%) and a low False Alarm Rate (FAR) were achieved. © 2016 Elsevier Ltd. All rights reserved. 1. Introduction Hydrocarbon pipelines are the major means of energy trans- portation. Pipeline leakage, however, can pose real risks to the environment and cause financial losses. Aging, incorrect installa- tion, material defects, and constructions near the pipeline are the major causes of the leakage fault. Several organizations have studied on oil and gas leak-related incidents. One of these studies, was on the subsea pipeline sys- tems, expresses that, in the Gulf of Mexico and Pacific regions a number of 80 pipeline leak incidents were reported in the 1996e2006 period (Murvay and Silea, 2012). Konersmann et al. (2009) stated that, between 2001 and 2005, only in the area of Alberta (in Canada), there have been 1326 reported gas leaks. Therefore, detecting and identifying the leak before it becomes a great economic and environmental problem is extremely important. Hardware- and software-based methods are the main categories of the leakage FDI systems (El-Shiekh, 2010). Hardware-based systems employ special equipment to isolate the faults directly. Ultrasonic and acoustic leakage detection methods, soil moni- toring, and thermal infrared imaging are examples of hardware- based methods (Bai and Bai, 2014). Software-based methods, which utilize ordinary sensors of the pipeline Supervisory Control and Data Acquisition (SCADA) system, are categorized into the model-based and data-driven methods (Valizadeh et al., 2009b). Examples of software-based methods are the Real Time Transient Modelling (RTTM), mass balance, and Negative Pressure Wave (NPW) methods (Zhang et al., 2015). The aforementioned FDI sys- tems are usually compared in the literature in terms of diagnostic speed, simple and cost-effective implementation, and the ability of accurate fault isolation and severity determination (Murvay and Silea, 2012). An appropriate FDI system is the result of compro- mise between these features. The signals are usually preprocessed before being fed into the FDI system. In the acoustic method, for example, the acoustic signal Abbreviations: ABC, artificial bee colony; ANN, artificial neural network; CCR, correct classification rate; DWT, discrete wavelet transform; FAR, false alarm rate; FDI, fault detection and isolation; LMSE, least mean square error; LPG, liquefied petroleum gas; MLPNN, multi-layer perceptron neural network; MSE, mean square error; NPW, negative pressure wave; PNN, probabilistic neural network; RTTM, real time transient modeling; SCADA, supervisory control and data acquisition; STFT, short-time fourier transform; SVM, support vector machine. * Corresponding author. E-mail addresses: zadkarami@put.ac.ir (M. Zadkarami), shahbazian@put.ac.ir (M. Shahbazian), salahshoor@put.ac.ir (K. Salahshoor). Contents lists available at ScienceDirect Journal of Loss Prevention in the Process Industries journal homepage: www.elsevier.com/locate/jlp http://dx.doi.org/10.1016/j.jlp.2016.06.018 0950-4230/© 2016 Elsevier Ltd. All rights reserved. Journal of Loss Prevention in the Process Industries 43 (2016) 479e487