Thesis Defence: Diversion Detection and Localization in Small HDPE Pipes Using GuidedUltrasonic Waves, Channel Estimation and Deep Learning

Abdullah Zayat, supervised by Dr. Anas Chaaban, will defend their thesis titled “Diversion Detection and Localization in Small HDPE Pipes Using GuidedUltrasonic Waves, Channel Estimation and Deep Learning” in partial fulfillment of the requirements for the degree of Master of Applied Science in Electrical Engineering.

An abstract for Abdullah’s thesis is included below.

Examinations are open to all members of the campus community as well as the general public. Registration is not required for in-person defences.

---

ABSTRACT

This thesis presents a comprehensive study on the use of advanced techniques for detecting and accurately locating diversions in small-diameter High Density Polyethylene (HDPE) pipes. The integrity of pipelines is critical for the safe and efficient transportation of fluids and gas, and the early detection of defects and unauthorized diversions is crucial for preventing environmental damage and economic losses. A novel technique is proposed in this thesis, which is based on the transmission of a specifically designed signal through the pipe using a custom-designed array of ultrasonic piezoelectric transmitters and receivers. The goal of this technique is to generate an estimate of the pipe channel characteristics, which can then be used for defect detection and localization.

The proposed technique combines ultrasonic guided waves, channel estimation, and deep learning to detect and localize defects in HDPE pipes. The first stage uses a supervised learning technique, a CNN-LSTM based architecture, for diversion detection. The second stage extends the first work by proposing an unsupervised learning technique that incorporates a transformer-based autoencoder as an anomaly detection algorithm, which is utilized to detect diversions as anomalies. Finally, we propose a transformerbased encoder for diversion localization, which allows us to accurately locate the diversions with high accuracy and low mean absolute error. The results of our simulations and experiments demonstrate the superiority of the proposed method over the state-of-the-art techniques and its ability to detect diverisions and localize them with high accuracy. The proposed technique is cost-effective and reliable and has the potential to be applied to real-world systems.