AIChE Meeting - Pipeline leak localisation

 Talks AICHE 2018 

AICHE 2017 Pittsburgh

Junyao Xie and Stevan Dubljevic, “Process Monitoring and Leakage Diagnosis for Distributed Pipeline System Based on Discrete Observer and Moving Horizon Estimation”, AIChE Annual Meeting, Pittsburgh, PA, October 31, 2018

Junyao Xie and Stevan Dubljevic, “Discrete Observer and Kalman Filter Design for Pipeline State Estimation and Leak Detection”, 68th Canadian Chemical Engineering Conference (CChEC), Toronto, October 29th, 2018.


This work is supported by Enbridge Inc

Talks AIChE 2017



Pipeline Big Data Analysis - Leak Localization

#4, 103F room Thursday, November 02, 2017 08:00 AM - 08:17 AM

Pipeline network in North America (Source Enbridge Inc)

Pipeline Energy and Environmental Economics

  • 55,377 km of gas pipelines
  • 28,181 km of liquid pipelines
  • Move 28% of crude oil produced North America and 23% of natural gas consumed in the United States
  • Workforce of 15,400 people
  • $27B+ in secured capital projects for growth
  • No.12 on the 2016 Newsweek Green Rankings
  • Environmental issues
  • Insurance and leak prevention (pipeline integrity)


Pipeline inspection

  • Inertial Measurement Unit
  • External signal sink-source communication

Leak position localisation

  • Measurements of IMU unit
    • Gyroscopes rotation angles $latex \omega_1$, $latex \omega_2$, $latex \omega_3$ - Euler angles
    • Linear translation motion $latex a_x, a_y, a_z$-acceleration
    • Magnetometer
  • Error of integration prevents accurate localization (known problem in aerospace engineering)
    \begin{equation} a_{x}=\frac{d^2x}{dt^2}\end{equation}
    the measurement error is accumulated by double integration of  $latex a_x $ signal
    \begin{equation} x(t)=\int \int a_{x}d\tau\end{equation}
  • Practical solution is dead-reckoning
  • The IMU chip is designed to be placed in the geometric center of the ball in order to prevent the error during rotation.
  • The position of the battery is designed to keep the mass balance of the robot.

Big Data Analysis

  • Cheaper swimbot (87$)
  • Large number of swimbot deployments (every batch of transfered fluid should have 2-3 swimbots)
  • Large number of data for statistical processing and mining
    • Deep Neural Network
  • Pipeline integrity and leak prevention

Big Data Processing

  • Problem of integration drift is the main issue in all IMUs
  • Algorithm is applied to fuse acceleration and and angular velocity to generate accurate positioning

    • Quaternions
    • Euler angels
    • Rotation Matrix
    • Fusion algorithm and Madgwick library
  • Artificial Neural Network and Deep Neural Network

Model of Deep Neural Network

Experimental Test Environment


Experimental Results

The experimental run takes 4-5 seconds in the recorded interval between 31 sec and 36 sec

Experimental Results





The experimental run takes 4-5 seconds in the recorded interval between 31 sec and 36 sec

Experimental Results

Position Calculated

Pipe length is 3.22 m and relative error is 5.3%

Acustic data


  • Leak localization is explored
    • Issues of drift associated with IMU are addressed
    • Fusioon algorithm
    • Deep Neural Network
  • Data exploration for large number of deployed swimbots

Future work:

  • Bayesian inference
  • Several agent swimbots with communication protocol

Author: Stevan Dubljevic

Created: 2017-11-06 Mon 18:48



This work is licensed under a Creative Commons Attribution 4.0 International License.