Pipeline Rupture Detection Using Real-Time Transient Modelling and Convolutional Neural Networks

2018 ◽  
Author(s):  
Joel Smith ◽  
Jaehee Chae ◽  
Shawn Learn ◽  
Ron Hugo ◽  
Simon Park
Keyword(s):  
Neural Networks ◽  
Real Time ◽  
Detection System ◽  
High Reliability ◽  
Leak Detection ◽  
Sudden Onset ◽  
False Alarms ◽  
Pipeline System ◽  
Transient Model ◽  
Social License

Demonstrating the ability to reliably detect pipeline ruptures is critical for pipeline operators as they seek to maintain the social license necessary to construct and upgrade their pipeline systems. Current leak detection systems range from very simple mass balances to highly complex models with real-time simulation and advanced statistical processing with the goal of detecting small leaks around 1% of the nominal flow rate. No matter how finely-tuned these systems are, however, they are invariably affected by noise and uncertainties in a pipeline system, resulting in false alarms that reduce system confidence. This study aims to develop a leak detection system that can detect leaks with high reliability by focusing on sudden-onset leaks of various sizes (ruptures), as opposed to slow leaks that develop over time. The expected outcome is that not only will pipeline operators avoid the costs associated with false-alarm shut downs, but more importantly, they will be able to respond faster and more confidently in the event of an actual rupture. To accomplish these goals, leaks of various sizes are simulated using a real-time transient model based on the method of characteristics. A novel leak detection model is presented that fuses together several different preprocessing techniques, including convolution neural networks. This leak detection system is expected to increase operator confidence in leak alarms, when they occur, and therefore decrease the amount of time between leak detection and pipeline shutdown.

Reality Check: Improving Real-Time Pipeline Monitoring Using Near Real-Time Fluid Data

2008 ◽  
Author(s):  
Joseph Jutras ◽  
Rick Barlow
Keyword(s):  
Real Time ◽  
Detection System ◽  
Leak Detection ◽  
Fluid Type ◽  
Transient Model ◽  
Oracle Database ◽  
Detection Model ◽  
The Past ◽  
Batch Data ◽  
Time Injection

MBS, the software based leak detection system employed by Enbridge, is a real time transient model and as such requires fluid characteristics of the various batches that enter the pipeline. In the past, of the 25 plus pipelines modeled, only 4 received fluid identifiers from the field. These fluid identifiers are a sub-string of the batch identifiers stored in flow computers located at custody transfer locations. On the remaining pipelines, Enbridge used fluid density from the field to infer fluid type and therefore characteristics. In the past whenever a number of fluids had the same density, MBS assigned a best-guess of fluid type. The ‘MBS Real Time Injection Batch Data’ project was proposed to bring fluid identifiers to MBS on the remaining lines with the purpose of improving MBS’ selection of fluid properties. Since injection points on the remaining lines were not custody transfer there were no flow computers at these locations. An existing application called Commodity Movement Tracking, or CMT, was used to provide fluid names to the leak detection model. CMT holds past, present, and future injection batch information in an Oracle database. Batch identifiers are queried, placed into the SCADA system, and forwarded on to MBS. This paper explores the new approach, introduced by the ‘MBS Real Time Injection Batch Data’ project, of providing MBS with batch identifiers.

Near Real-Time Automated Detection of Small Hazardous Liquid Pipeline Leaks Using Remote Optical Sensing and Machine Learning

2016 ◽  
Author(s):  
Maria S. Araujo ◽  
Shane P. Siebenaler ◽  
Edmond M. Dupont ◽  
Samantha G. Blaisdell ◽  
Daniel S. Davila
Keyword(s):  
Machine Learning ◽  
Crude Oil ◽  
Real Time ◽  
Optical Sensors ◽  
Environmental Conditions ◽  
Detection System ◽  
Leak Detection ◽  
False Alarms ◽  
Different Types ◽  
Pump Stations

The prevailing leak detection systems used today on hazardous liquid pipelines (computational pipeline monitoring) do not have the required sensitivities to detect small leaks smaller than 1% of the nominal flow rate. False alarms of any leak detection system are a major industry concern, as such events will eventually lead to alarms being ignored, rendering the leak detection system ineffective [1]. This paper discusses the recent work focused on the development of an innovative remote sensing technology that is capable of reliably and automatically detecting small hazardous liquid leaks in near real-time. The technology is suitable for airborne applications, including manned and unmanned aircraft, ground applications, as well as stationary applications, such as monitoring of pipeline pump stations. While the focus of the development was primarily for detecting liquid hydrocarbon leaks, the technology also shows promise for detecting gas leaks. The technology fuses inputs from various types of optical sensors and applies machine learning techniques to reliably detect “fingerprints” of small hazardous liquid leaks. The optical sensors used include long-wave infrared, short-wave infrared, hyperspectral, and visual cameras. The utilization of these different imaging approaches raises the possibility for detecting spilled product from a past event even if the leak is not actively progressing. In order to thoroughly characterize leaks, tests were performed by imaging a variety of different types of hazardous liquid constitutions (e.g. crude oil, refined products, crude oil mixed with a variety of common refined products, etc.) in several different environmental conditions (e.g., lighting, temperature, etc.) and on various surfaces (e.g., grass, pavement, gravel, etc.). Tests were also conducted to characterize non-leak events. Focus was given to highly reflective and highly absorbent materials/conditions that are typically found near pipelines. Techniques were developed to extract a variety of features across the several spectral bands to identify unique attributes of different types of hazardous liquid constitutions and environmental conditions as well as non-leak events. The characterization of non-leak events is crucial in significantly reducing false alarm rates. Classifiers were then trained to detect small leaks and reject non-leak events (false alarms), followed by system performance testing. The trial results of this work are discussed in this paper.

Real Time Statistical Leak Detection on the RRP Crude Oil Network, Holland

2002 ◽  
Author(s):  
Joep Hoeijmakers ◽  
John Lewis
Keyword(s):  
Crude Oil ◽  
Real Time ◽  
Detection System ◽  
Leak Detection ◽  
Field Application ◽  
Normal Operation ◽  
False Alarms ◽  
Pipeline Network ◽  
Oil Pipeline ◽  
Detection Systems

Prior to the year 2000, the RRP crude oil pipeline network in Holland and Germany was monitored using a dynamic leak detection system based on a dynamic model. The system produced some false alarms during normal operation; prompting RRP to investigate what advances had been made in the leak detection field before committing to upgrade the existing system for Y2K compliance. RRP studied the available leak detection systems and decided to install a statistics-based system. This paper examines the field application of the statistics based leak detection system on the three crude oil pipelines operated by RRP. They are the 177 km Dutch line, the 103 km South line, and the 86 km North line. The results of actual field leak trials are reported. Leak detection systems should maintain high sensitivity with the minimum of false alarms over the long term; thus this paper also outlines the performance of the statistical leak detection system over the last year from the User’s perspective. The leak detection experiences documented on this crude oil pipeline network demonstrate that it is possible to have a reliable real-time leak detection system with minimal maintenance costs and without the costs and inconvenience of false alarms.

Real time leak detection system applied to oil pipelines using sonic technology and neural networks

2009 ◽  
Author(s):  
Alvaro M. Avelino ◽  
Jose A. de Paiva ◽  
Rodrigo E. F. da Silva ◽  
Gabriell J. M. de Araujo ◽  
Fabiano M. de Azevedo ◽  
...  
Keyword(s):  
Neural Networks ◽  
Real Time ◽  
Detection System ◽  
Leak Detection ◽  
Oil Pipelines

Optimization of Averaging Window Length and Alarming Hold Time in Volume Balance RTTM Leak Detection to Minimize False Alarms and Spill Volume

2018 ◽  
Author(s):  
Shawn Learn ◽  
Ehsan Shahidi
Keyword(s):  
Detection System ◽  
Hold Time ◽  
Leak Detection ◽  
False Alarms ◽  
Pipeline System ◽  
Window Length ◽  
Post Processing ◽  
Volume Balance ◽  
The Common ◽  
Moving Averaging

Reliability and sensitivity are two main performance metrics of leak detection systems as defined by API 1130 [1]. Proper thresholding scheme is one of the primary factors in having a sensitive and reliable leak detection system with timely detection. In RTTM leak detection, if not dealt with properly, severe pipeline pressure transients can degrade the performance of the leak detection system. One of the common basic methods of reducing the effect of pressure transients is using moving averaging windows; having looser thresholds on the shorter averaging windows, while maintaining tighter thresholds on the longer ones. The thresholds are typically set to meet the API 1149 [2] curve for the pipeline. While the post-processing of filtered data and alarm assessment has been explored via different methods such as sequential probability ratio test, to the authors’ knowledge, there is currently no systematic way of selecting the averaging windows to minimize false alarms prior to the post-processing of the average-filtered data. Moreover, to be able to maintain tight thresholds, especially in shorter averaging windows, one of the common methods is to apply dynamic thresholds, i.e. temporarily expanding thresholds when transients occur. While effective in some scenarios, the main disadvantage of this method is that the imbalance caused by a transient may not clear until the entire averaging window period is passed. This causes either extended periods of degraded performance, or more false positives. This paper utilizes an alarming hold time (also referred to as alarm persistence [3]) to remedy this problem where the averaging window length is reduced while maintaining detection time and sensitivity. To find the optimal set of threshold values, hold times, and averaging window lengths, a Particle Swarm Optimization (PSO) is performed. The ‘fitness function’ of the optimization algorithm is designed to minimize total spill volume for leak scenarios and have minimum false alarms for no-leak scenarios. The former is achieved via setting the objective function to the spill volume and the latter is enforced via applying constraints to the optimization algorithm. For no-leak scenarios, the historical operational data of a pipeline system is used. For leak scenarios, the historical data is modified by introducing a bias in the inlet volume of the section to simulate a leak. The result of the PSO provides a set of alarming parameters, threshold value, averaging window length, alarm hold time, and clearing threshold that provide the minimum false alarm rate and spill volume for different detectability ranges. The optimization method proposed in this paper can be applied to any mass or volume balance-based leak detection system that utilizes moving averaging windows. However, the leak detection parameters found with this method depend on the pipeline system.

Real-time Detection of Aortic Valve in Echocardiography using Convolutional Neural Networks

2020 ◽  
Vol 16 (5) ◽  
pp. 584-591 ◽  
Author(s):  
Muhammad Hanif Ahmad Nizar ◽  
Chow Khuen Chan ◽  
Azira Khalil ◽  
Ahmad Khairuddin Mohamed Yusof ◽  
Khin Wee Lai
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Heart Disease ◽  
Aortic Valve ◽  
Real Time ◽  
Convolutional Neural Networks ◽  
Valvular Heart Disease ◽  
Detection System ◽  
Processing Unit ◽  
Real Time Detection

Background: Valvular heart disease is a serious disease leading to mortality and increasing medical care cost. The aortic valve is the most common valve affected by this disease. Doctors rely on echocardiogram for diagnosing and evaluating valvular heart disease. However, the images from echocardiogram are poor in comparison to Computerized Tomography and Magnetic Resonance Imaging scan. This study proposes the development of Convolutional Neural Networks (CNN) that can function optimally during a live echocardiographic examination for detection of the aortic valve. An automated detection system in an echocardiogram will improve the accuracy of medical diagnosis and can provide further medical analysis from the resulting detection. Methods: Two detection architectures, Single Shot Multibox Detector (SSD) and Faster Regional based Convolutional Neural Network (R-CNN) with various feature extractors were trained on echocardiography images from 33 patients. Thereafter, the models were tested on 10 echocardiography videos. Results: Faster R-CNN Inception v2 had shown the highest accuracy (98.6%) followed closely by SSD Mobilenet v2. In terms of speed, SSD Mobilenet v2 resulted in a loss of 46.81% in framesper- second (fps) during real-time detection but managed to perform better than the other neural network models. Additionally, SSD Mobilenet v2 used the least amount of Graphic Processing Unit (GPU) but the Central Processing Unit (CPU) usage was relatively similar throughout all models. Conclusion: Our findings provide a foundation for implementing a convolutional detection system to echocardiography for medical purposes.

Pipeline Diagnostics With Ultrasonic Meters

2016 ◽  
Author(s):  
Nicole Gailey ◽  
Noman Rasool
Keyword(s):  
Real Time ◽  
Measurement Errors ◽  
Flow Measurement ◽  
Detection System ◽  
Leak Detection ◽  
The United States ◽  
Time Data ◽  
Critical System ◽  
Detection Systems ◽  
Transient Models

Canada and the United States have vast energy resources, supported by thousands of kilometers (miles) of pipeline infrastructure built and maintained each year. Whether the pipeline runs through remote territory or passing through local city centers, keeping commodities flowing safely is a critical part of day-to-day operation for any pipeline. Real-time leak detection systems have become a critical system that companies require in order to provide safe operations, protection of the environment and compliance with regulations. The function of a leak detection system is the ability to identify and confirm a leak event in a timely and precise manner. Flow measurement devices are a critical input into many leak detection systems and in order to ensure flow measurement accuracy, custody transfer grade liquid ultrasonic meters (as defined in API MPMS chapter 5.8) can be utilized to provide superior accuracy, performance and diagnostics. This paper presents a sample of real-time data collected from a field install base of over 245 custody transfer grade liquid ultrasonic meters currently being utilized in pipeline leak detection applications. The data helps to identify upstream instrumentation anomalies and illustrate the abilities of the utilization of diagnostics within the liquid ultrasonic meters to further improve current leak detection real time transient models (RTTM) and pipeline operational procedures. The paper discusses considerations addressed while evaluating data and understanding the importance of accuracy within the metering equipment utilized. It also elaborates on significant benefits associated with the utilization of the ultrasonic meter’s capabilities and the importance of diagnosing other pipeline issues and uncertainties outside of measurement errors.

scholarly journals Leak Detection Systems for Short Pipelines

2000 ◽  
Author(s):  
Renan Martins Baptista
Keyword(s):  
Detection System ◽  
Leak Detection ◽  
Low Complexity ◽  
Transient Model ◽  
Service Companies ◽  
Detection Systems ◽  
Volume Balance ◽  
Pipeline Networks ◽  
Phase Liquid ◽  
Flow Detection

This paper describes procedures developed by PETROBRAS Research & Development Center to assess a software-based leak detection system (LDS) for short pipelines. These so-called “Low Complexity Pipelines” are short pipeline segments with single-phase liquid flow. Detection solutions offered by service companies are frequently designed for large pipeline networks, with batches and multiple injections and deliveries. Such solutions are sometimes impractical for short pipelines, due to high cost, long tuning procedures, complex instrumentation and substantial computing requirements. The approach outlined here is a corporate approach that optimizes a LDS for shorter lines. The two most popular implemented techniques are the Compensated Volume Balance (CVB), and the Real Time Transient Model (RTTM). The first approach is less accurate, reliable and robust when compared to the second. However, it can be cheaper, simpler, faster to install and very effective, being marginally behind the second one, and very cost-efective. This paper describes a procedure to determine whether one can use a CVB in a short pipeline.

scholarly journals Real Time Object Detection Canbe Embedded on Low Powered Devices

2019 ◽  
Author(s):  
Jimut Bahan Pal
Keyword(s):  
Neural Networks ◽  
Object Detection ◽  
Open Source ◽  
Real Time ◽  
Detection System ◽  
Computing Power ◽  
Convolution Neural Networks ◽  
Real Challenge ◽  
New Algorithms

It has been a real challenge for computers with low computing power and memory to detect objects in real time. After the invention of Convolution Neural Networks (CNN) it is easy for computers to detect images and recognize them. There are several technologies and models which can detect objects in real time, but most of them require high end technologies in terms of GPUs and TPUs. Though, recently many new algorithms and models have been proposed, which runs on low resources. In this paper we studied MobileNets to detect objects using webcam to successfully build a real time objectdetection system. We observed the pre trained model of the famous MS COCO dataset to achieve our purpose. Moreover, we applied Google’s open source TensorFlow as our back end. This real time object detection system may help in future to solve various complex vision problems.

Sign in / Sign up

Export Citation Format

Share Document