You are here

Adding intelligence to pipeline monitoring

07th February 2019

Technologies such as artificial intelligence, data analytics and edge computing are now allowing pipeline safety and security to be improved

Pipeline companies use a wide variety of methods to monitor pipelines – from highly advanced technology to patrolling the pipeline right-of-way. Visual inspections are done regularly – either by walking, flying or using drones – and the industry also uses electronic monitoring from high-tech control rooms and patrols inside the pipeline.

But technology is now coming online that can advance the practice, moving from traditional alarm-based systems to a more autonomous agenda utilising the benefits of Artificial Intelligence (AI). There are a myriad of challenges facing pipeline operators complicated by the remote nature of much of the infrastructure and the ageing pipeline system that has a lack of continuous visibility. Operators need to safeguard this infrastructure against vandalism and damage as well as ensuring safety and environmental compliance.

One company at the forefront of asset intelligence for remote infrastructure is California-b ased Atomiton. Its system can continuously monitor remote pipeline integrity through multi-sensor data integration that allows modelling of vital equipment such as flowmeters and valves. By utilising AI and edge computing it can detect and alert central operations to potential issues such as leaks, corrosion, freezing damage or vandalism. It also allows operators to capture and analyse remote operations through image analytics. But above all it adds intelligence by predicting and optimising  pipeline maintenance and integrity.

“Detecting something happening to the pipeline has been a solution sought for a long time mostly because they are remote,” Jane Ren, founder and CEO of Atomitonexplains. “We do not always want to send
people there: when something happens, it is usually very srious and sometimes can have catastrophic consequences.”

She explains that one of the biggest issues with pipeline detection is that  traditionally it is threshold based. “You are trying to see an anomaly and if your sound or acoustic signal exceeded a certain threshold
then there is going to be an alarm on the traditional system.

“The biggest problem with that system is false alarms which means that 80, 90 per cent of the time you get an alarm that isnothing significant. Which in some instances can cause alarm fatigue, where people do not

Adding intelligence
That scenario is now changing thanks to the maturation of AI. “From the technology perspective we see a growing number of sensors used in pipelines which means there is a growing volume of data which AI can help us interpret and gain insights from,” Ren says. “We recently applied our technology to one of the large natural gas companies that run pipelines here in the US. For this we did not rely on the traditional simple threshold-based system, but we used artificial intelligence of pattern recognition to differentiate the different vibrations: threedimensional vibration patterns of different things happening on the pipeline.”

There are numerous factors that cause false alarms on pipeline monitoring systems but one of the biggest culprits is rain. Other common causes are animals walking on the pipe or heavy trucks driving by, causing vibration patterns on the pipe that trigger an alarm. But alarms cannot always be ignored as there are events that need to be detected and acted upon such as vandalism, drilling or hammering or other damaging impacts on the pipeline.

“When we apply accelerometer and vibration AI on the edge for pipelines, the algorithm can detect that these threedimensional images or patterns are very different between the types of impacts and can detect with high confidence on over 80 per cent of the incidents,” Ren says. “Not just, is something happening, but what is happening? Is it raining or vibration or is it something or someone trying to damage the pipeline? That is a significant difference between how those things can be detected using AI compared to traditional sensing and monitoring methodologies.”

Although there is a great deal of commonality between monitoring requirements, there are also some challenging differences which means that any solution aimed at this sector does require the technology or solution to be versatile to allow it to integrate a variety of different sensors. “If you want to just read the meter at the regular level you can simply utilise a camera along with a visual analytic tool, but if you want to detect a gas leak you may need to deploy some specific sensors that can detect gas molecules,” Ren explains. “On the other hand, if you wanted to detect vandalism you might want to use a vibration sensor, or an accelerometer. For flow analysis you would need to use sonic or ultrasonic sensors. You may even want to triangulate with multiple signals to make sure your detection is validated with more than one type of data.”

When it comes to sensors, the majority are added as part of the solution, although existing ones can be integrated. “We choose multiple different types of sensors depending on the customer need, the dimension or the nature of the pipeline and other requirements such as cost,” Ren adds. “We integrate them into our technology and provide the complete solution to the customer.” 

Adding a digital profile
When it comes to improving visualisation through modelling devices Atomiton utilise a digital twin concept. However, Ren is keen to emphasise that it is not a static digital twin, but a dynamic digital profile or model. “This means that you can get the latest information from gauges and meters as frequently as you want, and continuously update the profile” she says. “This can be by second, minute, day or by week, you can control it. This continuous profiling provides real-time operations intelligence.

Older or analog equipment can also be digitized. There is no need for people to read and record the data manually, because AI on the edge solution handles it. It uses image analytics to learn how to read those gauges and meters, which is often far more accurate than human readings. Whenever there are errors in alarms systems it is often humans’ mistakes in terms of the reading capabilities.

“Interestingly this is not a single dimensional problem either, because there are numerous kinds of meters and gauges out there utilising dials and numerical displays. The AI programme can first tell which type of meter or gauge it is and then read the numbers or dials and create a digital profile. It will decide what significance that data has and where it needs to be sent; some data feeds down into the billing and ordering system and others to operations or maintenance.”

Image analytics
There are several different applications where you would want to use imaging technology to add intelligence to remote assets and operations. One of them is simply to inspect the infrastructure which includes the pipelines to read the meters and gauges. The second type is more involved and is where you can use images to monitor any deformation of the infrastructure. In this scenario the image can define the baseline profile of the infrastructure, such as a tower or pipeline. “This goes back to the AI algorithms that we deploy and use,” Ren explains. “The software can learn the shape and the colour. After a certain level of intensive training, or machine learning, once the shape and colour changes significantly, it can detect the difference. You maintain the baseline, it can detect what has been compromised or different. Of course, image analytics can also be used for people. You can use the technology to confirm that the right people are in the right places, wearing the correct safety equipment and operating in a safe manner.

“One further benefit is the ability to optimise field personnel visits and schedule with real-time visibility of infrastructure integrity. The real-time visibility of infrastructure integrity is critical to safe operations, as are reducing the cost of inspection and reliability of the infrastructure inspection and reducing human errors in meter and gauge reading. But the real upside of that value, is the availability of real-time product flow and imagery information that will be available for analytics.

“For image analytics where we identify people, certainly, that benefit is with the safety and security and just the impact of the workers for the industrial operations. The other benefit of intelligent infrastructure is to avoid environmental compliance issues or violations. It’s not just about interpreting. For example, we’ve deployed a digital silt fence at a construction site for one of our customers because if any violation happens, the fine from the environmental agency here is about $40,000 a day. That’s a very tangible economic benefit when you have a continuous visibility in your infrastructure.”

Beyond natural gas pipelines, there are other pipeline operations that have similar problems. For example, in an oil pipeline, you would want to analyse the flow and improve corrosion, as well as identify and pinpoint potential leaks. Different sensor types and devices may be integrated with the software to build digital profiles, and then used to predict and optimize pipeline maintenance and integrity.

The benefits of leveraging AI for remote intelligent infrastructure is clear, offering the ability to optimise performance, monitor integrity and reduce both safety risk and environmental impact. Combining intelligent algorithms with an end to end solution that integrates sensors with real-time edge computing for profiling of assets and activities is the future for pipelines and it is here today.