Seeing is believing: Visual Analytics in the Oil Industry

IMG 1605

Seeing is believing: Visual Analytics in the Oil Industry

What if a camera could not only record what it sees, but could also interpret and then take action based on what it sees? Could this capability be exploited in oil and gas?

Visual data grows up

Machine interpretation of visual data has been around for what seems like forever. Airports rely on machine readable luggage tags (those bar codes on the paper tag) to route your belongings through the maze of conveyor belts that connect your luggage to your destination. Modern cash registers scan retail item tags and figure out what you’re purchasing, its weight, its price, any applicable discounts, etc. You become the checkout clerk. These are pretty mundane examples today, but back a few years they were considered pretty revolutionary.

Overtime these visual systems have evolved to record and interpret other data. One of my more embarrassing encounters had to be when I picked up a moving vehicle violation in Calgary. The camera was dash mounted inside a ghost vehicle owned and operated by Calgary’s finest, who were hunting for people in my situation – speeding. The cool part was that the system read the license plate from a distance, looked me up on the vehicle registration system, and mailed me the ticket, along with the photo of my car.

Next generation visual analytics

It should be no surprise that these visual data capture and interpretation systems are evolving rapidly, enabled by several digital technologies are combining and recombining to create a seriously clever new category of analytics that works with visual data.

• Cheap low energy sensors (in this case, optical), that see in all light and weather conditions. Other sensors can read other “signatures” like heat and emissions;

Cloud computing, that connects multiple such sensors to the cloud where data storage and compute services are effectively unlimited and practically free, and,

Machine learning, a technique where software learns, in this case how to interpret what it sees by being fed thousands of similar images.

Add in decision support, and viola! We have a new kind of solution that can “watch” the real world and take independent “action” based on what it “sees”.

Better still, make it low energy (perhaps incorporating a solar powered device with an integrated battery for night time use), and give it an option to transmit stills as well as video, just when needed so that bandwidth is minimised. Make it light weight as payload on a drone (aerial, underwater or on a self driving vehicle), in addition to pole mounted or located in tight spaces.

It turns out that these systems can be “taught” to recognise virtually anything, from intruders to contractors to wildlife to equipment. Tesla’s self driving car is also based on the super human interpretation of visual data, but specifically for vehicles.

I’ve never thought about the world in these terms before – I’ve only ever assumed that man-made symbols, like the digits of a license plate, could be interpreted by computers. But a system that can take in any scene and interpret it correctly? That’s revolutionary.

Show me the money

Moreover, these new visual analytic systems, interconnected via the web, demonstrate many of the same attributes of other exponential technologies. Multiple cameras connected to a single machine learning engine mean that each camera learns instantly from the experiences of the other cameras, creating a solution whose learning ability is super-human in terms of its range of interpretable situations and decisions it can correctly and reliably take.

Today, these systems are able to recognise and interpret with 90% accuracy what they see, and they will only get better over time.

They will also fall in cost. In fact, they are already ridiculously lower cost than paying for full time camera watching dudes in a control room. A single camera might be in the $0.15-$0.25/hour to operate, as compared to $25/hour for a squad of guys (say $60k each all in, 1 dude per shift, 3 shifts, gross up 30% to provide 24/7 coverage).

Cameras are more reliable too. Dudes need bathroom breaks, vacations, training and supervision, and dudes are easily bored with watching screens that don’t change frequently.

Really compelling use cases

Where would such visual analytics find a home in oil and gas? Here’s just a few examples:

Augment control room operations staff

Have you ever toured control rooms of oil processing plants? These on-site bunkers (although the newest ones are off site, usually in some downtown tower where people actually want to work and live), have the usual bank of operators at the SCADA controls, and sometimes feature camera feeds from key positions of the facilities, such as entrances, fuel tanks and storage yards.

Typically there’s not a lot of cameras – they’re expensive because they need dedicated broadband or optical cabling to backhaul the image to the control room monitor screens. There’s also not a lot of real estate in control rooms to allow for too many monitors. Therefore, the monitors might rotate through several views, meaning something might be missed by the human handlers.

But visual analytic systems could monitor hundreds of cameras at a time, and take action based on what is happening in the real world. Only unusual situations would need to be handed off to a human to take action.

Visual analytics would simplify control rooms, allow for further consolidation of control facilities, and potentially reduce the number of operations staff.

Improve compliance cost and effectiveness

Oil and gas facilities need to demonstrate compliance with regulations, and be able to demonstrate that operations have effective compliance regimes in place. Some compliance activities in some jurisdictions will require “eyes-on” inspections of assets and facilities to detect and report on operating state and condition. Some incidents will require demonstration that compliance monitoring was in effect and operational.

Of course, there will always be some regulations that insist on “person-on-site” to carry out compliance activities, but regulators are increasingly open to acceptable alternatives if that might help improve compliance and expand production. Eyes-on no longer means person-on-site.

Visual analytic systems, with their GPS and date/time stamping, still photos and video, are treated with greater confidence than some time sheet that reportedly shows that an operator walked the perimeter.

I expect to see new compliance requirements to monitor green house gas (GHG) emissions, a task that could be accomplished with camera sensors that can “read” and record rogue emissions, once the camera system knows what to look for.

Since compliance costs money, and no one is prepared to “pay for” compliance as a feature, companies should be aiming to reduce their cost of compliance to the lowest possible level and still be compliant. Visual analytics systems have it over the human alternative.

Improve safety outcomes

Imagine a camera system that could look at a field worker and recognise that she is not wearing high visibility clothing, or he is smoking in a hazardous area, or she didn’t make use of safety harnesses or hand grips.

The system could send a gentle reminder (“please hold the handrail”) only when it needs to, not all the time, or could alert the contractor that his people are out of compliance, along with photo evidence. The system could monitor yards, intersections, rights of way to identify emerging unsafe conditions such as heavy equipment in close proximity to people.

Safety outcomes, near misses, incidents and compliance with safety protocols should all improve with visual analytic safety.

Manage field services

Imagine a supervisory engineer who has contracted for services to a well site. Using visual analytics, the system could support the supervisory engineer by automatically opening and closing gates, logging arrival and departure times, and monitoring site activities. This could move supervisory engineers out of field and provide them with greater leverage (ie, monitoring more services at more wells in parallel).

Visual analytics should improve the effectiveness of field services and reduce the friction associated with contracting for services.

Improve security

The most obvious, but probably least impactful, use case is in security monitoring. A visual analytics system could monitor access points to facilities (such as gates), and interpret if a visitor is wildlife (a deer at the gate), an expected and authorised service team, or an unknown and presumed hostile intruder.

The system could automatically execute tactics depending on visitor (such as broadcasting danger sounds to animals, greetings to approved visitors, and warnings to trespassers).

Security costs should decline.

Remove staff from the field

Some remote and off shore assets still have high levels of human presence. Visual analytics systems offer the potential to take staff, including some kinds of permanent staff as well as the services contractors, out of the field, which improves the productivity of the workforce in general, as well as reduces safety concerns stemming from travel.

Combining visual analytics with submersible and aerial drones creates the possibility of automated routine inspections in particularly difficult zones of operations.

Moving ahead

Visual analytics are not going to revolutionize oil and gas extraction and production, but will lower operating cost, improve compliance, and keep headcount from rising again. That should be good enough.

No Comments

Post A Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.