28 May The Promises and the Problems of the Internet of Things
One of the oldest ideas in industrial automation is to hook up some device to a computer system to keep an eye on it remotely. What happens when we scale this idea up?
Thingys and the Net
A lot of money is going to be spent on sensor technologies that connect to the internet. The vendor world is full of breathless hype about the power of a future network of physical devices, vehicles, equipment, valves, pumps, clothing and many other things encrusted with electronics, software, sensors, actuators, and connectivity which enables these objects to connect and exchange data.
Digital technology companies describe these sensor-enabled devices with their uber-buzzphrase the “industrial internet of things (IIoT)”.
Oil and gas companies have decades of experience with sensor-enabled things – it’s called SCADA or Supervisory Control and Data Acquisition. Way back in the day (think 1950’s up to the 1970’s), factory floors had lots of people who stood in front of machines such as pumps and motors, looked at the dials, made manual adjustments and recorded measurements on clipboards. It was dangerous and costly. Quality suffered. Advances in sensors, computers and communications enabled central control rooms to pull real time data from operating equipment and present it to an operations team. Payback was pretty swift – just a couple of years.
With the exponentially falling costs of sensors, chips, analytics and data communications, the cost barriers to adding technology to things have fallen away. Suppliers of industrial gear (such as pumps and valves), will be highly motivated to add digital smarts to their wares as a means of competitive differentiation, and to be in the game to capture value from data.
I’ve been approached by many different suppliers of smart things, including:
- Wrist watches that can provide precise where-abouts of mine workers for safety purposes
- Hard hats with embedded biometric readers to help identify when employees are suddenly prone
- Eyeglasses that behave as augmented reality displays for field services work
- Pumps that send operating data like pressure, temperature and vibration to tablets and phones
- Tank gauges that send tank volumetric measures to the cloud
- Rail car sensors that capture rail car location and cargo measures for rail monitoring anywhere in the world
- Vibration monitors for bridge decks that also serve as weigh scales
There is no practical limit to where internet-enabled devices could be deployed in industry. The chips are so small, feature-packed and low power that it’s not out the question that they could become “disposable“. Observers, usually the ones selling this technology, believe that there are some 8 billion internet-visible things in 2017, which will grow to between 20 billion and 50 billion in the next 5 years. Clearly, managers will need to get in front of this wave lest it get out of hand.
What could go wrong?
These sensors are almost pointless without an internet connection. To get the benefit of the data from the devices, the industry’s various assets (such as off shore platforms, plants and refineries, wells, batteries, tanks, pipelines, jetties), need some measure of access to the internet that lets their data exist in the cloud.
- Problem #1. Most of these old assets were never set up for internet access. SCADA yes, but Wifi? No.
- Problem #2. Oil and gas is rarely found in the midst of civilisation. The likelihood of an available internet service provider is low.
Perhaps it won’t be necessary for robust, always on and high capacity connectivity – some digital devices will use only the barest minimum of network capacity to get the job done. New wireless technologies can blanket a large plant relatively inexpensively, and without the need to trench conduit for wireline.
Australia’s vast and empty bush country doesn’t have internet connections. An innovative sensor company with a digital tank gauge uses satellite uplinks to squirt a tiny data packet of tank data at only certain times of the day when specific satellites are overhead. Check out Hydip.
As end point devices, things will generate significant data volumes. Not all of this data will be of equal value. A camera equipped with visual analytics (a kind of artificial intelligence), need not dispatch all video, but only that moment of video, or even just a frame of video, that needs AI interpretation. Exactly how much computing and analytics should be at the edge is an important design question, along with the question of how important is the data that end point devices generate.
- Problem #3. Most sensor data is going to be pretty useless, but some sensor data will be very valuable. We don’t know which is which.
- Problem #4. Sensors on devices will need some undefined ability to store data, and interpret data locall
Securing the end points from cyber threats is suddenly really important. If there’s one thing about digital that has fully sunk in at the Board level, it’s that cyber matters. Device makers from the SCADA era have not historically incorporated into their designs the kinds of cyber protections that are common to things like mobile phones and servers. Authentication, authorization, patch upgrades and other mundane work typical of commercial IT are pretty foreign to the world of operations where such processes have not been necessary.
One of Australia’s natural gas companies experienced a failure of a SCADA computer that supervised a large battery of gas wells. It took some 48 hours to track down the engineer who was responsible for keeping the system running. As an operations system, it was designed not to fail, so there was no process in place for when it did fail. It had no ability to log an incident, contact a support team, escalate the failure, and patch the equipment, all of which are standard fare for help desks for commercial IT systems. Frankly, this situation is likely to unfold more regularly as sensor-enabled devices find themselves in the field.
- Problem #5. SCADA environments that were not designed to cope with failure aren’t designed for sensors that might fail.
- Problem #6. SCADA systems were not designed to be patched, which is suddenly really important.
Buyers of sensor-enabled devices need to be wary of unintentional supply chain capture. Unlike the broader digital industry with its open source orthodoxy, technology innovation in oil and gas has often been “closed”. Proprietary technologies from the traditional suppliers to the industry don’t integrate well with other technologies. Oil and gas technology has a long R&D cycle, measured in years to progress from innovation to widespread adoption. Suppliers need to recover their investments and therefore protect their innovations with patents. They avoid open source designs.
The integrators of oil and gas technology (generally, the large equipment players), have built out their product families to try to satisfy the demand for equipment that works well together. In doing so, they have inadvertently (or explicitly), created walled gardens within which their technologies thrive, but other technologies do not.
Compare this situation with defence and military technology companies. Military tech must be kept open and fluid to allow for innovation and advancement of military kit, such as smart bombs, improved radar, better protective clothing, encrypted field communications, and so on. Military buyers would not invest in a billion dollar equipment purchase if that equipment would be frozen in time and unable to take advantage of the latest technologies. Military grade equipment must be plug and play. Indeed, ExxonMobil retained a defence contractor to oversee its next oil infrastructure SCADA project to bring that kind of approach to the design.
- Problem #7. The motivations of the various providers of technology to the industry are not necessarily aligned with the owners.
Buyers should also be mindful of the lack of data standards for sensors and devices. The International Body of Standards recently reported that there are over 50 different bodies that set standards in the energy industry, and that there are over 400 different standards that cover information and data. This confusing state creates barriers to achieving interoperability. Since it takes time and costs money to investigate a standard and figure out how to comply with it, and with so many possible standards, digital innovators are understandably reluctant to engage with standards bodies. This has not been the case with mechanical and electrical standards – industry has fewer standards bodies and compliance with those standards is usually a mandatory requirement.
- Problem #8. Data standards in the world of sensors is immature.
Companion technologies, like artificial intelligence and machine learning, become mandatory. Things will generate enormous quantities of data and tools like Excel are not robust enough to process that volume. A single drone flight will generate much more data than can be comfortably analyzed by even a big desktop computer. As the number of installed devices grows, and as users progress from just monitoring devices to analyzing the data from devices, companies will need to invest in analytics.
- Problem #9. New digital sensors will overwhelm traditional data management practices.
Who owns the data from the sensors? Historically, a company that sold assets like pumps cared not a wit what the customer did with the data from the sensors on that pump. Today, I sense a mega battle for data supremacy brewing between the suppliers of strap-on sensor technology, the legacy SCADA systems providers where the sensors are likely to plug in, the original equipment manufacturers whose assets incorporate sensors, the owners of the assets, and the third party analytics and AI companies who interpret the data. Data may be the new oil.
- Problem #10. The ownership of the data from sensors is up for grabs.
Every company in the oil and gas value chain is going to be either a passive victim or an active combatant in the battle for that data. Now is not the time to sit idly by and wait for this market to land.