This is a long post. There is a lot to understand on this topic and this is the primer you’ll need. Please do follow the links. It will take a while but it’ll be worth it.
There is much being discussed around the acceleration of technology and how exponential development in multiple areas is converging and how this will impact on industrial production. About time too really because productivity is failing to pick-up, interest rates at or below zero, cash being hoarded by companies and investment rates are low [link]. This exponential increase is actually not that new – Moore’s law’s been around since… well since Moore launched Intel, and it was a one-way flow before even Schottky was on the scene. These trends are not new and if you don’t believe me here’s a great talk from 11 years ago by Ray Kurzweil with lots of evidence and predictions [Link]. Even Bill Gates saw this in a book he published in 1999 called Business @ the Speed of Thought [link], though these days that speed would be seen as a little too slow.
In my opinion the Oil industry harnessed many of the aspects of this movement in upstream exploration during the 1990’s. It was an early adopter in the process of FINDING reserves. Then the process of adoption stopped. Productivity per geoscientist and the complexity of the information they deal with is orders of magnitude better than it was during the last oil-bust of 1986, so much so that we now have more fields and deposits than we know what to do with. We’re pretty good at finding the stuff. But we’re quite rubbish at developing and operating it at low cost – especially small deposits, which we are so good at spotting now. Development, Operations & Maintenance has ossified – contracts and work practices are stuck. From an operator approach, the production of hydrocarbons has barely moved since the early 90’s (FPSO concepts aside). In my opinion through outsourcing, procurement, short-termism and misalignment of incentives it has become positively petrified.
If the Fourth Industrial Revolution is really going to have an impact we’ll need to address development on four fronts: Economic, Social, Political, and Technical
If this is going to happen then it has got to make sense for the bottom line. That means productivity: outputs, inputs and the cost of technology. McKinsey, recently cited in Industries of the Future, by Alec Ross [Link], suggested that the manufacturing sector could raise productivity by 2.5% to 5% and save more than $1Trillion in cost annually.
In one example McKinsey says “To capture the potential, manufacturers can consider three moves. Primarily, companies can gather more information and make better use of it. An oil-exploration company collected more than 30,000 pieces of data from each of its drilling rigs—yet 99 percent of that data was lost due to problems of data transmission, storage, and architecture. The tiny trickle of data it did capture was incredibly useful for managers. But so much more can be done. The executives we surveyed said that correcting these data inefficiencies should improve productivity by about 25 percent. [Link]”
Of course some traditional economists think we’re doomed to no innovation and permanent low growth – such as Robert Gordon [Link]. There are many in the old-guard of Oil and Gas that would agree. Most of them have their secretaries print their emails out for them, and refuse to carry a smart phone. Good luck chaps, I think you’ll find the millennials don’t care what you think anymore. Others say differently [Link]
There are many big hitters with some very big numbers, they’re all pointing in the same direction. I’m backing the future, not the past. And I think that Industry 4.0 will feature in the future of Oil and Gas. There are challenges but the prize is big enough that we will overcome them.
The way that many of us work is going to have to change. Luckily the Millennials are already preparing for this shift with their search for meaningful work, emphasis on creativity and individuality; and understanding that they can blend their work and leisure time in ways that the crumbly generation see as slacking and entitled. Forbes have a top ten ways in which the work place will be influenced [Link] and Linda Grattan has her views here [Link]. For me it just seems an obvious way to work. But then I’ve never been very good at dealing with routine, structure and command-and-control. It’ll be interesting to see how we can blend the command-and-control requirements of operations with the caffeine fuelled micro-attention span of people even more “wired” than me.
We’re also seeing cyber-social developments such as the creative commons movement [Link] and open source projects like the Arduino [Link] all of which are fuelling exponential cross-fertilisation of ideas. We are witnessing the rise of the sharing economy [link] and temporary configurations of people who move about often. These are all challenging assumptions about ownership and permanence that are at odds with our current ownership-model for resources.
The Guardian in Nov 2015 reported that ” this revolution could leave up to 35% of all workers in the UK, and 47% of those in the US, at risk of being displaced by technology over the next 20 years, according to Oxford University research cited in the report, with job losses likely to be concentrated at the bottom of the income scale.” [link]
With modern communications and the ability to mobilise quickly we’ve already seen massive changes in the way the people (or, in Greek, demos) interact with conventional democratic systems and capitalism. This is very thoughtful piece by Yanis Varoufakis the recently deposed Greek finance minister [link]. Whether that’s the Arab spring, so-called ISIS, Brexit, the mass-migration of populations or the astonishing rise of Donald Trump, things are getting decidedly odd in traditional politics. There’s a lot of complaint and not a lot of traditional power that can be exercised in public anymore [link]. Just take a look at the mass-mobilisation of a Brazilian flash-mob to protest graft allegations levied against the establishment [link]
Cyber-politics is a whole new dimension. Whether cyber aggression is aimed at accessing private information, denying or altering the dissemination of information or compromising the physical integrity of machine-based systems the ability of people to alter the course of events through “hacking” has never been so great. China has its infamous PLA unit 61398 [Link] one of over 20 cyber-military units it controls, North Korea doesn’t like Sony much as the 2014 hack showed [Link], Iran might be the land of the rising Shamoon that hit Aramco [Link], Ukraine has got on the wrong side of Russian Hackers who shut off their power grid [Link], and who knows who might have written Stuxnet that took out the Iranian centrifuges while telling the control room all was normal [Link]? Now the actors are not only nation-states, but also corporations and little boys alone in their bedrooms [link]
We have the Geneva convention that is supposed to stop states shooting the red cross, bombing civilians, gassing troops and firing mercury-filled dumb-dumbs. We have the international court in The Hague (funded by Andrew Carnegie incidentally [link]) that prosecutes war criminals. I’m not sure who I should call if North Korea invades my X-Box or steals my Bitcoins. And if you are a corporation with cross-border operations you don’t either.
There are a number of technologies that are developing exponentially at the moment and they’re feeding into changed ways-of-working that will bring about the fourth industrial revolution. Ultimately this will help you plan to build better plant and it will help you operate what you have better. Optimising operations is a sense-and-respond problem. Prepare for the future, know what’s going on right now and do things to make it better. Technology that helps falls into four areas that increase:
- learning about what’s possible;
- what’s going on right now and situational awareness;
- knowledge of interdependence, decision options and consequence; and
- ability to execute quickly and accurately
Increased learning about what is possible
Big data has gained traction in the last decade. Grab lots of data from everywhere, apply some Bayesian stats, set a base-level and determine the probability of correlation. Works really well when you buy a book from Amazon and it suggests that you might want to buy some reading glasses to go with it. Works pretty well in finding potential hidden relationships and developing predictive algorithms for equipment failure too [link] [link]
Like a lot of developments, this area is moving fast. How do you know what’s even possible these days? It’s so hard to keep up. Data overload, over-stimulation, who even has time to read this stuff?
I remember Schlumberger creating an amazing “portal” called the hub [link], Other companies did similar [link]. Initiatives were started to capture the learnings from each employee and make them available to all other employees. I even heard a talk once describing the use of retiree mentors to help existing employees [link]. People were planning for “The Big Crew Change” when the aging workers retire and new low people come on board [link]. This all tied into concepts like “Hive Minds” which were popular in the 90’s [link].
Well “The Big Crew Change” became the “The Big Layoff” when oil prices crashed in 2014. All that experience and knowledge was not on the balance sheet but was on the P&L. So it was fired without financial impairment and write-off. But the fundamental problem remained, and probably got worse. So much to know, so much to learn and no time to do it. Welcome to one of the drivers that will build demand for machine learning.
Machines can analyse masses of information much quicker than humans can. Up until recently, however, doing that in context and to derive meaning from them has been hard. Development has been showcased by game-playing computers such as Deep Blue for Chess [link], then Watson for Jeopardy [link], and most recently a Google built machine – AlphaGo for GO [Link].
Combine learning algorithms with connected systems, however, and things get really interesting. Learning requires teaching. Unlike programming in Fortran, learning machines construct their own programs by being taught and from the situations they encounter. Distributed and cloud-connected learning is exponential, one machine learns something somewhere and every other machine knows it. Forever. Perhaps we should blow the dust off those long-forgotten “portal” promises around knowledge bases, institutional learning and corporate memory?
What’s going on right now and situational awareness
Machines that learn what matters and suggest how to respond can eliminate operator overload by removing the trivial and hiding noise. Automated actions can be taken to keep things running. I recently heard an analogy about the difference between the information received by a pilot of a typhoon (arguably the worlds most advanced fighting machine) and a world-war 2 Spitfire. The Spitfire pilot had dials telling him the airspeed, engine speed etc. All just data. The Typhoon pilot, however, could not possibly cope with all the data available. So this data is assembled to show him only what he needs to know in the current situation and that depends on context.
Paul Smith, former UK RAF Pilot says “Bring all these [sensor and interpretation] elements together and it becomes clear why we talk about Eurofighter Typhoon operators having the ‘Combat Edge’ – the situational awareness and a suite of flexible weapons options that offer pilots a real advantage in the battlespace.” [Link]
Building on this, if the aircraft systems detect a heat-seeking missile closing, it launches flares automatically and tells the pilot afterwards – no point in raising an alarm and waiting! Same for the oil and gas sector, why do automatic fault development detection systems write a report and wait. Why don’t they just order the parts, consolidate shipments and schedule engineers for the next maintenance activity? It’s a small example but Amazon is already letting washing machines re-order soap powder [link], Imagine what Amazon-like logistics would do for the Oil and Gas industry.
In order to know what’s going on right now requires a lot of sensors talking to each other and reporting back. Too much detail for a human system to ever cope with properly. The data needs to be reporting to systems that learn what’s important, what actions it should take and how it should present its findings to its operators. The system needs to learn how to behave. These systems need to be aware of the situation and act accordingly. Cloud computing is also an important vector this mix, where systems are connected to each other through internet, and keep each other in synch sharing learning and preparing information so that it can be shared widely, securely and scalably. Google are letting developers play with their learning platform [link]. This is an area where we will see rapid innovation that Oil and Gas can benefit from.
Of course there are some very boring building blocks that will be needed. Connecting systems together will of course require a lot of plumbing – don’t underestimate the size of this problem, here is an example of the type of architecture you might need [link]. Companies like Eigen (www.eigen.co), Tibco (www.tibco.co.uk), BEA (www.bea.com) are active in this area. And it’s important that we really know that our data is correct – as in this case when a demolition crew targeted the wrong house and blamed it on google maps [Link]. So companies like datum360 (www.datum360.com) and Informatica (link).
Decision making & Interdependence
Knowing what’s going on is great, but what do you need to do to make your situation better? That’s the question that quickly arises once teams get sight of data and information in context. Firstly it’s important to know what the options could be – but also how choices in one system effect another.
Simulation is one of the keys to understanding the consequences of decisions – that’s why chess computers work out 100’s of moves ahead and choose the best one to use now. To simulate the decisions on a plant requires a digital model of the plant and its behaviour against which to run tests. The digital model is sometimes called a “Digital Twin” and this allows you to make a change or react to a fault condition and see what the knock-on consequences of selected actions will be in the future. This can be used to test options and optimise outcomes. Hit the model with a series of possible actions in an automated way and it’s possible to uncover the best sequence of actions and back-calculate why, rather than the normal forward progression. It’s very powerful – here are some articles discussing simulation of plant [link]
Integrated planning enables you to make a decision about sequencing events in such a way as to minimise down-time by running jobs in parallel within real-world constraints. This might mean being prepared and ready-to-act when an opportunity unexpectedly arises. Understanding system-wide effects is the key to getting this right, and with more complex interconnected systems with cross-ownership (like present in the UK sector of the North Sea) it not easy – and the owner stakes in oil fields can lead to misalignment of financial interests. Bain has a good article on integrated planning [link].
So far it seems that human + machine combination provides the best mix for solving problems. The creativity of the human is key and augmented decision making with rapid feed-back loops from simulation enables optimisation of decisions. From the first simple spreadsheets that appeared in business the testing of “what-if” scenarios has meant that we have been able to tune procedures across many areas of operations. The combination is not a new concept, here is a very relevant paper from Carnegie Mellon 1998 [link]
Increased speed and accuracy of execution
One of the issues that I’ve come across is the “precision” approach of some operators in the field. The best plans are of no use if they are not executed properly, if parts aren’t damaged and if the wrong parts were not fitted. It happens. Sometimes, of course, the instructions make no sense and the field have to modify them to make them work. That modification of instruction is rarely fed back into the system so little learning takes place. Sometimes the plant is updated and records not updated. All this leads to mismatch between what is recorded and what the plant operators “know”.
Sometimes the physical effort required to perform an inspection means that it cannot be done as often as you’d like, or perhaps is skipped by a crew unwilling or unable to schedule. Autonomous vehicles are in use for inspection activities firstly replacing deep divers and latterly, as costs have gone down they are found in inspections roles as Drones taking cameras into inaccessible places. Perhaps it won’t be long until we have small UAV’s mapping plant and equipment in huge detail. Here is a TED talk that demonstrates what’s already possible [link]
On-site machining of parts may soon be replaced by on-site manufacturing. Additive manufacturing (a broader term than 3D printing) is finding its way not only into printing of small intricate parts but emerging are the start of large-scale construction. It’s not there yet, but imagine what this would mean for logistics or construction in hostile environments. Here is an example of a team in Amsterdam who are in the process of printing a Steel bridge over a canal. That could change some of our approaches to Maintenance and Modification one day. [link]
And, of course, there will always be people involved. But multi-skilled and informed. Augmented reality displays – identifying parts, performing on-the-fly risk assessments and acting as advisors. This will change the way that operators will be able to apply basic skills augmented with real-time instruction and feedback. Meron Gribetz demonstrates here a virtual reality system that could revolutionise the on-shore-off-shore interface, as well as providing just-in-time information. Here is his TED talk [link]
And if you don’t think a Robot can replace people on platforms – have a look at this [link]