Today and The Future
Industrial operations in the future will occur very differently than they do today. As examples, let’s look at a mine or a chemical refinery:
Today, the machinery at a plant is operated in a half manual/half automated fashion. But regular operation is just one phase through which every piece of industrial equipment must pass, and the additional phases of planning, inspection, maintenance, upgrading, and tear-down are all completely manually. People in hardhats climb ladders and walk down tunnels in order to squeeze through dangerous places just to reach the piece of equipment they need to build, examine, repair or replace.
In the future, not only will regular operations be managed from a control room full of laptops anywhere in the world, but all other stages that go into running a plant will be dispatched from the same control room. Rarely, if ever, will a human enter the maze of tunnels and pipes. This is not to say that everything in the automated plant will run smoothly on its own. No no, human intervention will always be needed. Instead the removal of humans from the interior of an operation will be the result of everything one might want to know about the operation, or everything one might want to change about the operation, being turned into data (“data-ized”). Data that can be accessed from anywhere in the world.
This data will be collected from a bewildering array of sensors in the plant, sent over the internet, and visualized and analyzed in real time on screen. Decisions will be made, and command data will be sent to robots and other internet-connected actuators for execution.
What separates today from the future? Put simply, humans have some amazing capabilities that are prohibitively expensive, or outright impossible, to replicate with a machine:
- Humans are great at moving around. Not only can humans move along a pathway without bumping into things, they can move quickly and surely in areas they’ve never seen before. Then they can transition to ladder climbing mode, then stop and hold on to stretch wayyyyyy out to reach just the right switch.
- Humans are dexterous. Not only can humans squeeze through tight spaces, they can turn tiny screws and big valves, and use any tool in addition to their hands.
- Humans are clever and creative. If their path is blocked, they go around. If they can’t reach a tiny little screw, they bend a paperclip and fish it out!
An industrial plant is not just a sum of vast steel hulks of boilers and conveyor belts, but more than the sum of a billion-and-one little dial tune-ups, trips to turn a valve, or fine alignment adjustments done with a sledge-hammer.
The true gulf between today and the future is this vast multitude of tiny modifications, driven by an even more enormous number of decisions, allowed by a truly stupendous number of observations. All of which have to be captured and modeled as data points. The basic, daily production operation of the plant is only a minor contributor to this barely countable number of data points; its all the “meta-operations” surrounding and enabling the main operations that generate 99% of these data points.
From this perspective, the task of reaching the future becomes not just daunting, but overwhelming. Automating the actual visible pieces of machinery that we associate with the operations of the plant would be hard enough, but what about the invisible billion-and-one little operations? Impossible.
Well, whenever a task is impossible, we can break the task down into smaller and smaller tasks until we find something that is possible.
Visual Inspection Observations
Let’s think of the plant as a collection of data points, and organize these to the phases:
Our rule of thumb is that software is easy, and hardware is hard. This is to say that interacting with the physical world is much harder than tapping keys on a keyboard. If I were to sort these categories according to their quantity of relative interaction with the physical world, (from most to least) the list would be:
- (Tie) Construction, Tear-down
Planning has the least interaction with the physical world, followed by inspection. However, planning requires too much domain knowledge and will forever remain an art of choosing between trade-offs. One could say that in the future, when all subsequent phases are automated, the only criterion separating success from failure will be the art of planning. This is too big a bite to chew!
What about inspection? Inspection does indeed require interacting with the physical world. You have to get out and actually inspect elements of the plant. However, while some inspections might require physically touching or interacting with the element under inspection, many inspections do not. In fact, all visual inspections do not. These inspections follow a “window-shopping” philosophy where you look, but don’t touch. If you can navigate within the physical world to get into position for a visual inspection, after that point, it’s just easy software.
So let’s focus on visual inspection.
In any phase, actions will be preceded by decisions, which will be informed by observations. Observations are always the first step, and the quality and quantity of observations limits the possibilities and effectiveness of all subsequent decisions and actions. On the flip-side, high quality and quantity observations broaden the possibilities for, and enhance the quality of, any subsequent steps. Observations are the critical foundation of every action, and focusing on observations dove-tails nicely with focusing on visual inspection.
So let’s focus on observations.
Put these two together, and that’s why we are the visual inspection observation crew!
New Technologies, New Benefits
Recent developments in technology have made navigating the physical world dramatically easier. Quadcopter drones with advanced cameras and onboard processing capabilities can fly to targets while avoiding obstacles, and get into position for a visual inspection of any surface reachable from the sky.
Recent developments in technology have also made data storage and analysis dramatically better. Cloud computing offers both the scale and the cost-efficiency to allow systematic, daily, full-plant observations of every nook and cranny. New machine-learning techniques allow these systematic observations to be statistically analyzed in ways that save time and yield undiscovered insights.
We are an industrial exploration technology firm, focused (but not exclusively) on mining. We seek to provide novel capabilities for visual inspection observations, using quadcopter drones and large-scale statistical computing.