There’s no shortage of Silicon Valley companies eager to wire up the world, and no shortage of sensor manufacturers which can supply new measurements of process variables. The systems to shape and present all this data have benefitted massively from the ‘big data gold rush’.
This is leading to a new problem – and one that’s quite distinct from the familiar ‘information overload’. If a new situation occurs – how does the operator know how to respond? This may arise from a real fault occurring, or it may just be that a source of information is incorrect. A classic example is the Air France flight 447 disaster, where the pilots were faced with conflicting information from their instruments. This not only lead to a conflict between them about what to do, but it lead to the autopilot disengaging so their decisions became critical to the plane’s safety.
Many industries could now be facing a similar problem. The desire to have real-time, continuous (and by implication highly efficient) control of their facilities is leading a desire for sensing, control and automation. This leads to an assumption that replacing knowledge which is normally on the factory floor can be replaced with ‘a computer’. The temptation is all the stronger when knowledge is leaving the industry.
How can we use the latest techniques to support human operators and assist their decisions?
A good example is the first recognisable spreadsheet: VisiCalc. It became the ‘killer app’ for the desktop computer because it allowed small business to ask the key question: ‘what if?’ This allowed them to predict the effect of things outside their control and study what to do without undue effort. This became hugely popular because analytics was supporting rather than usurping control.
An example is the work we’ve done on analytics on London Underground travel data. The maths is able to discern patterns in the data – interesting enough in itself. But, the user is able to play the same ‘what if’ games by closing stations at peak times to see where the traffic re-routes. This allows controllers to learn about the expected behaviour so they can respond and be ready when unexpected events occur.
This is the control system of the future – there’s still a ‘man in the loop’ who has the final say, but he has the ultimate advantage: the chance to experiment and learn.