The Blog of operational excellence and energy management for industry

Browse our articles, expert advices and clients testimonials: our experience save your time!
In metallurgy, Big Data and big furnaces make a pair

In metallurgy, Big Data and big furnaces make a pair

Electricity-intensive industries didn’t wait for global warming to become an issue before worrying about energy efficiency. Take metallurgy, for example: during the steel-working stages, of course (melting, refining and recycling), but also for extractive metallurgy (the initial ore processing), which uses impressively large and powerful furnaces. These very costly facilities are already designed to be highly efficient, so there is little room for improvement in the actual design. The next step towards reducing their energy consumption involves Big Data.


Heavy industry… but finely tuned

Let’s briefly go over the pyrometallurgical process. This is one of the transformations of the crude ore, which is generally extracted in the form of an oxide alloy, ready for use by steelmakers or chemical industries. Deoxidation (in the chemical sense, which consists in eliminating the oxygen by making it react with carbon) is performed by heating the ore to a very high temperature (between about 1400 and 1800 °C). This is often carried out in electric arc furnaces that are over 10 metres in diameter, have tens of megawatts of power (20 to 80 MW, if not more) and operate around the clock.

In principle, managing these furnaces consists in analysing the incoming ore (quality and composition), then adjusting the process parameters (quantities of reducing agent, instantaneous heating power, etc.) to suit the specifications. In practice, though, it is far more complicated than that! First, it is not easy to know the exact composition of the ore at that time. Next, there are several electrodes in a furnace and sometimes they have to be adjusted individually to obtain an even distribution of the energy. Moreover, setting aside the energy objectives, the process can become unstable, which raises safety issues for the staff and the risk of producing an alloy that is not to spec.


Sparse technical information but copious know-how

You might think that installing a few sensors would suffice to monitor all that. While there is some truth in that, in practice the situation is more complicated. It is virtually impossible to install instruments inside the refractory lining because of the high temperatures. Not only that but the high electrical currents and the electric arcs generate electromagnetic fields that make electrical measurements complicated. Lastly, the sensors become fouled, like the furnace itself, and its behaviour changes over time and with use. Does this mean that the interior of the furnace can be read like a black box? Not entirely: some (several hundred) remote temperature sensors, electrical data (resistance, reactance, etc.) and information about the alloy produced (composition, casting temperature) can be used to deduce the furnaces’ operating condition. Not to mention the skill and experience of the steelworkers, who know not only how to operate their furnaces but also how to detect risky situations.


A furnace 4.0… and steelworkers are still as important as ever

The fact remains that some situations are discovered late, and that energy optimisation very often proves to be too fine-tuned and/or too complicated to be left solely to the teams’ instinct, no matter how brilliant and experienced they are. Some steelworkers use expert systems that cross-reference the electrical parameters and the composition of the ore with business rules. However it is still complicated to document knowledge that is seldom recorded, that is more about know-how, and which is often specific to a particular ore/furnace combination.

There is, however, a solution on the horizon: Big Data. The existing data is sufficient, in fact, to construct statistical models that can be used to predict how a given furnace will behave. The results are promising. For example, it can predict an event with 70% probability, 10 minutes before it occurs. In the short term, though, it will not be possible to automate furnace operation: the steelworker’s opinion will continue to carry the most weight for a long time yet. Like the vehicle diagnostics readers in cars, these tools should be seen as driving aids. “Caution: you are entering a control zone!”