Smart buildings predict when critical systems are about to fail

By Chris Baraniuk

Imagine a building that tells you – before it happens – that the heating is about to fail. Some companies are using machine learning to do just that. It’s called predictive maintenance.
Software firm CGnal, based in Milan, Italy, recently analysed a year’s worth of data from the heating and ventilation units in an Italian hospital. Sensors are now commonly built into heating, ventilation and air conditioning units, and the team had records such as temperature, humidity and electricity use, relating to appliances in operating theatres and first aid rooms as well as corridors.
They trained a machine learning algorithm on data from the first half of 2015, looking for differences in the readings of similar appliances. They then tested it on data from the second half of the year – could it predict faults before they happened? The system predicted 76 out of 124 real faults, including 41 out of 44 where an appliance’s temperature rose above tolerable levels, with a false positive rate of 5 per cent.

Advertisement

.

“We started with the hospital because the heating, ventilation and air conditioning system is critical,” says Carlo Annis of building management firm eFM, which worked with CGnal on the experiment. These predictive algorithms could help fix faults before facilities crash –  avoiding unnecessary work at the same time.

Advertisement: Replay Ad

Ads by ZINC

“It’s a nice technique that can be applied to existing telemetry [remote sensing] datasets,” says David Shipworth of University College London. But the differences in function may become less distinct as the units age, he says. “The barrier between normal and faulty operation will become more and more blurred.”

Sounds like a fault

Other companies are also crunching this kind of data. Finnish start-up Leanheat puts a wireless temperature, humidity and pressure sensor into apartments to remotely control heating and monitor appliance health. Its system is now installed in nearly 400 apartment blocks, says chief executive Jukka Aho.
“Once we had these sensors in place, very quickly there was evidence that buildings were not controlled optimally,” says Aho.
Instead of adjusting heating simply based on the outside temperature, Leanheat’s models take into account how the weather has changed. Has the temperature fallen to zero from 10 degrees – or risen from 10 below?
In the US, start-up Augury installs acoustic sensors in machines to listen for audible changes in function and spot potentially imminent failures. CEO Saar Yoskovitz says Augury has “diagnosed” machines in facilities including hospitals, power plants, data centres and a university campus.

As the cost of sensors continues to fall, Shipworth says there will be more systems like this on the market. “We’ll see a whole bunch of different machine learning approaches thrown at this over the next few years,” he says.