Unexpected downtime is terrible for any facility. Whether it’s a factory, data center or an office building, the impacts of unexpected downtime are wide-ranging and difficult to measure.
When a key system suddenly fails; earnings may drop to zero, costs may remain high, stock may spoil, compensation may be due, customers may be lost, and reputation may be tarnished. When these events occur, the clock starts ticking and every minute costs. This forces management and maintenance teams into frantic and stressful scenarios, where rash and expensive decisions become the best course of action.
It doesn’t have to be this way, however, smart predictive maintenance systems are able to reduce the vast majority of unanticipated downtime by alerting maintenance teams before a system fails.
Typically, predictive maintenance for building systems involves scheduled servicing on certain dates or after a certain number of operating hours. Similar to your car, where simple historical data has determined that it is best to service the vehicle after a certain number of miles/km; there is little concern for all the other variables that may affect performance and potential for breakdown. Similar to your car, your building systems, despite regular servicing, can and probably will breakdown every now and then, often at the worst times.
A recent research paper from CGnal promotes a dissimilarity-based approach to predictive maintenance with application to heating ventilation and air conditioning (HVAC) systems. The Milan based software firm analysed a year’s worth of HVAC data from an Italian hospital. Using records of temperature, humidity and electricity use for each system, across the hospital, the researches sought to better predict the different maintenance demands of similar appliances.
“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. “We tested our dissimilarity-based features over one year of historical data from 17 Heating, Ventilation and Air Conditioning systems… obtaining good accuracies”, the report reads.
The research team trained a machine-learning algorithm to predict faults before they happened. Their method 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. Concluding that “the attained results motivate us to further research the topic”.
In Finland, smart technology start-up Leanheat has installed wireless temperature, humidity and pressure sensors in approximately 400 apartment blocks. Their technology enables the remote control of heating but also the monitoring of appliance health and performance through big data analysis. “Once we had these sensors in place, very quickly there was evidence that buildings were not controlled optimally,” says chief executive Jukka Aho.
The contemporary internet of things (IoT) mindset, and the related drop in the cost of sensors, encourages us to think outside the information gathering box. Beyond the traditional time, hours, miles approach, and even the individual temperature, pressure and humidity variables, in order to predict faults. And it sounds like one US start-up is doing just that.
New York based Augary are bring predictive maintenance to new markets by listening to a machine to tell if it is working properly or has a malfunction. By using acoustic sensors in machines to hear the audible changes in function, they claim they can even predict a machine’s future failures. The acoustic data is fed into an algorithm to identify what’s wrong and predict what’ll break next.
“We focus on three types of machines: pumps, fans, and chillers or compressors. Within these three families we can diagnose any of the machines we encounter,” said Saar Yoskovitz, CEO of Augury. “We don’t need to generate an algorithm for this pump versus others. The sounds these machines make have a unique fingerprint and we can detect it across any model… a fan belt sounds like a fan belt.”
That IoT mindset opens up a world of possibilities for sensing machine health in innovative new ways. We could soon see vibration sensors, internal video cameras or even smells signaling the problems with machines before they happen. “We’ll see a whole bunch of different machine learning approaches thrown at this over the next few years,” according to David Shipworth of University College London.
For the buildings sector these developments represent real value. The rise of the smart building has brought about energy optimization based on occupancy, user comfort alongside power generation and storage. It is now time to unlock the value of machine health for ideal performance and accurate pre-emptive maintenance.
“For residential high-rise buildings, as well as other commercial and institutional facilities, predictive maintenance plays a key role in maintaining property value and reducing risk for the building owner,” states Rebecca Kim is president and CEO at San Diego-based Urban Property Services Inc. “Whereas higher property value is often perceived as a result of attractive building appearance or amenities, the building systems behind the façade have the greatest impact on value.”
[contact-form-7 id="3204" title="memoori-newsletter"]