As the CEO of Smart Component Technologies, Nick Koiza knows all about adjusting technology to best serve customer needs. He discusses the importance of paying attention to the data, and why it’s essential not to replace human intelligence with AI.
If there is one thing that characterises technology businesses like ours, it’s that we are always learning. At SCT we are constantly learning from our experiences, the data our IoT smart monitoring devices return and from our customers, and their data.
In an earlier insight in this series, I mentioned that remote condition monitoring becomes the eyes of the infrastructure manager. By wirelessly returning data each time an asset is used or, in the case of switches and crossings on a railway line, each time rolling stock passes the asset, a picture of the asset’s performance is built up. Over time, the volume of data collected and transmitted to the cloud for analysis multiplies and the picture goes from being an outline sketch to a full visualisation of the asset over time. Within the data is real insight into the condition of the asset and the substructure around it.
Having installed a large number of our SWiX devices on the rail network in the UK and around the world, we are constantly gathering large volumes of data. And, because they have been in situ for a long period of time, the data provides intelligent insight into far more than the individual asset.
Value in sharing data
The amount of data collected over a long period can sometimes indicate cases that we can’t always explain. When that happens, we work with customers to understand what the data is telling us. That’s when a collaborative approach works best, when customers share their data with us. For example, data from the SWiX might throw up an exception, and when the customer shares with us the type of train, its tonnage and line speed passing that S&C asset at a particular time, we can better understand the results. By sharing datasets and discussing them we uncover the reason for certain unexpected behaviour.
An example of this happening was a case where we could see, from the data, that an asset wasn’t behaving as expected. Our data analytics team saw the data behaving in an unusual way, and we alerted the customer that there were high impact forces around a particular crossing. There could be several explanations for the data readings, the asset could be at fault, there could be high displacements, or some other reason. As we are in daily contact with our customers about their data, we could follow up with their track investigation and learn that the crossing was indeed cracked. From that, we can go back to the data and understand that this type of behaviour within the data indicates a possible cracked crossing. It is a kind of machine plus human learning or ‘explainable AI’ – AI with added HI (human intelligence).
AI with added HI
Applying human intelligence to the data provides far greater insight than AI alone. Our data scientists can combine data with lateral thinking and the lived experience of our customers to get a fuller, more nuanced view. From that particular example, we could go back weeks, months or even years to see when small changes to the data started to occur. We can then model those findings against data from other assets to get a better understanding of the asset lifecycle, or the substrate condition. This alerts the customer much earlier so that preventative measures can take place. Again, remote monitoring tells the track engineers whether the maintenance intervention was effective as the data would return to normal.
Data drives development
Another by-product of collecting and analysing data over a long time means that we can achieve far more with SWiX than it was initially developed for. More data has given us more insight into the potential applications for SWiX. Initially, we were calculating vertical displacement, but now that we can understand the rate of measurement of movement, we can better pinpoint and categorise potential faults that may be forming on the horizon. This gives customers the ability to forecast and schedule maintenance interventions in a more cost-effective and safer way.
Flexibility to work your way
Data from the SWiX remote monitoring devices is transmitted wirelessly to the cloud, AWS in our case, and machine learning algorithms are applied. Our data analytics team are very hands on and have a constant dialogue with customers. Not all customers use our interface. Some routes have built their own user interface to derive value from their data. Others have tweaked what we have developed.
It’s important that we don’t prevent customers from making a tool that is very specific to their needs, because at its heart is the crucial information – data they can trust from people who talk their language.
To find out what we can do for you, get in touch with us on +44 (0)1223 827160 or at info@smartcomptech.com
Comments