A 7 step approach for using smart phones to provide a smart remote factory tour
Before we can offer recommendations on condition monitoring, it’s always useful to visit a prospect’s site where we can take a close look at the machines in question and interview the people that know them best. However, the restrictions we face in light of the ‘new normal’ mean that, in most cases, we’re not currently able to carry out such a visit.
But there are alternatives. When a physical visit isn’t possible, smartphones and video-conferencing technology, accompanied by certain basic information, can allow us to tour a prospect’s site remotely and learn everything we need to make the right recommendations for its machines.
Here’s our 7-step guide:
1. Injecting knowledge
Our approach to condition monitoring is to understand the machines as best we can so we can effectively inject our engineering knowledge into their machine learning. Rather than following the classical data science model, in which data is thrown at algorithms in the hope that they learn about a machine, we try and give those algorithms a head start by making sure that condition monitoring knowledge is applied in the right manner from the start.
We avoid the need for algorithms to learn about a machine’s on/off states each time it switches off, for example, as this is a “known”. After all, a human looking at a machine or graph will know intuitively just to look at the “on” state. The difference with machines is that they lack context. It’s up to us, therefore, to provide that context, and ensure that machine learning algorithms aren’t spending time concerned with things we already know. Instead, they can look for the things we don’t know, and the things we want to find when looking for issues at scale.
To make this happen, we begin by looking at their machines and components. Ideally, we’d visit the prospect’s site to see their machines, read their data, and speak to their operators “in the flesh”. In these unprecedented times, of course, we need to do this remotely, and this requires - at least - basic contextual information.
2. Keep it simple
Video is the near-ideal means of studying a client’s machines, as it allows us to take a 360-degree view. And in most cases, recording that video on a smartphone will do the job. Factory floors can often be noisy environments, with typically poor wi-fi access. Customers with the best intentions might employ live video or AR solutions to offer a remote tour of their machines only for the connection to continuously drop out, ultimately making it unusable.
By sharing a recording - hosted on the prospect’s servers or simply emailed – the viewer can pause or scroll back to a point of interest if needed. What’s more, sharing a video recording means presenters can take their time, going into detail without the pressure and stress of giving a live commentary.
In a perfect world, the video footage would run concurrently with data captured at the same time. But we know what a complex and time-sensitive operation this can be. So, for our purposes, a screenshot of high-frequency data, appropriate to the process we’re looking at, generally allows us to see how that data relates to specific actions.
3. Know your stuff
The ideal person to lead us on a tour of a machine is the person that knows that machine best. Someone like a maintainer or maintenance expert, who interacts with the machine all the time, is often better qualified than a data expert to talk us through what the machine’s doing at any given moment, commenting on what the inputs and outputs are, or the points at which problems are often found.
4. Never too much
We ask our clients to collect as much footage as possible - we’d always rather have too much than not enough. Lingering over a particular section to show it running is recommended, especially if it’s performing particular movements.
If a machine is particularly troublesome, filming it on a “down” setting during a maintenance shift, will still allow us to start to form an understanding, even if it’s not actually in operation. And if a machine’s concealed in an enclosure, it’s far more useful to us to see what’s moving inside that box rather than just the enclosure itself.
5. Changing states
Reliable condition monitoring requires observing a machine in known states. Although the same component may repeatedly carry out the same action, factors operating speed and material can often change. It’s important, therefore, that we’re able to see the different things that go into a machine, and the different things that come out, allowing us to see the details during each of that machine’s states. It’s important to get visibility and understanding of these changes.
6. Respecting privacy
There are almost always security issues to consider when touring a client’s site, especially when it involves proprietary equipment. As a result, we wouldn’t be able to photograph such specialist machines during a physical visit or discuss them with anyone outside of Senseye.
In the current situation, these issues can be overcome by asking the clients themselves to take the video footage themselves and host on their own secure servers. By not sharing it, or allowing us to record anything, the client maintains full control over the video, thereby protecting their IP.
7. Context is king
Machine learning is increasingly capable of some seriously heavy lifting. But, by injecting additional knowledge, we can make it work harder, saving it from telling us what we already know. Site tours are essential to gathering that knowledge. Current circumstances make it hard for us to carry out those tours in person but, armed with a smartphone, good condition monitoring data, and an expert commentator, we’re able to work remotely and effectively with clients.
Interested in learning more about Senseye PdM? We've just launched ROI Lock® - guaranteeing financial returns on your predictive maintenance project with Senseye. Find out more and arrange your digital factory tour below!