PHYSICS, MACHINES AND DATA
At Tignis, we help you increase the reliability of your connected mechanical systems so you can be smarter and more innovative in the ways you do your job. Our “secret sauce” is using physics-based machine learning algorithms along with a database replica of your physical environment—a digital twin—to yield concise analysis, tailored to your specific environment’s needs.
In future blog posts, we’ll go into more detail about what all that means. For today, let’s start out by looking at a few key ways you can modernize your asset monitoring processes to make them more reliable and efficient. These three focus areas are a best practice for any asset manager.
1. Make the most of your existing sensors
The internet of things (IoT) promises ultimate connectivity across all things in every business environment, including the individual parts of the mechanical systems you manage. This advent of 24/7 monitoring capabilities encouraged facilities to equip mechanical systems with a broad range of sensors. As you know, using sensors can help extend your condition-based maintenance through early detection of events that might impact the reliability and efficiency of your systems.
That’s great—it’s certainly worth gathering relevant data about your systems—but, even putting aside that word “relevant” for a minute (because not all data is relevant, after all) installing sensors takes time and money. There’s not just the cost of the equipment itself, but also the infrastructure needed to make it useful, from wiring the electrical connections (which can be especially expensive in manufacturing and industrial environments) to the more high-tech components, such as the network and IoT software platform you use for collecting and analyzing the sensor data. There’s also the time it takes just to engage your IT team and then figure out where the sensors actually need to go. And don’t forget, fitting new sensors into any hard-to-reach places can pose life safety risks that take longer to mitigate and can run up your budget even further.
The good news? If you have some automation in your current system, you already have a head start. With the right software, you can leverage the data collection capabilities of your existing sensors and apply physics-based machine learning models using a digital twin to yield the results you need. In many environments, you can use this approach to detect and manage a rich set of maintenance conditions with a surprisingly low number of sensors.
By applying this technique and working with us to develop the right solution that meets their needs, our customers transform their condition monitoring and expand coverage while overcoming the typical limitations of cost, life safety, IT project churn, and uncertainty about new sensor placement. By calculating simple algorithms based on known physical principles, we help companies do away with extraneous sensor equipment in places where it’s expensive and problematic to add and maintain.
2. Find faulty sensors more easily
Once you’ve evaluated the sensors in your environment to see how you can do more with less, you need to make sure the sensors you have in place are continuing to add value over time. Assuming a given sensor is providing data that is useful to your systems analysis, you need to keep the sensor operational in ways that ensure that data is accurate.
For most maintenance teams responding to a preventive action request, this entails lots of hands-on physical testing or cross-validation with other sensors. The reality is that most of us don’t have time to do this work proactively, so we end up testing sensors only when we suspect a fault—and that’s usually when something much larger has already gone wrong.
Again, the solution comes down to having the right data and knowing what to do with it. Just as physics-based calculations can help you determine where sensors are unnecessary, similar calculations can tell you whether the data being reported by a sensor is suspected of being inaccurate. Sometimes this is just a warning, and can clue you in to the need for proactive maintenance; and sometimes, when the physics just don’t make sense, it’s a clear sign that a sensor has failed—or else there’s a serious problem with your systems.
As with #1 above, these calculations only work if you’ve scripted your machine learning algorithms to apply basic physical laws, and you have previously mapped your entire constellation of system variables into a central database (digital twin).
3. Weed out the false positives
Having lots of sensors producing data and lots of system rules processing the data leads to lots of alerts going off, calling your attention to supposedly urgent conditions. Some of these are real, but many are not. It’s no surprise to any seasoned asset manager that rules are fallible, and very often alerts come across whose business relevance is basically zero. At Tignis, we refer to these generally as false positives. We also just sometimes call them noise.
Here are the three most typical responses to a false positive alert:
- “The alert is based on faulty data, which I might or might not recognize as faulty in the moment.”
- “The alert is based on accurate data, but I’m pretty sure it’s reporting a condition that doesn’t exist.”
- “The data is accurate, and the condition exists—but, for my business purposes, I just don’t care.”
And actually, “I don’t care” is really the logical response to all of these alerts. You shouldn’t care, because you have more important things to focus on in your day. It’s like having someone run into your office every five minutes with a new, urgent message that actually has no pertinence to the business, and all you want to do is do your work.
You can say enough’s enough, and shut and lock the door so that person can’t run in any more. But then, what happens when they’re no longer crying wolf, and there actually is a serious problem?
Such is the danger of having too much “noise” in your day—it blurs the edges between the things you actually care about and the stuff you don’t. By modernizing your asset monitoring systems to report only relevant alerts, you can be more intentional and clear-headed about the alerts you respond to, and more impactful in your job.
