Navigate the Predictive Maintenance, IoT and AI at InnoTrans 2022 — a practical checklist
by Christian Sprauer, CEO and founder of Railnova July 27, 2022
Predictive maintenance and IoT have been all the rage at InnoTrans for the last couple of years.
As InnoTrans 2022 is around the corner, I thought to update a checklist I created to help you navigate these subjects.
We understand. As a railway operator and transit authority, the temptation is irresistible. You will be able to save millions on maintenance and downtime costs. You can be alerted before assets break down. You can stress less and watch your new real-time dashboards give you all the relevant information.
Predictive maintenance and IoT at InnoTrans: reality and fiction
However, one of my biggest challenges in early talks with clients is to explain the difference between reality and fiction. Decision-makers are flooded with “IoT and AI can do everything” messages. It can become hard to see what can be done and what is, simply put, fiction.
Often faced with various questions, I decided to create a checklist for InnoTrans. I hope this helps you cut to the chase. I also hope to simplify your decision-making by outlining what questions are relevant in the railway industry when it comes to digitalisation.
1. Is it a live software demo or an InnoTrans-specific PowerPoint presentation?
This should be the first red flag. Usually, when faced with a PowerPoint presentation, this means that the software is not production ready.
This is what we mean when we say fiction. The information on that PowerPoint may not be wrong. It is simply not yet a reality. The question is whether you want to wait for the product to be ready. Time is money, as they say.
To put off a live showing can also mean avoiding in-depth questions. Whether it is because there are still bugs in the software or simply because of an unknown answer.
Be sure to ask for a live demo to see if the solution fits your data and use case out of the box!
2. Is the offer a free POC (Proof of Concept) or a clear step forward?
Free might sound better, but this isn’t always the case. I have nothing against free things, generally speaking. But sometimes, free can hold you back from true potential.
Proof of Concept will clarify if the solution works in your context. The willingness to pay for this is proof that your organisation is ready for change. A paid-for PoC commits your internal resources and obligates the vendor to deliver.
Additionally, paid PoCs help you clarify objectives. This minimizes risk when the decision time comes.
If you are getting a PoC, check how accurately anomalies are detected. Ask how easily you can configure new alerts and data sources. Don’t forget to see how easily data and alerts can be consumed by users and other systems as well.
Now, if you are looking to take it a step further, look for a Proof of Value (POV). POV takes a deeper dive into the value of a solution or product for your organisation. So you can justify adopting it and better measure success after the fact.
3. Are requirements to your technical specification being added in the InnoTrans booth?
You may be facing a software consultancy business that generates revenue from clients’ “wants” and “change requests” with little out-of-the-box functionality. Be wary if third parties add technical specifications unrelated to your business and your needs.
Check exactly how your use case will be achieved. Is it via custom developments or product configuration?
Be sure to ask for the typical revenue structure of a contract and check for licence revenues. A mix of 70% software licences (justified by a strong product and years invested in R&D) and 30% custom development (no more) is healthy and very likely to provide value.
4. Are the AI techniques proven?
Scanned invoice document recognition works thanks to advances in deep learning accuracy. However, automatic driving is still a far reality. When visiting all these fancy stands at InnoTrans, it can become tricky to know what works and what doesn’t for our industry.
When it comes to predictive maintenance in railways, you should be able to understand how the technology works and how accurate it is. Things should be explained in terms even people far removed from technology can understand.
Sometimes a boring physical model such as « the temperature is now 2x standard deviation greater than the moving average », validated on your fleet historical dataset, filtered for all corner cases and deployed on the real-time fleet data stream, can do miracles.
Plenty of resources are available online to understand the basics of AI. I’d recommend documenting yourself beforehand. You can start by watching this general video on AI, an oldie but goodie:
5. Does anyone check that the data covers your use case?
Some vendors will unscrupulously promise you that machine learning will infer abnormal events from unrelated data. They will likely mention the general context, the weather, and utilisation patterns.
This can be true in some specific cases. But special attention needs to be paid to the data sources. Especially concerning our physical world, complex failure modes, and normal wear and tear.
6. Is the promise “collect all data now, find patterns later”?
Although this might sound pragmatic at first, it might make your AI and IoT solution selection trickier.
Our experience at Railnova shows that you need to iterate several times between the data collection and pattern recognition techniques to deliver proper alert triggers. This can be done by adapting sampling rates, unit conversion, new signals from the MVB bus, and new sensors.
Collect data now for your use case and fine-tune it from the start. When new data is needed for new use cases, remotely update the software of your train-mounted device to obtain such new data.
Here is an example of how Railnova’s Railfleet does this:
7. Is there a clear safety and IT security approach for remote device software updates?
Retrofitting trains with hardware devices that support remote software updates can be challenging. Both from railway safety and from an IT security perspective. However, frequent remote software updates will be necessary.
If the solution is SIL4 and integrated into the train’s signalling system, such software updates will be impossible. Look for a solution that offers safe, ISA-approved, passive reading of train bus coupled with a clear IT security policy. Make sure the policy lists which staff members can access the device management platform to update the device software remotely.