Thank you! (We’re funded)
It’s been nine months since our last newsletter. In the intervening months, we’ve spent our time wisely, focusing on developing our core product: a solution that predicts when and why electric motors fail.
Moving forward, we’ll work with an amazing group of clients on further improving our technology. Backed by ENGIE Technical Services and UNIIQ, who support our journey and share our vision that in maintenance, reliability is key, we feel that we are in an excellent position to improve operational reliability at a lower cost. Most important, however, is that our team is eager, capable and driven to build great products.
When we started this venture, we believed that the maintenance domain offered great opportunities for innovative initiatives. We observed how deploying AI on time-series data would enable maintenance professionals to focus on what matters: failing equipment. Inspired by Clayton Christensen’s theory of “jobs to be done,” we believe that maintenance professionals first and foremost must ensure that equipment runs. This, of course, is not an easy task—with technological advances improving quality, speed and a host of other desirable attributes, machines are becoming increasingly more difficult to service. What’s more, the 24-hour economy, LEAN programs and ever-increasing competition mean that there is less time to perform maintenance tasks. Shrinking budgets and shrinking staff add insult to injury. All this results in a world where fewer staff members have less time to maintain increasingly complex equipment. What’s needed is a solution that empowers maintenance professionals to spend time on failing equipment, not checking up on healthy machines.
There are many ways to go about this. Traditional condition monitoring techniques range from intuition to vibration monitoring and beyond. However, many of these technologies are labor intensive, expensive and do not exploit the capabilities that machine-learning technologies provide. The fact of the matter is that a “smart” solution relies on data, and computers are fundamentally better at processing and analyzing the torrent of data that modern machines generate.
It seems like data is the new gold. We believe there is great truth in this, although it must be noted that we have a certain interest in subscribing to this view… Regardless, data analysis enables fact-based decisions. Better analysis of data offers better insight, enabling better decisions. But not all data is created equal. And not all data is preserved. Companies must develop policies around the creation of data and its preservation, especially in the maintenance domain, in order to benefit from advances in machine learning technologies.
When it comes to condition monitoring, the best results are achieved when working with (sufficiently) high-frequency data, usually related to vibration, current, pressure or a combination of these. The challenge here is that this type of data is very often not available at all, yet it is required in order to facilitate rapid learning cycles and timely warnings for upcoming failure. As algorithms learn from examples, we’ve found that we must create those examples ourselves. What’s more, it’s important to scale. More examples of both healthy and unhealthy data patterns improve the reliability of our algorithms. More examples require many sensors to be deployed. And for that to happen, clients must accept large scale deployments of our sensor on their turf. We are truly grateful that ENGIE, Strukton Rail and Worksphere, the Kaak Group, Inteqnion, the City of Rotterdam, Sitech, Royal Wagenborg, Tata Steel and several others have offered us the opportunity to perfect our craft. We plan to repay their trust by offering meaningful insight into when and why equipment fails.
Over the past couple of months, World Class Maintenance has provided a fertile ground for innovation. We have benefited tremendously from the support and network of Professor Henk Akkermans and his team. WCM is driving innovation in the maintenance domain and offers innovators the opportunity to work with corporates in a collaborative spirit. With the explicit goal of achieving tangible results, many successful innovations have been road-tested in the WCM-supported field labs.
Railforum, NVDO, Dinalog, firms such as IIR, and initiatives such as Rotterdam’s City Flows and Rotterdam Mobility Lab, iTanks and the Port of Rotterdam’s projects contribute to improving operational reliability through innovation, collaborative projects and knowledge sharing. It has been said that the Netherlands holds a spot in the global top 3 of players in the Maintenance World Championship series. Building on the proverbial finger in the dike, the Dutch have a history of doing more with less—or creating efficiencies—in keeping machines running. It’s amazing to be part of such an incredible network.
Our friends at Black Box Engineering, MaxGrip, Van Bodegraven Elektromotoren, Bakker Sliedrecht and Repair, FACTA, ENGIE, UNIIQ, Geckotech, AWS, Grimpe Communicatie, Districon, Gordian, Compris Consulting, Julio Karto, Makana Eyre and many others have been instrumental in helping us get where we are now. While we are not yet where we want to be, we do feel that we can build on a solid foundation.
A special thanks to the Yes!Delft community! The team in Delft creates an environment where collaborative innovation is the norm. At the start of our great endeavor, Yes!Delft provided guidance, support and a place to share ideas, frustrations and big wins. The latests ranking puts them at the #4 spot of incubators in Europe. For us, they are #1. By far.
As data scientists, it is incredibly rewarding to work with heavy equipment and tangible assets. The opportunity to learn what is signal and what is noise—knowing that it’s vital to be right—is what motivates us. With the investment from UNIIQ and ENGIE Technical Services we are keen to explore how we can further improve our predictions about when and why equipment fails. We believe that in 5 to 10 years, 0% unplanned downtime will be the new norm. We want to contribute our part. And to get there, we’ll need to get to work. So: let’s get on with it!
Resources38 See all resources
SAM4 is a predictive maintenance system that helps maximize asset uptime by detecting developing faults up to five months ahead. But there’s another plus: that same data provides concrete insights that enable energy and carbon reductions.
- White papers
In this report, we explain exactly how SAM4 detects common faults in hot strip mills using actual results from anonymized SAM4 data, to help steel engineers evaluate SAM4’s value for their own operations.
- Technical documents
- White papers
Semiotic Labs, a scale-up company based in Leiden, Netherlands, and SMS group have signed an agreement under which the two companies will cooperate in the field of predictive maintenance.