It’s all up from here! Machine Learning & Predictive Maintenance in Elevator Service
ThyssenKrupp’s MAX is on the rise
What do Inception, Elf, You’ve Got Mail, and Willy Wonka & the Chocolate Factory have in common? An iconic elevator scene, of course.
Elevator woes proliferate movies and television because they provide that perfect ‘wrench in the plan’ for the plot – from getting trapped inside an elevator or waiting an eternity for one to arrive, to the toil of the last-resort hike up the stairs (usually at the worst possible moment!).
In addition to the drama they stir up, we see these elevator debacles so often because they are common and relatable. In fact, experts estimate: “With more than 12 million elevators transporting over a billion people each day, elevator maintenance issues cause 190 million hours of downtime annually” [1].
One of the drivers of elevator downtime is the elevator service provider’s struggle to find the right cadence for regular maintenance. ICT Monitor Worldwide notes “…if you take systems offline to do maintenance too often, you reduce your production yields” [2]. The impact of reduced yield is extremely costly, as the nature of elevator service work necessitates expensive, highly skilled labor [3].
On the other hand, ICT poses, “If you wait until machinery breaks to fix it, you have downtime and unhappy customers” [4]. This downtime for an elevator breakdown can be lengthy and unpredictable – ranging from hours to days to weeks. The breakdown also causes a slew of labor inefficiencies. For example:
- Reduced yield: The technician may abandon an in-progress effort on another machine to address the breakdown, thus wasting the expended labor.
- Transportation time: The technician must move to the breakdown from his/her current location.
- Inventory uncertainty: The technician may arrive at the breakdown to find that he/she does not have the proper part, necessitating another visit.
- Safety risks: The pressure to move quickly to a breakdown can make technician missteps more likely.
To combat this issue, ThyssenKrupp – a top elevator provider – partnered with Microsoft and its Azure machine learning platform to develop the world’s first elevator predictive maintenance system, which they were rolling out by 2016 [5]. In his National Post article, Josh McConnell describes the functionality of the resulting product, “MAX”:
“Data from Max-connected machines such as door movements, trips, powerups, car calls and error codes are collected from around the world and then sent to the cloud to be analyzed by algorithms and machine learning. From there, operational patterns are picked up and the various components’ remaining lifetimes are calculated so technicians can replace parts before a breakdown occurs. Elevators can then be scheduled for maintenance during off-peak hours to [minimize] disruption and, therefore, increase efficiency.” [6]
ThyssenKrupp faced a number of challenges in developing and rolling out the product, from both a technical perspective (e.g. lack of internal analytics experience, software obstacles related to sensors in elevators) and operational perspective (e.g. tweaking technician training and dispatch methods, updating customer contracts in light of new value proposition) [7].
In spite of these challenges, MAX delivered excellent results. According to ThyssenKrupp, “it is capable of cutting elevator downtime in half” [8]. Additionally, MAX increased technician’s ability to fix the machine on the first visit, reduced inventory costs, and improved customer satisfaction [9]. Customers benefit from not only the reduced downtime, but also the increased transparency achieved through data on their elevator’s performance and parts via MAX’s mobile app [10].
Given the success of the project, ThyssenKrupp plans to further develop MAX’s capabilities. In particular, they hope to “create separate predictive models for doors, motors and other elevator components to increase analytical accuracy,” expanding upon general ‘call’ and door movements that are the focus today [11].
Looking forward, ThyssenKrupp should also focus on incorporating MAX algorithms in parallel as they develop new product lines such as “Multi,” an elevator whose “cabins can go sideways and aren’t limited to one per shaft” [12]. Further, they should consider the extent to which they rely on third-party vendors, diversifying their partners and developing the in-house capabilities to analyze and monitor (if not replace) vendors. As of 2017, they added a partnership with Cyient, predictive analytics expert – a step in this direction [13]. Finally, there is also an opportunity to quantify and market MAX’s ‘green’ and safety impact.
Beyond ThyssenKrupp, MAX raises questions for the elevator service industry at-large:
- How will the development of predictive technologies impact the competitive landscape? Will only players with the greatest scale be positioned to make the sizable capital investments required for this technology?
- What level of in-house capabilities (vs 3rd party support) is most efficient to support machine learning? How will that evolve?
- After optimizing frequency of parts replacement, will the next step be further improving the quality of the parts themselves or even 3D-printing replacements on the spot?
Time will tell, but in the meantime, we can enjoy more elevator drama from Tinsel town. (Word Count: 788)
________________________
Sources (MLA):
Photo courtesy of: Bourbeau, Jeremy Dean. “Businessman In Glass Elevator Going Up.” Burst.Shopify.com, 0AD, burst.shopify.com/photos/businessman-in-glass-elevator-going-up?q=elevator.
[1] Dr. Martin Hoelz – CIO and Head, Group Processes & Information Technology, ThyssenKrupp AG. Boardroom Insiders, Inc, San Francisco, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1932882578?accountid=11311.
[2] “9 IT Projects Primed for Machine Learning.” ICT Monitor Worldwide, Oct 13, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1950517394?accountid=11311.
[3] Dr. Martin Hoelz – CIO and Head, Group Processes & Information Technology, ThyssenKrupp AG. Boardroom Insiders, Inc, San Francisco, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1932882578?accountid=11311.
[4] “9 IT Projects Primed for Machine Learning.” ICT Monitor Worldwide, Oct 13, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1950517394?accountid=11311.
[5] “Thyssenkrupp Rolls Out MAX in Germany: World’s First Predictive Elevator Maintenance Service.” M2 Presswire, Apr 26, 2016. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1784105787?accountid=11311.
[6] McConnell, Josh. “The Future of Elevators; Moving People Sideways and Data to the Cloud.” National Post, Aug 08, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1927105305?accountid=11311.
[7] Dr. Martin Hoelz – CIO and Head, Group Processes & Information Technology, ThyssenKrupp AG. Boardroom Insiders, Inc, San Francisco, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1932882578?accountid=11311.
[8] Ibid.
[9] “9 IT Projects Primed for Machine Learning.” ICT Monitor Worldwide, Oct 13, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1950517394?accountid=11311.
[10] “Thyssenkrupp Rolls Out MAX in Germany: World’s First Predictive Elevator Maintenance Service.” M2 Presswire, Apr 26, 2016. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1784105787?accountid=11311.
[11] Dr. Martin Hoelz – CIO and Head, Group Processes & Information Technology, ThyssenKrupp AG. Boardroom Insiders, Inc, San Francisco, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1932882578?accountid=11311.
[12] McConnell, Josh. “The Future of Elevators; Moving People Sideways and Data to the Cloud.” National Post, Aug 08, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1927105305?accountid=11311.
[13] Dr. Martin Hoelz – CIO and Head, Group Processes & Information Technology, ThyssenKrupp AG. Boardroom Insiders, Inc, San Francisco, 2017. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/1932882578?accountid=11311.
In addition to the labor inefficiencies that this application of machine learning resolves, MAX’s ability to preempt elevator breakdowns transforms safety management from being reactive to proactive. You raise an interesting point about the potential to quantify safety impact — This technology has the ability to reduce several types of cost: 1) The monetary cost of repairing a broken elevator which is much greater than replacing an aging part, 2) The costs associated with human injury from a broken elevator accident, and 3) The reputational costs a company suffers if people are hurt in their elevators. I see many parallels between this application of machine learning and the use of AI to detect medical issues before they become dangerous. Both applications allows us to take a more proactive approach to diagnosis and reduce costs.
ThyssenKrupp’s exploration into Machine Learning is particularly interesting. Having worked in manufacturing for a number of years, I know the tremendous value that effective maintenance scheduling can provide an operation. By leveraging machine learning to better predict when elevators need maintenance, I am confident that ThyssenKrupp will unlock productivity & significantly reduce downtime (costs).
Instead of worrying about competitive responses to machine learning exploration, I would challenge ThyssenKrupp to look outside of this specific industry for future opportunities. I envision numerous applications for this technology/capability in industries that are capital-intensive. This could be a great opportunity for ThyssenKrupp to partner.
P.S. You should add a ‘Tower of Terror’ to your list of movies with iconic elevator scenes. Hopefully ThyssenKrupp can continue their commitment to this Machine Learning technology, as I would hate to be caught in my own rendition of ‘Tower of Terror’…
Like Jerzey#305, I agree that ThyssenKrupp should begin thinking about applications of this technology beyond the elevator industry. The statistic you presented, “MAX delivered excellent results. According to ThyssenKrupp, “it is capable of cutting elevator downtime in half”” is pretty impressive and could have a significant impact in other sectors. My first thought is this could be applied to machinery in production lines, increasing efficiency of processes and eliminating variability that could occur due to breakdowns. I found an interview to the CEO of a machine learning company based in Israel discussing this subject (ml in factory machine breakdowns). The piece makes a good point in noting the applicability of this technology in settings that, like with elevators, there is not a spare machine that can be used while a repair is taking place. According to the interviewee, adoption in the sector is growing fast: “the level or the rate of adaptation or adoption, is increasing and we’re now at a point that if industry before was like something to speak about, now we see how the market is doing it. We have request for a query, request for information, and for quotes from people who are actively looking for this and I expect it will increase in the next 18, 36 months. It will definitely increase.”[1]
[1] https://www.techrepublic.com/article/how-machine-learning-can-prevent-factory-machine-breakdowns-and-create-more-jobs/