Predictive maintenance: how far can we anticipate breakdowns?

Extending the lifespan of infrastructures is one of the pillars for moving towards cities and territories that use resources more sparingly. This effort requires, in particular, the development of new maintenance strategies, capable of better optimising equipment replacement and preventing failures. Pierre Naccache, President and Founder of Asystom, and Nicolas Hélas-Othenin, Operations Manager and Digital Transformation Lead at LISEA, meet to discuss the challenges of this emerging sector.

Founded in 2016, Asystom is a startup specialising in the development of solutions combining sensors and embedded intelligence to prevent industrial equipment failures. It is one of the ten companies selected to be part of Leonard’s CATALYST acceleration programme in 2022.

Subsidiaries of VINCI Railways (VINCI Concessions), LISEA and MESEA are respectively manager and maintainer of the first high-speed rail concession in France: the LGV Sud-Europe Atlantique (LGV SEA) linking Tours and Bordeaux. They are selected partners in our IA course to work on the issue of extending the life of railways. 

In this interview, Pierre Naccache, President and Founder of Asystom, and Nicolas Hélas-Othenin, Head of Operations and Digital Transformation at LISEA through the SEACloud project, meet to discuss the operational aspects related to the emerging sector of predictive maintenance, a crucial know-how that unites the two projects. 

How did your respective projects come about?

Pierre Naccache: Asystom was born as a continuation of our business of selling high-tech industrial equipment worldwide. When we received a call for support, we were systematically confronted with two problems: our customers were not experts in the equipment, which is quite normal, and the language barrier made things very complex. However, a precise perception of how a machine works is essential in order to be able to provide the best possible support to the customer. It requires having the right information, in sufficient quantity and continuously. And what could be more informative than listening (sound) and touching (vibrations)? This is how we developed our solution.

Nicolas Hélas-Othenin: After six years of construction and delivery ahead of schedule by the VINCI Group, the Sud Europe Atlantique high-speed line (LGV SEA) was commercially inaugurated in July 2017. For the first time, a private partner was confronted with the management of a high-speed rail infrastructure, and moreover for a period of fifty years. This is a real challenge but also an opportunity!

Thanks to VINCI Concessions’ experience in operating major mobility infrastructures, our foundations were solid. We therefore undertook work to equip ourselves with precise and effective tools for monitoring traffic, understanding incidents, measuring line performance and providing perfect knowledge of the condition of our facilities. This is essential in a context where the operation and maintenance of the line is carried out by light but robust structures.

It was in the search for this gain in efficiency and the desire to capitalise on the quantity of existing and continuously produced data that we moved towards automation (initially in the form of an Excel macro!), business intelligence and then AI. Since going live in 2017, we have demonstrated our ability, as a new player, to come up with new ideas. This great innovation dynamic has also enabled us to open one of VINCI Concessions’ centres of excellence for innovation, on the MESEA maintenance site in Villognon.

What does predictive maintenance mean to you?

Nicolas Hélas-Othenin: First of all, it means avoiding playing the fireman by intervening after the fact to put out fires (corrective maintenance), which is far too risky and also less efficient economically. Secondly, it means that preventive maintenance is sub-optimal. In order to avoid failure with the desired caution, preventive maintenance of even sound equipment is preferred, which again is not financially neutral. This caution is all the more important in the railway sector, where safety is the primary concern. Finally, predictive maintenance means analysing weak signals to understand the underlying causes of failures, then monitoring them in real time to predict the date of the next failure and plan maintenance accordingly.

Pierre Naccache: I can add that to do predictive maintenance, you need to have access to at least two elements. Firstly, detailed and sensitive information (precision sensors) to detect changes in operation as early as possible. Secondly, algorithmic processing to simulate the analysis process of an expert, via acoustic and vibratory analyses.

What are the main challenges with regard to the data you use?

Pierre Naccache: Our main objective is to be as flexible as possible to monitor any type of machine, whatever its use. To do this, the first challenge is to have independent access to relevant and quality data. This is why we have developed an intelligent and highly autonomous multi-sensor beacon.
The second challenge is to provide results that everyone can understand and use, regardless of their knowledge, function and expertise.

Nicolas Hélas-Othenin: I agree with Pierre. The first challenge is to obtain quality input data. This seems obvious but it often requires a lot of work, the added value of which is nil for the outside eye… Without this “Garbage In – Garbage Out”, as the Anglo-Saxons say, all analyses are good to go.
The second challenge is to desilter the data to make it accessible and usable by all; and also to ensure its dissemination in order to facilitate the analyses of business experts.
The third challenge is to govern the data in order to maintain or even increase its quality over time. This implies raising awareness at all levels of the importance of data, so that everyone takes ownership of it.

Does this imply changes for the teams on the ground?

Pierre Naccache: First of all, to believe that AI, which is generally the focus of fears in the public debate, will replace the maintenance teams is purely unrealistic even in the very long term. Let’s remain a bit pragmatic, the teams in the field need help and are very much in demand of diagnostic tools allowing them to respond to the requirement of constant improvement of the industrial process. The digitalisation of maintenance and – as far as Asystom is concerned – remote monitoring by artificial intelligence is a tool like any other, which makes it easier, smarter and faster to do things. The teams in the field have no doubt about it and are rapidly adopting it.

Nicolas Hélas-Othenin: Yes, it is the identification of pain points in the field and pockets of added value that is the shared genesis of our use cases. The fielword experts are MESEA’s central resource. Our first objective is therefore to relieve them of time-consuming tasks with little added value. They are then our ambassadors and our best asset in this change management.

Finally, can you imagine diversifying?

Pierre Naccache: The question of diversification would imply that the market is mature or restricted. But this is not the case at all, quite the contrary. Intelligent equipment monitoring is only just beginning and the technologies will continue to evolve. Predictive maintenance is still in its infancy. It is important to understand what is at stake for plants implementing these technologies. Firstly, these investments must bring immediate results and a return on investment. Secondly, the technologies must be able to support continuous improvement for increasingly predictive maintenance. And finally, this development must create effective corrective maintenance by anticipating repairs and scheduling them as early as possible.

Nicolas Hélas-Othenin: We are progressing gradually. Before diversifying, we would like to consolidate what we already have: test our predictive maintenance solutions over a longer period of time, increase our skills in data engineering to be more robust and more autonomous in the management of our data, take the time to document (which is not easy in such a changing environment), train our employees in the tools so that they can appropriate them, support them so that they can derive all the added value from them, and, in return, listen to them so that we can always better meet their needs.

At the same time, we are continuing our developments to extend predictive track maintenance to turnouts or other railway infrastructure sub-systems, such as signalling or telecoms… The universe of possibilities is already very vast at LISEA and MESEA. We are also open to collective projects, which would enable us to broaden our data base (geographically and temporally) and our expertise.

Contacts :

Nicolas Hélas-Othenin (LISEA) : nicolas.helas-othenin@lisea.fr

Pierre Naccache (Asystom) : p.naccache@asystom.com

Share this article on