Tierra S.p.A. and Politecnico di Torino presented the first promising results of a long series of joint activities with the Applied Research and Data Analytics teams.
On 17thJuly, Tierra’s Data Analytics team presented the results of the studies carried out by two students of the Politecnico di Torino, after the completion of an internship program on Machine Learning, in preparation for their master’s degree theses, as part of the ongoing collaboration.
Dena Markudova introduced the methodologies and logics behind her study on “Vehicle usage prediction towards smarter maintenance”. Dena worked on project Atene, which was carried out in collaboration with the Politecnico di Torino’s SmartData Research Center, coordinated by Prof. Marco Mellia.
“We have been collaborating with Tierra since a couple of years. We used Tierra’s data and provided our skills on algorithms and our team.”
Marco Mellia, Department of Electronics & Telecommunications SmartData@PoliTO coordinator
The goal of Dena’s study has been to achieve accurate maintenance scheduling through characterization of vehicle data and prediction of future utilization. The test method used for the research is the walk-forward validation, thanks to which a new model is created every day with updated data.
“We want to expand simple reports and integrate and evolve them with predictive abilities. All this requires a lot of work and must be done in collaboration with the customer. However, the tools that will enable the analyses have to be developed beforehand, and to do so we collaborated with the Politecnico di Torino for the statistical and mathematical instruments necessary to develop the data analysis algorithms we need.”
Riccardo Loti, Head of Research at Tierra S.p.A.
Lorenzo Perini explained his thesis on “Predictive Maintenance for off-road vehicles based on Hidden Markov Models and Auto encoders for trend Anomaly Detection”. Lorenzo performed predictive-oriented dataset construction to identify SPN (data values from the vehicle) that could be critical, with a focus on DM1 (diagnostic messages). Lorenzo developed his model in cooperation with DISMA, the Department of Mathematical Sciences of Politecnico di Torino, coordinated by Prof. Francesco Vaccarino and Prof. Luca Cagliero, who explained:
“DM1 are very heterogeneous and can vary from one manufacturer to another. The methodology used has to be adapted to different contexts, through more data and information that would improve the performance, also when dealing with different cases.”
The multiyear, ongoing collaboration between Tierra S.p.A. and the Politecnico di Torino is being successful under many aspects, one of which concerns personally Dena and Lorenzo who both graduated gaining first class honors and the publication of an article.
Dena and Lorenzo could count on the collaboration and real data made available by important customers of Tierra, among which Storti S.p.A.
Tierra’s CEO Paolo Traso thanked Dena, Lorenzo and Tierra’s Data Analytics and Applied Research departments for their work and their important contribution.
“We have been talking about data analytics for a very long time. The collaboration with the Politecnico di Torino has proved to be a winning strategy, following the results of these studies and of these analyses. The choice to move our headquarters closer to the Politecnico was not random. There is still a lot of work to do because this is an extremely complex but extremely powerful subject.”
Paolo Traso, CEO of Tierra S.p.A.
This is just the first of many projects carried out with the Politecnico di Torino.
Tierra will have other interesting news to share in the next future.