A robot suddenly failed in production. Could this failure have been predicted or even prevented? 20 students from the Albstadt-Sigmaringen University tried to find an answer to this question in the week-long Data Science Summer School 2018.
Together with partner DATATRONIQ, Leadec provided them with all the data required for an in-depth analysis. This included 36 different values such as temperature, torque and motor currents that are recorded every second. This resulted in 580,000 data sets from one week.
The analysis by the participants from the Data Science program clearly showed that the robot crash could have been foreseen. Values 10 or 20 times higher could already be measured hours before the crash. In the forecast model developed by the students, the anomalies can be identified early.
Science and practice
“A constant real-time analysis and the processing of data into useful information to gain predictive insights are required these days. The Summer School gives our students the opportunity to work on real cases and to develop new solutions in close exchanges with industry partners such as Leadec”, explains Prof. Stefan Ruf from the Albstadt-Sigmaringen University.
The project is not just an exciting experience for the Summer School participants. Leadec also benefits from the collaboration. “It is important that we can offer our customers clear added value with our services” said Michael Wojtas, Senior Project Manager at Leadec in the Product Excellence Cluster for Maintain. “The students’ analysis confirmed our own findings and we will incorporate this into further developments.
Leadec’s robotics center
The Leadec team and their partner DATATRONIQ cooperate in the field of robotics with Leadec’s Education Center in Chemnitz. There, failures and other atypical motion sequences will be simulated to train the algorithm of the robots for such events in future. It is anticipated these models will be scaled to robots and similar systems.
Alexander Döbelin, Head of the Global Product Excellence Cluster for Maintain & Support Services at Leadec, explains: “At Leadec, we collect and process digital data from our customers’ production processes and create digital business models. This means the processes are not only paperless but are also more efficient and transparent. Untapped potential can therefore be identified more quickly.”
Predictive analytics in real time
Why only intervene when it’s already too late? Ideally, to save customers time and money, it shouldn’t get to that point. Predictive maintenance can increase availability and reliability in production and ensure optimum use is made of resources for maintenance.