After his year-end internship at CARL, we are very pleased that Kevin Ducharlet will be able to continue working with the DRI teams to begin his Ph.D. (CIFRE thesis). Industry 4.0 leads to asset management using sensor-generated data. For CARL Software, this boils down to the development of a software platform that builds digital twins of these assets. However, the outliers and abnormal values contained in the sensor data make all the necessary calculations inaccurate. This may be due to faulty sensor behavior, a problem during signal transmission, or undesired sensor positioning.
This thesis topic is entitled: “Certification and confidence in sensor data: detection of outliers and abnormal values in time series.” will be directed by Louise Travé-Massuyès and Marie-Véronique Le Lann, within the DISCO team, LAAS-CNRS. In this respect, the objective of this project is to certify the quality of sensor data:
- As a first step, we want to calculate a confidence score for each sensor at each time step.
- Then, we want to detect anomalies in the whole system using the confidence data.
The result is the development of an anomaly detection software suggesting a diagnostic aid to be inserted in the paired digital software platform. The integration of the solution into a software platform that will be distributed to customers imposes certain requirements:
It must be run in a fully automatic way, as it is used in the background in the complete platform, which means that we cannot ask an operator to manually set certain parameters because it would be time-consuming for an expert to give a label (normal or abnormal) to each instance of data, the method must be unsupervised.
We are delighted to have Kevin with us and wish him good luck for the next 3 years!
Biography :
Kevin Ducharlet has been interested in computer science and artificial intelligence applications for several years. He wanted to continue his studies in this field with a Ph.D. after graduating from the IMT Lille Douai engineering school. He started working on this project through an internship that allowed him to become familiar with his environment and the state of the art of anomaly detection.