Satellites are designed to operate a mission defined through representative cases. At design time, various simulation models (thermal, power, mechanical) are processed to define a satellite design that fulfils those mission scenarios. Satellites are then operated during several years and observed through various sensors (voltage, temperature, mechanical deformation): about 500 sensors sampled at 0.1Hz for more than 10 years. Simulation model from design are used to forecast system performance.
Updating those models is already performed with different techniques depending on the disciplines addressed and the nature of the model (empirical law, PDEs resolution …). It is a specific and complex task to be executed, and it rarely makes use of the whole data available, limiting itself to some representative use cases.
We aim at making systematic the process to correlate models and data. And making use of all available data. The intern will deal with the use case of power generation and thermal regulation embedded in the satellite. The intern will make use of thermal and electrical simulation models, and will have access to archive of temperature and electrical sensors during lifetime of a satellite.
This 6-month internship is to be filled from March 2020 (date subject to some flexibility).
Tasks & accountabilities
The intern will implement state of the art method to correlate model and data, keeping as overall a requirement genericity of methodology. The intern will adapt or innovate methods to correlate models or correct models using machine learning, meaning exploiting the data to improve the models. Those methods will be implemented in a big data IT environment, data being stored in Hadoop and calculation to be executed through Spark.
The intern will demonstrate and benchmark different methods to better fit model to data; he/she will define and measure performance of each and will issue recommendation for an operational implementation for satellite monitoring and control.
Internship will be executed within Satellite Data Science team in Data Engineering entity within Space System Engineering Division.
The intern is in his/her last year to validate a master degree, with specialization in applied mathematics. He/she has knowledge of Python, Hadoop and Spark environment. He/she demonstrates knowledge of machine learning libraries, but also mathematical knowledge of implemented methods. He/she has experience in code best practices and tools (testing, versioning). He/she has knowledge of design engineering (common sense of multiphysic environment). Internship will be executed within Satellite Data Science team in Data Engineering entity within Space System Engineering Division.