Projecte llegit
Títol: Anomaly detection of satellite systems via Machine Learning
Director/a: OLMOS BONAFÉ, JUAN JOSÉ
Departament: TSC
Títol: Anomaly detection of satellite systems via Machine Learning
Data inici oferta: 10-12-2019 Data finalització oferta: 10-07-2020
Estudis d'assignació del projecte:
- GR ENG SIST AEROESP
Tipus: Individual | |
Lloc de realització: Fora UPC | |
Supervisor/a extern: Miguel Ángel Vázquez | |
Institució/Empresa: CTTC | |
Titulació del Director/a: Enginyer Telecomunicació | |
Paraules clau: | |
Anomaly detection, Satellite systems, Telemetry, Machine Learning, Deep Learning, CNN, LSTM | |
Descripció del contingut i pla d'activitats: | |
Satellites are continuously transmitting control data to the
gateway. This data corresponds to specific payload parameters such as orbit deviations, radio frequency signal power values, temperature, etc. The most widely extended approach for automatically detecting anomalous behavior in space operations is the use of out-of- limits (OOL) alarms. The OOL approach consists of defining an upper and lower threshold so that when a measurement goes above the upper limit or below the lower one, an alarm is triggered. Then engineers will inspect the parameter that is out of limits and determine whether it is an anomaly or not and decide which action to take. While out-of-limits approaches are useful and they successfully trigger alarms when parameter readings go out the defined thresholds, they suffer from some limitations. First, some behaviors are anomalous even if they are within the defined limits. It can happen that the parameter values reach the upper limit but do not hit it. Since engineers do not know when this will happen they have to monitor key telemetry parameters closely even if in most cases everything would be nominal. Paradoxically, sometimes the anomalous behavior is more in limits than the nominal one. The idea of this TFG is to collaborate in the design and development of a machine learning tool able to preemptively detect telemetry data anomalies beyond to the current out-of- limit approaches. Different anomaly events will be available for study, coming from real data operators. Eventually, a real proof of concept will be done considering a real time connection with the satellite teleport. In addition to this, the analysis of traffic data will be performed and beam congestion techniques will be studied. |
|
Overview (resum en anglès): | |
This project presents the development of a software that is able to detect punctual anomalies in satellite telemetry. The presented tool is a model-based method that has a modular architecture and consist in the prediction module and detection module.
The prediction module is the responsible of forecasting the telemetry given by the satellites and it is based on a deep learning techniques. Different deep learning approaches have been considered; namely, Convolutional Neural Networks and Long Short Term Memory Modules. The detection module is in charge of labeling and infer if the data instance has abnormal behaviour or not using a dynamic threshold computed from the mean and standard deviation of a set of the prediction errors. |