Using the Statistical Features of the Data to Detect Potential Failure of Unmanned Aerial Vehicles
| Main Authors: | Ahmad M. Alos, Zouhair Dahrouj |
|---|---|
| Format: | Article |
| Bahasa: | eng |
| Terbitan: |
, 2019
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/3557687 |
Daftar Isi:
- The Unmanned Aerial Vehicle is one of the most complex systems ever developed. Its complexity raises the chances of its failure. This paper focuses on predicting the potential failure of the UAV using the collected data from its previous missions. The collected data consist of discrete and continuous attributes. The proposed approach helps in determining the abnormal flights, and the contribution of the attributes in the potential faults. The values of the attributes are analyzed using some statistical parameters. The selected parameters are {Mean, Variance, Standard deviation, Kurtosis, Skewness, Minimum, and Maximum}. These parameters are used to build feature datasets to characterize the performed flights. Next, two algorithms can be used to extract the anomalies from the feature datasets. The used algorithms are the Principal Components Analysis-based anomaly detector and the One-Cluster KMeans. The conducted experiments showed that our approach is practical for detecting the faults and the contributed attributes using either discrete or continuous data.