A data-driven FCE method for UAV condition risk assessment based on feature engineering and variable weight coefficients

Evaluating the risk effectively is critical for the security and reliability of unmanned aerial vehicles (UAVs). With the improvement of related technologies, more and more condition monitoring (CM) parameters are collected from UAVs, which contains considerable information related to the condition risk. For the powerful capability to analyze these massive CM data, a data-driven fuzzy comprehensive evaluation method is proposed in this paper, which employs the feature engineering and the variable weight coefficients to achieve the accurate and timely condition risk assessment for UAVs. Given the CM data, the feature engineering is utilized to adaptively represent its historical normal status and provide the quantitative risk indications accurately reflecting its real-time risk. According to the real-time quantitative risk indications, the variable weight coefficients is utilized to dynamically adjust the initial weights of evaluating indices, which allows us to timely capture the slight condition risk of UAVs under the early abnormal status. At last, the risk membership vector of UAVs is obtained through the comprehensive evaluation to support the related decision-making. A case study using the real CM data of a UAV shows that the evaluation results provided by our proposed method are reasonable, comprehensive and interpretable.