Analysis Of Abnormal Root Cause Of Telemetry Data For Spacecraft

In the period of rapid development of the Internet,all walks of life rely on software to realize data operation,which cannot be separated from the work of operation and maintenance personnel.However,with the development of high and new technology,the increasingly huge software and hardware systems,big data and immeasurable changes all reflect that human decision making cannot cope with the major challenges in the current operation and maintenance.We hope to gradually reduce the proportion of human decision in operation and maintenance,so intelligent operation and maintenance came into being.Intelligent operation and maintenance are widely used in network services,but not in heavy machinery industries such as spacecraft.Moreover,the telemetry data sent back by the spacecraft has the characteristics of large data volume and complex data type,so it is necessary to apply the intelligent operation and maintenance method to ensure the normal operation of the spacecraft.In order to solve the problem of low operation efficiency and low accuracy when the existing algorithm to anomaly detection and root cause analysis,this paper takes the anomaly location under multi-dimensional telemetry data as the key research content,and conducts an in-depth study on the multi-dimensional time series anomaly detection and root cause analysis algorithm respectively.The main research contents and innovations are as follows:(1)In this paper,LSTM(Long short-term Memory network)is selected as the basic algorithm of deep learning method,and on this basis,a weighted multidimensional LSTM anomaly detection algorithm is proposed.Firstly,the multidimensional LSTM model is trained to generate prediction sequences.Secondly,the euclidean distance between the actual value and the predicted value is weighted and amplified.Finally,the threshold window is set to judge whether the spacecraft is abnormal.(2)According to the linear relationship between the telemetry data when abnormal events occur,the correlation coefficient matrix is selected for root cause analysis in this paper.The absolute correlation coefficient matrix was obtained from the normal data and the abnormal data respectively,and the correlation coefficient thermal diagram was observed to analyze the root cause set of the location of the relevant parameters with obvious variation of the correlation coefficient.In this paper,two years of spacecraft telemetry data provided by a space agency are used to carry out experimental analysis on the root cause analysis algorithm for anomaly detection proposed in this paper.In the aspect of abnormal detection,the F1 score rate of the algorithm in this paper both exceed 0.9,which proves the feasibility of the abnormal detection algorithm proposed in this paper in the actual operation and maintenance scenarios.In terms of root cause analysis,compared with existing root cause analysis algorithms,the algorithm in this paper can effectively locate the root cause of abnormal events,and has made great progress in both effectiveness and robustness.