Feature generation method for fault diagnosis of closed cable loop used in deployable space structures

In order to diagnose the fault of closed cable loop in deployable space structures, a method of feature generation from deployment angle is proposed. Existing reliability analysis of deployable space structure is usually based on estimated probability or expert knowledge, which are inaccurate and will attenuate the credibility of analysis result. In order to ground the knowledge of the reliability of state-of-the-art closed cable loop on engineering practices, we studied an approach that can identify the features of the data transmitted from spacecraft and automatically diagnose the faults in closed cable loop. The primary feature was identified as higher angle discrepancy in synchronized angles when faults in closed cable loop occur. To reduce the probability of erroneous diagnosis, the loss of velocity concordance in synchronized angles is used jointly with the primary feature. The mathematical expressions of the two features are specified as average difference of angular displacement and Gini concordance of angular velocity. The classifier of normal and faulty closed cable loop is generated by support vector machine, whose training data are produced via simulation or experiments. Case study of a three-panel solar array adopts the method to generate features from simulation result, which serves as the training data for support vector machine. The trained classifier is further applied in the diagnosis of angular signals from an experimental setup and the results validate the effectiveness and robustness of the proposed method.