A novel active sensing method for loosening detection of globe-cone joint considering assembly uncertainties

Globe-cone joint is widely used in satellite propulsion system and rocket engine. Loosening of globe-cone joints will trigger leakage failure, and even initiate catastrophic consequences. Effective loosening detection of globe-cone joints is a challenging issue due to assembly uncertainties. To address this issue, a novel active sensing method insensitive to assembly uncertainties is proposed. It contains three steps. In the first step, the concept of multiscale effort to compress (METC) is proposed and effective loosening features, i.e., the ETC values under different scales are constructed. Subsequently, the dimensions of these features are reduced by generalized discriminant analysis (GDA) method. Last, an ensemble learning model is built to accurately recognize different loosening states. Experimental results demonstrate that the detection accuracy is as high as 0.9429 using the proposed method. More experiments have been conducted to validate the superiorities of multiscale ETC and GDA for feature extraction and reduction. Furthermore, the advantages of the proposed method in effectively detecting loosening of globe-cone joints and have been verified by comparing with other common loosening detection methods.