Research On Dynamic Monitoring And Fault Diagnosis Of High-speed Automatic Machine Guns Based On Movement Morphology Decomposition And Multivariate EMD

High-speed automatic machine guns have complex movement characteristic,due to the harsh working environments such as high temperatures,high pressure and high impact,their performance deteriorates gradually,which seriously threatens the reliability and availability of weapon systems.Different from the periodic vibration deriving from rotating machinery,the complex movement characteristic corresponds to a series of short-term transient shocks that can solely be found in high-speed automatic machine guns.Owing to the particularity of the weapon systems,traditional techniques for condition monitoring and fault diagnosis of rotating machinery are hardly suitable for high-speed automatic machine guns.In practical application,there is an urgent need to develop a condition monitoring and fault diagnosis method that matches the complex system characteristic of high-speed automatic machine guns.this thesis proposes movement morphology decomposition combined with multivariate analysis methods to solve the problems of monitoring,forecasting and fault diagnosis of high-speed automatic machine guns,the effectiveness of the research methods is verified experimentally by analyzing the real measured short-term transient data.The main contributions of this thesis are briefly introduced as follows:(1)To cope with the particularity of high-speed automaton systems,a movement morphology decomposition method matching the complex system characteristics of high-speed automata is proposed.Through conducting movement morphology decomposition and timing analysis,the corresponding relationship between the movement morphology and the short-term transient shock series is established to separate the shock series corresponding to the target component.(2)For weak fault information is difficult to acquire due to superposition effect of state information of various components and strong background noise,multiscale CVA is proposed in this thesis and applied for the monitoring of dynamic performance of high-speed automatons.Dynamic monitoring models are built on each scale by taking serial correlations into account and the defective impact characteristics of high-speed automatons are extracted effectively.(3)To improve the sensitivity of CVA to early defects,canonical variate dissimilarity analysis is proposed for detection of incipient faults of high-speed automatons.The dissimilarity index is obtained through calculating the departure of the past and future projected data,which can be used as a supplement to the original indicators,the system of dynamic performance monitoring of high-speed automatons is improved.(4)For the problem of uncertain input-output relationship of multivariable predictive models,a multivariable LSSVM based on minimum state space learning is proposed and used for the prediction of the dynamic characteristics of high-speed automata.This method transforms the original multivariate data into the state space and utilizes the function relationship between state space and statistical indicators to establishes a multivariable predictive model,which provides a theoretical basis for the selection of input and output of multivariate predictive models.(5)A mode selection method based on space statistic is proposed,which builds data space with the same scale of IMFs,and the total variation of the IMFs in the data space is then counted to transform the multivariate data into a single dimension,providing a theoretical basis for optimal model selection of multivariable EMD.(6)The effect of mode alignment characteristics of multivariate EMD on feature extraction was studied,and a fault diagnosis method based on multivariate EMD and permutation entropy is proposed for high-speed automatons.The sample variables belonging to the same state are collected for establishing a data matrix,multivariate EMD decomposition facilitates information synchronization and data fusion among the sample variables.The permutation entropy of all IMFs is calculated as a feature vector to identify the failure pattern of the high-speed automaton.(7)The difference and relevance of high-speed automatons and rotating machinery are discussed,the research work of this thesis was extended to condition monitoring and fault diagnosis of rotating machinery,and the adaptability of multivariate statistic and analysis methods to rotating machinery was validated.