Research And Realization On Fault Feature Optimization For Critical Components Of Helicopter Transmission Chain

Helicopter usually works in severe environment, and its structure is very complicated. Transmission chain is the only channel for power transmission which can’t be redundantly designed, so the damage/fault in transmission system will threaten the safety of helicopter greatly. Therefore, it is particularly important to carry out condition monitoring and fault diagnosis for transmission chain. As is the basis of condition monitoring and fault diagnosis, it is of great significance to extract and optimize fault feature. The optimized feature vector is usually acquired from the multidimensional feature space though feature optimization, which is useful to improve the efficiency and accuracy of fault recognition. Therefore, it is particularly important to carry out the work of feature optimization.Funded by the project supported by National Natural Science Foundation of China “The nonlinear dynamics theory and method of enhancement detection of helicopter damage” and advance research project of ministerial level “The recognition of degradation state and the prediction of remaining life about the XX power transmission system”, fault feature optimization for critical components of helicopter transmission chain based on vibration analysis is studied in this thesis. The main contents include:(1) The failure modes and mechanisms of evolution of bearing and gear are researched, and the vibration features of bearing and gear are analyzed. 24 bearing features to be optimized are extracted from the two aspects of time domain and frequency domain. The feature extraction method of time synchronous average is researched as a key point, and there are 20 features of gear to extract.(2) The fault implanted experiment of bearing and gear is conducted. The bearing features of time domain and frequency domain are extracted, and the gear features of the time synchronous average signal are extracted. Analyze the activity of the features to recognize fault qualitatively.(3) The common dimensionless processing methods are researched, and based on the principle of keeping the variation coefficient invariable, the average dimensionless method is selected. Research the sensitivity index and stability index of features, and the traditional evaluation index is improved.(4) The sensitivity index and stability index are used to evaluate the feature set extracted to acquire the optimized feature vector. In order to validate the availability of feature optimization, found the fault diagnosis model of back-propagation neural network. The fault recognition accuracy of the optimized features and non-optimized features is compared and analyzed.(5) Design the fault diagnosis software for transmission chain. It consists of feature extraction software module, feature optimization software module, fault recognition software module of back-propagation neural network. The function of software is validated though the test.