Research On Anomaly Detection Method For Helicopter Rotor

Helicopter rotor as a lifting surface and controlling surface is vital to function realization and flight safety.It powers the helicopter forward and helps to complete the transition of various position.Because of complex structure and changing working condition,key parts of helicopter rotor system are vulnerable to failures.Once it appears breakdown,the consequences would lead to catastrophic accidents and major downtime.Carrying out online detection and failure diagnosis can ensure the safety of helicopter flight,and it brings enormous economic benefit and social benefit.This paper mainly solves the key problems in anomaly detection which are fault features extraction and anomaly detection for helicopter rotor system by using advanced signal processing techniques and pattern recognition method.Due to the complicated structure and working condition,the vibration data of rotor is non-stationary and non-linear.It is difficult to make an effective anomaly detection only through the analysis in time domain.This paper proposes a feature extraction method based on the singular value entropy of IMF with the help of Empirical Mode Decomposition.Empirical Mode Decomposition is a kind of time-frequency method,it can decompose the rotor vibration data into a series of IMFs order by frequency,then calculate the singular value entropy of IMF as failure feature.Compared to the time-domain statistical parameter features,the feature of singular value entropy of IMF has better detection results.The feature of singular value entropy of IMF has no consideration of the characteristics of the rotor itself,hence,the pertinence is not strong.Considering the characteristics of helicopter rotor itself,this paper proposes a feature extraction method based on Short Time Fourier transform(STFT).Short-time Fourier transform is applied on primitive vibration signal to obtain the joint distribution of time domain and frequency domain,then select amplitude spectrums of harmonic frequencies as rotor failure feature.The experiment results indicate that the anomaly detection based on the STFT feature achieves good performace and can directly reflect the state changes of helicopter.Hence,it can be used as the sensitive failure feature of helicopter rotor.Limited by own structural principle,data distribution and model parameters,for single intelligent detection algorithm,it is difficult to guarantee the accuracy and stability in face of complex application environments.In order to improve the robustness of the algorithm,this paper proposes a multi-model integrated anomaly detection algorithm.The multi-model integrated anomaly detection algorithm is built on three basis detector of different principle,k-means cluster detector,Gaussian model detector and LOF algorithm detector.The basis detector is promoted to strong detector using Boosting method,then combine these strong detectors to get the multi-model integrated anomaly detection algorithm.Compared to single detection algorithms,the multi-model integrated anomaly detection algorithm has better performance.