| Today,helicopters have become more and more widely used in various fields.How to ensure the safe operation of helicopters has become a top priority.As the main moving parts of helicopters,rotor helicopters hubs are prone to failure because of their high strength,long time stress,and complex force conditions.The real-time monitoring of the state of the rotor helicopters hub and the prediction of faults are of great significance for ensuring the stable and reliable operation of the helicopter.In recent years,acoustic emission technology has been one of the hot spots in the detection of crack initiation and expansion of the motor components,and the key to crack detection is de-noising preprocessing and feature extraction and analysis of acoustic emission signals.In this paper,based on the research needs of the detection of the cracks in the centerpiece of the rotor helicopter hub,a parallel processing method for denoising pre-processing and feature extraction of acoustic emission signals based on CPU/GPU heterogeneous architecture is proposed.The paper first studies the traditional wavelet threshold denoising method and analyzes the problems that the wavelet basis function,the time-frequency domain resolution,and the number of decomposition layers and thresholds are not suitable when dealing with acoustic emission signals,A set of solutions for de-noising pre-processing of acoustic emission signals is proposed,including the use of adaptive methods to construct optimal matching wavelet basis functions,lifting the resolution of the wavelet analysis by redundant wavelet packet decomposition,and using the cross-correlation method to identify wavelets break down the number of layers and thresholds.The denoising effect of the method is verified by MATLAB simulation.Based on this,this paper discusses in detail the method of the impact separation,event counting and extraction of simplified waveform parameter extraction,energy spectrum coefficient and other characteristic parameters of acoustic emission signals.Considering the huge amount of data under multi-channel conditions,This paper proposes a parallel processing scheme based on CPU/GPU heterogeneous architecture.,described in detail in the CUDA and C + + environment,multi-threaded acoustic emission signal denoising pre-processing,collision division,simplified waveform parameter extraction and other characteristic parameters of acoustic emission signals calculation specific structure and process.Finally,the application of the parallel processing scheme are programmed to realize the uppercomputer software of the multi-channel acoustic emission acquisition system.The software can realize the initial analysis and processing of the acoustic emission signal.At the end of this paper,the method of verifying the effect of the adaptive acoustic emission signal denoising preprocessing algorithm and parallel processing scheme is introduced.The adaptive acoustic emission signal de-noising pre-processing algorithm proposed in this paper has better SNR than the traditional de-noising method,and the RMSE is also lower,which can better preserve the characteristics of acoustic emission signals.In addition,the SNR and RMSE change with noise increased is also the smallest,and the environment adaptability is stronger.The parallel processing methods have greatly improved the calculation speed as compared with the traditional serial processing methods.The speedup ratio between the parallel scheme and the serial scheme of the denoising preprocessing algorithm is 121.99,and the speedup ratio between the parallel scheme and the serial scheme of the feature extraction algorithm is also 58.83.It lays the foundation for designing a data processing computer software system for multi-channel acoustic emission detection system. |
PDF Full Text Request