Study On Signal Denoising Method Of Internals Based On Time-Frequency Technology

On-line assessment and in-service maintenance of safety-related systems in nuclear power plants are basic measures to ensure the security of nuclear power plants.The internals vibration monitoring of the pressurized water reactor is used for early monitoring of the internals degeneration and failure,with the aim of long-term performance trend analysis of primary loop equipments and reactor internals.The current monitoring systems mainly use Fourier transform as the signal preprocessing method.However,the non-stationary characteristics of internals vibration signals and the low energy of early abnormal vibration make it difficult to identify the local features of the signals.In this paper,two time-frequency analysis methods including wavelet packet transform and empirical mode decomposition(EMD)were researched for the non-stationary characteristics of vibration monitoring signals.The denoising pre-processing of internals vibration signals was reached to enhance the condition monitoring and early fault diagnosis of internals.Firstly,the simulation vibration signals of reactor internals were calculated.Based on the research of the main structure of the pressurized water reactor internals,the finite element model of the upper internals was established,including the pressure vessel,the core barrel and the control rod guiding cylinder.The dry and wet modal analysis of the core barrel was carried out to research the natural frequency characteristic of the barrel under vacuum and water environment.At the same time,the natural vibration modes of beam type and the shell type of the barrel obtained by modal analysis were in accord with the actual engineering structure.For the established internals model,the calculation of the fluid between the pressure vessel and the barrel was performed to obtain the fluid pressure distribution,and the instantaneous pressure data of the main excitation position of the upper reactor internals was received.The transient dynamics was analysis to investigate the flow-induced vibration of the upper internals,and finally the acceleration vibration signals of the measuring point at the top of the pressure vessel were obtained.Secondly,the improved wavelet packet transform method was applied to preprocess the flow-induced vibration signals of the internals.The pivotal technologies affecting the denoising effect of wavelet packet threshold were improved:applying the parameter based on energy-entropy to select the optimal wavelet packet basis function,employing Gray coding ordering method to restore the correct order wavelet packet decomposition coefficients,separating the wavelet packet decomposition coefficient into multiple frequency bands,in order to perform multi-bands threshold selection method based on the energy comparison combined with an improved threshold function.The results demonstrated that the improved algorithm enormously improved the signal-to-noise ratio and effectively identified the useful information of the original signal.Finally,based on the EMD algorithm and the advantages of wavelet threshold denoising,EMD-wavelet threshold method was carried out for the preprocessing of flow-induced vibration signals in reactor internals.In the EMD algorithm,end effect is satisfactorily suppressed by the mirror extending method,making the decomposition more accurate and thorough.Subsequently,the intrinsic mode functions(IMFs)were filtered by the wavelet threshold method.It was demonstrated by the simulation test that the adaptation of EMD-wavelet threshold method not only allowed most of the noise to be filtered out,but also the noise and useful information was effectively identified,consequently the frequency characteristics of the structural vibration signals were successfully identified.