On Several Problems In FPGA Implementation Of Empirical Mode Decomposition

In industrial production fields,the electromotor types' rotary machines are common facilities.These machines often generate vibrations during their operations.The vibration signal usually contains the operation conditions and fault information of the machine.Therefore,the accurate analysis and treatment of vibration signal is very important for the operation condition monitoring,quality assessment and fault diagnosis of rotary machines.However,the vibration signals of the equipment are usually non-linear and non-stationary,the traditional analysis method has great limitation in processing these signals.The Hilbert-Huang transform(HHT),proposed by Huang and others in 1998,is a new processing technology for non-stationary signals.The HHT method is widely used in the collection of marine data,seismic signal and the detection and diagnosis of equipment status.The FPGA technology has strong power of parallel processing and computing.With the internal high-speed parallel management and the rich logic resources,the FFT etc.advanced algorithm can be hardware achieved in a FPGA chip.Combining FPGA technologies with HHT algorithm will realize the efficient,high speed and real time analyzing and processing of vibration signal.In this dissertation,we focus on several problems in FPGA implementation of Empirical Mode Decomposition(EMD),so as to complete the Hilbert-Huang transform within a FPGA chip.The FPGA based vibration signal processing circuits are designed.The main contents of our wok are as follows:1.The vibration signal processing methods are explored,the advantages and disadvantages of these methods are analyzing.Then Hilbert-Huang transform is introduced as a new time-frequency analysis method.Hilbert-Huang transform shows the effectiveness and adaptability in analyzing and processing the nonlinear and non-stationary signals.It's a potential method for non-stationary signal processing.2.During the circuits design,the operation quantity in HHT algorithm implementation,the data storage and data transmission during the collection process are overall considered.Reasonable chip selection,allocation of pins,component selection and task allocation are also accomplished.The main circuits are given in this dissertation.3.Several problems in FPGA implementation the EMD are focused,including the quickly envelopes fitting and the end effect in HHT.The comparing to several end effect processing methods is carried out by simulation experiments.Basing on the analyzing,the sawtooth wave transform in envelope fitting,and improved slope-based method in solving the end effect problem are adopted.The EMD method is well achieved.The main implementation process of EMD method is presented in this dissertation.4.The simulation signal and industrial vibration signal are used to verify thedesign.The analyzing results are compared with the results of MATLAB simulation.The correctness and effectiveness of the method is proved.