| In the field of mechanical fault diagnosis,the analysis of the mechanical equipment operating status through the characteristics of the frequency domain is a research hotspot in this field.With the development of modern measurement technology,it is very important that embedded spectrum computing equipment can obtain high-resolution spectrum information quickly and accurately.In spectrum calculation,the real-time calculation and spectrum resolution are contradictory.In order to solve the contradiction between the real-time calculation and the spectral resolution and improve the real-time calculation,this paper studies the more efficient spectrum calculation of long-sequence data in embedded devices.Related work includes the following:First,on the basis of studying the traditional FFT algorithm,the DFT algorithm is derived to obtain the Blocking FFT algorithm.The data sampled in chronological order are calculated independently by blocks,and the independent calculation results of each segment are superimposed to obtain spectrum result,which improves the real-time performance of the calculation.In order to solve the situation that the existing sliding DFT algorithm can only slide a few points,a multi-point sliding DFT algorithm that can slide a block of data is proposed.Newly sampled data,sliding out data,and original spectral values are used in this algorithm to calculate the spectral value after sliding.Secondly,on the basis of completing the algorithm derivation,the logical structure design and data transmission design of the Blocking FFT algorithm and the multi-point sliding DFT algorithm are completed in the embedded device ZYNQ-7000.The design of FPGA logic part mainly includes five modules: input buffer module,multiplication calculation module,FFT calculation module,logic control module and rotation factor generation module.The data communication transmission part design includes two parts from FPGA to ARM and from ARM to PC.It is transmitted to DDR3 through AXIS protocol inside ZYNQ,and it is transmitted from the Ethernet interface to the computer under the control of the ARM processor.Finally,the results of two algorithms for long-sequence spectrum in the embedded device ZYNQ-7000 are analyzed.The numerical accuracy,real-time calculation and hardware resource consumption of the Blocking FFT algorithm and multi-point sliding DFT algorithm are analyzed.The experimental results show that the Blocking FFT algorithm is better than the traditional FFT algorithm in terms of real-time performance,and the block FFT algorithm is similar to the FFT algorithm in numerical accuracy.Also,the Blocking FFT algorithm provides a stable and effective solution for the spectrum analysis of ultra-long sequence data.In addition,the multi-point sliding DFT algorithm still has better real-time performance when calculating the frequency spectrum when sliding block of points. |
