Research On Key Technologies Of Energy Dispersive X-Ray Fluorescence Spectrometer

Energy dispersive X-ray spectrometry is a multi-element analyzing method, which is capable :o measure the element types and element contents in the sample. However, the composition of energy dispersive X-ray spectrum is very complex, which has wide frequency band and overlapped beaks, making the analysis of the spectrum quite challenging. Thus, it is of great significance to pay nuch effort on the research of denoising, subtracting the background and decomposing the verlapped peaks.Noise in the spectrum would lead the spectrum-analyzing program to miss weak peaks, generate pseudo-peaks, and amplify the net-peak area error. A denoising method based on stationary wavelet transform is proposed in this dissertation. An optimized threshold is generated by the combination of cross-validation and stationary wavelet transform, which does not depend on the estimation of the noise variance. With this optimized threshold, the soft threshold function could brocess the stationary wavelet coefficients. After reconstructing the signal with the threshold coefficients, the spectrum has been denoised. Signal to noise ratio and root mean square error of the ienoised spectrum have been improved considerably.The continuous background in the spectrum has bad effect of the calculating of the peak bosition and peak width. A method based on dual-tree wavelet transform is presented to subtract the background. The theory about dual-tree wavelet transform is studied thorough. In the presented nethod, the spectrum is decomposed by the dual-tree wavelet transform and the coefficients at a high level is extracted to construct the estimated background. After subtracting the estimated background, a clean spectrum is obtained. This method would not introduce distortions to the spectrum.Most background-subtracting methods depend on the difference of the frequency bands of the characteristic peaks and the background. A method based on iterative wavelet transform is proposed to subtract the background. The spectrum is decomposed iteratively, thus reducing the peak contributions in the approximations,and the estimated backgrounds approach the real background iteratively. The optimal wavelet base is selected by wavelet entropy. Wavelet approximation energy is defined which would be the criterion to determine the end of the iteration.Overlapped peaks are difficult to decompose. Coefficients of Gaussian wavelet are used to simulate the deviation of the spectrum. The wavelets corresponding to the first and fourth deviations of Gaussian are used to decompose the signal. The coefficients would be used to simulate the first and fourth deviation of the spectrum. The peak-widths, peak number and peak positions are extracted with this method. This method has adjustable resolution and is not sensitive to noise.Wavelets coefficients are also used to form a compensating spectrum, which would be added to the spectrum. After this process, the spectrum has less overlapping peaks and maintains peaks area the same as well. The extracted information would be used as initial value for the fitting of the spectrum, which is used to calculate the net-peak area of each characteristic peak.An energy dispersive X-ray spectrometer is designed. The lower computer is composed with X-ray tube, Si-PIN detector and FPGA which would work as the multi-channel analyzer. The upper computer is designed by LabWindows/CVI, whose duty is to control the lower computer and analyze the data. The spectrometer is tested with<Calibration Specification for Energy Dispersive X-ray Fluorescence Spectrometers> and all the design requirements are met.Work of this dissertation has solved existing keys problems of energy dispersive X-ray spectrum, providing valuable inspiration for energy dispersive X-ray spectrum analysis and other spectra analysis.