种群算法在能量色散X荧光重叠谱拟合中的应用

在能量色散X荧光光谱分析中，常用的闪烁探测器如NaⅠ(Tl)探测器的能量分辨率都不高，均在8%左右。能量分辨率低下往往对谱数据分析带来较大的难题，特别是在高本底低计数的情况下剥离仪器谱重叠峰会受到很大限制，越是重叠严重的峰越是无法剥离，进而无法分辨峰值和峰面积，更无法进一步对元素进行定性定量分析。为此，结合遗传算法和免疫算法的优势建立新的种群算法应用在重叠谱分析上，该算法以欧式距离为进化的判断依据，以最大相对相似误差值为迭代准则进行迭代。利用高斯函数模拟不同重叠程度的仪器谱图，将种群算法应用在重叠峰分离和全谱模拟中，峰道址偏差在±3道以内，峰面积偏差不超过5%，证明该方法在能量色散X荧光重叠谱分析中有较好的效果。

Application of the Racial Algorithm in Energy Dispersive X-Ray Fluorescence Overlapped Spectrum Analysis

In the energy dispersive X-ray fluorescence spectrum analysis, scintillation detector such as NaI (Tl) detector usually has a low energy resolution at around 8%. The low energy resolution causes problems in spectral data analysis especially in the high background and low counts condition, it is very limited to strip the overlapped spectrum, and the more overlapping the peaks are, the more difficult to peel the peaks, and the qualitative and quantitative analysis cant be carried out because we cant recognize the peak address and peak area. Based on genetic algorithm and immune algorithm, we build a new racial algorithm which uses the Euclidean distance as the judgment of evolution, the maximum relative error as the iterative criterion to be put into overlapped spectrum analysis, then we use the Gaussian function to simulate different overlapping degrees of the spectrum, and the racial algorithm is used in overlapped peak separation and full spectrum simulation, the peak address deviation is in ± 3 channels, the peak area deviation is no more than 5%, and it is proven that this method has a good effect in energy dispersive X-ray fluorescence overlapped spectrum analysis.