激光诱导击穿光谱结合GA-BP-ANN检测炉渣中Ca

Laser induced breakdown spectroscopy combined with GA-BP-ANN for detection of Ca in slag

以激光诱导击穿光谱技术为基础,通过击穿炉渣中等离子体来获取炉渣光谱图,将遗传算法与BP神经网络进行结合,通过遗传算法对神经网络的权值和阈值进行优化建立基于遗传神经网络模型,对炉渣元素光谱图中的Ca元素含量进行定量检测,测得5种Ca元素#1、#2、#3、#4、#5的质量分数为29.4%、40.37%、37.13%、43.88%、38.68%,并计算检验样本相对误差分别为4.7%、5.2%、5.8%、4.1%、3.3%,相对误差均在6%以下,检测精度明显优于BP-ANN方法和光谱分析中常用的自由定标法,表明基于遗传神经网络对炉渣进行定量分析具有更好的检测效果。

Based on the laser induced breakdown spectroscopy technique,the slag spectrum is obtained by breaking the plasma in the slag,combining the genetic algorithm with the BP neural network,and optimizing the weight and threshold of the neural network by genetic algorithm.The neural network model quantitatively detects the content of Ca in the slag elemental spectrogram.The mass fractions of the five Ca elements#1,#2,#3,#4,#5 are 29.4%,40.37%,and 37.13%.43.88%,38.68%,and the relative error of the calculated test samples are 4.7%,5.2%,5.8%,4.1%,3.3%,and the relative error is below 6%.The detection accuracy is significantly better than the BP-ANN method and spectral analysis.The commonly used free calibration method shows that the quantitative analysis of slag based on genetic neural network has better detection effect.