为了快速定量测量柴油污染土壤中的柴油含量,扩大近红外光谱技术在污染土壤监测中的应用,制备了含柴油的高岭土样本集F1和含柴油的实际土壤样本集F2,对近红外光谱检测数据建立了BP神经网络、随机森林和支持向量机回归3种预测模型。结果表明,预测模型对样本集F1的预测精度良好,但无法准确预测样本集F2中的柴油含量。通过加标和重复加标处理,改善了实际土壤中柴油含量的预测效果。与直接建模相比,由柴油高岭土样本加入少量实际土壤样本后建立的近红外光谱模型,能更准确地预测柴油含量。

In order to quickly and quantitatively measure the diesel content in diesel-contaminated soil and expand the application of near-infrared spectroscopy technology in monitoring the contaminated soil,a diesel-containing kaolin sample set F1 and a diesel-containing actual soil sample set F2 are prepared,and their near-infrared spectra are detected,and three prediction models including BP neural network,random forest and support vector machine regression are established according to the detection data.The results show that the prediction models have good prediction accuracy for sample set F1,but cannot accurately predict the diesel content in sample set F2.Then,the prediction of diesel content in actual soil is improved through spiking and repeated spiking.The near-infrared spectral model established from diesel kaolin samples is predicted after adding a small amount of actual soil samples.Compared with direct modeling,it can predict diesel content more accurately.

叶嘉富(1998-),男,硕士生