传统的废旧塑料处理主要采用焚烧掩埋的方式, 不仅污染环境, 还造成了资源浪费, 废旧塑料回收对于推行循环经济、 促进可持续发展具有重要意义。 传统的塑料分类装置存在精度较低, 成本较高, 分类结果易受样品颜色影响及对操作人员身体健康造成威胁的缺点。 激光诱导击穿光谱技术(LIBS)具有多元素同时分析、 无需样品预处理, 检测速度快, 对样品损害小, 不受塑料颜色影响等优点。 将其与基于化学计量学的样品分类方法相结合, 应用到塑料样品检测领域, 有利于提升塑料分类精度。 但是, 现有分类方法需要更改参数多、 所构建模型普适性差。 通过自主搭建的LIBS实验平台, 研究激光器能量、 延时时间、 积分时间和光谱信号接收角度对光谱信号强度的影响, 获取最佳的实验条件。 并通过该平台分析11种塑料的2200个样品点, 选用偏最小二乘算法（PLS）对谱图数据进行建模分析, 研究最优的训练集和验证集比例及最佳的主成分数的选取方法, 实现样品类别与样品数据的最大相关性, 达到克服背景干扰、 提高分类精度的目的。 实验结果表明, 不替换干扰谱图时, 验证集样品的分类精度为99.80%, 测试集样品的分类精度为99.09%; 替换掉干扰谱图后, 验证集和测试集的样品分类精度均达到100%, 且分类时间减少。 由此可见, 激光诱导击穿光谱与偏最小二乘方法结合可以成功的用于塑料样品分类。

The traditional ways of waste plastics processing mainly use the burning landfill, which lead to environmental pollution and the waste of resources. Waste plastic recycling is very important on the circulation economy and the sustainable development.The traditional instruments have some shortcomings in plastic classification, such as lower precision, higher cost, the influence of the sample color and a serious threat to operating personnel’s health. Laser induced breakdown spectroscopy has many advantages, such as simultaneous multielement detection of elements, free from sample preparation, rapid and real-time analysis, slight damages to sample and no impact on the sample color. The method of Chemometrics combined with LIBS technique is applied to the plastic, which improves the accuracy of plastic classification. But at present,the classification has many problems, such as more parameters and the poor universality. Using on a self built LIBS instrument, we can study the laser energy, delay time, integration time and the angle of the optical fiber, which can achieve a better experiment condition. With the experimental platform, we analyze the 2 200 sample points and choose the partial least squares to analyze the spectral data. In order to achieve the correlation between the sample label and the data, we discuss the better ratio of the training set and validation set. The experimental results show that replacing the interference spectra, classification accuracy of all 11 plastic is increased to 100%, while the validation set’s accuracy is only 99.8% and the test set is 99.09% without replacing the interference spectra . It can be seen that the laser induced breakdown spectroscopy combined with partial least squares method can be successfully used for the plastic sample classification.