机器学习辅助光谱分析技术在环境微/纳塑料研究中的应用

Application of Machine Learning Assisted Spectral Analysis in Environmental Micro/Nano Plastics Analysis

微/纳塑料在环境中广泛存在,威胁着全球生态系统的平衡和稳定,因此有必要确定和评估微/纳塑料的环境赋存特征、环境行为和效应以及生态毒性效应。然而微/纳塑料具有的低浓度、小尺寸及容易吸附其他物质等特点,为其在复杂环境基质中的分析带来了巨大挑战。大多数分析方法在提供环境样本中微/纳塑料的定性定量信息方面存在成本高、准确性差、时间效率低等问题,而具有无损、高效、操作方便等优点的光谱分析技术可以弥补这些方法的不足。但采集的微/纳塑料光谱信号可能会受到环境样本中复杂成分的背景噪声干扰,亟需采用智能化手段提高分析的准确性和效率。机器学习具有强大的数据处理和自动化分析能力,准确性高、应用性广,适用于复杂光谱数据的分类和解析,将机器学习与光谱分析技术相结合有望成为微/纳塑料分析的可靠方法。首先对常用的机器学习辅助光谱分析技术进行综述,然后系统性地讨论了机器学习辅助光谱分析技术在微/纳塑料的环境赋存特征、环境行为和效应、生态毒理效应等研究中的应用,最后对该技术的发展前景进行了总结和展望。机器学习辅助光谱分析技术有望为环境中微/纳塑料的生态环境健康风险评估和污染防治提供重要数据支持。

The widespread presence of micro/nano plastics in the environment is threatening the balance and stability of global ecosystems,so it is necessary to determine and evaluate the environmental occurrence characteristics,environmental behavior,and ecotoxicology of micro/nano plastics.However,the characteristics of micro/nano plastics,such as low concentration,small size and easy adsorption of other substances,have brought great challenges to their analysis in complex environmental matrices.Most analytical methods have some problems such as high cost,poor accuracy and low time efficiency in providing qualitative and quantitative information of micro/nano plastics in environmental samples.Spectral analysis techniques with the advantages of non-destructive,high efficiency and convenient operation can make up for the shortcomings of these methods.However,in the process of spectral signal acquisition of micro/nano plastics,the complex components in environmental samples cause a lot of background noise interference,so it is urgent to combine intelligent means to improve the accuracy and efficiency of analysis.Machine learning has powerful ability of data processing and automatic analysis,high accuracy and wide application,and is very suitable for the classification and analysis of complex spectral data.Combining machine learning with spectral analysis techniques is expected to be a reliable method for micro/nano plastic analysis.The commonly used machine learn-assisted spectral analysis technology is first reviewed,and then the application of machine learning-assisted spectral analysis technology in the fields of environmental occurrence characteristics,environmental behavior and effects,and ecotoxicology of micro/nano plastics is systematically discusses.Finally,the development of this field are prospected.Machine learning-assisted spectral analysis techniques is expected to provide important data support for ecological health risk assessment and pollution prevention of micro/nano plastics in the environment.