基于红外光谱聚类分析的纳滤膜污染动态发展行为研究

污水再生利用是解决水资源短缺问题的有效对策。纳滤技术由于能够生产高质量的再生水，成为污水深度处理、再生利用的有效方法之一。然而，在纳滤过程中存在复杂的、动态的膜污染现象，会导致产水通量、产水质量下降等问题。研究膜污染动态发展的行为，对于膜污染的分阶段针对性控制具有重要意义。有机物是污染层动态发展过程的重要指示性成分，红外光谱是表征污染层发展过程中表面有机物官能团变化情况的重要手段。但由于红外光谱中峰的数量多，系列样品之间峰强度的差别较小（尤其是当膜污染过程中的采样间隔较小时），利用直观观察不易甄别不同样品间的谱图差异及其变化趋势，在此水平上难以对膜污染阶段进行准确识别、对各阶段特征进行有说服力的分类概括。为探索膜污染的动态发展过程，本研究将傅里叶变换红外光谱与统计学聚类分析相结合，对膜污染过程中不同时间点的膜样本进行红外光谱分析，再对红外光谱数据进行一系列预处理和系统聚类分析，从而客观解读膜污染动态发展过程中系列样品红外光谱分阶段变化规律。考虑到类别间距离度量方法、红外吸收峰强度标准化、峰之间自相关性、峰与样本之间交互作用等因素的影响，研究采用对应分析对红外数据进行预处理，提取各样本在主要维度上的得分，随后基于标准化欧式距离对各样本进行聚类。在为期一个月的城市污水深度处理纳滤试验过程中，由于污染物在膜表面累积，纳滤膜发生了较为严重的污染。通过对13个不同时间点的膜样本进行红外光谱聚类发现，膜污染可清晰划分为如下阶段：空白膜、阶段Ⅰ（3 h～8 d）、阶段Ⅱ（10～15 d）和阶段Ⅲ（20～30 d）。采用红外聚类，得到膜表面X射线光电子能谱（XPS）和三磷酸腺苷（ATP）含量分析等方法的交互验证。结果表明，随着膜污染的发展，膜表面有机物成分与共存微生物量发生协同变化，各阶段大致特征为：阶段Ⅰ各类有机污染物初步覆盖，微生物开始富集；阶段Ⅱ多糖类污染物比例上升，微生物的富集趋于稳定；阶段Ⅲ整体污染趋于成熟，有机污染物氢键特征更加明显。该研究通过对红外数据进行聚类分析，能够灵敏地探测各红外图谱之间的差别，有助于对红外光谱规律的深度挖掘，为膜污染阶段的识别和划分提供了一种客观、自动、可量化的辅助性方法，并且有助于归纳出不同阶段的污染层特征，可作为膜污染时序特征的侦查手段。此外，除了膜污染的研究，在材料、吸附等领域，只要有一系列变化的红外光谱，均可尝试采用红外光谱聚类分析方法，获取基于红外特征的定类信息或分阶段规律。

Tracking the Dynamic Evolution of NF Membrane Fouling Through Clustering Analysis Based on ATR-FTIR Spectra

Sewage reclamation is one of the effective countermeasures for solving the water shortage problem. Nanofiltration (NF) process is an effective method for reclaiming secondary effluent since it can provide high-qualitied water. However, during the nanofiltration process, complicated and dynamic membrane fouling occurred, which can cause the decrease in flux and effluent quality. Tracking the dynamic evolution of NF membrane fouling in water treatment is important for controlling the membrane fouling depending on different fouling stages. Organic matters are important indicative contaminant for the dynamic evolution of fouling layer. Attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) is one of the most significant methods to characterize the change of functional groups in fouling layer. However, the peaks of ATR-FTIR are complex and the variations between different samples are tiny, especially when the fouled membrane samples are similar in time. It is difficult to directly discriminate the variation and trend of different ATR-FTIR spectra and cannot be the convictive evidence for fouling stages recognition. To investigate the dynamic evolution of membrane fouling, this study categorized the NF membranes samples obtained at different fouling time through combining the ATR-FTIR spectra and clustering analysis. Considering the influence of distance measurement method between categories, normalization of ATR-FTIR peak absorbance, correlation between peaks, and interaction between peaks and samples, this study utilized the correspondence analysis as the pretreatment of the ATR-FTIR spectra to obtain the scores of different membrane samples along main dimensions and then clustered the samples based on normalized Euclidian distance. During the 1-month NF experiment using sewage secondary effluent, because of the deposition of foulants, membrane fouling occurred and 13 fouled membranes were obtained at different time. Based on the hierarchical clustering of ATR-FTIR spectra, the fouling process can be clearly divided into the stages of: virgin membrane, stage Ⅰ(3 h～8 d), stage Ⅱ (10～15 d), and stage Ⅲ (20～30 d). The results of the clustering analysis was further interpreted by X-ray photoelectron spectroscopy (XPS) and adenosine triphosphate (ATP) content on the membrane surface. It was shown that with the evolution of membrane fouling, the organic composition and coexisting microorganism amount on the membrane surface changed concordantly. The characteristics of different stages may be interpreted as: in stage Ⅰ, the membrane was initially covered by organic foulants, and microorganism began to gather; in stage Ⅱ, the proportion of polysaccharide-like substance increased and the gathering of microorganism became stable; in stage Ⅲ, the membrane fouling became mature and the hydrogen bond characteristics of organic foulants became more evident. In this study, the variation of different ATR-FTIR spectra was detected sensitively through clustering analysis. The study provides an objective, automatic and measurable auxiliary method for recognizing and characterizing membrane fouling stages. Besides, it is meaningful for investigating the ATR-FTIR spectra of a series of samples in not only membrane fouling research but also other fields such as materials science and adsorption research.