近红外光谱与多元统计方法用于生产过程实时分析

Near Infrared Spectroscopy and Multivariate Statistical Process Analysis for Real-Time Monitoring of Production Process

近红外漫反射光谱快速、无损且适用于实际复杂样品分析等特点被广泛应用于在位和在线分析。基于近红外光谱漫反射技术和多元统计过程分析方法建立了一种适用于生产过程分析与在线实时监测新方法。该方法实时采集生产过程中物料的近红外光谱,并利用已建立的模型计算统计量Hotelling T2对生产过程进行实时评价。模型的建立采用了主成分分析,统计量根据实时光谱在主成分模型上投影进行计算。实际应用表明,利用该方法可以根据统计量实时监测生产过程中物料的变动情况,并根据不同批次统计量的进一步统计实现批次间的相对稳定性分析。因此,该方法可作为生产过程质量控制的良好手段。

Near infrared diffusive reflectance spectroscopy has been applied in on-site or on-line analysis due to its characteristics of fastness, non-destruction and the feasibility for real complex sample analysis. The present work reported a real-time monito- ring method for industrial production by using near infrared spectroscopic technique and multivariate statistical process analysis. In the method, the real-time near infrared spectra of the materials are collected on the production line, and then the evaluation ofthe production process can be achieved by a statistic Hotelling T^2 calculated with the established model. In this work, principal component analysis （PCA） is adopted for building the model, and the statistic is calculated by projecting the real-time spectra onto the PCA model. With an application of the method in a practical production, it was demonstrated that a real-time evaluation of the variations in the production can be realized by investigating the changes in the statistic, and the comparison of the products in different batches can be achieved by further statistics of the statistic. Therefore, the proposed method may provide a practical way for quality insurance of production processes.