摘要
柠檬酸发酵液化清液生产过程监控对整个柠檬酸生产至关重要,近红外光谱能够通过不同波长下分子的振动多方面地反映过程的运行状况,包含了大量的过程信息。但是,现有方法往往是建立近红外光谱与总糖总氮等质量变量的回归模型,通过判断质量变量是否超过阈值实现对过程运行状态的事后报警,忽略了近红外光谱内部的很多有用信息,监控效果较差。该文充分利用和分析近红外光谱的统计特性,提出一种基于近红外光谱生产过程的统计监控方法,首先建立近红外光谱和总糖总氮的概率偏最小二乘模型(probability partial least squares,PPLS),然后基于模型对不同的信息设计监控指标,能够充分利用近红外不同波长上的信息,实现故障的事前预警。结果表明,采用该方法得到漏报率为9.68%,错报率为25.81%,可以有效地对柠檬酸发酵液化清液生产过程进行监控。
The monitoring of liquefied process of clear solution for citric acid fermentation is crucial to the citric acid production.Near-infrared spectroscopy can reflect the operation status through the vibration of molecules at different wavelengths,and contains a lot of process information.However,the existing methods often develop regression models between near-infrared spectroscopy(NIR)and quality variables such as total sugar total nitrogen,then the process is monitored afterwards by judging whether the quality variable exceeds its threshold.These methods often ignore useful information in NIR,resulting in poor monitoring performance.In this paper,the statistical property of near-infrared spectroscopy was fully utilized and analyzed,and a statistical monitoring method was proposed based on NIR.Firstly,a probability partial least squares model was developed for the estimation of total sugar total nitrogen by NIR.Then,based on this model,monitoring indicators were designed to achieve advance warning of faults by taking full use of different information at different wavelengths of the near infrared.The results showed that this proposed method could effectively monitor the liquefied process of clear solution for citric acid fermentation,the false negative rate was 9.68%and the false positive rate was 25.81%.
作者
郝超
赵忠盖
刘飞
HAO Chao;ZHAO Zhonggai;LIU Fei(Key Laboratory of Advanced Process Control for Light Industry,Ministry of Education,Jiangnan University,Wuxi 214122,China)
出处
《食品与发酵工业》
CAS
CSCD
北大核心
2020年第20期214-220,共7页
Food and Fermentation Industries
基金
国家自然科学基金(61833007
61573169)。
关键词
近红外光谱
概率偏最小二乘
柠檬酸
过程监控
波长选择
near infrared spectroscopy
probability partial least squares
citric acid
process monitoring
wavelength selection
