基于小波包分析和K-OPLS的集成方法在颗粒粒径分布检测中的应用  

Integrated method of wavelet packet analysis and K-OPLS and its application in measurement of particle size distribution

作　　者：陈敏[1] 贺益君[1] 王靖岱[1] 阳永荣[1] 

机构地区：[1]化学工程国家重点实验室(浙江大学),浙江大学化学工程与生物工程学系,浙江杭州310027

出　　处：《化工学报》2010年第6期1349-1356,共8页CIESC Journal

基　　金：国家自然科学基金项目(20776124, 20736011);中央高校基本科研业务费专项资金资助项目(2009QNA4028)~~

摘　　要：颗粒粒径分布的实时在线检测对于调控气固流化床中颗粒的流化特性具有重要意义。基于混料均匀设计法安排实验,以声发射(AE)技术为检测手段,结合小波包分析,提出采用K-OPLS方法构建颗粒粒径分布的声信号预测模型,定量描述小波包能量特征与颗粒粒径分布的非线性变化规律。实验结果显示,留一交叉验证法的均方根误差(RMSE)仅为0.063,表明基于K-OPLS的颗粒粒径分布声信号预测模型能准确测量气固流化床中颗粒的粒径分布,具有良好的工业应用前景。Real-time and on-line measurement of particle size distribution （PSD） in gas-solid fluidized bed is one of the important issues in multiphase flow processes.In this work,a mixture uniform design method was adopted for the experiment.Acoustic emission （AE） technique was applied to determine the particle size distribution.Wavelet packet analysis was introduced to characterize the energy spectrum feature in the AE signals.Furthermore,K-OPLS was proposed to construct the PSD prediction model based on AE signals,so that a quantitative non-linear relationship between the energy feature and PSD was established.The leave-one-out cross-validation error is 0.063,indicating that the model based on the K-OPLS method could be used to predict PSD with high precision and the AE method could be considered as an effective PSD measurement technique in industrial gas-solid fluidized beds.

关 键 词：气固流化床 颗粒粒径分布 声发射技术 小波包分析 K-OPLS 

分 类 号：TQ021.1[化学工程]