论文部分内容阅读
针对间歇过程的在线故障诊断需要预测过程变量的未知输出问题,提出了一种数据展开和故障分类器数据选择相结合的方法。首先,对包含批次信息的三维数据进行数据展开,对间歇过程的多阶段分别建立PCA模型并进行过程的故障监测;然后,选取故障发生时刻之后的部分长度采样时刻的数据进行故障的特征提取,离线建立LSSVM的故障分类器模型;最后,通过故障分类器进行在线故障诊断,实现故障分类并确定发生了某类故障。该方法提高了间歇过程在线故障诊断的实时性和准确性,通过青霉素发酵仿真过程的应用,进一步验证所提方法的可行性和有效性。
Aiming at the problem that the on-line fault diagnosis of intermittent process needs to predict the unknown output of the process variable, a method combining the data unwinding and the data selection of the fault classifier is proposed. Firstly, the data of three-dimensional data containing lot information is expanded, PCA model is established for each phase of the batch process and the process of fault monitoring is performed. Then, the data of partial length sampling time after the fault occurrence time is selected to extract the fault feature , The LSSVM fault model is established offline. Finally, online fault diagnosis is carried out by the fault classifier to classify the fault and determine that some kind of fault has occurred. The method improves the real-time performance and accuracy of on-line fault diagnosis of batch process. The feasibility and effectiveness of the proposed method are further verified by the application of penicillin fermentation simulation.