Abstract: Oscillations are a common type of plant-wide disturbances in chemical process, which may impact overall process performance.Normally, the fault diagnosis methods only based on the amplitude domain of data have a poor performance on these problems.The method introduced, time delay analysis, based on the time domain knowledge of data, could establish the fault propagation path and identify the root cause of the oscillation.The innovation of this paper was to combine the Hopfield network and time delay analysis method and create a complete fault isolation strategy.The approach was applied to the Tennessee Eastman (TE) model and the result proved its good quality for industry application.

Key words: time delay analysis, Hopfield network, traveling salesman problem, fault diagnosis

摘要： 振荡是化工过程中常见的对全流程运行性能有显著影响的故障类型,仅基于数据幅值域知识的故障诊断方法对这一类故障诊断性能不佳。时滞分析基于数据信号时域知识,根据波形相关性分析变量之间因果关系,通过得到的因果模型确定故障完整传播路径,可进一步识别出扰动发生的根本原因。将Hopfield网络与时滞分析相结合,解决了时滞分析当变量数众多时,从变量对的因果关系难以得到故障传播路径的问题,并同时讨论了时滞分析数据窗选取、对称时滞确立等的原则,提升了故障传播路径建立的准确度,建立了基于时滞分析的完备的故障诊断策略,最后通过TE模型验证了方法的优越性。

关键词: 时滞分析, Hopfield网络, 旅行商问题, 故障诊断