Abstract:

In chemical processes, fault propagation pathways and root cause identification could be discovered though analysis of interactions and time delay relationships among different process variables. Because of its importance in improving process safety and operation profit, fault discovery has been a popular and challenging research topic. Common methods such as correlation and entropy transfer functions, which usually cannot get accurate time delay and interaction strength between variables by limited applicability for linear and weak nonlinear systems or high computation demand, have experienced many disadvantages in actual application. Recently, a new convergent cross mapping(CCM) algorithm in ecology has been considered suitable for causality analysis and time delay identification for nonlinear coupling process variables. However, CCM fails to find application in externally disturbed chemical processes because it cannot establish stable embedded flow from process data. An improved CCM, disturbance filtered cross mapping method(DFCM), overcame many challenges of creating stable embedded flow by analyzing external disturbance, filtering disturbed process data, and applying filtrated data to CCM calculation. Case studies showed good results of time delays and causality analysis, thus DFCM could be applied to chemical processes under external disturbance.

Key words: CCM, causality analysis, time-delay detection, algorithm, process system, safety

摘要：

在化工过程中，可以通过分析变量间的相互作用和时滞关系，推理故障的传播路径和网络，指出故障的根原因。这对提高过程安全性，增强经济效益具有重要意义，是研究的热点和难点。常见的互相关函数和传递熵等方法，由于只适用于线性或弱非线性系统，或计算量较大，往往无法准确地获得变量间的时滞信息和作用强度，在实际应用中存在不足。近年来，在生态领域研究中提出的交叉收敛映射（CCM）算法，被认为是一种适用于非线性耦合过程因果分析的方法，可适用于耦合变量间时滞关系的检测。但对于带有外部扰动的化工过程，CCM无法根据随时受到扰动的过程数据构造出稳定的嵌入流形，导致了时滞和因果分析失败。而基于CCM进行改进的扰动过滤交叉映射（DFCM）方法，通过分析外部扰动对系统的影响，预先筛选扰动数据，再将过滤后的数据代入交叉映射的计算中。算例表明，这种处理能有效地避免扰动下嵌入流形不稳定的问题，适用于处于扰动下的化工过程，并得到良好的时滞和因果关系分析效果。

关键词: CCM, 因果分析, 时滞检测, 算法, 过程系统, 安全