基于收敛交叉映射的化工过程扰动因果分析方法

摘要

When there is a fault in the chemical plant, the operators may not be able to find the root cause of the fault. Data-driven method can be a great help to reduce the uncertainty of root cause, even to locate the root cause. By analysis of chemical process data, the disturbance propagation way can be detected, which can help to locate the root cause. In this article, convergent cross mapping (CCM) with the characteristic of time is proposed to detect the causality in dynamic chemical process. Furthermore, the usage of Akaike information criterion is proposed to determine the most proper embedding dimension. To prove the effectiveness of the method, the new method is applied to the ecosystem examples, causality testing benchmark model and CSTR model. Comparing the result calculated by the original CCM, the effectiveness of the new method is found.%当化工装置出现异常情况，操作工人往往无法及时准确地定位故障发生的原因。基于数据的方法能够通过化工过程数据，分析异常工况中的扰动传播路径，确定异常工况出现的根本原因。针对化工动态系统，提出了具有时间特性的收敛交叉映射方法(CCM)，和基于赤池信息准则的维度选择方法。为了验证提出算法的有效性，在简单的生态系统，因果检测基准系统和全混釜反应器（CSTR）中进行验证，并与原有的收敛交叉映射算法进行对比，体现出具有时间特性的收敛交叉映射算法的优越性。