基于奇异值分解的铅锌冶炼过程故障二次分离

Pb-Zn Smelting Process Fault Isolation Based on Singular Value Decomposition

铅锌冶炼过程是非常复杂的反应工艺,工况变化较大,在线监控和故障诊断的难度很大。铅锌冶炼过程故障分离的目的是根据异常监控变量来判断故障类型,由于铅锌冶炼过程故障的特殊性,不同类型的故障可能导致相同的监控变量发生异常,采用传统的故障分离方法不能有效地判断故障类型。本文提出一种基于奇异值分解的故障二次分离的方法,通过进一步分析监控变量和故障类型的对应关系,解决故障分离不彻底的问题。试验结果证明了方法的有效性。

Because Pb-Zn smelting process is accompanied with fast changes and remarkable sophisticated characteristics, it is hard to on-line monitor it and diagnose its fault. The objective of fault isolation is to determine the fault types based on the abnormal surveillance variables. However the fault isolation can not be effectively carried out because of the specificity of Pb-Zn smelting process that different faults may cause the same surveillance variables to be abnormal. To solve this problem, singular value decomposition （SVD）-based 2-step fault isolation approach is proposed to achieve effective fault isolation in Pb-Zn smelting process by further .analyzing the relationships between fault types and abnormal variables. The feasibility of the proposed method is testified by experiments.