基于相对化变换的信息增量矩阵故障诊断方法

Fault diagnosis based on relative transformation of information increment matrix

已有的信息增量矩阵故障诊断方法,虽然能有效地克服传统主元分析方法因存在严重的模式复合效应而不能辨识故障的不足,但是其自身忽略了量纲的影响,常造成在一些系统中起重要作用的变量因其自身绝对值较小而不能检测出绝对值更小的故障,而这些重要变量的很小故障通常又对系统的安全和稳定起着非常关键的作用。依据各变量在实际生产过程中的重要程度,在对系统进行相对化变换的基础上,提出一种改进的信息增量矩阵的故障诊断方法。通过仿真验证了新方法的有效性。研究结果表明:该方法能有效检测出原有故障,同时,又能检测出原系统一些虽然较小但却起重要作用的变量所发生的故障。

The existing information incremental matrix fault diagnosis methods can effectively overcome the shortcomings that traditional principal component analysis method cannot identify fault because of its serious pattern composition effects,but it ignores the impact of differences in scale for system variables,which often induces the important variables with small absolute value undetected when faults with smaller absolute values happened.While the small fault usually plays a key role for the security and stability of the system.Thus,according to the different important levels of different variables in the actual production process,on the basis of the relative transformation,a modified information incremental matrix fault diagnosis method was proposed.The simulation results show the effectiveness of the proposed method.The results show that this new method not only can keep the same detectability as the original system,but also can detect the fault of some small but important variables.