基于重构的故障分离方法能够抑制残差污染影响，有效降低误诊率，但该类方法计算量会随系统维度和故障变量数量呈指数级上升，难以直接应用于高维复杂工业过程的在线故障诊断。因此，提出的主成分分析法是一种基于序列特征选择算法的重构主成分分析故障诊断方法，该方法基于历史数据建立主成分分析监测模型，利用综合指标对实时数据进行故障检测，在故障分离过程中引入序列特征选择方法来定位故障变量，并〖JP2〗采用数学仿真算例和实际工程算例对该方法的诊断性能进行验证。结果表明：所提方法可以在较小计算量的情况下保证高诊出率和低误诊率，在诊断精度和诊断效率之间达到良好平衡，能够有效处理高维系统复杂故障，满足了在线诊断需求。

The reconstructionbased fault isolation method can suppress the influence of the smear effect and effectively reduce the false alarm rate. However, the computational cost of these methods will increase exponentially with the system dimension as well as the number of fault variables, making it challenging to directly apply in the realtime fault diagnosis of highdimensional complex industrial process. Therefore, a new fault diagnosis method integrating the sequential floating forward selection aided reconstructionbased principal component analysis (SFFSRBPCA) was proposed, the PCA monitoring model was established based on historical samples, and the combined index was used to detect the faults of realtime data. Then, the sequence feature selection method was introduced to locate fault variables in the fault isolation process. Furthermore, a simulation example and a practical industrial case were employed to verify the diagnostic performance of the proposed method. The results show that the proposed method can ensure a high fault detection rate and a low false alarm rate by consuming a small amount of computation, achieving a good balance between diagnostic accuracy and diagnostic efficiency. The proposed method can effectively deal with the complex faults of highdimensional systems and meet the online diagnosis requirements.

TK221

“十四五”国家重点研发项目(2022YFB4100702)