Multi‐Parameter Correlation Dynamic Warning and Abnormal Parameter Identification Method for Equipment Fault Diagnosis Under Complex Working Conditions

Industrial machinery typically exhibits complex structures and variable operating conditions, rendering single‐parameter monitoring insufficient for comprehensive status assessment. Conventional fixed threshold alarm systems often suffer from increased false negatives. This paper proposes a multi‐parameter dynamic early warning and abnormal parameter identification method based on multivariate state estimation (MSET) to address issues such as false alarms, missed alarms, and abnormal parameter localization. The methodology comprises two principal stages: offline parameter optimization utilizing grey relational analysis to identify condition‐sensitive variables, followed by construction of a state memory matrix from these parameters; and online implementation of a dynamic warning model incorporating alert indicators with adaptive threshold modules. The sliding‐window based dynamic threshold enables real‐time equipment status evaluation through computed warning indicators, while cumulative anomaly contribution rate calculations facilitate precise parameter localization. The experimental results demonstrate that the proposed method emits a warning 196 min earlier compared to conventional approaches, enabling the early identification of abnormalities in oil pumps.