Valve Stiction Detection via Median Ensemble Empirical Mode Decomposition: A Robust Framework

Valve stiction is a prevalent cause of oscillations in control systems, posing significant risks to the safety and stability of production processes. While decomposition-based methods for detecting valve stiction have gained popularity due to their operational independence from prior knowledge, they frequently suffer from mode mixing issues that undermine the accuracy and robustness of stiction detection. Motivated by the above challenge, this work proposes a robust framework for stiction detection based on median ensemble empirical mode decomposition (MEEMD). The framework leverages the robust performance of MEEMD in handling mode mixing, enabling precise extraction of oscillatory components from complex system signals. Stiction detection is then accomplished by identifying the fundamental and harmonic relationships among these components. Simulation experiments, along with results from real industrial cases, demonstrate that the proposed framework effectively identifies stiction-induced oscillations, thereby ensuring a reliable operation of the system.