Oscillations are a common abnormal phenomenon in the process control system. They can degrade control performance or even cause plant shutdown. It is crucial to accurately detect and characterize the process oscillations. In this paper, a novel oscillation detector is proposed by combining the particle swarm optimization (PSO) and nonlinear chirp mode decomposition (NCMD). Because the performance of NCMD relies on the selection of mode number Q and bandwidth parameter α, PSO is utilized to search the optimal parameter pairs. Then, the multiple oscillations contained in the process variables (PV) can be extracted by NCMD with the optimal parameters. The normalized correlation index and sparseness index are used to discarding the spurious modes and quantifying the degree of oscillations, respectively. After detecting, by utilizing the time-frequency information provided by the oscillating modes, multiple oscillation types can be accurately characterized. Comparisons are provided to show the advantages of the proposed method. The effectiveness and utility are validated by simulations as well as various industrial cases. INDEX TERMS Oscillation detection, nonlinear chirp mode decomposition, particle swarm optimization, control performance assessment.

No abstract

Nonlinearity-caused oscillations are a frequent issue in process control systems. Its incidence degrades the product quality, stability and safety of the plant. Therefore, it is important to detect and analyze the nonlinearity-caused oscillations to maintain the control performance. In this study, we propose a novel oscillation detection and analyze method based on an improved variational nonlinear chirp mode decomposition (VNCMD) algorithm. Specifically, the original VNCMD needs to manually set the mode number in advance, which is a challenging task in practice. To tackle this problem, an improved VNCMD (IVNCMD) is proposed by utilizing the approximate entropy of instantaneous frequency. Then a novel IVNCMD-based detector is developed to detect and analyze the nonlinearity-induced oscillations by revealing the harmonic content of process variable. Besides detecting the nonlinearity problem, the IVNCMD-based method can contribute in locating the root cause for nonlinearity-caused unit-wide oscillations. The proposed method is model-free and data-driven thus requiring no prior knowledge about the process dynamics. Compared with the latest related methods, the proposed method is able to process nonstationary oscillations and providing corresponding time-frequency information. The effectiveness and advantages are demonstrated through simulations as well as industrial applications.