Recovery of Control-Loop Oscillations in Industrial Time Series with Missing Values

Recovery of Control-Loop Oscillations in Industrial Time Series with Missing Values

연구 분야: Infrastructure

학회: International Conference on Cyberspace Simulation and Evaluation

Poor control performance in industrial processes is a pervasive and escalating issue, with oscillation phenomena being the primary form of control performance deterioration. Therefore, accurate recovery of oscillations is crucial for process monitoring, as it ensures process stability, enhances product quality, and maximizes operational efficiency. However, real-world industrial data often suffer from various issues, with data loss being particularly prevalent. This loss compromises critical oscillation information, such as frequency, amplitude, and phase, thereby hampering oscillation detection and controller performance assessment. Despite the significance of this issue, research on handling missing values in process monitoring remains limited. To address this challenge, this work proposes a masked optimized decomposition algorithm (MOD). The algorithm leverages a mask to impute missing values and recovers oscillations by minimizing the loss of each component during signal decomposition. We conducted performance comparison study through simulations and industrial example. The results demonstrate that the proposed work is a promising too for control-loop oscillation data preprocessing.