Root-cause Identification Of Plant-wide Oscillations Based On PDC

Process industry is not only the producers of energy and all kinds of raw materials, but also a main consumer of energy and resources. Energy saving and lower consumption are the effective way of raising economic benefits in enterprises and promoting the low carbon development strategy of the nation. Because of some faults such as poor controller tuning in control-loops, control valve stiction, process nonlinearity, poor process and control system design, external input disturbance and other reasons, the plant-wide oscillations occur regularly in process industries. Estimating the performance of the multi-loop control system, detecting the abnormal process conditions and then accurately locating and diagnosing the root-cause of oscillations are becoming more and more necessary now. When the root cause of oscillations are detected and located, the maintenance of fault source can make great sense to imporve enterprise product quality and production efficiency.In this paper, a new method used for identifying the propagation path of plant-wide oscillations and locating the root-cause is proposed. The partial directed coherence analysis(PDC) has been integrated with the traditional methods of fault diagnosis, and a great amount of historical process data which are sampled and collected in the distributed control system(DCS) are utilized and analysed. Based on the traditional power spectrum and spectral independent component analysis (spectral ICA) algorithm, the process variables of oscillatory control loops with primary energy and the dominant oscillation frequency are extracted, and then the selected data sets are analyzed by partial directed coherence algorithm. The direction of oscillations propagation and relationship intensity between different controlled variables are identified and visualized in the form of the PDC causality digram. Integrated with a proposed weight threshold search algorithm based on depth first search/breadth first search(DFS/BFS) and the understanding of process knowledge, the PDC causality diagram is simplified, and then the root-cause of the oscillations can be intuitively identified.With the proposed method, simulation and industrial history data sets are used to diagnose the root-cause of oscillations, the diagnosis results demonstrate the feasibility of the proposed method.