Poster Session

Detection and Diagnosis of Plant-Wide Oscillations using the Spectral Envelope Method

Hailei Jiang, PhD student in Chemical Engineering (Process Control)

Supervisor - Dr. Sirish Shah and Dr. Biao Huang

Oscillations are a common type plant-wide disturbance in process industry and often observed due to a variety of reasons: sticky valves, etc. In a real industrial setting, oscillations generally have a source and then propagate through the whole process, and hence a number of the measurements often have same frequency oscillations. It is important to detect the oscillations and categorize the measurements based on the frequencies they exhibit. In this presentation, a multivariate statistical method called spectral envelope is successfully used to detect and categorize the oscillations of a set of industrial process data. This spectral envelope method is introduced as a useful tool in frequency domain to solve problems in process industry.